KR20090005998A - Systems and methods of generating diagnostic images for structural health monitoring - Google Patents

Abstract

A system and a method for producing a diagnosis image for structure soundness monitoring are provided to determine the structure state reliably by integrating calculated tomography algorithm to the index of different structure state. A method for producing a diagnosis image for structure soundness monitoring comprises a step of acquiring a plurality of damage indices about the network which is combined in a main structure and has a plurality of diagnosis network patches(DNP), where each patch can operate as at least one of transmitter patch and sensor patch, a step of producing damage index distribution about the surface by using the obtained damage indices, and a step of formatting distribution into one or more tomographic images.

Description

SYSTEM AND METHOD FOR GENERATION OF DIAGNOSTIC IMAGES FOR SUSTAINABILITY SURVEY

The present invention relates to diagnostics of structures, and in particular to methods of monitoring the health of structures. This application is a partial application of US Application No. 10 / 942,714 entitled "Method for Monitoring the Health of Structures," filed September 16, 2004, which was filed on September 22, 2003. Claims the benefit of US provisional application 60 / 505,120 entitled “Sensors and Systems for Integrity Monitoring of Structures,” which is hereby incorporated by reference in its entirety.

All structures in use (bridges, aviation, space, unmanned aerial vehicles, refineries / chemical installations, ships, vehicles, high-rises, etc.) require proper inspection and maintenance to prolong their life or to prevent accidental breakage. The integrity and integrity of the structure should be monitored. It is clear that the health monitoring of structures has become an important topic in recent years. To date, numerous inspection methods have been used, including conventional visual inspection and non-destructive methods such as ultrasound and eddy current scanning, acoustic emission and X-ray inspection to identify defects or damage to the structure. In these conventional inspection methods, it is necessary to at least temporarily detach the structure from the state of use for inspection. Although the conventional methods are still used for the monitoring of isolated sites, they require a lot of time and money.

As sensor technology advances, new diagnostic techniques for monitoring the complete state of the structure have been significantly improved. Typically, these new diagnostic techniques utilize sensor systems consisting of suitable sensors and actuators mounted on the main structure. However, these problem solutions have several drawbacks and do not provide an effective online diagnostic method that provides a reliable sensor network system and / or precision monitoring method that can diagnose, classify and predict the condition of a structure with minimal manpower. For example, U. S. Patent No. 5,814, 729 to Wu et al. Discloses a method for detecting a change in attenuation characteristics of vibration waves in a structure to find a plate-shaped crack region in a laminated composite structure in the structure. A piezoceramic device is used as the actuator for generating the vibration wave, and an optical fiber cable having different grating locations is used as a sensor for receiving the wave signal. The drawback of this system is that it cannot accommodate multiple actuator arrays, so each actuator and sensor must be mounted separately. The defect detection is based on a change in the vibration wave traveling along the direct path between the actuator and the sensor, so this detection method is present around the boundary of the defect and / or structure existing outside the direct path. The defect cannot be detected.

Other defect detection methods can be found in US Pat. No. 5,184,516 to Blazic et al. This patent discloses an embedded conformal circuit for the health monitoring and evaluation of structures. This right angle circuit consists of a series of stacks of strain sensors, each of which measures strain changes at corresponding positions to identify defects in the right angle structure. The right angle circuit is a passive system, for example, does not have an actuator for signal issuance. A similar passive sensor network system can be found in US Pat. No. 6,399,939 to Mannur, J. et al. The patent discloses that a piezoceramic-fiber sensor system has planner fibers embedded in a composite structure. The drawback of these passive methods is the inability to monitor the delamination and damage between the sensors. In addition, these methods can detect the state of the main structure only in the local region where the embedded circuit and the piezoelectric fiber are attached.

US Pat. No. 6,370,964 to Chang et al. Discloses one method for detecting damage in a structure. This patent discloses a sensor network layer called the Stanford Multi-Actuator-Receiver Transduction (SMART) Layer. The SMART ^layer® includes a piezoceramic sensor / actuator in which a piezoceramic sensor / actuator (or simply, piezoceramic) is disposed at equal intervals, and the flexible dielectric film is bonded to the piezoceramic sensor / actuator via the piezoceramic sensor / actuator. The actuator generates acoustic waves, and the sensor receives the sound waves and converts them into electrical signals. To connect the piezoceramic to an electronic device, the plated wire is etched using conventional flexible circuit techniques and laminated between a plurality of circuit boards. As a result, a significant amount of flexible circuit board area is required to cover the plated wire area. In addition, the SMART layer ^® should be adhered to the main structure made of a laminated composite layer. Due to the internal stress caused by the high temperature cycle during the fixing process, the piezoceramic in the SMART layer ^® can cause microdestruction. In addition, the substrate of the SMART layer ^® can be easily separated from the main structure. In addition, since the SMART layer ^® is very difficult to insert or attach to the main structure having the fastening part, the plating wire is easily bent due to the compression load applied to the fastening part. Broken piezoelectric ceramics and bent wires are susceptible to electromagnetic interference noise and can cause induction of electrical signals. In heavy conditions, such as heat stress, battlefield shock and vibration the SMART Layer may not have ^® it does not strongly reliability as a tool for monitoring structural health. It is also expensive because the main structure must be dismantled when replacing a damaged and / or defective actuator / sensor.

Defect detection of other structures is disclosed in US Pat. No. 6,396,262 to Light et al. This patent discloses a magnetostrictive sensor for investigating damage to the structure, the sensor having a ferromagnetic strip and a coil disposed proximate the strip. The main drawback of this system is that internal damage between the sensor cannot be detected because it cannot be designed to accommodate the sensor array.

Due to the above disadvantages, the data analysis methodology used in the conventional monitoring system has a limitation in accurately and efficiently monitoring the main structure. Therefore, a new and efficient methodology for analyzing and interpreting data obtained from the main structure system is needed for the determination of the state of the structure and the prediction of failure.

Accordingly, it is an object of the present invention to provide an accurate method for determining the state of a structure using different methods such as dividing, crossing and adaptive neural fuzzy-inference positioning of network paths. It is for. This method is integrated with convex-set interpolation.

Another object of the present invention is to provide a reliable method for determining the structure state by incorporating a computed tomography algorithm for indices of different structure states.

It is still another object of the present invention to provide a method for analyzing a structure state using a hyperspectral tomography cube and a structure state manifold.

Another object of the present invention is to provide a technique for classifying structure states using a codebook-template based classifier. This method is integrated with multilayer perception for tomography of the surface.

It is still another object of the present invention to provide a prediction method for predicting a structural state by modeling a diagnostic network system and updating its parameters. This method is integrated with system identification and surveillance learning algorithms.

These and other objects and effects are achieved by the structural health monitoring software, which includes interrogation, processing, classification and prediction modules and analysis data from diagnostic network patch (DNP) systems attached to the main composite structure and / or main metal structure. Is achieved. The DNP system includes a plurality of actuators / sensors, and provides an internal wave-ray communication network in the main structure by transmitting acoustic wave pulses (or equivalent ram waves) between the plurality of actuators / sensors. .

According to one aspect of the invention, a computer-implemented method for generating tomographic images for the health monitoring of the structure, the method comprising obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNP), wherein Each of the patches may operate as at least one of a transmitter patch and a sensor patch, wherein the damage index value is an amount affected by damage in the main structure; Generating a distribution of damage index values for a surface using the obtained damage index values; And formatting the distribution as at least one tomographic image using a computer process.

According to another aspect of the present invention, a computer-implemented method for generating tomographic images for monitoring the health of a structure includes obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNPs), Wherein each of the patches can act as at least one of a transmitter patch and a sensor patch, each of the damage index values being a signal generated by one of the patches in response to an impact applied to a main structure of the network. Associated with; Generating a distribution of damage index values for a surface using the obtained damage index values; And formatting the distribution as at least one tomographic image using a computer process.

According to another aspect of the invention, a computer readable medium for executing one or more sequence instructions for the health monitoring of a structure, the execution of one or more sequence instructions by one or more processors is performed by the one or more processors, a plurality of diagnostic networks Acquiring a plurality of damage index values for a network having a patch (DNP), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch, each of the damage index values The amount to be affected by damage in the main structure of the network; Generating a distribution of damage index values for a surface using the obtained damage index values; And formatting the distribution as at least one tomographic image using a computer process.

According to another aspect of the invention, a computer readable medium for executing one or more sequence instructions for generating a tomographic image for the health monitoring of a structure, wherein execution of the one or more sequence instructions by one or more processors is performed by the one or more processors. Shiji, obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNP), wherein each of the patches can act as at least one of a transmitter patch and a sensor patch, the damage Each of the index values is associated with a signal generated by one of the patches in response to an impact on a main structure of the network; Generating a distribution of damage index values for a surface using the obtained damage index values; And formatting the distribution as at least one tomographic image using a computer process.

According to another aspect of the invention, a system for generating tomographic images for the health monitoring of a structure is coupled to a main structure, the network having a plurality of diagnostic network patches (DNP), wherein each of the patches is a transmitter Can operate as at least one of a patch and a sensor patch; Means for obtaining a plurality of corruption index values for the network; Means for generating a distribution of damage index values for a surface using the obtained damage index values; And means for formatting the distribution as at least one tomographic image using a computer process.

According to the configuration of the present invention, since the data obtained from the main structure system can be easily and conveniently analyzed and interpreted for the determination of the state of the structure and the prediction of failure, the structure can be accurately and efficiently monitored. have.

Advantages and features of the present invention as described above will be apparent to those skilled in the art based on the following detailed description of the present invention.

The following description includes many details for purposes of illustration, but those skilled in the art will recognize that many other changes and modifications to the details below are possible within the scope of the invention. Accordingly, the following embodiments of the present invention are presented without loss of the generality of the claims and without limiting the claims.

The above publications are provided for the purpose of the present application prior to the filing date of the present application, and the present invention is not to be construed that the present invention is not qualified by the prior art. Also, the release dates provided may differ from the actual release dates, which need to be verified independently.

1A is a schematic partial ablation plan view of a patch sensor 100 according to an embodiment of the present invention. FIG. 1B is a schematic cross-sectional view of the patch sensor 100 taken along the A-A direction of FIG. 1A. As shown in FIGS. 1A-1B, the patch sensor 100 includes: a substrate 102 configured to be attached to a main structure; Hoop layer 104; Piezoelectric devices 108 for generating and / or receiving signals (especially Lamb waves); A buffer layer 110 for providing mechanical impedance matching while reducing thermal stress mismatch between the substrate 102 and the piezoelectric device 108; Two electric wires 118a-b connected to the piezoelectric device 108; A molding layer 120 for fixing the piezoelectric device 108 to the substrate 102; The coating layer 106 for protecting and sealing the molding layer 120 is provided. The piezoelectric device 108 includes a piezoelectric layer 116, a lower conductor 112 connected to the electric wire 118b, and an upper conductor 114 connected to the electric wire 118a. The piezoelectric device 108 may operate as an actuator (or equivalently, a signal generator) when a predetermined electric signal is applied through the wires 118a-b. Upon application of an electrical signal, the piezoelectric layer 116 is deformed to produce a lamb wave. In addition, the piezoelectric device 108 includes a receiver for detecting a vibration signal, converting the vibration signal applied to the piezoelectric layer 116 into an electrical signal, and transmitting the electrical signal through the wires 118a-b. Can operate as

The substrate 102 may be attached to the main structure using a structural adhesive such as conventional casting agent thermosetting epoxy such as butyralthenolic, acrylic polyimide, nitriale phenolic or aramid. The substrate 102 may consist of an insulating layer against thermal and electromagnetic interference, protecting the piezoelectric device 108 attached thereto. In some cases, the dielectric substrate 102 needs to cope with a temperature of 250 ° C or higher. It is also possible to have a dielectric constant for minimizing signal transmission delay and crosstalk between the piezoelectric device 108 and the main structure, and high impedance for reducing power loss at high frequencies.

The substrate 102 may be made of various materials. Kapton ^® polyimide manufactured by DuPont (Wilmington, Delaware) is preferred for general use, and Teflon perfluoroalkoxy (PFA), poly p-xylylene (PPX), and polybenzimidazole (PBI) can be used for special purposes.

For example, PFA films can have good dielectric properties and low dielectric losses suitable for low voltages and high temperatures. PPX and PBI can provide stable dielectric strength at high temperatures.

The piezoelectric layer 116 may be made of piezoceramic, piezoelectric crystal, or piezoelectric polymer. Piezoelectric crystals, such as PZN-PT crystals manufactured by TRS Ceramics, Inc. of Pennsylvania State, are preferably employed in the design of piezoelectric device 108 due to its high strain energy density and low strain hysteresis. For small patch sensors, such as PZT ceramics manufactured by Fuji Ceramic Corporation, Tokyo, Japan, or APC International, Ltd., Mackeyville, Pennsylvania. Piezoceramic may be used for the piezoelectric layer 116. The upper and lower conductors 112 and 114 may be made of a metal such as chromium or gold, and may be attached to the piezoelectric layer 116 by a conventional sputtering process. In FIG. 1B the piezoelectric device 108 is shown having only a pair of conductors. However, it will be apparent to those skilled in the art that the piezoelectric device 108 may include a plurality of layers of conductors having various thicknesses for optimizing the performance of the piezoelectric layer 116 in the generation / detection of the signal wave. The thickness of each conductor can be determined by the thermal limits and mechanical loads applied to the particular main structure to which the patch sensor 100 is attached.

