WO2020255446A1

WO2020255446A1 - Operation sound diagnosis system, operation sound diagnosis method, and machine learning device for operation sound diagnosis system

Info

Publication number: WO2020255446A1
Application number: PCT/JP2019/047278
Authority: WO
Inventors: 優一木津; 孝志甲斐; 英俊池田
Original assignee: 三菱電機株式会社
Priority date: 2019-06-17
Filing date: 2019-12-03
Publication date: 2020-12-24
Also published as: WO2020255206A1

Abstract

This operation sound diagnosis system (10) comprises: an operation sound detection unit (11) for detecting the operation sound of an actuator or driven machine, an operation state detection unit (12) for acquiring a time series of the driving position or driving speed of the actuator or the force produced by the driving, a sound-vibration-time-series-spectrum acquisition unit (15), a feature point extraction unit (16), and a cause determination unit (18). The sound-vibration-time-series-spectrum acquisition unit outputs a time series spectrum associating frequency and time with the power of a frequency spectrum calculated from operation sound time series data. The feature point extraction unit outputs feature point data for feature points on the time series spectrum in sets comprising the frequencies and times at the feature points, as well as the driving position or driving speed of the actuator at those times or the force produced by the driving at those times. The cause determination unit determines the cause of the operation sound by comparing the values of the feature point data with numerical ranges for each phenomenon using the combination of the frequencies or times for the feature points for that phenomenon and the operation data during those times as multidimensional data.

Description

Drive sound diagnostic system, drive sound diagnostic method and machine learning device for drive sound diagnostic system

The present invention relates to an actuator or a drive sound diagnostic system for diagnosing the drive sound of a driven machine driven by the actuator, a drive sound diagnosis method, and a machine learning device for the drive sound diagnosis system.

In general, in a device having a power source such as a motor, it is known that the drive sound of the power source or the machine to which the power source is driven contains a lot of information about the power source and the state of the drive target. There is. For example, when some abnormality occurs in the power source or the driving target, a sound or vibration different from the normal state is generated. Therefore, a diagnostic device for diagnosing whether or not sound or vibration is abnormal is known. However, when the device is moved for the first time, or when the device is driven immediately after the mechanical configuration and drive pattern of the device are changed, it is difficult to immediately determine whether the drive sound is normal or not. Therefore, there is a need for a technique for easily identifying the cause of the driving sound by acquiring the driving sound, which is the sound or vibration generated by the device, with a sensor. In particular, there is a demand for a technique for performing diagnosis using a driving sound more easily and versatilely without limiting the mechanical configuration and settings of the device.

Patent Document 1 discloses a technique of measuring the sound or vibration generated by a device including a rotating device and identifying the presence or absence of an abnormality or the cause of the abnormality of the device. According to this technology, first, before the actual operation of the device, the power source when abnormal noise is generated and the set of frequency and time characteristics of the sound emitted by the driven machine, which is the machine driven by the power source, are acquired in advance for each abnormality cause. I will do it. The characteristic of the frequency is the frequency at which the apex is generated from the time change of the spectrum obtained by performing a short-time Fourier transform or the like on the time series data of the measured driving sound. The time feature is the time interval that produces vertices for each frequency feature. Then, during the actual operation of the device, the presence / absence and the cause of the abnormality are identified by comparing the characteristics related to the frequency and time acquired during the actual operation by the same method with the characteristics related to the frequency and time acquired in advance. ..

Japanese Unexamined Patent Publication No. 10-274558

However, in the technique described in Patent Document 1, the presence or absence and the cause of the abnormality of the device are specified only from the data obtained by measuring the driving sound generated by the actuator or the driven machine. Therefore, there is a problem that it is difficult to identify the cause of the abnormality depending on the position or speed of the actuator or the driven machine.

The present invention has been made in view of the above, and an object of the present invention is to obtain a driving sound diagnostic system capable of identifying the cause of an abnormality depending on the position or speed of an actuator or a driven machine.

In order to solve the above-mentioned problems and achieve the object, the drive sound diagnosis system according to the present invention includes a drive sound detection unit, an operating state detection unit, a sound vibration time series spectrum acquisition unit, and a feature point extraction unit. , A factor determination unit, and so on. The drive sound detection unit detects a drive sound that is a sound or mechanical vibration generated by an actuator or a driven machine driven by the actuator. The operation state detection unit acquires the drive position, drive speed, or force generated by the drive of the actuator in time series. The sound vibration time series spectrum acquisition unit calculates a frequency spectrum corresponding to each time of the sound vibration data which is the time series data of the detected driving sound, and sets the power of the calculated frequency spectrum in association with the frequency and the time. Outputs the time series spectrum set to. The feature point extraction unit extracts points as feature points whose waveforms with respect to the power frequency and time of the time series spectrum satisfy the specified conditions, and the actuator at the frequency, time, feature point waveform, and feature point time of the feature point. Outputs feature point data that is a set of operation data that is the drive position, drive speed, or force generated by the drive. For each phenomenon that occurs in the actuator or the driven machine that is the cause of the driving sound, the factor determination unit includes at least one frequency and time including the feature points that occur with the phenomenon, and the driving position of the actuator at that time. Detected by comparing the driving speed or the operation data, which is the force generated by driving, with the factor judgment condition that defines the first numerical range when the combination is multidimensional data, and the numerical value of the feature point data. Determine the cause of the driving noise.

The drive sound diagnostic system according to the present invention has an effect of being able to identify the cause of an abnormality depending on the position or speed of the actuator or the driven machine.

A block diagram showing an example of the functional configuration of the drive sound diagnostic system according to the first embodiment. The figure which shows an example of the hardware composition at the time of applying the drive sound diagnosis system which concerns on Embodiment 1 to an elevator. A block diagram showing an example of the hardware configuration of the arithmetic processing unit according to the first embodiment. A block diagram schematically showing an example of the functional configuration of the arithmetic processing unit of FIG. 2 according to the first embodiment. A flowchart showing an example of the processing procedure of the driving sound diagnosis method according to the first embodiment. The figure which shows an example of the sound vibration data by Embodiment 1. The figure which shows an example of the operation data by Embodiment 1. A flowchart showing an example of a procedure for synchronizing sound vibration data and operation data according to the first embodiment. The figure for demonstrating the procedure for synchronizing the sound vibration data and the operation data by Embodiment 1. The figure which shows an example which divided the operation data of FIG. 7 by the operation mode. A diagram showing an example of time-series spectral data obtained by frequency-converting the sound vibration data of FIG. 6 in a three-dimensional graph with three axes of time, frequency, and power. A diagram showing an example of sound vertices extracted from the time-series spectral data of FIG. 11 in a three-dimensional graph with three axes of time, frequency, and power. The figure which shows an example of the hardware configuration at the time of applying the drive sound diagnosis system which concerns on Embodiment 2 to a picking unit. A block diagram schematically showing an example of the functional configuration of the server device according to the second embodiment. The figure which shows an example of the record of the apparatus data by Embodiment 2. A block diagram schematically showing an example of the functional configuration of the user terminal according to the second embodiment. Block diagram showing an example of hardware configuration of server device and user terminal A block diagram showing an example of the functional configuration of the factor determination unit in the drive sound diagnosis system according to the third embodiment. A block diagram showing an example of the functional configuration of the machine learning device of the drive sound diagnostic system according to the fourth embodiment. A block diagram schematically showing an example of the functional configuration of the server device according to the fourth embodiment. A block diagram schematically showing an example of the functional configuration of the user terminal according to the fourth embodiment.

The drive sound diagnosis system, the drive sound diagnosis method, and the machine learning device of the drive sound diagnosis system according to the embodiment of the present invention will be described in detail below with reference to the drawings. The present invention is not limited to these embodiments.

Embodiment 1.
FIG. 1 is a block diagram showing an example of the functional configuration of the drive sound diagnosis system according to the first embodiment. The drive sound diagnosis system 10 includes a drive sound detection unit 11, an operation state detection unit 12, a time synchronization unit 13, an operation mode extraction unit 14, a sound vibration time series spectrum acquisition unit 15, and a feature point extraction unit 16. A driving vibration extraction unit 17 and a factor determination unit 18 are provided.

The drive sound detection unit 11 detects the sound or vibration generated by the actuator and the driven machine, which is a machine driven by the actuator. In the following, the sound or vibration generated by the actuator and the driven machine will be referred to as the driving sound. The drive sound detection unit 11 is a sensor that detects sound or vibration. A vibration sensor such as a microphone or an acceleration sensor that detects sound is an example of the drive sound detection unit 11. The drive sound detection unit 11 may detect the sound emitted by the drive device that drives the actuator.

The operation state detection unit 12 acquires the operation state of the actuator connected to the drive device in chronological order. The operating state of the actuator includes the drive position, drive speed or force generated by the drive of the actuator. The motor is an example of an actuator. The encoder that detects the rotation angle of the motor, the linear scale that detects the position of the linear motor, the position sensor, the displacement meter, the distance sensor, the speed detector, the current detector, the acceleration sensor, the gyro sensor, and the force sensor detect the operating state. This is an example of part 12.

The time synchronization unit 13 is composed of sound vibration data, which is time-series data of the drive sound detected by the drive sound detection unit 11, and operation data, which is time-series data of the operation state detected by the operation state detection unit 12. Synchronize the time. In one example, the time synchronization unit 13 obtains the difference between the reference times of the sound vibration data and the operation data, corrects the time of the sound vibration data or the operation data using this difference, and among the sound vibration data and the operation data. Synchronized sound vibration data and operation data are acquired by extracting the data between the commonly acquired times.

An example of operation data is time series data of current, generated torque, position, and speed. The operation data may be information on the driving device or may be an estimated value based on some detectable data. The drive device of the motor stores the position command, speed command, torque command, current command, and voltage command of the motor, and the drive sound diagnosis system 10 may acquire these from the drive device of the motor and use them as operation data as they are, or detect them. The operation data may be obtained by a comparison operation with possible data or an estimation operation including filtering.

Also, the operation data is not limited to analog signals. The binary data that is turned on when the speed reaches the command speed is an example of operation data. In addition, the operation data does not have to be a set of time and value. The function that takes the time as an argument, which represents the speed pattern of each time, is an example of operation data.

The operation mode extraction unit 14 divides the operation state into two or more sections separated by time based on the operation data detected by the operation state detection unit 12. Specifically, the operation mode extraction unit 14 determines the type of the operating state of the actuator by analyzing the operation data, and extracts the interval of the time when the operating state of the specific type of actuator is set as the operation mode. In the operation mode, the operation state may be determined from the operation data after synchronization with the sound vibration data by the time synchronization unit 13, or the operation may be performed from the operation data before synchronization with the sound vibration data by the time synchronization unit 13. The state may be determined.

The sound vibration time series spectrum acquisition unit 15 acquires a time series spectrum obtained by frequency-converting the sound vibration data synchronized by the time synchronization unit 13 and obtaining the spectrum of the sound vibration data corresponding to each time. .. The time series spectrum is a set in which the power of the calculated frequency spectrum is associated with the frequency and the time. When performing frequency conversion, the sound vibration time series spectrum acquisition unit 15 selects the time at which frequency conversion is performed by frequency-converting only the section of a specific operation mode from the extraction results of the operation mode extraction unit 14. You may. As an example, the operation mode in which the actuator, which is unlikely to measure the driving sound, is stopped can be excluded from the target of frequency conversion. In this case, since the amount of data to be subjected to frequency conversion processing is reduced, it is possible to reduce the memory used during calculation and the processing time.

The feature point extraction unit 16 extracts the point as a feature point when the waveform with respect to the power frequency and time of the time series spectrum obtained by the sound vibration time series spectrum acquisition unit 15 satisfies a predetermined condition. The feature point extraction unit 16 sets the frequency, time, power, waveform of the feature point, and operation data at the time of the feature point as feature point data.

The operation vibration extraction unit 17 extracts vibration components from the time-synchronized operation data and acquires vibration data including the frequency or amplitude thereof.

The factor determination unit 18 determines the factor determination condition, which is a numerical range including the feature points generated in association with the phenomenon determined for each phenomenon that occurs as a factor of the driving sound, and the feature extracted by the feature point extraction unit 16. The cause of the driving sound is determined by comparing with the point data. At this time, the factor determination condition may include a numerical range including vibration data generated in association with the phenomenon that is predetermined for each phenomenon that is a factor of the driving sound. In this case, the factor determination unit 18 compares the factor determination condition with the feature point data extracted by the feature point extraction unit 16 and the vibration data extracted by the driving vibration extraction unit 17 to obtain the driving sound. Estimate the cause.

If the numerical range including the feature points is specified for each abnormality of the device, the range of the numerical values including the feature points according to the cause of the abnormality of the device must be obtained for all types of driven machines. It doesn't become. Further, even when the configuration of the driven machine is changed, it is necessary to obtain the range of numerical values including the feature points for each cause of abnormality of the device. Since there are many types of driven machines and there are many variations of changes in the configuration of the driven machines, it is a reality to find the numerical range that includes the characteristic points for each cause of abnormality of the device for all of them. Not the target.

However, in the first embodiment, the numerical range including the feature points is defined not for each abnormality of the device but for each phenomenon that occurs as a factor of the driving sound. That is, the feature amount is not defined by a specific drive pattern such as the value of the rotation speed of the motor, but is expressed by a conditional expression that does not depend on the drive pattern, for example, a conditional expression that uses the rotation speed of the motor as a variable. Therefore, even if the driven machine is different or the configuration of the driven machine is changed, if the factors of the driving sound are the same, the phenomenon that occurs is the same, regardless of the type or configuration of the driven machine. The same factor determination conditions can be used. As a result, the time required for preparation for diagnosis can be shortened as compared with the case where the range of numerical values including the feature points is specified for each abnormality of the device, and the diagnosis of the driving sound can be performed for general purposes. Can be done.

The feature point data is a set of the frequency, time, power of the feature point, the waveform of the feature point, and the operation data at the time of the feature point. Therefore, it includes not only information about sound or vibration, but also information about the position or speed of the actuator or driven machine. That is, since the factor determination unit 18 also determines the position or speed of the actuator or the driven machine when determining the factor of the generation of the driving sound, it depends on the position or speed of the actuator or the driven machine. It is possible to easily identify the cause of the abnormality.

Further, in FIG. 1, the time synchronization unit 13, the operation mode extraction unit 14, and the operation vibration extraction unit 17 may be appropriately included in the drive sound diagnosis system 10 or driven depending on the performance or device configuration required for diagnosis. It may be removed from the sound diagnostic system 10. In one example, when sound vibration data and operation data are acquired by the same measuring instrument or device, the acquired data is synchronized by using AD (Analog to Digital) converters with the same acquisition timing. be able to. That is, the sound vibration data and the operation data can be synchronized without providing the time synchronization unit 13. In this case, by removing the time synchronization unit 13, it is possible to reduce the memory used in the drive sound diagnosis system 10 and the processing time.

FIG. 2 is a diagram showing an example of a hardware configuration when the drive sound diagnosis system according to the first embodiment is applied to an elevator. As shown in FIG. 2, the diagnosis target 100 includes a motor 110 which is an actuator, a driven machine 120, and a drive device 130 for driving the motor 110. Further, the diagnosis target 100 is provided with a drive sound diagnosis system 10A including a drive sound detection unit 11, an operating state detection unit 12, and an arithmetic processing unit 140.

The motor 110 is a servomotor that controls the current of the armature winding based on the difference between the drive command and the rotation information. The motor 110 may be an actuator that generates power by receiving energy or an electric signal from the device that drives the motor 110. In this example, the motor 110 is controlled by the drive device 130. However, the state quantity of the motor 110 controlled by the drive device 130 is not limited to the current. Hydraulic pressure, pneumatic pressure, heat, and ultrasonic waves are examples of the state quantities of the motor 110 controlled by the drive device 130. Further, the motor 110 is not limited to one that generates a rotational force, and may be a linear motor that drives in the translation direction.

The driven machine 120 is an elevator that conveys a drive target in the vertical direction according to the rotation of the motor 110. The driven machine 120 converts the rotation of the motor 110 into vertical motion, the bracket 121 for fixing the motor 110 to the gantry, the gearbox 122 for amplifying the rotational force generated by the motor 110 and transmitting it to the ball screw 123. The ball screw 123 and the coupling 124 connecting the gearbox 122 and the ball screw 123 are provided.