In order to sustain the temperature cycle, each layer of the piezoelectric device 108 needs to have a coefficient of thermal expansion similar to that of the other layers. In addition, the thermal expansion coefficient of the conventional polyimide including the substrate 102 is approximately 4-6 × 10 ⁻⁵ K ⁻¹ , and the thermal expansion coefficient of the conventional piezoelectric ceramic / crystal including the piezoelectric layer 116 is approximately 3. X 10 ^-6 K ^-1 . This mismatch in thermal expansion may be the maximum cause of failure of the piezoelectric device 108. If the piezoelectric device 108 is broken, the patch sensor 100 of the main structure should be replaced. As described above, the buffer layer 110 may be used to reduce the adverse effect of the mismatch in thermal expansion coefficient between the piezoelectric layer 116 and the substrate 102.

The buffer layer 110 is preferably made of a conductive polymer or metal, particularly aluminum having a thermal expansion coefficient of 2 × 10 ⁻⁵ K ⁻¹ . The buffer layer 110 may be replaced or added to one or more buffer layers made of alumina, silicon, or graphite. In one embodiment, the thickness of the aluminum buffer layer 110 is approximately equal to the thickness of the piezoelectric layer 116, which is about 0.25 mm thick, including two conductors 112 and 114, each about 0.05 mm thick. It can be done. In general, the thickness of the buffer layer 110 may be determined by the material properties and the thickness of the adjacent layer. The buffer layer 11 may provide durability and dual function for thermal load of the piezoelectric device 108. In other embodiments, the piezoelectric device 108 may stack another buffer layer on the upper conductor 114.

Another function of the buffer layer 110 is the amplification of the signal received by the substrate 102. As the lamb wave signal generated by the patch sensor 100 propagates along the main structure, the strength of the signal received by the other patch sensor 100 attached to the main structure decreases as the distance between the two patch sensors increases. do. When one RAM signal reaches the position where the patch sensor 100 is installed, the substrate 102 receives the signal. Next, the strength of the received signal may be amplified at a specific frequency depending on the material and thickness of the buffer layer 110. The piezoelectric device 108 then converts the amplified signal into an electrical signal.

Moisture, mobile ions, and poor environmental conditions may deteriorate the performance of the patch sensor 100 and reduce the lifetime, so that two layers of protective coating layers, that is, the molding layer 120 and the coating layer 106 may be used. . The molding layer 120 may be manufactured by a conventional manufacturing method using epoxy, polyimide, or silicon-polyimide. In addition, the molding layer 120 is made of low thermal expansion polyimide, and may be deposited on the piezoelectric device 108 and the substrate 102. Since the passivation of the molding layer 120 does not form a conformal hermetic seal, the coating layer 106 may be laminated on the molding layer 120 to seal the sealing layer. The coating layer 120 is made of a metal such as nickel, chromium or silver, and may be laminated by conventional techniques such as electrolysis or e-beam deposition and sputtering. In one embodiment, an epoxy film or polyimide film may be further coated on the coating layer 106 to provide a protective layer against scratching and cracking.

The hoop layer 104 may be made of a dielectric insulating material such as silicon nitride or glass, and surrounds the piezoelectric device 108 on the substrate 102 so that the conductive components of the piezoelectric device 108 are not electrically shorted. .

FIG. 1C is a schematic plan view of a piezoelectric device 130, which is a conventional type known in the art and may be substituted for the piezoelectric device 108. As shown in FIG. FIG. 1D is a schematic cross-sectional view of the piezoelectric device 130 taken along the B-B direction of FIG. 1C. As shown in FIGS. 1C-D, the piezoelectric device 130 includes a lower conductor 134; Piezoelectric layer 136; An upper conductor 132 connected to an electric wire 138b; A connecting member 142 connected to the electric wire 138a; And a conductive piece 144 for connecting the connecting member 142 to the lower conductor 134.

1E is a schematic partial ablation plan view of a patch sensor 150 according to another embodiment of the present invention. FIG. 1F is a schematic side cross-sectional view of the patch sensor 150 shown in FIG. 1E. As shown in Figure 1e-f, the patch sensor 150, the lower substrate 151; Upper substrate 152; Hoop layer 154; Piezoelectric device 156; Upper and lower buffer layers 160a-b; And two electric wires 158a-b connected to the piezoelectric device 108. The piezoelectric device 156 includes a piezoelectric layer 164; A lower conductor 166 connected to the electric wire 158b; And an upper conductor 162 connected to the electric wire 158a. The function and material of the patch sensor 150 are similar to those of the counterpart of the patch sensor 100. Each buffer layer 160a-b may include one or more lower layers, and each lower layer may be made of a polymer or a metal. The upper substrate 152 may be made of the same material as the material of the substrate 102.

The patch sensor 150 may be attached to the main structure to monitor the health of the structure. In addition, the patch sensor 150 may be inserted into the stack. 1G is a schematic cross-sectional view of a composite laminate 170 via a patch sensor 150. As shown in FIG. 1G, the main structure includes a plurality of stacks 172; And at least one patch sensor 150 fixed to the plurality of stacks 172. In one embodiment, the inside of the stack 172 may contain an adhesive material such as an epoxy resin before the curing process. During the curing process, the adhesive material from the stack 172 fills the cavity 174. In order to prevent accumulation of such an adhesive material, the hoop layer 154 may have a structure filling the cavity 174.

FIG. 1H is a schematic side cross-sectional view of another embodiment 180 of the patch sensor 150 of FIG. 1E. As shown, the patch sensor 180, the lower substrate 182, the upper substrate 184, the hoop layer 198; Piezoelectric device 190; Upper and lower buffer layers 192 and 194; And a piezoelectric device 196. For simplicity, a pair of wires connected to the piezoelectric device 190 is not shown in FIG. 1H. The piezoelectric device 190 may include a piezoelectric layer 196; Lower conductor 194; And an upper conductor 192. The function and material of the components of the patch sensor 180 are similar to those of the counterpart of the patch sensor 150.

The hoop layer 198 may have one or more lower layers 197 of various dimensions such that the outline of the outside thereof matches the shape of the cavity 174. The filling of the cavity 74 into the lower layer 197 prevents the adhesive material from accumulating during the curing treatment of the laminate 170.

2A is a schematic partial ablation plan view of a hybrid patch sensor 200 according to an embodiment of the present invention. FIG. 2B is a schematic cross-sectional view of the hybrid patch sensor 200 taken along the C-C direction of FIG. 2A. As shown in FIGS. 2A-B, the hybrid patch sensor 200 includes a substrate 202 configured to attach to a main structure; Hoop layer 204; Piezoelectric device 208; An optical fiber coil 210 having both ends 214a-b; Buffer layer 216; Two wires 212a-b connected to the piezoelectric device 208; Molding layer 228; And a coating layer 206. The piezoelectric device 208 includes a piezoelectric layer 222; A lower conductor 220 connected to the wire 212b; And an upper conductor 218 connected to the electric wire 212a. In other embodiments, the piezoelectric device 208 may be the same as the piezoelectric device 130 of FIG. 1C. The optical fiber coil 210, an optical fiber cable winding 224; And a coating layer 226. The components of the hybrid patch sensor 200 may be similar to those of the counterpart of the patch sensor 100.

The optical fiber coil 210 may be configured as a Sagnac interferometer to receive a lamb wave signal. The elastic strain on the surface of the main structure caused by the lamb wave can be added on the existing strain of the fiber optic cable 224 caused by bending and stretching. As a result, the frequency / phase change amount in the movement of light through the optical fiber cable 224 depends on the total length of the optical fiber cable 224. In one embodiment, the optical fiber coil 210 may be used as the main sensor and the piezoelectric device 208 may be used as the auxiliary sensor in consideration of being not affected by electromagnetic interference and vibration noise.

The optical fiber coil 210 uses the Doppler effect to the frequency of the motion of light through the optical fiber cable winding 224. In each of the optical fiber coils 210, the inside of the optical fiber winding is under compressive force and the outside is under tensile force. These compressive and tensile forces cause strain on the fiber optic cable 224. The vibrational displacement or strain of the main structure caused by the lamb wave is added to the strain of the fiber optic cable 224. According to the birefringence equation, the angle of reflection on the cladding surface of the fiber optic cable 224 is a function of the strain caused by the compressive and / or tensile forces. Thus, the inside and outside of each fiber winding produces a different angle of reflection than the inside and outside of the straight fiber, so that the frequency of light is dependent upon the relative flexural displacement of the lamb wave when light is transmitted through the fiber coil 210. Therefore, it is displaced from the center input frequency.

In one embodiment, the optical fiber coil 210 comprises 10 to 30 fiber optic cables 224 and has a minimum winding diameter 236 (di) of at least 10 mm. A gap 234 (dg) may exist between the innermost winding of the optical fiber coil 210 and the outer circumference of the piezoelectric device 208. This gap 234 depends on the minimum winding diameter 236 and the diameter 232 (dp) of the piezoelectric device 208 and about two or three times the diameter 230 (df) of the optical fiber cable 224. It is desirable that the dimension be larger than the diameter 232 by.

The coating layer 226 is preferably composed of a metal or polymer material, in particular epoxy, to increase the sensitivity of the optical fiber coil 210 to the deflection displacement or strain of the lamb wave guided by the main structure. In addition, a controlled tensile force may be added to the optical fiber cable 224 to add a tensile stress during the winding process of the optical fiber cable 224. The coating layer 226 may maintain the internal stress of the optical fiber cable winding 224 and allow uniform in-plane displacement with respect to the deflection displacement of the lamb wave for each optical fiber winding.

The coating layer 226 may be made of another material such as polyimide, aluminum, copper, gold or silver, and the thickness thereof may be in the range of about 30% to twice the diameter 230. The coating layer 226 formed of the polymer material may be formed in two ways. One embodiment is to install the optical fiber cable winding 224 on the substrate 202, and then sprayed by a device such as a biodot spay-coater, another embodiment is the optical fiber cable winding 224 is formed by dipping in a molten bath of coating material.

The coating layer 226 made of metal may be formed by a conventional metal coating method such as magnetron reaction sputtering or plasma sputtering as well as electrolysis. In particular, zinc oxide may be used as a coating material of the coating layer 226 to provide piezoelectric properties to the coating layer 226. Coating zinc oxide on the upper and lower sides of the optical fiber cable winding 224 causes the optical fiber coil 210 to contract or expand concentrically in a radial direction corresponding to the electrical signal. In addition, a silicon oxide or tantalum oxide coating material may also be used to control the refractive index of the optical fiber cable winding 224.

Silicon oxide or tantalum oxide can be deposited using direct / indirect ion beam deposition or electron beam vapor deposition. It is also known that the coating layer 226 can be formed on the optical fiber cable 224 using other methods without departing from the invention.

The piezoelectric device 208 and the optical fiber coil 210 may be attached to the substrate 202 using a physically solidified adhesive instead of the usual polymer. The physical coagulation adhesives include, but are not limited to, butyl acrylate-ethyl acrylate copolymer, styrene-butadiene-isoprene terpolymer, polyurethane alkyd resin, and the like. The adhesion properties of these materials can be kept constant during and after the coating process due to the lack of crosslinking of the polymer structure. In addition, these adhesives can be optimized for a wide range of wetting of the substrate 202 without sacrificing sensitivity to various analytes, compared to conventional polymers.

2C is a schematic partial ablation plan view of a hybrid patch sensor 240 according to another embodiment of the present invention. 2D is a schematic side cross-sectional view of the hybrid patch sensor 240 shown in FIG. 2C. As shown in Figure 2c-d, the hybrid patch sensor 240, the lower substrate 254; Upper substrate 242; Winding layer 244; Piezoelectric device 248; An optical fiber coil 246 having both ends 250a-b; Upper and lower buffer layers 260a-b; And two electric wires 252a-b connected to the piezoelectric device 248. The piezoelectric device 248 includes a piezoelectric layer 264; A lower conductor 262 connected to the wire 252b; And an upper conductor 266 connected to the electric wire 252a. The optical fiber coil 246 may include an optical fiber cable winding 258; And a coating layer 256. The components of the hybrid patch sensor 240 are similar to those of the counterpart of the hybrid patch sensor 200.

As in the case of the patch sensor 150, the hybrid patch sensor 240 can be attached to the main structure and / or embedded in the composite laminate. In one embodiment, the winding layer 244 may be configured similarly to the winding layer 198 to fill the cavity formed by the patch sensor 240 and the composite laminate.

3A is a schematic partial ablation plan view of an optical fiber patch sensor 300 according to an embodiment of the present invention. 3B is a schematic side cross-sectional view of the optical fiber patch sensor 300 taken along the D-D direction of FIG. 3A. As shown in FIGS. 3A-B, the optical fiber patch sensor 300 includes a substrate 302; Hoop layer 304; An optical fiber coil 308 having both ends 310a-b; Molding layer 316; And a coating layer 306. The optical fiber coil 308, an optical fiber cable winding 312; And a coating layer 314. The material and function of each element of the optical fiber patch sensor 300 is similar to that of the counterpart of the hybrid patch sensor 200 of FIG. 2A. The diameter 313 of the innermost winding of the fiber optic cable 312 can be determined by the material properties of the fiber optic cable 312.

FIG. 3C is a schematic partial ablation plan view of an optical fiber coil 308 housed within the optical fiber patch sensor of FIG. 3A, illustrating a method of winding the optical fiber cable 312. As shown in FIG. 3C, the outermost winding of the optical fiber coil 308 starts at one end 310a and the innermost winding ends at the other end 310b. 3D is a schematic partial ablation plan view of another embodiment 318 of the optical fiber coil 308 shown in FIG. 3C. As shown in FIG. 3D, the optical fiber cable 322 is folded and wound such that its outermost windings begin to form at both ends 320a-b. The optical fiber cable winding 322 may be coated with a coating layer 319.

The optical fiber coils 308 and 318 shown in Figs. 3C-D can be directly attached to the main structure and used as optical fiber coil sensors. For this reason, the terms "optical fiber coil" and "optical fiber coil sensor" will be used synonymously below. 3E-F illustrate another embodiment of the fiber coil 308. As shown in FIG. 3E, the optical fiber coil 330 includes an optical fiber cable 334 having both ends 338a-b and wound in the same form as the cable 312; And a coating layer 332. The coil 330 may include a hole 336 for receiving a fastener, which will be described later. Similarly, the optical fiber coil 340 of FIG. 3F includes an optical fiber cable 344 having both ends 348A-B and wound in the same form as the cable 322; And a coating layer 342. The coil 340 may include a hole 346 for receiving a fastener. 3G is a schematic side cross-sectional view of the optical fiber coil 330 taken along the D-D direction of FIG. 3E.

It should be noted that the sensor shown in FIGS. 3A-G can be embedded in a stack in a form similar to that shown in FIG. 1G.

4A is a schematic partial ablation plan view of a diagnostic patch washer 400 in accordance with one embodiment of the present invention. 4B is a schematic side cross-sectional view of the diagnostic patch sensor 400 taken along the E-E direction of FIG. 4A. As shown in FIGS. 4A-B, the diagnostic patch sensor 400 includes: an optical fiber coil 404 having both ends 410a-b; Piezoelectric device 406; A support element 402 for bonding and receiving the optical fiber coil 404 and the piezoelectric device 406 with an adhesive material; A pair of wires 408a-b connected to the piezoelectric device 406; And a covering disk 414 configured to cover the optical fiber coil 404 and the piezoelectric device 406.

The materials and functions of the optical fiber coil 404 and the piezoelectric device 406 are similar to those of the optical fiber coil 210 and the piezoelectric device 208 of the hybrid patch sensor 200. In one embodiment, the piezoelectric device 406 is similar to the piezoelectric device 130 except that there is no hole 403. The optical fiber coil 404 and the piezoelectric device 406 may be attached to the support element 402 using a conventional epoxy. The support element 402 may have a notch 412 through which both ends 410a-b of the optical fiber coil 404 and a pair of wires 408a-b may pass.

In FIGS. 4A-B, the diagnostic patch sensor 400 can act as an actuator / sensor and can include an optical fiber coil 404 and a piezoelectric device 406. In other embodiments, the diagnostic patch washer 400 may act as a sensor, and may include only an optical fiber coil 404. In other embodiments, the diagnostic patch washer 400 may act as an actuator / sensor and may include only a piezoelectric device 406.

As shown in FIGS. 4A-B, the diagnostic patch washer 400 may have a cavity 403 for receiving another fastener such as a bolt or rivet. 4C is a schematic diagram of an exemplary bolted structure 420 using a diagnostic patch washer 400 in accordance with one embodiment of the present invention. In the bolted structure 420, a pair of structures 422a-b such as plates may be fixed using a conventional bolt 424, a nut 426 and a washer 428. It is well known that the stress of the structure tends to concentrate on the bolted portion 429 and damage the structure. The diagnostic patch sensor 400 may be embedded in the bolted structure 420 and used to detect such damage.

4D is a schematic cross-sectional view of an exemplary bolted structure 430 using a diagnostic patch washer 400 in accordance with another embodiment of the present invention. In the bolted structure 430, the honeycomb / laminated structure 440 may be fixed using a conventional bolt 432, a nut 434, and a pair of washers 436 and 438. The honeycomb and laminate structure 440 includes a composite stack 422 and a honeycomb portion 448. In order to detect damage of the structure near the bolting region, a pair of diagnostic patch washers 400a-b may be inserted into the honeycomb portion 448, as shown in FIG. 4D. A sleeve 446 is required to support the top and bottom patch washers 400a-b against the composite stack 442. In addition, a heat shield annular disk 444 may be interposed between the composite stack 422 and the diagnostic patch washer 400b to protect the washer 400b from destructive heat transfer.

As shown in FIG. 4B, the covering disk 414 has the outer circumference 415 separating the optical fiber coil 404 and the piezoelectric device 406 from excessive contact load due to torque applied to the bolt 424 and the nut 426. It may have an angle of inclination to form a protective locking mechanism.

5A is a schematic diagram of a diagnostic system 500 having a sensor / actuator in accordance with one embodiment of the present invention. As shown in FIG. 5A, the diagnostic system 500 includes a sensor / actuator device 502 for generating and / or receiving a lamb wave signal; Two conductor wires 516; An adjusting device (508) for processing signals received by the device (502); An analog-to-digital (A / D) converter 504 for converting the analog signal into a digital signal; A computer 514 for managing the entire system 500; Amplifier 506; A waveform generator 510 for converting a digital signal into an analog lamb wave signal; And a relay switch module 512 configured to switch the connection state between the device 502 and the computer 514. In general, one or more devices 502 may be connected to a relay switch 512.

The device 502 can be configured with one of the sensors shown in FIGS. 1A-2D and 4A-D, including a piezoelectric device that generates a lamb wave 517 and simultaneously receives a lamb wave generated by another device. . To generate the lamb wave 517, the waveform generator 510 receives the digital signal of the waveform excited from the computer 514 (especially the analog output card included in the computer 514) via the relay switch module 512. Can be received. In one embodiment, the waveform generator 510 may be configured as an analog output card.

The relay switch module 512 may be configured as a conventional plug-in relay board. By means of a "cross-talks" linker between the actuators and the sensors, the relay switches included in the relay switch module 512 can cooperate with the computer 514 to select each relay switch in a particular order. have. In one embodiment, the analog signal generated by waveform generator 510 may be transmitted to other actuator (s) via branching wire 515.

The device 502 can function as a sensor for receiving Lamb waves. The received signal is sent to a regulator 508 that can adjust the signal voltage and filter the electrical noise to select a meaningful signal within the appropriate frequency band. The filtered signal is then sent to an analog-to-digital converter 504 that can be configured as a digital input card.

5B is a schematic diagram of a diagnostic system 520 with sensors in accordance with another embodiment of the present invention. The diagnostic system 520 includes a sensor 522 having an optical fiber coil; Optical fiber cable 525 for connection; A laser light source 528 for providing a carrier input signal; A pair of modulators 526 and 534; Acoustooptic modulator (AOM) 530; A pair of couplers 534 and 532; A photodetector 536 for detecting an optical signal transmitted through the optical fiber cable 525; A / D converter 538; Relay switch 540; And a computer 542. The sensor 522 may be configured as one of the sensors shown in FIGS. 2A-4D having an optical fiber coil. In one embodiment, coupler 524 may connect fiber optic cable 525 to another fiber 527, which may be connected to another sensor 523.

The sensor 522, in particular the fiber coil included in the sensor 522, can act as a laser Doppler velocitimeter (LDV). The laser light source 528, preferably a diode laser, may emit an input carrier optical signal to the modulator 526. The modulator 526 may be configured as a heterodyne modulator, and splits a carrier input signal into two signals, one signal for the sensor 522 and the other for the AOM 530. do. The sensor 522 displaces the input carrier signal by the Doppler frequency corresponding to the lamb wave and transmits it to the modulator 534 which can be configured as a heterodyne synchronizer. The modulator 534 may demodulate the transmitted light to remove the carrier frequency of the light. The photodetector 536, preferably a photodiode, converts the demodulated optical signal into an electrical signal. The A / D converter 538 then digitizes the electrical signal and transmits it to the computer 542 via the relay switch module 540. In one embodiment, the coupler 532 may connect an optical fiber cable 546 connected to another sensor 544.

6A is a schematic diagram of a diagnostic network patch system (DNP) 600 installed in a master structure 610 in accordance with one embodiment of the present invention. As shown in FIG. 6A, the system 600 includes a patch 602; Transmission link 612; At least one bridge box 604 connected to the transmission link 612; Data collection system 606; And a computer 608 for managing the DNP system 600. The patch 602 may consist of a device 502 or a sensor 522, where the format of the transmission link 612 may be determined by the type of patch 602, and may include wires, fiber optic cables, or both. It includes. Typically, the main structure 610 may be made of composite or metal.

The transmission link 612 terminates at the bridge box 604. The bridge box 604 connects a plurality of patches 602 to receive a signal from an external waveform generator 510 and transmits the received signal to an external A / D converter 504. The bridge box 604 can be connected via an electrical / optical cable and at the same time can accommodate an electronic regulator 508 for adjusting operating signals, filtering received signals, and converting optical fiber signals to electrical signals. Using the relay switch module 512, the data collection system 606 connected to the bridge box 604 relays a plurality of patches 602 and simultaneously receives signals received from the patches 602 in a predetermined order. Therefore, multiplex the channel.

It is well known that the generation and detection of lamb waves is influenced by the installation positions of actuators and sensors on the main structure. Thus, the patches 602 should be properly paired within the network structure to maximize the use of the lamb wave to identify damage.

6B is a schematic diagram of a diagnostic network patch system 620 having a strip network structure in accordance with one embodiment of the present invention. As shown in FIG. 6B, the patch system 620 may be attached to the main structure 621 and includes a patch 622; A bridge box 624 connected to the computer 626; And a transmission link 632. The plurality of patches 622 may be configured as a device 502 or a sensor 522, the format of the transmission link 632 may be determined by the format of the patch 622. The transmission link 632 may be configured of an electric wire, an optical fiber cable, or both.

The computer 626 may coordinate the operation of the patch 622, which may function as an actuator and / or a sensor. Arrow 630 indicates the propagation of the lamb wave generated by patch 622. In general, defects 628 in the main structure 621 may affect the transmission pattern in the form of wave scattering, diffraction, and transmission loss of the lamb wave. The defects 628 include damage, cracks, exfoliation of composite structures, and the like. The defect 628 can be monitored by detecting a change in the transmission pattern of the lamb wave captured by the patch 622.

The network structure of the DNP system is important in the health monitoring system of a lamb wave-based structure. In the network structure of the DNP system 620, the wave-ray communication path should be uniformly randomized. The uniformity of the communication path and the distance between the patches 622 can determine the minimum detectable dimension of the defect 628 in the main structure 621. An optimized network structure with an appropriate path structure can improve the confirmation accuracy of the damaged portion without increasing the number of patches 622.

Another structure for forming the crosstalk path between the patches may consist of the pentagonal network shown in FIG. 6C. 6C is a schematic diagram of a diagnostic network patch system 640 having a pentagonal network structure in accordance with another embodiment of the present invention. The system 640 may be installed in the main structure 652, and may include a patch 642; A bridge box 644 connected to the computer 646; And a transmission link 654. The plurality of patches 642 may be comprised of an apparatus 502 or a sensor 522. In the system 630, the patch 642 can detect the defect 650 by sending and receiving a lamb wave indicated by arrow 648.

6D is a schematic perspective view of a diagnostic network patch system 660 installed in rivet / bolt coupled composite laminates 666 and 668 according to another embodiment of the present invention. As shown in FIG. 6D, the system 660 includes a patch 662; And a diagnostic patch washer 664, each washer connected by a pair of bolts and nuts. In FIG. 6D, the bridge box and the transmission link are not shown for simplicity. The plurality of patches 662 may be comprised of an apparatus 502 or a sensor 522. In the system 660, the patch 662 and the diagnostic patch washer 664 can detect the defect 672 by transmission and reception of the lamb wave as indicated by arrow 670. Normally, a defect 672 occurs near the fastener hole. The diagnostic patch washer 664 may be in communication with another adjacent diagnostic patch 662 disposed in the strip network structure, as shown in FIG. 6D. In one embodiment, the optical fiber coil sensors 330 and 340 may be used as a substitute for the diagnostic patch washer 664.

6E is a schematic perspective view of a diagnostic network patch system 680 installed in a composite laminate 682 that may be repaired by an adhesive patch 686 in accordance with one embodiment of the present invention. As shown in FIG. 6E, the system 680 includes a plurality of patches 684, which may be configured as an apparatus 502 or a sensor 522. 6E does not show the bridge box and the transmission link for simplicity. In the system 680, the patch 684 detects a defect 688 located between the repair patch 686 and the composite stack 682 by transmitting and receiving the lamb wave indicated by the arrow 687. Can be.