Further, the lifting machine has a slider 125 driven in the vertical direction by the rotation of the ball screw 123, a stage 126 fixed to the slider 125 on which the workpiece is mounted, and a linear guide for slidably guiding the movement of the slider 125 in the vertical direction. A guide 127 and a bracket 128 for rotatably fixing the ball screw 123 to the linear guide 127 via a bearing are provided.

Here, the case where the driven machine 120 is an elevator having a ball screw 123 is illustrated, but the driven machine 120 may be a machine that generates sound or vibration according to the rotation of the motor 110. A machine with screws, belts, gears, cams, linkages, bearings or seals, or a combination of these elements is an example of a driven machine 120.

The drive device 130 is connected to the motor 110 via a cable 151. The drive device 130 includes a motor drive 131, a motor control device 132, and a display 133. The motor drive 131 supplies the power for driving the motor 110 to the motor 110. Further, the motor drive 131 drives the motor 110 in accordance with a drive command transmitted from the motor control device 132.

The motor control device 132 controls the amount and timing of the current supplied by the motor drive 131 to the motor 110 by sending an electric signal such as a command position or a command speed to the motor drive 131. The display 133 displays various states of the entire system including the drive sound diagnosis system 10A and the diagnosis target 100, and notifies the user.

In the first embodiment, the motor drive 131 is connected to the drive sound detection unit 11 and the operation state detection unit 12, and relays communication between the drive sound detection unit 11 and the operation state detection unit 12 and the motor control device 132. Has the function of

Further, the motor control device 132 is a controller that gives a drive command to the motor drive 131 including the position or speed pattern of the motor 110. The motor control device 132 is a control device including a PLC (Programmable Logic Controller), a motor driving CPU (Central Processing Unit), a DSP (Digital Signal Processor), a pulse generator, and the like.

Further, the display 133 acquires the state of the entire system including the drive sound diagnosis system 10A from at least one of the motor drive 131, the motor control device 132, or the arithmetic processing unit 140 by communication, and displays it in a format that is easy for the user to see. I do. The display 133 may have a built-in liquid crystal display. By including the display 133 in the system, after the arithmetic processing unit 140 executes the diagnosis of the driving sound, the result of the diagnosis of the driving sound is displayed on the display 133 via communication, and the user is notified in an easy-to-understand manner. Can be done.

The drive device 130 may be a device that drives at least one motor 110, and may be configured by combining a part or a plurality of the present embodiments.

The arithmetic processing unit 140 is a processing device capable of executing diagnostic processing of the drive sound diagnostic system 10A described later as software. In the first embodiment, the CPU of the microcomputer built in the motor control device 132 realizes the function of the arithmetic processing unit 140. FIG. 3 is a block diagram showing an example of the hardware configuration of the arithmetic processing unit according to the first embodiment. The arithmetic processing unit 140 includes a processor 141 and a memory 142. The processor 141 and the memory 142 are connected via the bus line 143. The CPU and GPU (Graphics Processing Units) are examples of the processor 141.

FIG. 4 is a block diagram schematically showing an example of the functional configuration of the arithmetic processing unit of FIG. 2 according to the first embodiment. The arithmetic processing unit 140 includes a time synchronization unit 13, an operation mode extraction unit 14, a sound vibration time series spectrum acquisition unit 15, a feature point extraction unit 16, an operation vibration extraction unit 17, and a factor determination unit 18. , Provided as software executed by the CPU of a microcomputer. The same components as those in FIGS. 1 and 2 are designated by the same reference numerals, and the description thereof is omitted.

FIG. 2 shows a configuration in which the arithmetic processing unit 140 is built in the motor control device 132, but the embodiment is not limited to this configuration. The arithmetic processing unit 140 may be built in the motor drive 131, which is another device of the drive device 130, or may be built in the display 133. Further, another microcomputer may be attached to the motor control device 132 to serve as the arithmetic processing unit 140. Furthermore, it may be connected as a separate device independent of the drive device 130.

The drive sound detection unit 11 is a microphone that detects the drive sound emitted by the diagnosis target 100. The detected drive sound is transmitted to the arithmetic processing unit 140 via the cable 152 and the drive device 130. In the example of FIG. 2, the drive sound detection unit 11 is fixed to the elevator. However, the drive sound detection unit 11 only needs to be able to detect the drive sound of the diagnosis target 100, and the installation method is not limited to being fixed to the driven machine 120. The drive sound detection unit 11 may be fixed to the motor 110, or may be installed away from the driven machine 120 at a distance such that the drive sound can be detected.

If the sounding location of the phenomenon to be diagnosed is limited, the sound can be collected within the range or sound can be collected by arranging the microphone adjacent to the sounding location or by using a directional microphone. The range may be limited. By limiting the sound collection, it is possible to reduce ambient noise that interferes with the diagnosis and reduce misdiagnosis.

Further, a plurality of microphones can be used as the drive sound detection unit 11. By collecting the same sound or vibration with multiple microphones, false diagnosis due to noise can be reduced.

Further, a recorder may be used to record the driving sound emitted by the diagnosis target 100 in time series as sound vibration data. A smartphone or voice recorder is an example of a recorder. In addition, a moving image may be taken with a camera or the like, and only the sound part may be extracted and used as sound vibration data.

Alternatively, the drive sound detection unit 11 may be provided with a storage unit that stores the drive sound emitted in time series as sound vibration data, and may transmit the stored sound vibration data to a higher-level arithmetic unit as needed. In this case, when the drive sound diagnosis is not performed, communication between the drive sound detection unit 11 and the arithmetic processing unit 140 is not performed, and the time series stored by the drive sound detection unit 11 when the drive sound diagnosis is performed is performed. Sound vibration data is transmitted together. With such a configuration, it is possible to reduce the processing related to the communication between the drive sound detection unit 11 and the arithmetic processing unit 140.

The operating state detection unit 12 is an encoder attached to the motor 110 to detect the rotation angle of the motor 110. The rotation angle data detected by the encoder is transmitted to the arithmetic processing unit 140 via the cable 153 and the drive device 130 as an operating state.

The installation location of the operating state detection unit 12 is not limited to the case where it is fixed to the motor 110. The operating state detection unit 12 may be fixed to the drive unit of the driven machine 120, which is the machine driven by the motor 110. Similar to the drive sound detection unit 11, the operation state detection unit 12 is provided with a storage unit that stores the detected time-series operation state as operation data, and transmits the stored operation data to a higher-level arithmetic unit as needed. It may be a method.

The arithmetic processing unit 140 receives the drive sound detected by the drive sound detection unit 11 and the operation state detected by the operation state detection unit 12 via the

cables

152 and 153, and diagnoses the cause of the drive sound.

The arithmetic processing unit 140 is connected to the drive sound detection unit 11 and the operation state detection unit 12 in a high-speed network without data delay in order to exchange appropriate data between the drive sound detection unit 11 and the operation state detection unit 12. The environment is desirable.

Next, the operation of the diagnosis target 100 will be described. The user of the elevator of the driven machine 120 inputs a drive command of the motor 110 to the motor control device 132 in order to convey a work (not shown) by the elevator. The motor control device 132 transmits a drive command to the motor drive 131 based on the information including the drive pattern and the operation timing input by the user. The motor drive 131 controls the motor drive current and drives the motor 110 in accordance with the received drive command. In the elevator, the ball screw 123 rotates as the motor 110, which is a power source, rotates, and the slider 125 connected to the ball screw 123 and the stage 126 connected to the slider 125 move up and down.

In one example, the user of the elevator mounts the work on the stage 126 when the position of the stage 126 is moving vertically downward of the linear guide 127 by the above operation. The stage 126 moves vertically upward of the linear guide 127 as the motor 110 rotates. Along with this movement, the elevator conveys the work mounted on the stage 126.