6F is a schematic diagram illustrating one embodiment of a wireless data communication system 690 for controlling a network patch system for remote diagnostics in accordance with one embodiment of the present invention. As shown in FIG. 6F, the system 690 includes a bridge box 698; And a terrestrial communication system 694 that can be operated by the ground controller 692. The bridge box 698 may be connected to a diagnostic network patch system implemented on a host structure, such as an aircraft 696, which requires extensive structural health monitoring.

The bridge box 698 can operate in two ways. In one embodiment, the bridge box 698 may operate as a signal transmitter. In this embodiment, the bridge box 698 is an RF telemetry capable of transmitting micro miniature transducers and health monitoring information of the structure to the terrestrial communication system 694 via radio signals 693. It may have a microprocessor of the system. In another embodiment, the bridge box 698 may operate as a receiver of electromagnetic waves. In this embodiment, the bridge box 698 may have an assembly for receiving power from the terrestrial communication system 694 via a radio signal 693. The received power can be used to operate the DNP system installed in the structure 696. The assembly may comprise a micromachined silicon substrate having a stimulating electrode, a complementary metal oxide semiconductor (CMOS), a bipolar power regulation circuit, a hybrid chip capacitor, and a receiving antenna coil. have.

The structure of the bridge box 698 may be configured similarly to the outer layer of the main structure (696). In one embodiment, the bridge box 698 may have a plurality of honeycomb laminated structure. Here, a plurality of microstrip antennas are embedded in the outer surface of the honeycomb laminated structure of the plurality of layers, and the antennas operate as a right angle load antenna. The multi-layer honeycomb laminate has a honeycomb core and a multi-layer dielectric made of organic and / or inorganic materials such as e-glass / epoxy, kevlar / epoxy, graphite / epoxy, aluminum, or steel. A laminate can be provided. As the integrated micromachining technology rapidly develops, the dimensions and production costs of microstrip antennas can be further reduced, which can lead to a reduction in the operation / production costs of the bridge box 698 without any functional impairment.

The scope of the present invention is not limited to the use of standard Wireless Application Protocol (WAP) and wireless markup languages for structure sound wireless surveillance systems. By using the mobile Internet toolkit, the system provides a stable site that allows WAP-enabled cell phones, Pocket PCs with HTML browsers, or other HTML-enabled devices to accurately access structural status monitoring or infrastructure management. Can be rescued.

Just as the microphone array can be used to find the direction of a moving source, the sensor assembly array can be used to detect damage by measuring the difference in signal arrival time. 7A is a schematic diagram of a diagnostic network patch system 700 having sensor assemblies in a strip network structure in accordance with one embodiment of the present invention. As shown in FIG. 7A, the system 700 may be installed in the main structure 702 and includes a sensor assembly 704 and a transmission link 706. Each sensor assembly 704 has two receivers 708 and 712 and one actuator / receiver device 710. Each receiver 708 and 712 can be configured with one of the sensors shown in FIGS. 1A-4D, and the actuator / receiver device 710 is one of the sensors shown in FIGS. 1A-2D and 4A-D. A piezoelectric device for lamb wave generation can be provided. When the actuator / receiver 710 of the sensor assembly 704 transmits a lamb wave, the adjacent sensor assembly 704 uses all three elements, namely the actuator / receiver device 710 and the receivers 708 and 712. To receive a lamb wave. By using all three elements as the receiving unit, each sensor assembly 704 can receive a more refined lamb wave. In addition, by measuring the time difference of arrival between the three elements, the direction of the defect 714 can be extracted with high accuracy.

7B is a schematic diagram of a diagnostic network patch system 720 having sensor assemblies in a pentagonal network structure in accordance with another embodiment of the present invention. As shown in FIG. 7B, the system 720 can be installed in the main structure 722 to detect a defect 734 and includes a sensor assembly 724 and a transmission link 726. Each sensor assembly 724 is similar to the sensor assembly 704.

8A is a schematic diagram of a sensor assembly 800 having an optical fiber coil in series connection according to an embodiment of the present invention. The sensor assembly 800 is similar to the sensor assembly 704 of FIG. 7A and includes two sensors 804 and 808 and one actuator / sensor 806. In this structure, the input signal can enter the sensor via one end 810a, and the output signal from the other end 810b is the contribution of the input signal and the three sensors 804, 806 and 808. Can be the sum of. In one embodiment, the signal output from each sensor can be separated from other signals using a wavelength-based de-multiplex technique.

8B is a schematic diagram of a sensor assembly 820 having optical fiber coils connected in parallel in accordance with one embodiment of the present invention. The sensor assembly 820 is similar to the sensor assembly 704 of FIG. 7A and includes two sensors 824 and 828 and one actuator / sensor 826. In this structure, the input signal can enter the three sensors through the three ends 830a, 832a, and 834a, respectively, and the output signals from the other ends 830b, 832b, and 834b are input signals and three sensors. 824, 826, and 828 can be the sum of the contributions.

In FIGS. 8A-B, the sensors 804, 808, 824, and 828 are shown as an optical fiber coil sensor 308. However, for those skilled in the art, each sensor 804, 808, 824, and 828 can be configured as one of the sensors shown in FIGS. 1A-4D, and the intermediate sensors 806 and 826 are shown in FIGS. 1A-2D and 4A-D. It can be configured as one of the sensors, it is obvious that it has a piezoelectric device for generating a lamb wave. In addition, the sensor assemblies 800 and 820 may be interposed in the composite laminate in the same form as shown in FIG. 1G.

9 is a curve diagram 900 of actuator and sensor signals in accordance with one embodiment of the present invention. In order to generate the lamb wave, the actuator signal 904 may be applied to an actuator such as the patch sensor 100. The actuator signal 904 may be configured as a toneburst signal having a plurality of wave peaks having the highest amplitude in the middle of the waveform. The actuator signal 904 can be designed by using a Hanning function for various waveforms and has its intermediate frequency of 0.01 MHz to 1.0 MHz. The actuator receives the actuator signal 904 to generate a lamb wave having a specific excitation frequency.

Signals 912a-n represent sensor signals received by the plurality of sensors. As shown, each signal 912 is separated by wave packets 926, 928 separated by signal extracting windows (or signal extracting envelopes) 920, 922, and 924, respectively. And 930. These wave bundles 926, 928 and 930 may have different frequencies due to the dispersion mode at the position of the sensor. The signal separation window 916 is applied to identify a lamb wave signal from each sensor signal. The wave fluxes 926, 928, and 930 correspond to the basic symmetry mode S ₀ , the reflection mode S _{0 _ ref} , and the basic asymmetry mode A ₀ , respectively. The reflection mode S _{0_ref} represents lamb wave reflection from the boundary of the main structure. Basic shear modes S ₀ ′ and other higher modes can be observed. However, these are not shown in FIG. 9 for the sake of simplicity.

Portions 914 of sensor signal 912 are electrical noise due to toneburst actuator signal 904. In order to separate the portion 914 from the sensor signal 12, a masking window 918, which is a sigmoid function that is postponed during the operation time, is used as a threshold function. Can be applied to). Next, wave speeds 926, 928, and 930 may be extracted from the sensor signals of 912 using moving wave-envelope windows 920, 922, and 924 along the time series of each sensor signal. The envelope windows 920, 922, and 924 apply a hill-climbing algorithm that searches for peaks and valleys of the sensor signal 912, and interpolates the searched data points on the time axis. can be determined by interpolating. If the size of the nearest data point is smaller than the size of the current data point until the size comparison of the wave in the forward and backward direction is made continuously for all data points of the wave signal, the size of the data point in the wave signal And location can be stored. Once the envelope of the wave signal is obtained, each envelope is divided into sub-envelope windows 920, 922 and 924 at a time interval corresponding to that of the lamb wave mode. The server envelope windows 920, 922, and 924 can be applied to extract wave speeds 926, 928, and 930 by moving along the entire time history of each measured sensor signal 912.

Applying the DNP system to the host structure, the structure health monitoring software initiates processing of the DNP system. The monitoring software may include a survey module, a processing module, a classification module and a diagnostic module. 10 is a flow diagram 1000 illustrating an exemplary process of a survey module in accordance with one embodiment of the present invention. The inspection module monitors the detection of defects, identification of impacts and repaired-bonding-patch performance of the main structure. In step 1002, the survey module divides the diagnostic patch of the DNP system into a plurality of subgroup sets and assigns one actuator for each subgroup. Each diagnostic patch may be switched to function as a sensor after temporarily functioning as an actuator. 11A illustrates an example of an actuator network structure 1100 that includes a plurality of subgroups divided by survey modules in accordance with one embodiment of the present invention. Each actuator 1102, 1104, 1106, and 1108 may also function as a sensor, so that various combinations of subgroups may constitute the actuator. Arrow 1110 indicates the propagation of a lamb wave signal between actuators 1102, 1104, 1106, and 1108. Table 1 shows the possible subgroups, each group having one actuator. For example, subgroup 1 has one actuator A1 1102 and two sensors A2 1104 and A4 1108.

Subgroup formed by the four patches of FIG. 11A Subgroup number Actuator sensor One A1 A2, A4 2 A2 A1, A3, A4 3 A3 A2, A4 4 A4 A1, A2, A3

11B illustrates another example of an actuator / sensor network structure 1120 that includes subgroups divided by survey modules in accordance with another embodiment of the present invention. As shown in FIG. 11B, four subgroups 1122, 1124, 1126, and 1128 may be created using four actuators / sensors 1132a-1132d and 13 sensors 1130a-1130m. Table 2 shows the elements of each subgroup formed by the patch of FIG. 11B.

Subgroup formed by the seventeen patches of FIG. 11B Subgroup number Actuator sensor 5 A1 A2, S1, S2, S3, S4 6 A2 A1, A4, S3, S4, S5, S6, S7 7 A3 A2, S6, S7, S8, S9, S10 8 A4 A2, S11, S12, S13

The thirteen sensors 1130a-1130m of FIG. 11B can also function as actuators. However, while only one patch of each subgroup temporarily functions as an actuator, the other patches are synchronized to act as sensors. As in the case of FIG. 11B, one sensor (eg, s3) belongs to one or more subgroups (groups 5 and 6). 11A-B show only four actuators / sensors and 13 sensors for simplicity. However, it is apparent to those skilled in the art that the present invention can be practiced without limiting the number of patches.

The network structure of the diagnostic patch system as shown in FIGS. 11A-B may be configured to maximize the performance of the entire network with a minimum number of actuators and sensors. The diagnostic network may be represented by an undirected graph G = (N, E). Here, node N and edge E represent patch positions and wave-communication paths, respectively. The graph G may be configured as a diagnostic network communication relationship, where the node points 1102, 1104, 1006, and 1108 of FIG. 11A represent elements of the actuator and sensor set, and the solid line that is the edge 1110 of FIG. 11A is shown in Table 1 Ordered pairs of relevance of the actuator set and the sensor set. The graph G is connected if at least one path exists between all node pairs i and j . In the exemplary optimal design for network path uniformity, the symbols defined as follows are used. n is the number of nodes; x _ij ∈ {0, l} is a decision variable representing a path between nodes i and j ; And _{x (= {x 12, x} 13, ..., x n -1, n}) is a topological structure of the network design. R (x) is the constraint of the network design, such as the number of patches; c _ij is a value variable of the network design, such as the propagation distance of the lamb wave, and multiplies the number of intersection points on each network path by a sensitivity factor for another network path or excitation frequency. The optimal design of the diagnostic network is arg max

It can be represented as. Here, the optimal problem is the variable x (= { x ₁₂ , x ₁₃ , ..., x _{n -1, n} that satisfies the limit R _min while minimizing the objective function Z (x) representing the uniformity of the network path. }) Must be solved for the value of.

In another example of an optimal group design, each sensor of a network subgroup is associated with one actuator of that group as shown in FIG. 11B. The performance of the network depends on the location and number of actuators and sensors in each subgroup. For group placement of this patch, the matrix of actuators / sensors is considered. Here, each element (i, k) of the matrix is 1 if the i-th sensor is associated with the k-th actuator, and 0 otherwise. In such a group design, a joint integer programming formulation consisting of the assignment of the following various indications and constraints may be applied. Each actuator is assigned to only one subgroup. Here, each sensor may be assigned to one or more subgroups. x _ic is 1 if the i-th actuator is assigned to subgroup c, and 0 otherwise; y _ic is 1 if the jth sensor is assigned to subgroup c, and 0 otherwise. The two constraints are

Where k is the number of specified subgroups, and m and n are the number of actuators and sensors, respectively.

Returning to FIG. 10, the genetic algorithm implemented in the survey module designs the network and signalpath in step 1004. One or more artificial defects are applied to the main structure, such as removable patches, to simulate the defects. Each actuator then sends a signal to one or more sensors of the divided subgroups. Based on the signal received by the sensor, the genetic algorithm determines the optimal network, signal path and sequence of operation of the actuator to accurately detect the location and type of the artificial fault. Depending on the geometry and material of the main structure, the determination of the subgroup set may include adjusting the number of actuators / sensors in the communication network.