The diagnosis target 100 emits a driving sound when it is driven by the motor 110. Specifically, the diagnosis target 100 includes the torque pulsation of the motor 110, the translational and torsional rigidity of the ball screw 123, the connection rigidity of the coupling 124, the meshing rigidity of the gear, the movement, deformation or collision of the machine, and the linear guide 127. A driving sound is emitted due to sliding friction between the screw and the slider 125. The drive sound emitted by the diagnosis target 100 is detected by the drive sound detection unit 11, the rotation angle of the motor 110 is detected by the operation state detection unit 12, and the motor drive 131, the motor control device 132, and the calculation are performed via the

cables

152 and 153. It is transmitted to the processing unit 140. However, the transmission method does not necessarily have to be wired, and may be wireless or via a recording medium.

The display 133 appropriately communicates with the motor control device 132, and requires a user including the rotation angle of the motor 110 detected by the operation state detection unit 12 and the presence or absence of an abnormality in the entire system including the diagnosis target 100. Various information to be displayed is displayed on the display 133.

In the first embodiment, the time synchronization unit 13, the operation mode extraction unit 14, the sound vibration time series spectrum acquisition unit 15, the feature point extraction unit 16, the operation vibration extraction unit 17, and the factor determination unit 18 are one piece of hardware. Although it is configured on the processing unit 140, it may be divided into separate hardware. As an example, the time synchronization unit 13 can be separated from the arithmetic processing unit 140 and configured as software on the microcomputer of the motor drive 131. In this case, the motor drive 131, which performs control processing at a higher speed than the motor control device 132, performs time synchronization processing with strict time requirements, and the arithmetic processing unit 140 of the motor control device 132 is relatively time-constrained. The processing of the operation mode extraction unit 14, the sound vibration time series spectrum acquisition unit 15, the feature point extraction unit 16, the operation vibration extraction unit 17, and the factor determination unit 18 is performed. As a result, the required performance of the arithmetic processing unit 140 can be reduced, and it is possible to diagnose the cause of the driving sound even in a device having a low arithmetic processing capacity.

In addition, the realization of each function is not limited to the realization by software on the CPU of the microcomputer, and uses electronic circuits such as ASIC (Application Specific Integrated Circuits), FPGA (Field Programmable Gate Array) or CPLD (Complex Programmable Logic Device). You may.

Next, the procedure for diagnosing the driving sound in the driving sound diagnosis system 10A targeting the sound or vibration emitted by the diagnosis target 100 will be described. FIG. 5 is a flowchart showing an example of the processing procedure of the driving sound diagnosis method according to the first embodiment. In the first embodiment, as described above, when the driven machine 120 is driven by the drive of the motor 110, a driving noise is generated due to the movement, deformation, collision, rigidity, friction, etc. of the machine. Further, the drive sound differs in the magnitude and frequency of the sound pressure depending on the drive method of the motor 110. Therefore, in the first embodiment, a method of diagnosing the driving sound generated by the driven machine 120 will be described.

The drive sound detection unit 11 of the drive sound diagnosis system 10A detects the drive sound (step S11), and the operation state detection unit 12 detects the operation state of the motor 110 (step S12).

The time synchronization unit 13 captures the drive sound from the drive sound detection unit 11 as time-series sound vibration data (step S13). FIG. 6 is a diagram showing an example of sound vibration data according to the first embodiment. In this figure, the horizontal axis represents time and the vertical axis represents amplitude.

Further, the time synchronization unit 13 captures the operation state from the operation state detection unit 12 as time-series operation data (step S14). FIG. 7 is a diagram showing an example of operation data according to the first embodiment. In this figure, the horizontal axis represents time, and the vertical axis represents rotation angle, rotation speed, and current. In FIG. 7, the rotation angle Po of the motor 110, the rotation speed w of the motor 110, and the motor current i from the reference point of the ball screw 123 are shown as operation data.

Here, it is necessary to take in the data between the same time for the acquisition of the sound vibration data and the acquisition of the operation data. However, the acquisition time of the data to be captured may be different as long as it is between the same time, or the acquisition time of some data may be different from the same time interval.

Further, regarding the frequency of sound or vibration emitted by the diagnosis target 100, if the band is predicted by some knowledge, it is desirable to thin out the sampling of the sound vibration data and the operation data to the target frequency band. The thinning process can reduce the amount of memory used and the processing time. When the thinning process is performed, the frequency conversion of the sound vibration data is performed in the process after the processes of the drive sound detection unit 11 and the operating state detection unit 12, so it is desirable that the filter process is performed by the noise removal filter. The low-pass filter, high-pass filter, band-pass filter, notch filter, and band-eliminate filter are examples of noise reduction filters. Further, as the noise removal filter, one or more of the illustrated filters may be applied. By doing so, the effect of reducing aliasing noise due to thinning can be expected.

Next, the time synchronization unit 13 performs synchronization processing of sound vibration data and operation data (step S15). FIG. 8 is a flowchart showing an example of a procedure for synchronous processing of sound vibration data and operation data according to the first embodiment.

First, the difference between the reference times of the sound vibration data and the operation data is acquired (step S31). Here, the reference time is a time set to time 0 in each data. When the sound vibration data and the operation data are not synchronized, the timing of setting the time 0 of each data is set individually, so even if the data acquisition time is the same, the actual acquisition timing may be different. is there. Therefore, the time of each data is corrected by obtaining the difference between the actual times of the timings at which the time is set to 0 in each data.

As a method of obtaining the difference in the reference time, there is a method in which the timing of acquiring data is designed to be a constant time difference in advance and the constant time difference is used as the difference in the reference time. Another method is to access a common master clock in advance so that the acquisition timing of each data is the same and the difference in the reference time is zero. Alternatively, there is also a method in which the time in a common master clock at the timing of starting data acquisition is stored as a time stamp, and the difference between the time stamps of the sound vibration data and the operation data is used as the difference in the reference time.

FIG. 9 is a diagram for explaining a procedure for synchronizing the sound vibration data and the operation data according to the first embodiment. In the example of FIG. 9, it is assumed that the time 0 of the sound vibration data Da is the data acquisition time and the time 0 of the operation data Do is the data acquisition time. It is assumed that the time 0 of the sound vibration data Da is t1 at the reference time, and the time 0 of the operation data Do is t2 at the reference time. In this case, the difference Δt of the reference time is t2-t1.

Then, the acquisition time of each data is corrected based on the difference in the obtained reference time (step S32). Specifically, among the sound vibration data and the operation data, the difference between the obtained reference times is added as a correction to the acquisition time of each data of the data for which the acquisition is started earlier.

In the example of FIG. 9, the data that started to be acquired first is the sound vibration data Da. Therefore, the reference time t2 of the operation data is obtained by adding the difference Δt of the reference time to the acquisition time t1 of the sound vibration data Da.

Returning to FIG. 8, the time interval acquired by the sound vibration data and the operation data in common is calculated (step S33). An example of a specific calculation method is common from the acquisition start time of the data that was started to be acquired later in the above process to the acquisition time of the earlier timing of the last acquisition time of the sound vibration data and the operation data. This is the method of setting the time between the acquired times. In the example of FIG. 9, from the acquisition start time t2 of the operation data Do to the last acquisition time t3 of the sound vibration data Da is the time interval Δct that is commonly acquired.

Then, the data at the non-common time that is not between the calculated common and acquired times is discarded (step S34). In the example of FIG. 9, the sound vibration data Da before the time t2 and the operation data Do after the time t3 are discarded. As described above, the sound vibration data and the operation data can be synchronized, and the synchronization process of the sound vibration data and the operation data is completed.

Returning to FIG. 5, the operation mode extraction unit 14 estimates the operation state of the driven machine 120 based on the time-synchronized operation data, and extracts the operation mode (step S16). In the first embodiment, the operation mode of the driven machine 120 is divided into three, which are stopped, constant speed operation, and acceleration / deceleration.

<Stopped>
The section in which the rotation of the motor 110 is stopped is defined as being stopped. Here, the fact that the rotation of the motor 110 is stopped means that the sum of the changes in the rotation angle data of the motor 110 detected by the operating state detection unit 12 falls within a specific threshold value for a certain period of time. At this time, a section in which the sum of the changes in the rotation angle data is within a specific threshold value is defined as a section in which the position is stopped. The fixed time included in the condition may be determined according to the average value of the detection cycles of the sound vibration data and the operation data.

<During constant speed operation>
A section in which the rotation speed of the motor 110 is constant and is not stopped is defined as constant speed operation. Here, the constant speed means that the sum of the changes in the speed data falls within a specific threshold value for a certain period of time as in the case of stopping. At this time, a section in which the sum of the changes in the speed data is within a specific threshold value is defined as a section in which the velocity is constant and longer than a predetermined time.

<During acceleration / deceleration>
The section that is neither stopped nor constant speed operation is defined as being accelerated or decelerated. With respect to the driving direction, when the acceleration is positive, it may be distinguished from accelerating, and when the acceleration is negative, it may be distinguished from decelerating. FIG. 10 is a diagram showing an example in which the operation data of FIG. 7 is divided by the operation mode. In this example, in the operation mode, acceleration / deceleration is in progress from time 0 to t11, constant speed operation is in progress from time t11 to t12, acceleration / deceleration is in progress from time t12 to t13, and after time t13. Is stopped.

Returning to FIG. 5, the sound vibration time series spectrum acquisition unit 15 performs frequency conversion on the sound vibration data for which the time synchronization unit 13 has time-synchronized, and calculates the spectrum of the sound vibration data at each time (step S17). ). The spectrum of the sound vibration data is time-series spectrum data in which the power of the calculated frequency spectrum is paired with the frequency and time.

It is desirable to perform frequency conversion by short-time Fourier transform (STFT) or wavelet transform. Since the time series spectrum is obtained by these methods, it is possible to reduce the labor and processing such as filter design.

Here, regarding the frequency of the driving sound emitted by the diagnosis target 100, if the band is predicted by some knowledge, a filter for extracting the frequency component in the band predicted before the frequency conversion is applied to the sound vibration data. You may. By applying the filter, it becomes possible to diagnose the driving sound more accurately.

In the first embodiment, the sound vibration time series spectrum acquisition unit 15 frequency only the sound vibration data corresponding to the period during constant speed operation extracted by the operation mode extraction unit 14 among the time-synchronized sound vibration data. Converts and discards data for other periods.

FIG. 11 is a diagram showing an example of time-series spectrum data obtained by frequency-converting the sound vibration data of FIG. 6 in a three-dimensional graph with three axes of time, frequency, and power. In this figure, the vertical axis represents power and the two orthogonal axes in the plane perpendicular to the vertical axis represent time and frequency.

Returning to FIG. 5, the feature point extraction unit 16 determines the frequency and time of the power of the spectrum of the sound vibration data represented by the time series data obtained by the sound vibration time series spectrum acquisition unit 15. The feature points that are the conditions are extracted (step S18). In addition, the feature point extraction unit 16 generates feature point data for the extracted feature points (step S19). The feature point data is a set of operation data at the frequency, time, power, waveform, and time of the feature point.

An example of the conditions used when the feature point extraction unit 16 extracts the feature point is a vertex, a peak, a ridgeline, and a saddle point. The case of acquiring vertices as feature points will be described.

First, since the spectrum of time-series sound vibration data contains noise, a low-pass filter is applied to the time axis and frequency axis. The time constant of the low-pass filter is determined by the sampling period of the sound vibration data and the frequency conversion resolution. Alternatively, instead of applying a low-pass filter, a Hilbert transform may be applied to the spectrum of the time series.

Next, for the time-series spectrum after applying the low-pass filter, the power threshold value is determined and divided into a region where the power is larger than the threshold value and a region where the power is smaller than the threshold value. The median spectrum of the time series is an example of a power threshold.

Then, the point power is greater than the threshold value _{_{x = (t x, f x}} , p x) from among, to get the next four points. Here, t _x, f _x, p _x, the time at each point x, represents the frequency and power.
(1) Point x1 (t _{x + 1} , f _x , p ₁ ) at the time t _{x + 1} next to the acquisition time t _{x when} the frequency is f _x
(2) at a frequency f _x, the previous one acquisition time t _x the time t _x-1 of the point _{x2 (t x-1, f} x, p 2)
(3) at the acquisition time t _x, the next frequency f _{x + 1} of point x3 frequencies _{_{_{f x (t x, f x}}} + 1, p 3)
(4) at the acquisition time t _x, the frequency of the previous one f _x frequency f _x-1 of the point _{_{x4 (t x, f x-}} 1, p 4)

Next, the powers of a total of five points, which are the above four points x1, x2, x3, x4 plus the original point x, are compared. When the point x has a higher power than the other four points among the five points, that is, the point x surrounded by the other four points x1, x2, x3, x4 among the five points has the maximum power. When, the point x is a candidate for the vertex. In this example, a point x, two points adjacent to the time axis direction and two points adjacent to the frequency axis direction around the point x are used to extract the point having the maximum power. However, the embodiment is not limited to this. That is, when a plurality of points surrounding the point x on the plane formed by the time axis and the frequency axis and the point x are used and the power of the point x surrounded by the plurality of points is maximized, the point x Any method may be used as long as the method is used to extract the point with the maximum power.

Finally, the candidate vertices are arranged in descending order of power, and a certain number of points are extracted as the maximum from the largest.

Here, the candidate points may be arranged in descending order of power, and a certain number of points obtained from the largest may be used as vertices, or all candidate points whose power exceeds a predetermined threshold value may be used as vertices. Good. Moreover, you may combine these methods.

By acquiring the vertices of the waveform with respect to the frequency and time of the power of the time-series spectrum as feature points in this way, it is possible to reduce the amount of calculation by the factor diagnosis described later by a method with a small amount of processing.

If the operation data is not acquired at the time of the feature point, the feature point can be used by using the operation data of the time closest to the time of the feature point, or based on the operation data of multiple times before and after the time of the feature point. By interpolating the operation data at the time of, the operation data at the time of the feature point can be determined.

FIG. 12 is a three-dimensional graph with three axes of time, frequency, and power, showing an example of sound vertices extracted from the time-series spectrum data of FIG. This figure shows the

vertices

1, 2, and 3 extracted by the method described above.

Returning to FIG. 5, the operation vibration extraction unit 17 extracts vibration components from the time-synchronized operation data and acquires the vibration data (step S20). The frequency, amplitude or phase of the vibration component is an example of vibration data. This process may be performed independently of the sound vibration time series spectrum acquisition process in step S17.

An example of the vibration component extraction method is a method of finding the time from the top of a wave peak to the top of the next peak in the operation data during constant speed operation.

Next, the factor determination unit 18 compares the feature point data generated in step S19 and the vibration data acquired in step S20 with the factor determination conditions registered for each factor of the drive sound to generate the drive sound. The factor is determined (step S21). Here, each factor determination condition registered for each factor of the driving sound is a feature that occurs in association with each phenomenon that occurs in the motor 110 or the driven machine 120 that is the factor of the driving sound for the feature point data. Numerical range when the combination of at least one of the frequency and time including the points and the driving position, driving speed, or driving data of the force generated by the driving of the motor 110, which is the actuator at the above time, is regarded as multidimensional data. Is defined. Regarding the vibration data, for each phenomenon that occurs in the motor 110 or the driven machine 120 as a factor of the driving sound, a numerical range including a vibration component generated in association with this phenomenon is defined.

For the feature point data, the factor determination unit 18 determines the cause of the sound vibration data by comparing the acquired feature point data with the numerical range for each factor. Further, regarding the vibration data, the factor determination unit 18 compares the acquired vibration data with the numerical range for each factor, that is, compares the vibration component of the operation data with the peak of the sound to generate the vibration data. Determine the factors.

As an example, when dirt adheres to the linear guide 127 and the friction between the slider 125 and the linear guide 127 increases to generate a scraping noise, the scraping noise generated is within a specific position section of the slider 125. Occurs only when it is in. Therefore, when the positions of the motor 110 at the time when the feature points are generated are concentrated in a specific position or a section P having a certain width, the diagnosis target 100 determines that sound is generated in the section P. 18 determines. Here, the width of the section P can be determined by the ratio to the difference between the maximum value and the minimum value of the position data in the time series operation data. Further, the factor determination condition may be a condition in which it is possible to inspect that the distribution of the positions of the motor 110 at the time of the feature points is concentrated on a specific position. For example, the number or ratio of feature points deviating from the section P of a certain width, the average and variance of the positions of the feature points at the time, or the correlation coefficient of the frequency and position at the feature points are calculated, and the calculated value is calculated in advance. It may be inspected that it is within the specified range.