In step 1006, the actuators in the i th subgroup operate to generate a lamb wave signal in accordance with the operating sequence from the relay switch array module 512 (see FIG. 5A). Next, in step 1008, the signal SCI carrying the state information of the structure is measured by the sensor of the j-th subgroup. Here, the j th subgroup includes the i th subgroup. In step 1010, the question module calculates a deviation of the measurement signal from the reference signal. Here, the reference signal is formed by executing steps 1004 and 1006 in the absence of artificial defects. Next, in step 1012 the survey module stores the deviation and measurement signal as a document formatted in Extensible Markup Language (XML) in an appropriate signal database storage (e.g., computer 514). do. In addition, the irradiation module may store setting information data including coordinates of the actuator and the sensor, the operation period, the number of the actuator and the sensor, the voltage level, the type of the patch, and the operational failure state. In step 1014, the irradiation processing of the irradiation module is completed.

The irradiation module may perform steps 1006, 1008, 1010 and 1012 at a set of discrete excitation frequencies. Here, the actuator of the DNP system is operated at an excitation frequency to generate a lamb wave. The processing module then processes the stored sensor signals to determine the structure's state index (SCI) for each network path at the excitation frequency. The SCI of the network path between a pair of actuators and sensors indicates the quantitative state of the effect of a defect on the main structure, and thus the degree of change in the state of the structure located inside the main structure. The SCI may include the arrival time of the lamb wave mode, the spectral energy of the lamb wave mode distributed in the time-frequency range, or the highest amplitude of the sensor signal. However, the inclusion target of the SCI is not limited to this. 12 illustrates an exemplary processing procedure 1200 for the identification and determination of the arrival time of a lamb wave mode in accordance with one embodiment of the present invention. In step 1202, the processing module loads a set of sensor signal data from a signal database storage such as computer 514. Here, each sensor signal data may be measured at one excitation frequency. The excitation frequency then represents the frequency at which the actuator of the DNP system operates to generate a lamb wave. The stored configuration information data of the network patch system is checked, and the number of network paths is checked to check whether an appropriate actuator and sensor are assigned to the path of each network link. Next, in step 1204, the processing module detrends each loaded sensor signal data to remove non-stationary signal components. Next, in step 1206, the electrical noise 914 caused by the tone burst signal 904 is removed by applying a masking window 918 to each detrended signal data. Next, in step 1208, a short-time Fourier transform or wavelet transform is performed on the de-noise signal to obtain a time-frequency signal energy distribution for the center excitation frequency bandwidth along the time base.

In step 1210, the processing module accumulates the entire set of time-frequency signal energies to produce a multibandwidth energy distribution on the time-frequency plane. Next, in step 1212, the processing module extracts the protrusion curve from the multiband energy on the time-frequency plane. The protrusion curve extracted from this energy distribution can show the trajectory curve of each wave mode and provide a local maximum along the frequency axis. In the extraction of the protrusion curve, the investigation of the local maximum can be carried out on a fixed value of the time axis. Here, the maximum value of the distributed data column is compared with two new columns given by one step displacement in both directions, and the maximum value is stored when the value is larger than the newly determined threshold value. In step 1214, based on the protrusion curve, the processing module identifies the trajectories of the S ₀ , S _{0 _ ref} and A ₀ mode waves (926,928 and 930 in FIG. 9) on the time-frequency plane. Next, in step 1216, the confirmation processing of the processing module ends.

As described below, the S ₀ , S _{0 _ ref} and A ₀ mode waves determined in step 1214 can be used to design moving envelope windows of various time intervals for the mode wave. By using the protrusion curve extraction method, it is possible to accurately determine the arrival time of each mode wave so that the phase velocity and the time difference of arrival between these modes can be calculated accurately without using the dispersion curve of the structure. The scope of the present invention is not limited to using wavelet transform in the time-frequency analysis method.

13A-B are a flow chart 1300 illustrating an exemplary processing procedure for calculating an SCI value (or equivalent defect index value) in accordance with one embodiment of the present invention. To calculate the SCI value, the processing module may use the sensor signal dataset measured at a set of excitation frequencies. In step 1302, the processing module loads a plurality of sensor signal datasets. Here, each sensor signal dataset is measured at one excitation frequency, and each sensor signal of the dataset such as signal 912 corresponds to the network path of the DNP system. Next, in step 1304, one of the plurality of sensor signal datasets is selected. Next, in step 1306, a sensor signal is selected from the selected sensor signal dataset. In operation 1308, the selected sensor signal is detrended by applying a moving-average filter, and at the same time, the masking window 918 is applied to divide the selected sensor signal into an operation unit 914 and a receiving unit 916. In step 1310, the sensor signal is decomposed into a plurality of lower bandwidth wave packets 926, 928, and 930 by a wavelet decomposition filter using a Daubechies wavelet filter coefficient. For lower bandwidth wave packet decomposition, binary filters for DowBets wavelet filter coefficients are designed to provide high resolution and low resolution, and high and low recovery filters. The decomposition filter decomposes the detrended signal into wavelet coefficients for multiple decomposition levels. Next, in step 1312, the processing module synthesizes a new lower bandwidth wave packet within the associated frequency range. Here, the lamb wave signal includes waves of S ₀ , S _{0 _ ref} and A ₀ modes within the frequency bandwidth. The frequency range of the synthesized signal is a protrusion curve extraction method such that the reconstruction filter includes a range of lower bandwidth change of each wave signal so as to correspond to a bandwidth of the synthesized signal including a wave signal in S ₀ , S _{0 _ ref} and A ₀ modes Determine using. Next, the synthesized signal is generated using a reconstruction filter and wavelet coefficients of signal decomposition. Next, in step 1314, the processing module performs a signal extraction window (or equivalent moving envelope) on the synthesized lamb wave signal to extract the S ₀ , S _{0 _ ref} and A ₀ mode waves 926, 928 and 930 as independent waveforms. Windows) 920, 922 and 924. Each S ₀ , S _{0 _ ref,} and A ₀ mode wave 926, 928, and 930 corresponds to the envelope of each wave mode. In step 1316, the processing module determines the maximum width of each of the envelope windows 920, 922 and 924, the width of the intermediate position, and the interval width on the time axis. Next, in step 1318, SCI for the selected sensor signal is calculated. In one embodiment, the SCI is based on the spectral energy change of each wave in the S ₀ , S _{0 _ ref} and A ₀ modes. In this embodiment, the processing module determines the spectral energy of each wave in S ₀ , S _{0 _ ref} and A ₀ modes. The processing module then calculates the sum of the spectral energies of S ₀ , S _{0 _ ref} and A ₀ modes and calculates the summed energy difference between the reference and the defect state of the main structure. As a result, the spectral energy difference is used as the SCI value of the selected sensor signal. In another embodiment, the processing module selects a change in the maximum and intermediate positions of the envelope window as the SCI value.

In addition, if the diagnostic measurement system employs traditional vibration sensors such as accelerometers, displacement transducers or strain gauges, the processing module may be configured to obtain a natural frequency, attenuation ratio or mode shape from a set of vibration signal data obtained at multiple vibration sensor locations. The dynamic parameters of the structure can be calculated. In another embodiment, the processing module uses the variation of the dynamic parameters of the structure as the SCI value when using the traditional vibration sensor signal instead of the lamb wave signal.

After the processing module calculates SCI data for all network paths, it removes abnormal sensor signals that may be included in the two data sets of the sensor signals of the reference and defective state structures. To this end, the processing module calculates as a probability whether each sensor signal has a reasonable signal amplitude distribution. In step 1320, the processing module determines a discrete probability density function (DPDF) of the signal amplitude and for the amplitude distribution p ( x _i ) .

Estimate the second, third, and fourth moments of. In step 1322, the covariance δ, skewness factor η , and flatness factor κ of the DPDF are used to calculate the steady state constant α of each sensor signal from the estimate of the amplitude distribution. The steady state factor is defined as the product α = δ ^3/2 η ⁻² κ ^3/4 of these factors with exponents. In step 1324, the processing module examines whether all sensor signals included in the selected sensor signal dataset are considered. As shown in FIG. 13, if the answer to decision 1324 is negative, the process proceeds to step 1306, and if yes, the process proceeds to step 1326.

In step 1326, the processing module calculates a second PDF of the SCI dataset that includes the SCI value of the sensor signal included in the selected sensor signal dataset. Next, in step 1328, an abnormal value of the SCI value outside the 3-cisma of the SCI distribution is extracted based on the second PDF. By examining the steady state constants of the SCI anomalies, the processing module removes the SCI values of the anomalies from the SCI dataset for more reliable health monitoring of the structure.

Since the temperature change during the measurement of the sensor signal may affect the sensor signal of the lamb wave, the SCI value obtained from the lamb wave sensor signal should be modified to compensate for the temperature difference between the reference and the defective structure state. The processing module checks whether the reference measurement temperature is different from the measurement temperature of the defective structure state. The processing module creates a temperature reference table for lamb waves. To prepare the temperature reference table, calculate time widths and maximum values of the S ₀ -mode envelope of all network paths of the reference structure, and average the time widths for the 95% network paths within the envelope maximum distribution. do. Using the temperature reference table, the processing module may calculate the temperature adjustment parameter as an average ratio of the time width of the reference structure and the value of the temperature reference table corresponding to the time of the defective structure. In step 1330, the processing module compensates for the influence of the temperature change of the sensor signal by adjusting the SCI data of the defective structure using temperature adjustment parameters. Next, in step 1332, the processing module stores the SCI dataset as an Extensible Markup Language (XML) formatted document. Next, in decision step 1334, the processing module examines whether an SCI data set of each excitation frequency has been generated. If the answer to the determination step 1334 is negative, the process proceeds to step 1304;

FIG. 14A is a flow diagram 1400 illustrating an exemplary sequence for generating a tomographic image identifying areas in which changes occur in the state or defects of structures in accordance with one embodiment of the present invention. In step 1402, the processing module loads the diagnostic patch coordinate data, and the SCI value of the network path determined by the diagnostic patch. In step 1404, for any i-th network path line, the coordinates of the actuator and sensor, {1/2 of the minimum distance of the path line abutting the surface of the structure, {

} And {

}) Computes the bisection of the network path. Next, the SCI value of the i th network path is assigned to the bisection of the i th network path. Next, in step 1406, the processing module calculates intersections of the network paths. The processing module

Slope, its inverse

For, and i paths

And

Calculate the constant of. Next, the processing module has a slope m _i

Coordinates on the i th path line for all other k th path lines crossing the i th path line in a state corresponding to

Calculate In step 1408, the processing module computes the product of the SCI values of the ith and kth network paths to allocate a new SCI on each intersection. If there is no intersection point, the designated SCI is half of the SCI value of the i-th path line, and the intersection point is equal to the bisection point. Thus, the SCI values considered as z-axis data on the coordinate planes of the actuators and sensors of the network path are assigned to all bisections and intersections. In one embodiment, the SCI data of all of these binary and intersection points are stored in the SCI data storage as an Extensible Markup Language (XML) formatted document.

For any i-th path line, the processing module is configured to perform a z-axis Gaussian function or generalized bell function in a plane perpendicular to the path line direction such that the maximum value at the center of the Gaussian function is the SCI value of the path. ). In step 1410, this z-axis function is used to form a three-dimensional block on a network path coordinate plane such that the cross section of the Gaussian function extends parallel to the path line from the start point of the path line to the end point. In fact, the three-dimensional function of the i-th path line intersects by overlapping with the three-dimensional function of any other k-th path line. The SCI value in the intersection region is estimated by the product of the intersecting Gaussian SCI functions on the network path coordinate plane. The width of the three-dimensional function in the cross section plane is the shortest distance in all path lines, and this value is multiplied by the ratio of the SCI value of the i th path to the shortest path line. The processing module calculates SCI values on the network plane for all network paths. In step 1412, the processing module interpolates the SCI dataset with respect to the three-dimensional Gaussian function overlapping points on the mesh grid formed by dividing each binary point, intersection point, and the entire area of the structure into small grid elements. . In the interpolation, the processing module uses Delaunay triangulation of convex-hull for grid data of SCI values.

The processing module further refines the SCI distribution on the network path plane by applying a genetic algorithm to accurately detect defects of the main structure. In step 1414, the processing module establishes an initial population of chromosomes and simultaneously assigns each chromosome to one of the corresponding mesh grid points. Next, in step 1416, the processing module evaluates and ranks the chromosomes by correlation with the SCI distribution data of adjacent mesh grid points. In step 1418, the processing module selects the parent from the population using a randomization method that is considered to be most likely to reproduce the parent with the highest rating. The processing module also regenerates the children from the combination of parent such that possible random mutations of the children occur. Next, at step 1420, the parent chromosome is replaced with children chromosome. Steps 1416-1420 are repeated for several generations until a new children population is formed in step 1422. In step 1422 the children are evaluated and the entire population of parents is replaced to be the parent itself. Next, in step 1424, the processing module takes the subdivided SCI on the lattice point having the composition of the final population of the chromosome.