Further, as another example, it is known that when the centers of the two axes connected by the coupling 124 are deviated, a characteristic sound is generated at a frequency twice the rotation speed of the coupling 124. .. Therefore, the rotation speed of the coupling 124 is calculated by multiplying the speed of the motor 110 in the operation data at the time when the feature point is generated by the conversion ratio of the gearbox 122, and the calculated rotation speed and the frequency of the feature point are 2. The factor determination unit 18 determines that the centers of the two axes connected by the coupling 124 are deviated when the relationship is in direct proportion to the fold. At this time, instead of the speed of the motor 110, the current of the motor 110 may be acquired as operation data and the values may be accumulated to substitute for the speed.

As an example of yet another condition, when the machine resonates mechanically by driving the motor 110, a resonance sound having a specific resonance frequency f is generated at a specific speed v that excites the mechanical resonance. Therefore, the rotation speed of the motor 110 at the time when the feature point is generated is concentrated in a specific speed or a section V having a certain width, and the frequency of the feature point is concentrated in a specific frequency or a section F having a certain width. Occasionally, the factor determination unit 18 determines that the diagnosis target 100 is generating a sound having a frequency f due to mechanical resonance at a rotation speed v.

Further, when the frequency of the vibration acquired by the operating vibration extraction unit 17 and the time period of the feature point where the frequency f due to mechanical resonance is generated are the same, the rotation of the motor 110 causes the motor 110 to operate at regular intervals. The factor determination unit 18 determines that the vibration component excites mechanical resonance. That is, by comparing the vibration component of the operation data with the peak of the sound, the factor determination unit 18 determines the phenomenon that the mechanical resonance is periodically excited by the vibration component of the drive and the resonance is excited.

In this way, the factor determination unit 18 uses the registered factor determination conditions for the feature point data extracted by the feature point extraction unit 16, or the feature point data and the vibration data extracted by the operation vibration extraction unit 17. On the other hand, by inspecting, the factor of the driving sound depending on the position or speed of the actuator or the driven machine 120 is determined. As for the factor determination conditions to be registered, similar conditions may be arranged in advance and inspected as a binary tree search. By doing so, the number of conditions to be inspected in determining the factor can be reduced, and the discrimination time can be shortened. This completes the process of the drive sound diagnosis method.

The drive

sound diagnosis systems

10 and 10A according to the first embodiment synchronize the time-series sound vibration data of the drive sound emitted by the actuator or the driven machine 120 with the time-series operation data of the operating state of the actuator. , Extract the operation mode. In addition, feature points were extracted from the spectrum of time-series sound vibration data obtained by analyzing the sound vibration data over time, and the operation data at the frequency, time, power, waveform, and time of the feature points were combined. Generate feature point data. Then, the cause of the driving sound was discriminated by comparing the feature point data with the factor discriminating condition prepared in advance. In this way, since the cause of the driving sound is diagnosed based on the sound vibration data and the operation data, the location where the driving sound is generated due to the adhesion of foreign matter to the driven machine 120 can be specified, and the actuator or the driven machine can be identified. The cause of the abnormality depending on the position or speed of 120 can be easily determined. Further, it is possible to determine the cause of the frequency of the generated sound changing according to the speed of the actuator or the driven machine 120. Furthermore, even when the entire device is vibrating, it is possible to determine whether the sound vibration is due to driving.

In addition, the drive

sound diagnosis systems

10 and 10A diagnose the cause of sound or vibration based on the sound vibration data and operation data of the feature points and their generation time. Therefore, even if there is a device that emits a loud operating noise next to the diagnosis target, erroneous diagnosis can be suppressed from the causal relationship between the occurrence time and the operation data.

Further, since the drive

sound diagnosis systems

10 and 10A diagnose the sound generation factor by combining the frequency spectrum of the sound vibration data and the operation data, the operation pattern changes when determining the change of the sound vibration data spectrum. The influence of the change in the spectrum of the sound vibration data due to the above can be removed. In particular, even when the operation pattern changes depending on the work conditions, the influence of the sound vibration data on the spectrum can be removed by the operation data, so that the drive

sound diagnosis systems

10 and 10A perform appropriate diagnosis. be able to.

Further, the drive

sound diagnosis systems

10 and 10A diagnose the cause of the drive sound by substituting the feature amount obtained by time-frequency analysis of the sound vibration data into a conditional expression that does not depend on the drive pattern. Therefore, it is not necessary to prepare operation data at the time of normal or abnormal time in advance. Therefore, even if the configuration of the drive device or the machine is changed, it is possible to perform a general-purpose and immediate diagnosis of the drive sound.

A case where the centers of the two axes connected by the coupling 124 are deviated will be described as an example. It is known that this sound is characteristic of a frequency twice the frequency of the rotation speed of the coupling 124. It is assumed that the drive pattern is changed from the rotation speed of the motor 110 at 10 rotations per second to the rotation speed of the motor 110 at 20 rotations per second due to the change in the usage of the machine. In the technique described in Patent Document 1, the threshold value of the peak frequency of sound when detecting an abnormality must be set for each drive pattern. An example of the peak frequency threshold value is represented by the following equation (1) before the drive pattern is changed and by the following equation (2) after the drive pattern is changed.
Peak frequency before change = 10 x conversion ratio of gearbox 122 x 2 ± error [Hz] ... (1)
Peak frequency after change = 20 x conversion ratio of gearbox 122 x 2 ± error [Hz] ... (2)

On the other hand, in the present embodiment, the peak frequency, which is a feature amount of the sound generated by the deviation of the coupling 124, is represented by the following equation (3).
Peak frequency = frequency of rotation speed x conversion ratio of gearbox 122 x 2 ± error [Hz] ... (3)

That is, the peak frequency of the sound generated by the deviation of the coupling 124 including the equations (1) and (2) is expressed by a conditional expression with the rotation speed of the motor 110 as a variable, and for each drive pattern, in this case. It is not necessary to prepare a conditional expression in advance for each rotation speed of the motor 110. Since the frequency of the rotation speed can be obtained from the acquired operation data, it is possible to determine the deviation of the coupling 124 in any operation pattern by using the equation (3). .. That is, the frequency, which is a feature amount of sound, is determined by substituting it into a conditional expression that does not depend on the drive pattern. Here, the sound generated when the centers of the two axes connected by the coupling 124 are deviated has been described, but similarly, for other factors, the feature amount obtained by analyzing the vibration data over time and frequency is obtained. By substituting into a conditional expression that does not depend on the drive pattern, the cause of the drive sound can be diagnosed.

Furthermore, the cause is diagnosed based on the driving sound emitted by the driven machine 120. By diagnosing using the driving sound, it is possible to make a diagnosis regardless of the machine even when the mechanical rigidity is low.

Further, the drive

sound diagnosis systems

10 and 10A diagnose the cause of sound or vibration by using the result of time-frequency analysis of the sound vibration data. By performing time-frequency analysis, it is possible to reduce erroneous sensor detection or factor diagnosis error due to noise.

Further, the drive

sound diagnosis systems

10 and 10A diagnose the cause of the drive sound based on the feature amount obtained by analyzing the sound vibration data over time and frequency. As a result, the data to be diagnosed can be reduced and the processing time can be reduced. Further, by using the frequency, it is possible to estimate the cause of the driving sound.

Further, in the drive

sound diagnosis systems

10 and 10A, the operation mode extraction unit 14 estimates the operation state of the device based on the operation data and extracts the operation mode. Then, frequency conversion can be performed only on the sound vibration data corresponding to the period during constant speed operation, and the sound vibration data during acceleration / deceleration in which the actuator drive is difficult to stabilize can be discarded. Therefore, compared to the case where the sound vibration data during acceleration / deceleration is also used, the accuracy of the diagnosis is improved, and the sound vibration data unsuitable for the diagnosis is discarded to increase the amount of memory and the processing time used in the arithmetic processing. Can be reduced.

Further, the drive sound

diagnostic systems

10 and 10A determine the period during which the driven machine 120 is not operating based on the operation data, and omit the sound vibration data during the period when the driven machine 120 is not operating, so that the sound is not driven. It is possible to prevent false detection due to. Further, the processing cost can be reduced by omitting the frequency conversion in the state where the driven machine 120 is not operating.

Furthermore, when investigating the cause of the driving sound, the influence of the driving sound due to other phenomena can be eliminated by specifying the operation mode including the driving sound to be investigated. In particular, when a plurality of actuators are provided, it is possible to estimate the actuator that causes the driving noise.

Further, the drive

sound diagnosis systems

10 and 10A determine that the operation mode extraction unit 14 is in constant speed operation for a certain period of time when the sum of the changes in the speed data, which is the operation data, falls within a specific threshold value. By doing so, when the speed of the actuator is within a certain width, the driving sound generated by the driven machine 120 is stable and tends to be in a homogeneous state, so that the factor of the sound can be determined more accurately. .. Further, since the sound in a stable state can be extracted without specifying the speed, the same factor diagnosis can be performed even if the drive pattern is different during constant speed operation. Further, it is possible to extract the sound when the actuator or the driven machine 120 stabilizes at a plurality of speeds, which is desirable for factor estimation.

Further, since the driving

sound diagnosis systems

10 and 10A include a driving vibration extraction unit 17 that extracts a vibration component from the driving data, the factor determination unit 18 can compare the vibration component with the occurrence time of the feature point. As a result, it is possible to determine the phenomenon in which mechanical resonance is periodically excited by the vibration component of the drive and the resonance is excited.

Further, the drive sound

diagnostic systems

10 and 10A detect that the distribution of the positions of the motor 110 at the feature points and the time of the feature points is concentrated at a specific position. By doing so, it is possible to determine whether or not the detected feature point is a feature point generated by a factor that depends on the operation data.

Further, the drive sound

diagnostic systems

10 and 10A extract the feature points as the feature points where the power of the time-series spectrum is the apex of the waveform with respect to the frequency and time. By reducing the number of points for diagnosis by extraction, the processing performed by the factor determination unit 18 can be reduced. By calculating the correlation coefficient between the vertices of the spectrum and the operation data, it is possible to determine whether or not the detected vertices are caused by a factor that depends on the operation data.

Further, the drive

sound diagnosis systems

10 and 10A determine the cause of the drive sound emitted by the driven machine 120, and notify the user of the elevator through the display 133. By doing so, the user of the elevator can detect that there is a problem in the driving sound of the driven machine 120 and take an appropriate action based on the diagnosis result.

Embodiment 2.
FIG. 13 is a diagram showing an example of a hardware configuration when the drive sound diagnosis system according to the second embodiment is applied to the picking unit. In this example, the diagnosis target 200 is a picking unit that transfers the work 291 flowing on the belt conveyor 225 to another belt conveyor 226.

As shown in FIG. 13, the diagnosis target 200 includes a plurality of motors 211,212,213,214, actuators 215, a driven machine 220, and a driving device 230. The drive sound diagnosis system 10B is provided on the diagnosis target 200. The drive sound diagnosis system 10B includes a drive sound detection unit 11, an operating state detection unit 12, a wireless network device 240, a server device 250, a user terminal 260, and a network device 270. The server device 250 and the user terminal 260 are connected via a communication line 280. In the driven machine 220, the vertical direction is the Z direction, the extending direction of the

belt conveyors

225 and 226 is the Y direction, and the directions perpendicular to the Y direction and the Z direction are the X directions in the plane perpendicular to the Z direction. And.

The driven machine 220 is a picking unit. The driven machine 220 includes a linear rail 221, a ball screw 222, a linear guide 223, a head 224, and two

belt conveyors

225 and 226.

The linear rail 221 is a rail that fixes the drive direction of the motor 212, which is a linear motor. In this example, the linear rail 221 is arranged so as to extend in the Y direction on the upper portions of the two

belt conveyors

225 and 226 arranged in parallel. A head 224 is connected to the linear rail 221 via a motor 212, a ball screw 222, and a linear guide 223. The movable direction of the motor 212 when the motor 212 is driven is limited to the extending direction of the linear rail 221. In the example of FIG. 13, the movable direction of the motor 212 is limited to the X direction.

The ball screw 222 is connected to the motor 211, and the rotation of the motor 211 drives the head 224 in the Z direction. The linear guide 223 is a guide that limits the driving direction of the ball screw 222 to the Z direction. The linear guide 223 is fastened to the motor 211 and the ball screw 222, and is fixed to the movable portion of the motor 212. As a result, the ball screw 222 is driven along the linear rail 221 by the drive of the motor 212.

The head 224 is a stage driven in the Z direction by the drive of the ball screw 222. The head 224 has an actuator 215 having a shape extending in the Y direction and having a mechanism for holding the work 291 at the lower portion. The actuator 215 is a vacuum pad that holds the work 291 by a vacuum suction mechanism. The work 291 held by the actuator 215 can be moved up and down by the ball screw 222. The work 291 is a picking target of the picking unit.

The belt conveyor 225 is a feeder that supplies the work 291 from the positive side to the negative side in the Y direction. A motor 213 is connected to the belt conveyor 225. The work 291 on the belt is transported by driving the built-in belt by the rotation of the motor 213.

The belt conveyor 226 is an unloader that transports the work 291 from the positive side to the negative side in the Y direction. A motor 214 is connected to the belt conveyor 226. The work 291 on the belt is transported by driving the built-in belt by the rotation of the motor 214.

The motor 211 is connected to the ball screw 222. The motor 211 is a servomotor that receives the current controlled by the drive device 230 and rotates the shaft.

The motor 212 is connected to the linear rail 221. The motor 212 is a linear servomotor that receives the current controlled by the drive device 230 and drives the motor 212 in the X direction.

The motor 213 is connected to the belt conveyor 225, and the motor 214 is connected to the belt conveyor 226. The

motors

213 and 214 are stepping motors that receive a pulsed electric signal from the drive device 230 and rotate the shaft.

The actuator 215 is a plurality of vacuum pads that receive an electric signal from the drive device 230 and attract or detach the work 291. The work 291 to be picked is held or released by the action of suction or desorption.

The drive device 230 has a plurality of motor drives 231, 232, 233 and a motor control device 234.

The motor drive 231 is connected to the motor 211 with a cable, and supplies power to drive the motor 211 while referring to the rotation angle of the motor 211. The motor drive 232 is connected to the motor 212 with a cable, and supplies power to drive the motor 212 while referring to the position of the motor 212. The motor drive 233 is connected to the

motors

213 and 214 with a cable, and transmits a pulse-like command to the

motors

213 and 214.

The motor control device 234 controls the motor drives 231,232,233 and controls the drive of each motor 211,212,213,214. The motor drives 231, 232, 233 and the motor control device 234 are connected by a cable, and information can be exchanged with each other by communication. Further, the motor control device 234 is connected to the actuator 215 by a cable (not shown), and the suction and attachment / detachment of the work 291 by the actuator 215 are controlled by an electric signal. The vacuum pump is an example of the actuator 215 because it attaches or detaches to the work 291. When the vacuum pump operates, the work 291 is sucked, and when the vacuum pump does not operate, the work 291 is attached and detached.

The drive sound detection unit 11 is a microphone that is arranged adjacent to the driven machine 220 and detects the sound emitted by the diagnosis target 200. The drive sound detection unit 11 is provided with a communication unit 241A. An example of the communication unit 241A is a wireless communication device. The drive sound detection unit 11 sets the detected drive sound with a time stamp which is the time when the sound is detected, and transmits the detected drive sound to the server device 250 via the communication unit 241A.

The operating state detection unit 12 is a logger that is connected to the driving device 230 via a cable and acquires information on the driving device 230 as an operating state. The operation state detection unit 12 appropriately communicates with the drive device 230 to acquire and record various data including the operation state. The operation state detection unit 12 acquires the speed command of the

motors

211, 212, 213, 214 and the suction state of the actuator 215 as the operation state of the diagnosis target 200, and outputs the speed command to the time synchronization unit 13. The binary data of adsorption or desorption is an example of the adsorption state.

The operation state detection unit 12 is provided with a communication unit 241B. An example of the communication unit 241B is a wireless communication device. The operation state detection unit 12 sets the acquired operation state with a time stamp which is the time when the operation state is acquired, and transmits the acquired operation state to the server device 250 via the communication unit 241B.