The SCI distribution on the mesh grid point corresponding to the final chromosome indicates the degree of structural state change of the main structure. The structural state change or defect area of the main structure can be accurately identified from the subdivided SCI distribution. In step 1426, the processing module may provide a genetic tomographic image using the interpolated SCI distribution to identify the structural condition or defect of the main structure. Further, by repeatedly performing the above steps 1402-1426 at a set of excitation frequencies, a set of tomographic images is obtained.

FIG. 14B is a flowchart 1430 illustrating an exemplary sequence for generating tomographic images for identifying regions with structural changes or defects in accordance with another embodiment of the present invention. In step 1432, the processing module loads a time-of-arrival dataset in lamb wave mode, such as S ₀ mode. As described above, the arrival time of the lamb wave mode can be used as the SCI. In step 1212, the processing module accurately calculates the time difference of arrival between lamb wave modes for all network paths using the extracted protrusion curve. Next, in step 1434 a conventional algebraic reconstruction technique is applied to the loaded time of arrival dataset for extensive investigation of the defects in the main structure. A tomographic image of the entire area of the main structure is then generated based on the time of arrival data reconstructed in step 1436. In one embodiment, steps 1432-1436 are repeated to generate tomographic images of the entire area. Wherein each tomography image is based on a time-of-arrival dataset measured at different excitation frequencies. By stacking the tomographic images, a hyperspectral tomographic cube of the entire area is obtained.

In addition, the treatment module may use a simultaneous iterative reconstruction technique to investigate the characteristics of the defects in the areas of potential defects in the main structure. In step 1438, the network paths are rearranged to focus on areas of potential defects. Next, in step 1440, the processing module may apply a simultaneous iterative reconstruction technique to the loaded time-of-arrival dataset to examine the characteristics of the defect in the probable area. Next, a tomographic image of a probable area is generated based on the data set restored in step 1442. In one embodiment, steps 1432-1442 are repeated to generate a set of tomographic images of areas of potential defects. Here, each tomographic image is based on a time-of-arrival dataset measured at different excitation frequencies. By stacking the tomographic images, a hyperspectral tomographic hexahedron of a defective area is obtained.

In another embodiment, the SCI distribution can also be obtained using a genetic distribution on the time-of-arrival dataset of the network path integrated with a projection curve extraction method for short-time-Fourier-transformation (STFT) of sensor signals. Is calculated and a tomographic image is generated. In this embodiment, the tomography image is different from the tomography image of step 1436 or 1442. Tomographic image generation based on well-known scattering-operator-eigenfunctions can be used as a method of extraction and genetic distribution of protrusion curves for data of time of arrival of lamb waves.

When the processing module displays a color tomographic image, the range of colors is adjusted to improve the visibility of the background color of the hot spot where the defect is present. In addition, the tomographic image may include a color mark and a colored dotted line indicating a network path line displayed on a two-dimensional or three-dimensional image of the position of the actuator and the sensor and the shape of the structure. The processing module stores the tomographic image and its color range in tomographic database storage. 14C shows an example of the tomographic image 1450 represented by the gray scale obtained in step 1426, where the region 1452 represents a defect portion.

14D illustrates a hyperspectral tomographic cube 1460 in accordance with one embodiment of the present invention. As shown in FIG. 14D, the hyperspectral tomographic cube 1460 includes a plurality of two-dimensional tomographic images 1462, 1464, and 1466. Here, each picture is generated at an excitation frequency, and the z axis represents an excitation frequency. For simplicity, FIG. 14D shows only three layers 1462, 1464 and 1466. However, it is apparent to those skilled in the art that the hyperspectral tomographic cube 1460 may comprise a plurality of tomographic layers that are generated in the continuous excitation frequency range.

14E illustrates a three-dimensional defect generation manifold 1470 showing a state change of a structure in accordance with one embodiment of the present invention. Like the hyperspectral tomographic cube 1460, the manifold 1470 includes a plurality of two-dimensional tomographic images stacked in the z direction. Here, each image is generated by applying a plurality of vibration repetition cycles corresponding to the z value to the main structure. In addition, each tomography image shows only the part showing the state change of the structure. Thus, each slice of the three-dimensional defect generation manifold 1470 represents a state change or defect occurrence in the structure.

As described above, in step 1410 of FIG. 14A, the processing module calculates an SCI value near an intersection point of a network path. In addition, a classification module including a neuro-fuzzy inference system can also calculate the SCI value at the intersection. 15 is a schematic diagram 1500 illustrating a sequence of neurofuge inference systems for providing a structured system state index (SCI) at the intersection of network paths in accordance with an embodiment of the present invention. As described in step 1408, each intersection of the network paths has two crossing path lines with their SCI values and distances. In order to obtain the structured SCI value at the intersection, the distance of the two crossing path lines is used in a fuzzy if-then rule system that cooperates with the neural network. The expert system then produces an output of the SCI value of the cross path.

For any n intersections (P1-Pn) 1502, each of the two crosswalk line distances 1504 is three fuzzy membership functions 1506 ( A ₁ / B ₁ ) subject to short, medium, and long distances. , A ₂ / B ₂ , A ₃ / B ₃ ). For the membership function, generalized bell function

Is used in conjunction with the control parameters ( a, c ) to cover the input area of the pathline distance standardized in the specification of the structure. In layer 1508, all nodes are fixed nodes, denoted by ∏, and these are input signals.

:

Is the product of. Each node output represents the firing strength of the logic rule. Any i-th node N of layer 1510 is the output of that layer 1510.

The ratio of the firing strength of the i th logical rule to the firing strength of all the logic rules so that this is the normalized firing strength.

) Is calculated. Further, the SCI value of step 1408 of the cross path of the layer 1512 is input to a multilayer perception or neuro network. At layer 1514, each node is a node function

Applies to here,

Is the simple backward propagation multilayer perception and the SCI value of the input layer 1512.

This is the result of a representation of a network layer that can be compared with. here,

Requires two SCI values of the cross-path line as input. If all three neurons 1514 and one neuron 1516 have the identity function of FIG. 15A, the presented neurofuge is equivalent to a Sugeno (TSK) fuzzy inference system that performs a linear fuzzy IFDEN logic rule. Adjusting the relevant connection strengths or weighting factors on the neural network link according to the error distance initiates the adaptation of the neural network. In one embodiment, a sigmoidal function may be used as the neuron function of the result layer 1514. In another embodiment, the neuronal network layer may utilize backward propagation multilayer perception and radial basis function networks. In layer 1516 the node as the output of result layer 1514 is

Compute the sum of the input signal, such as to generate an output 1518 including the SCI value at the intersection.

15B is a schematic diagram illustrating an exemplary processing sequence of a cooperative hybrid expert system for simulating an SCI distribution on a mesh grid point (or equivalent grid point) of a structure from an SCI distribution on an intersection point according to an embodiment of the present invention. (1519). For artificial defects with rubber patches of various dimensions having known information about the location of the defect and the extent of the defect on the structure, the classification module may be configured to provide a first SCI chromosome on the grid point after steps 1418-1426.

Produces an output 1528. If the input 1518 to the cooperative hybrid expert system is an SCI distribution on an intersection given by the coordinates and dimensions of a rubber patch, the final output 1540 of the cooperative hybrid expert system is destined for steps 1534, 1536 and 1538. Using the derived adaptive SCI chromosome set 1524 results in an SCI distribution of hot spots for rubber patches of various dimensions. In addition, the neurofuzzy inference system of FIG. 15A is also applied to the intersection and the SCI value 1518 of the artificial defect, and adapted SCI chromosome

1524 illustrates the output of the neurofuzzy inference system of FIG. 15A.

From step 1418-1426. In step 1534, the root mean square norm is calculated by calculating the difference between the two chromosomes.

, j = 1, ..., n x m is obtained. Where n x m is the dimension of the lattice point. Adaptive value of each chromosome in the step 1536 (fitness value) is the calculated difference (fitness = exp (-E) ). Genetic engineering is then performed for crossover and mutation of the chromosomes in step 1538. Here, the operation plan of the classification module uses a genetic algorithm. The classification module then provides an SCI chromosomal distribution 1524 on the lattice point that is optimal for the artificial defect. With these SCI chromosomes, in step 1526 the unsupervised neural network is trained to achieve clustering or classification of the set of SCI distributions on the grid points. However, the classification module is repeated to adapt to the hybrid expert system while the processing module performs processing to update the SCI distribution for each excitation frequency.

The classification module continues to classify the type of defect (or equivalent hot spot area) from the SCI distribution 1540 on the grid point. FIG. 16A is a schematic diagram 1600 illustrating Gabor jets applied to a hot spot area according to an embodiment of the present invention. As shown in FIG. 16, hot spot region 1610 is recognized and segmented from background SCI distribution 1602 on grid points. In general, the shape and location of the hotspot region 1610 varies with the excitation frequency and the number of network paths. In addition, the variety of physical properties and shapes of the structure increases the difficulty in classifying the defects. In one embodiment of the invention, the classification module uses multilayer perception (MLP) or a forward neural network to classify defects in the hotspot area 1610 of the structure. The classification module uses the Gabor wavelet feature 1606 to polymerize that feature into the MLP. The Gabor wavelet feature 1606 is derived from the Gabor wavelet transform of the SCI distribution with different orientations 1608 and multiple resolutions 1604. The Gabor wavelet function is

Is defined as

Is a positional parameter specifying the position of the wavelet relative to the selection,

Is a modulation parameter for orienting the wavelet in the selection direction, and

Is a scale parameter. Using a series of coefficients called Gabor Jets, the classification module computes Gabor projects for multiple orientations and resolutions in a given hotspot area 1610. Each Gabor Jet includes a plurality of coefficients corresponding to a plurality of orientations and resolutions to be configured with an orientation and logons of different scales. The classification module may capture a local SCI distribution structure of each hotspot region by computing a series of Gabor jets at a plurality of points in the hotspot region to obtain an input feature.

16B is a schematic diagram 1620 illustrating a multilayer perception (MLP) for classifying defect types according to one embodiment of the present invention. As shown in FIG. 16B, the MLP 1624 includes an input feature layer 1628 that receives three layers: a Gabor Jet; Hidden layer 1630; And an output classification layer 1632 for determining the type of defect of the hot spot 1610. The plurality of neurons of the output classification layer 1632 may be nodes indicating the format of the state of the structure.

16C is a schematic diagram 1640 illustrating a classifier in a fully connected state classifying the state of a structure in accordance with one embodiment of the present invention. As shown in FIG. 16C, a series of Gabor jets 1644 are created using an SCI distribution 1643 that includes three hotspot regions 1641. MLP 1644 is similar to MPL 1624 and categorizes the type of defect in hotspot area 1641 as one of categories CO-C5 1646. For simplicity, FIG. 16C shows only three hotspot regions 1641 and six categories, but it is apparent to those skilled in the art that there is no limit to the number of hotspot regions and categories.

16D is a schematic diagram 1650 illustrating a modular network classifier that classifies the state of a structure in accordance with one embodiment of the present invention. As shown in FIG. 16D, a series of Gabor jets 1652 are generated for each hotspot region 1641 of the SCI distribution 1643. Each MLP 1654 is similar to the MPL 1624 and categorizes the type of defect in each hotspot area 1641. Next, nonlinear transformation and mixing 1655 is applied to the result of the MLP 1654 before classifying the defect. The state of the structure is trained for different conditions or different defects of the structure such that the peak of the output node is taken in one of the forms of the state of the structure.

According to one embodiment of the invention, the diagnostic classification module sets reference templates as codebooks for the state or defect of each type of structure. The codebook for each defect type consists of a dataset of cluster points of different versions of the SCI distribution described in FIG. 17B or a dataset of wavelet transform coefficients of the SCI distribution. The template or SCI distribution of the hot spot region is aggregated by K-mean and learning vector quantization (LVQ) clustering algorithms. The K-average algorithm divides n sets of vectors into c groups (G _i , i = 1, ..., c ), and centers the clusters in each group to minimize the cost function of dissimilarity measures. Detect. This algorithm uses the unsupervised learning data clustering method to locate multiple clusters without using classification information. Once the cluster of the SCI distribution of the hot spot area on the lattice point is determined by the K-average algorithm, the marker before moving to the second stage of supervised learning to determine the position of the plurality of cluster centers in the aggregated data. To give. During the supervised learning, the cluster center is fine-tuned to access the desired decision hypersurface. The learning method is simple. First, the cluster center c nearest to the input vector x must be found. Next, if x and c belong to the same classification, c is moved toward x ; If not, it is shifted to form the input vector x . The LVQ algorithm can classify an input vector by assigning the input vector to the same classification as an output unit having a weight vector closest to the input vector. Therefore, the LVQ network uses the classification information of the SCI value to fine-tune the cluster center to minimize misclassification.

FIG. 17A is a flowchart 1700 illustrating an exemplary processing procedure of a K-average method / LVQ algorithm for developing an aggregated 'codebook' according to an embodiment of the present invention. The classification module starts the first K-average method clustering process as the unsupervised learning data clustering method in step 1702. Where the cluster center

Is initialized by randomly selecting c points from the SCI data of the hot spot area. In step 1704, the classification module is

The membership matrix S is calculated by 0 or 0. Here, the binary membership matrix S is

Defines a partition group of c , where x is a randomly selected input vector. Next, in step 1706 the classification module is a cost function

And

Calculate Where the SCI vector

And corresponding cluster centers

Euclidean distance is selected as a measure of dissimilarity between. Next, in step 1780, the cluster center is

Is updated in accordance with step 1710 to determine whether the cost is below a certain allowable value. If the answer to step 1710 is YES, go to step 1714, or go to another decision step 1712 to determine if the newly calculated cost is less than the previous one. If the answer to step 1712 is NO, go to step 1714; otherwise go to step 1704. Next, in step 1714, the classification module initiates a second LVQ clustering process to fine tune the cluster center to minimize misclassification. Here, the cluster obtained from steps 1702-1708 is given a marker by a voting method (i.e., if the cluster has data points belonging to class i as a majority in the cluster, it is given a mark of class i). In step 1716, the classification module randomizes the training input vector (x),

Find i to be the minimum. Next, in step 1718, the classification module is

And

If it belongs to the same classification

To

As many times as

Update as many times. here,

Is a positive small constant that decreases with each learning rate and repetition. In operation 1720, the classification module may generate a codebook including the SCI cluster center of the SCI distribution of the hot spot region of the grid point.

FIG. 17B is a schematic diagram 1730 illustrating an exemplary processing sequence of a classification module for constructing a defect classifier using the codebook generated by the processing step of FIG. 17A according to an embodiment of the present invention. The defect is located in the hot spot area on the grid point of the diagnostic network path. The SCI distribution 1734 of the hot spot area for the state of each structure is used to design a codevector for the state or defect of the structure. Each SCI distribution 1734 is obtained at the operating frequency. For the network signal measured at different excitation frequencies, another block template 1738 can be obtained from the collection 1734 on the SCI distribution of the hotspot region. The code vector is given by a series of cluster centers of block templates of the SCI distribution of the hot spot area. Next, the operating frequency is differentiated to obtain a classification codebook 1738 including a series of optimization block templates according to the state or defect criteria of each structure. In order to establish a codebook based classifier that takes into account the operating frequency, frequency multilayer perception 1740 must be given as a codevector of the codebook 1738 corresponding to a series of operating frequencies. The output of the frequency multilayer perception 1740 is input to the neural network input layer 1741. Next, using the output from neural network input layer 1741, another multilayer perception 1742 also classifies the state or defect of the structure to polymerize the output of the frequency multilayer perception. In one embodiment of the present invention, the wavelet transform of the Fourier coefficients and the SCI values may be used as an input of the K average method algorithm of FIG. In another embodiment of the present invention, major component analysis integrated with Fisher linear discriminant analysis or eigenspace separation transformation can be used to provide different codebooks that are highly sensitive to the type of defect. It can be used in LVA clustering method based on PCA for SCI distribution or wavelet modified SCI distribution.

Structures are subject to damage such as aging, defects, wear and degradation of their operational / service capabilities and reliability. Thus, it is necessary to look at the whole life of the structure having various stages, from precise machining to obsolescence. For a given network patch system, it obeys different timescales during defect formation to investigate the structure of the time varying characteristics of the current wave transmission structure of the network patch system. 18A is a schematic diagram 1800 of three generating regions of inability / use of a structure, dynamics of a network patch, and a network system matrix in accordance with one embodiment of the present invention. As shown in Fig. 18A, a slow-time coordinate τ representing a structure defect occurrence is introduced, and a fast-time coordinate n describing the current network dynamics for the wave transformation is introduced. .

In a fast timeframe nested over a long-term lifetime, a black box model identified from an input operating signal and an output detection signal, wherein the dynamic system of the diagnostic network patch system is a self-recovery with a state-space model or an exogenous input. It can be described by an autoregressive moving average with exogenous input (ARMAX) model. Instead of using an ARMAX model that can be incorporated into a fault diagnostic system for investigating the functionality of components of an embedded system, one can use the state-space dynamics of the network patch system at a fixed lifetime (τ). . For the convenience of explanation, the state space power model considered in the non-distribution domain

, Where status vector

Is the wave transform state vector of the network system,

Is the input force vector of the actuator in the network patch.

Are the system matrix and the input matrix, respectively. The excitation force for generating ram waves in all network patches is

Is considered unchanged over its lifetime. The measurement formula of the network sensor is

, Where,

Is the sensor signal vector,

Is the system observation matrix. System Matrix of the Diagnostic Network Patch System

Is considered independent of the fast time coordinate.

In order to model the network dynamics of the diagnostic patch system, the diagnostic module uses a system matrix (eg, subspace system recognition method) to reconstruct the dynamic system from the measured actuator / sensor signals of the network patch.

) Is calculated. "Evaluation of Normal Modes and Other System Parameters of Composite Laminate Plates" published in Composite Structure 2001 by Kim et al., Incorporated herein by reference, and "Structural Dynamics of Vibration Structures," published in the DSMC Report, ASME, 2003. System Rebuild Method " can be employed to establish a rebuilt dynamic system model that utilizes multiple inputs and outputs of the sensor network system.

The base quantity for monitoring and diagnosis is one indication contained in the sensor signal measured from the time-varying system. State changes or defects in a structure represent changes in the wave transformation or dynamic properties of the structural system, which inherently comprise a network of a plurality of sensors and actuators. The system matrix

Can be considered as an indication because it is observable and sensitive to changes in the state of the structure. The system matrix as one indication can be considered as one of the dynamic properties related to proper damage, such as natural vibration, damping ratio, and vibration mode shape, which represent, for example, the state change of the structure as a sensitive amount to the defect / shock / aging of the structure. have. Thus, the status index of the structure on the diagnostic network patch

System matrix at lifetime

Non-linear functions with variables of (

) May be described. An example of a similar approach is the acoustic and vibration journal published by Kim, "Identifying Damages Using the Reconstructed Residual Frequency Response Function," by Kim, Composite Structures, and "Bending of Disassembled Honeycomb Sandwich Beams," published in 2002. Stiffness and natural frequency, and composite structures, published in 2003, "a model for reducing the natural frequency of fatigue damage governed by a matrix of composite laminates," which is incorporated herein by reference in its entirety.

To determine the status of a near future structure in the defect area, the diagnostic module uses the current trend of the system matrix as a temporary indication related to the defect / shock of the main structure. If the temporary signs indicate signs of deterioration, such as changes in signs related to defects / shocks that increase with time τ, the diagnostic module predicts the behavior of the hotspot area with respect to the remaining life of the structure and activates an early warning. Thus, the system matrix generated by the network dynamics of lamb wave transmission of the structure

Future trends will enable us to predict the fault / shock condition of the structure. System matrix of the future

In order to evaluate the diagnostic module, the SCI vector

Because of the high nonlinear nature of, it is desirable to use a training method of Recursive Neural Networks (RNN) with previous dynamic reconstruction models determined from simulated sensor signals. In other embodiments, a feedforward neural network (FFN) may be used. Curves 1802 and 1810 are SCI vectors (

) And matrix (

), The end of life of the structure (

; 1804). Sensor signal 1808 is time (

; Measured to access the state of the structure at 1806.

18B schematically illustrates the structure of a recursive neural network 1830 for predicting future system matrices in accordance with an embodiment of the present invention. As shown in FIG. 18B, the structure of the RNN 1830 has four input nodes 1836, an additional feedback path node 1838, four hidden nodes 1834, and one output node 1832. do. The input dataset consists of a series of discrete time delayed system matrix series. The output layer consists of one neuron 1832 corresponding to system matrix elements that are predicted at a future first time setting. In the RNN 1830, the current operating state of the output is a function of the previous operating state as well as the current input. At time τ, the output node (output signal at τ + 1) is the previous time step

Calculated by the operation of the hidden node 1834 at. Therefore, each training pattern is

, The previous three time delay values

And an additional input from an additional feedback loop 1840, wherein the output

Is a one-step preceding prediction. The network may provide an estimate of the next future system matrix based on current and previous system matrix values. As an operation function of a node included in a hidden layer and an output layer

A sigmoid function consisting of is used. The nodes must operate within the range of the operating function, and all element data in the system matrix of the operating state are graduated at intervals of [−0.5 0.5]. The learning level of the RNN is determined by the prediction error between the actual output of the network and the target output corresponding to the input dataset. The prediction error is used to adjust the weight until the actual output matches the target value. The RNN of the diagnostic module terminates the learning process when the number of repetitions of the training reaches the set number of times and the error is within the allowable value.

Future system matrix

Using the state-space model of, the diagnostic module generates a diagnostic sensor signal in the hot spot area of the structure from the input of the same actuator signal. Next, one-step leading SCI vector

The recognition and classification method described with reference to FIGS. 9-18B may be applied to the diagnostic sensor signal in order to calculate. Finally, the diagnostic module displays the diagnostic tomographic image and stores it in the diagnostic tomographic image storage device.

As noted above, the monitoring software includes a survey module, a processing module, a classification module, and a diagnostic module. These modules use Extensible Markup Language (XML) to store their processed data and / or images in a database based on structured-query-language (SQL), and the apparatus of the state monitoring system of the structure. Read criteria and system data for location, network path and parameters. Each XML formatted document is described by data and tags created by the structure's surveillance system. In addition, each module can interpret the XML document to read the data input to the other module. A tag in an XML document consists of a root element in the outermost node and a child element in a nested node and has attributes that appear as a name / value pair following the name of the tag.

The health monitoring software of the structure may also have a Simple Object Access Protocol (SOAP) or Remote Procedure Call (RPC) -XML. These are lightweight protocols for exchanging SCI data and images in a distributed structural computing system for monitoring the status of structures. In the distributed server system, every module is also an XML web service capable of communication and remote operation between networks using XML-RPC including an open standard SOAP of structure state information for all structured structure systems, or an XML-formatted document. Can be configured. In order to provide an XML Web service for monitoring the health of a structure, the modules are abstracted by compiling into a Common Object Module (COM), and then a SOAP wrapper such as Microsoft's SOAP Toolkit ^™. ) Is wrapped. The modules may use a low-level application programming interface (API) for control that runs directly on the SOAP process for the COM object.

While the present invention has been described with respect to specific embodiments, the above description relates to preferred embodiments of the invention, and the invention is susceptible to various modifications without departing from the spirit and scope of the invention as set forth in the appended claims. Understand it is possible.

1A is a schematic plan view of a partial ablation of a patch sensor according to an embodiment of the present invention.

FIG. 1B is a schematic side cross-sectional view of the patch sensor shown in FIG. 1A.

1C is a schematic plan view of a typical piezoelectric device that may be used in the patch sensor of FIG. 1A.

1D is a schematic side cross-sectional view of the exemplary piezoelectric device of FIG. 1C.

1E is a schematic plan view of a partial ablation of a patch sensor according to another embodiment of the present invention.

FIG. 1F is a schematic side cross-sectional view of the patch sensor shown in FIG. 1E.

1G is a schematic cross-sectional view of a composite laminate including the patch sensor of FIG. 1E.

1H is a schematic side cross-sectional view of another embodiment of the patch sensor of FIG. 1E.

2A is a schematic plan view of a partial ablation of a hybrid patch sensor according to an embodiment of the present invention.

FIG. 2B is a schematic side cross-sectional view of the hybrid patch sensor shown in FIG. 2A.

2C is a schematic plan view of a partial ablation of a hybrid patch sensor according to another embodiment of the present invention.

FIG. 2D is a schematic side cross-sectional view of the hybrid patch sensor shown in FIG. 2C.

3A is a schematic plan view of a partial ablation of an optical fiber patch sensor according to an embodiment of the present invention.

3B is a schematic side cross-sectional view of the optical fiber patch sensor shown in FIG. 3A.

FIG. 3C is a partially cutaway schematic plan view of the optical fiber coil housed in the optical fiber patch sensor of FIG. 3A. FIG.

FIG. 3D is a partially cutaway schematic plan view of another embodiment of the optical fiber coil shown in FIG. 3C.

3E-3F are partial ablation schematic plan views of another embodiment of the optical fiber coil of FIG. 3C.

3G is a schematic side cross-sectional view of the optical fiber coil of FIG. 3E.

4A is a schematic plan view of a partial ablation of a diagnostic patch washer according to an embodiment of the present invention.

4B is a schematic side cross-sectional view of the diagnostic patch washer shown in FIG. 4A.

4C is a schematic diagram of an exemplary bolted structure using the diagnostic patch washer of FIG. 4A in accordance with an embodiment of the present invention.

4D is a schematic diagram of an exemplary bolted structure using the diagnostic patch washer of FIG. 4A in accordance with another embodiment of the present invention.

5A is a schematic diagram of a diagnostic system having a sensor / actuator device in accordance with one embodiment of the present invention.

5B is a schematic diagram of a diagnostic system having a sensor in accordance with one embodiment of the present invention.

6A is a schematic diagram of a diagnostic network patch system applied to a host structure in accordance with one embodiment of the present invention.

6B is a schematic diagram of a diagnostic network patch system having a strip network structure according to an embodiment of the present invention.

6C is a schematic diagram of a diagnostic network patch system having a pentagonal network structure in accordance with one embodiment of the present invention.

6D is a schematic perspective view of a diagnostic network patch system mounted in a rivet / bolt coupled composite laminate in accordance with one embodiment of the present invention.

6E is a schematic perspective view of a diagnostic network patch system mounted in a composite laminate repaired with an adhesive patch in accordance with another embodiment of the present invention.

Figure 6f is a schematic diagram showing an embodiment of a wireless communication system for controlling a remote diagnosis network patch system according to an embodiment of the present invention.

7A is a schematic diagram of a diagnostic network patch system having sensor assemblies in a strip network structure in accordance with one embodiment of the present invention.

7B is a schematic diagram of a diagnostic network patch system having sensor assemblies in a pentagonal network structure in accordance with another embodiment of the present invention.

8A is a schematic diagram of a sensor assembly including an optical fiber coil in series connection according to an embodiment of the present invention.

8B is a schematic diagram of a sensor assembly having optical fiber coils in parallel connection according to another embodiment of the present invention.

9 is a curve diagram of an actuator and a sensor signal according to an embodiment of the present invention.

10 is a flowchart illustrating an exemplary processing procedure of a survey module according to an embodiment of the present invention.

11A is a schematic diagram of an exemplary actuator network structure including subgroups in accordance with an embodiment of the present invention.

11B is a schematic diagram of a network structure having actuator / sensor subgroups in accordance with another embodiment of the present invention.

12 is a flowchart illustrating an exemplary processing procedure for identifying a lamb wave mode according to an embodiment of the present invention.

13A-B are flow charts illustrating an exemplary processing procedure for calculating SCI values in accordance with one embodiment of the present invention.

FIG. 14A is a flowchart illustrating an exemplary process for generating a tomographic image to identify a region having a state change or defect of a structure in accordance with one embodiment of the present invention.

FIG. 14B is a flowchart illustrating an exemplary process for generating a tomographic image to identify a region having a state change or defect of a structure in accordance with another embodiment of the present invention.

FIG. 14C is a tomogram generated by the processing procedure of FIG. 14A.

14D is a hyperspectral tomographic cube in accordance with an embodiment of the present invention.

14E is a three-dimensional defect generation manifold showing a state change of a structure in accordance with one embodiment of the present invention.

15A is a schematic diagram illustrating an exemplary processing sequence of a neurofuzzy inference system for providing a state index (SCI) distribution of a structured system at the intersection of a network path in accordance with an embodiment of the present invention.

15B is a schematic diagram illustrating an exemplary processing sequence of a cooperative hybrid expert system for simulating SCI distribution on a grid point of a structure in accordance with an embodiment of the present invention.

FIG. 16A is a schematic diagram illustrating Gabor jets applied to a hot spot area according to an embodiment of the present invention. FIG.

16B is a schematic diagram illustrating a multilayer perception (MLP) for classifying defect types according to one embodiment of the present invention.

16C is a schematic diagram illustrating a network classifier in a fully connected state for classifying a state of a structure according to an embodiment of the present invention.

16D is a schematic diagram illustrating a modular network classifier for classifying a state of a structure in accordance with one embodiment of the present invention.

FIG. 17A is a flow chart showing an exemplary processing procedure of a K-Means / Learning Vector Quantization (LVQ) algorithm for creating a codebook according to one embodiment of the present invention.

FIG. 17B is a schematic diagram illustrating an exemplary processing sequence of a classification module for building a defect classifier using the codebook generated by the steps of FIG. 17A according to an embodiment of the present invention.

18A is a schematic diagram of three generating regions of inability / use of a structure, dynamics of a sensor network system, and a network system matrix in accordance with one embodiment of the present invention.

18B is a schematic diagram showing the structure of a recurrent neural network for predicting a future system matrix according to an embodiment of the present invention.

Claims (43)

A computer-implemented method of generating tomographic images for monitoring the health of a structure, Obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNPs), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch, and the damage index values Is an amount affected by damage in the main structure; Generating a distribution of damage index values for a surface using the obtained damage index values; And Formatting the distribution as at least one tomographic image using a computer process. The method of claim 1, wherein generating the distribution comprises: Designating a plurality of points on the surface; And And assigning a damage index value to the designated point using the acquired damage index value. The method of claim 2, wherein assigning the damage index value comprises: Using an expert system to determine a damage index value at said designated point. 4. The neuro fuzzy inference system of claim 3, wherein the expert system is based on a fuzzy if-then rule for each distance of the transmission path of the network and cooperates with a neural network. fuzzy inference) computer running method characterized in that. 5. The computer-implemented method of claim 4, wherein the neural network is a back propagation multiplayer perception with radial basis function networks. The method of claim 2, wherein generating the distribution comprises: Creating a mesh-grid point on the surface; And Determining the damage index value at the mesh grid point using the assigned damage index value. 7. The computer-implemented method of claim 6, wherein determining the damage index value at the mesh grid point comprises using interpolation to calculate the damage index value at the mesh grid point. 7. The computer-implemented method of claim 6, wherein generating a mesh grid point on the surface comprises performing Delaunay triangulation to form the mesh grid point. 7. The method of claim 6, wherein determining the damage index value at the mesh grid point comprises applying an algebraic reconstruction technique. 7. The computer-implemented method of claim 6, wherein determining the damage index value at the mesh grid point comprises applying a simultaneous iterative reconstruction technique (SIRT). 7. The method of claim 6, wherein determining the damage index value at the mesh grid point comprises applying a method based on scattering-operator-eigenfunction. The method of claim 6, Training the cooperative hybrid expert system with artificial damage; And Using the cooperative hybrid expert system to filter out false hot sot regions within the distribution. The method of claim 6, Purifying the damage index value at the mesh grid point; And And using a genetic algorithm to assign the purified damage index value to the mesh grid point to produce an updated distribution of damage index values. The method of claim 6, And converting the data of the mesh grid point and the damage index value at the mesh grid point into a document formatted in Extensible Markup Language (XML). The method of claim 1, wherein generating the distribution comprises: Determining an intersection point at which signal transmission paths of the network cross each other; And For each of said intersections, assigning to said intersections a product of two impairment index values associated with two intersections signal transmission paths, respectively. The method of claim 15, wherein generating the distribution comprises: Determining a bisection of a portion of the signal transmission paths, wherein each of the portions of the signal transmission paths do not intersect other signal transmission paths; And For each of these binary points, assigning a damage index value associated with the corresponding path to the binary points. 16. The computer-implemented method of claim 15, wherein the product of the two damage index values is calculated using a three-dimensional Gaussian function. The method of claim 1, And storing the tomographic image in a reservoir. The method of claim 1, Repeating obtaining a plurality of damage index values to format the distribution at a plurality of excitation frequencies to produce a plurality of tomographic images; And Stacking said tomographic images to create a hyperspectral cube. The method of claim 1, Repeating the step of obtaining a plurality of damage index values to format the distribution at a plurality of consecutive time points of the damage state to produce a plurality of tomographic images; And Stacking the tomographic images to create an evolution evolution manifold, wherein the damage evolution manifold represents an expanded state of the structural state of the main structure. A computer-implemented method of generating tomographic images for monitoring the health of a structure, Obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNPs), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch, and the damage index values Each of is associated with a signal generated by one of the patches in response to an impact applied to a main structure of the network; Generating a distribution of damage index values for a surface using the obtained damage index values; And Formatting the distribution as at least one tomographic image using a computer process. The method of claim 21, wherein generating the distribution comprises: Designating a plurality of points on the surface; And assigning a damage index value to the designated point using the acquired damage index value. The method of claim 22, wherein generating the distribution comprises: Creating a mesh grid point on the surface; And Determining the damage index value at the mesh grid point using the assigned damage index value. A computer readable medium for executing one or more sequence instructions for monitoring the health of a structure, Execution of one or more sequence instructions by one or more processors is accomplished by the one or more processors, Obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNPs), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch, and the damage index values Each of is an amount to be affected by damage in the main structure of the network; Generating a distribution of damage index values for a surface using the obtained damage index values; And And formatting the distribution as at least one tomographic image using a computer process. The method of claim 24, wherein generating the distribution comprises: Designating a plurality of points on the surface; And assigning a damage index value to the designated point using the obtained damage index value. The method of claim 25, wherein generating the distribution, Creating a mesh grid point on the surface; And Determining the damage index value at the mesh grid point using the assigned damage index value. The method of claim 24, wherein generating the distribution comprises: Determining an intersection where signal transmission paths of the network cross each other; And For each of said intersections, assigning to said intersection a product of two impairment index values each associated with two crossing signal transmission paths. The method of claim 27, wherein generating the distribution comprises: Determining a bisection of a portion of the signal transmission paths, wherein each of the portions of the signal transmission paths do not intersect each other with the other signal transmission paths; And For each of the binary points, assigning the binary index value to the binary index associated with the corresponding path. The method of claim 24, wherein execution of one or more sequence instructions by one or more processors is performed by the one or more processors. Repeating obtaining a plurality of damage index values to format the distribution at a plurality of excitation frequencies to produce a plurality of tomographic images; And And stacking the tomographic images to produce a hyperspectral cube. The method of claim 24, wherein execution of one or more sequence instructions by one or more processors is performed by the one or more processors. Repeating the step of obtaining a plurality of damage index values to format the distribution at a plurality of consecutive time points of the damage state to produce a plurality of tomographic images; And Stacking the tomographic images to create an evolution evolution manifold, wherein the damage evolution manifold represents an expanded state of the structural state of the main structure. 25. The device of claim 24, wherein the one or more sequence instructions comprise a cellular phone that enables a Wireless Application Protocol (WAP), a Pocket PC with an HTML browser, or another device that enables HTML. A computer-readable medium for executing a wireless communication method of WAP or Wireless Markup Language (WML) for Internet Web Access. A computer readable medium for executing one or more sequence instructions for generating tomographic images for health monitoring of a structure, the method comprising: Execution of one or more sequence instructions by one or more processors is accomplished by the one or more processors, Obtaining a plurality of damage index values for a network having a plurality of diagnostic network patches (DNPs), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch, and the damage index values Each of is associated with a signal generated by one of the patches in response to an impact on a main structure of the network; Generating a distribution of damage index values for a surface using the obtained damage index values; And And formatting the distribution as at least one tomographic image using a computer process. 33. The method of claim 32, wherein generating the distribution comprises: Designating a plurality of points on the surface; And assigning a damage index value to the designated point using the obtained damage index value. 34. The method of claim 33, wherein generating the distribution comprises: Creating a mesh grid point on the surface; And Determining the damage index value at the mesh grid point using the assigned damage index value. 33. The method of claim 32 wherein execution of one or more sequence instructions by one or more processors is performed by the one or more processors. Repeating obtaining a plurality of damage index values to format the distribution at a plurality of excitation frequencies to produce a plurality of tomographic images; And And stacking the tomographic images to produce a hyperspectral cube. 33. The method of claim 32 wherein execution of one or more sequence instructions by one or more processors is performed by the one or more processors. Repeating the step of obtaining a plurality of damage index values to format the distribution at a plurality of consecutive time points of the damage state to produce a plurality of tomographic images; And Stacking the tomographic images to create a damaged extinguishing manifold, wherein the damaged extinguishing manifold represents an evolution of the structural state of the main structure. In the system for generating a tomographic image for the health monitoring of the structure, A network coupled to the host structure, the network having a plurality of diagnostic network patches (DNPs), wherein each of the patches can operate as at least one of a transmitter patch and a sensor patch; Means for obtaining a plurality of corruption index values for the network; Means for generating a distribution of damage index values for a surface using the obtained damage index values; And Means for formatting the distribution as at least one tomographic image using a computer process. 38. The apparatus of claim 37, wherein the means for generating the distribution is Means for designating a plurality of points on the surface; And Means for assigning a damage index value to the designated point using the obtained damage index value. 38. The apparatus of claim 37, wherein the means for generating the distribution is Means for generating a reticulated lattice point on the surface; And Means for determining a damage index value at the mesh grid point using the assigned damage index value. 38. The apparatus of claim 37, wherein the means for generating the distribution is Means for determining an intersection point at which signal transmission paths of the network cross each other; And And for each of said intersections, means for assigning to said intersections a product of two impairment index values each associated with two intersections signal transmission paths. 41. The apparatus of claim 40, wherein the means for generating the distribution is Means for determining a bisection of a portion of the signal transmission paths, wherein each of the portions of the signal transmission path do not intersect other signal transmission paths; And And for each of these binary points, means for assigning to said binary points an impairment index value associated with a corresponding path. The method of claim 37, wherein Means for operating means for obtaining a plurality of damage index values, means for generating a distribution of damage index values, and means for formatting the distribution at a plurality of excitation frequencies to produce a plurality of tomographic images; And And means for stacking the tomographic images to create a hyperspectral cube. The method of claim 37, wherein Means for operating means for obtaining a plurality of damage index values, means for generating a distribution of damage index values, and means for formatting the distribution at a plurality of consecutive points of time in a damaged state to generate a plurality of tomographic images; And And means for stacking the tomographic images to create an evolution evolution manifold, wherein the damage evolution manifold represents an expanded state of the structural state of the main structure.

Images