FIG. 13 shows a case where one drive sound detection unit 11 and one operation state detection unit 12 are provided for the

multi-axis motors

211, 212, 213, 214 and the actuator 215. A plurality of drive sound detection units 11 or a plurality of operation state detection units 12 may be provided. In particular, in the case of the driven machine 220 having a configuration for eliminating interference due to the driving sound of other motors by shielding or the like, the driving sound detecting unit 11 is provided for each of the

motors

211, 212, 213, 214 and the actuator 215. By providing it, more accurate diagnosis can be expected.

The wireless network device 240 is a device that performs wireless communication with the

communication units

241A and 241B. The wireless network device 240 has a communication unit 241C. A wireless LAN (Local Area Network) is constructed by the wireless network device 240 and the

communication units

241A, 241B, 241C, and the wireless network device 240 serves as an access point for the

communication units

241A, 241B, 241C. The wireless network device 240 also has a role of a router at the same time, and relays communication between each terminal and the network by the communication line 280. The

communication units

241A, 241B, 241C and the wireless network device 240 synchronize the time between the wireless devices including the

communication units

241A, 241B, 241C and the wireless network device 240, if necessary. This is because the time used by the drive sound detection unit 11 and the operating state detection unit 12 is set to an accurate time.

In the second embodiment, the function processing unit for determining the factor using the sound vibration data and the operation data is distributed to the server device 250 and the user terminal 260. The server device 250 and the user terminal 260 are connected via a communication line 280.

When the user terminal 260 instructs the server device 250 to execute a factor determination process for the driving sound of the driven machine 220, the server device 250 receives sound vibration data and operation which are time-series data of the driving sound acquired from the driven machine 220. A process of generating device data including feature point data is performed using operation data which is time-series data of states.

FIG. 14 is a block diagram schematically showing an example of the functional configuration of the server device according to the second embodiment. The server device 250 is an information processing device installed on the network by the communication line 280. The server device 250 has the functions of a time synchronization unit 13, an operation mode extraction unit 14, a sound vibration time series spectrum acquisition unit 15, and a feature point extraction unit 16. That is, the server device 250 includes a time synchronization unit 13, an operation mode extraction unit 14, a sound vibration time series spectrum acquisition unit 15, and a feature point extraction unit 16 as applications on the server device 250. In the second embodiment, the configuration in which the operating vibration extraction unit 17 is omitted is illustrated.

Further, the server device 250 includes a communication unit 251 that communicates between the drive sound detection unit 11 and the operation state detection unit 12 and between the user terminal 260.

Further, the server device 250 includes a device data storage unit 252. The device data storage unit 252 is a database such as RDBMS (Relational Database Management System) or Not only SQL. The server device 250 collects device data including the drive sound and the operation state sent to the network from the drive sound detection unit 11 and the operation state detection unit 12 in a database, processes the device data as necessary, and then stores the data. As an example, as described in the first embodiment, the sound vibration data and the operation data are synchronized, the time-series spectrum data of the sound vibration data is acquired, and the feature points are extracted from the spectrum data to obtain the feature points. The device data storage unit 252 stores device data including feature point data, which is a set of operation data at frequency, time, power, waveform, and time of feature points.

FIG. 15 is a diagram showing an example of a record of device data according to the second embodiment. One record of device data includes a registration time, an operation mode, feature point data, and operation data. By analyzing the data collected in the server device 250 by deep learning or the like, it is possible to determine the factors with higher accuracy.

The user terminal 260 instructs the server device 250 to determine the cause of the driving sound of the driven machine 220, performs the factor determination process using the device data received from the server device 250, and displays the result. The user terminal 260 is an information processing device possessed by the user of the driven machine 220. An example of a user terminal 260 is a laptop or desktop personal computer.

FIG. 16 is a block diagram schematically showing an example of the functional configuration of the user terminal according to the second embodiment. The user terminal 260 has the function of the factor determination unit 18 described in the first embodiment. That is, the user terminal 260 includes a factor determination unit 18 as an application on the user terminal 260.

Further, the user terminal 260 includes a communication unit 261, an input unit 262, and a display unit 263. The communication unit 261 communicates with the server device 250 via the communication line 280. As an example, an instruction to execute a factor determination process input from the input unit 262 by the user is transmitted to the server device 250, and various information including device data is received from the server device 250. Further, the communication unit 261 can be connected to the drive sound detection unit 11 and the operation state detection unit 12 to acquire sound vibration data in which the drive sound is converted into time series data and operation data in which the operation state is converted into time series data. ..

The input unit 262 is an input interface with the user. The keyboard or mouse is an example of the input unit 262. The input unit 262 inputs an instruction to execute the factor determination process of the driven machine 220.

The display unit 263 displays information necessary for executing the factor determination process of the driven machine 220. The liquid crystal display is an example of the display unit 263. The display unit 263 displays the result of factor determination or device data acquired via the network.

The network device 270 is a communication device that relays communication between the drive sound detection unit 11, the operating state detection unit 12, the server device 250, and the user terminal 260 provided in the driven machine 220. The router is an example of the network device 270.

As shown in FIG. 13, by configuring the drive sound diagnosis system 10B via the network, the drive sound can be diagnosed even in a remote place different from the factory where the driven machine 220 is installed. ..

The server device 250 and the user terminal 260 for diagnosing the driving sound are realized by the information processing device as described above. FIG. 17 is a block diagram showing an example of the hardware configuration of the server device and the user terminal. The server device 250 and the user terminal 260 include an arithmetic unit 401, a memory 402, a storage device 403, a communication device 404, an input device 405, and a display device 406. The arithmetic unit 401, the memory 402, the storage device 403, the communication device 404, the input device 405, and the display device 406 are connected via the bus line 407.

The arithmetic unit 401 is a processor including a CPU that performs arithmetic processing. The memory 402 functions as a work area for storing data used by the arithmetic unit 401 in the middle of arithmetic processing. The storage device 403 stores computer programs, information, and the like. The communication device 404 has a communication function with other devices connected to the network. The input device 405 receives an input from the operator. The input device 405 is a keyboard, a mouse, or the like. The display device 406 outputs a display screen. The display device 406 is a monitor, a display, or the like. A touch panel in which the input device 405 and the display device 406 are integrated may be used.

The functions of the time synchronization unit 13, the operation mode extraction unit 14, the sound vibration time series spectrum acquisition unit 15, and the feature point extraction unit 16 shown in FIG. 14 are such that the arithmetic unit 401 reads out the computer program stored in the storage device 403. It is realized by executing.

Further, the function of the factor determination unit 18 shown in FIG. 16 is realized by the arithmetic unit 401 reading and executing the computer program stored in the storage device 403.

Next, the operation of the diagnosis target 200 will be described. The user of the picking unit inputs the drive commands of the motors 211,212,213,214 and the actuator 215 to the motor control device 234 in order to convey the work 291 by the picking unit. As an example, the user inputs a command from the user terminal 260 to the motor control device 234. Then, a command is transmitted to the motor control device 234 via the network device 270, the communication line 280, the server device 250, the wireless network device 240, the communication unit 241C, and the communication unit 241B. The motor control device 234 generates and transmits a drive command to the motor drives 231, 232, 233 according to the input command.

The motor drives 231,232,233 control the motor drive current according to the received drive command to drive the motors 211,212,213,214 and the actuator 215.

When the motor 211 is driven, the ball screw 222 connected to the motor 211 rotates, and the head 224 and the actuator 215 connected to the ball screw 222 move in the Z direction.

When the motor 212 is driven, the linear guide 223 connected to the motor 212, the motor 211, and the device connected to the motor 211 move in the X direction.

By driving the motor 213, the belt conveyor 225 rotates, and the work 291 on the belt conveyor 225 moves from the positive side to the negative side in the Y direction. Similarly, when the motor 214 is driven, the belt conveyor 226 rotates, and the work 291 on the belt conveyor 226 moves from the positive side to the negative side in the Y direction.

The actuator 215 holds the work 291 by attracting the work 291 moving on the belt conveyor 225 at the timing of contacting directly under the actuator 215. As a result, even when the actuator 215 is moved upward by driving the motor 211 after suction, the work 291 can be kept in contact with the actuator 215.

After that, the motor 211 rotates to move the actuator 215 upward. At this time, since the work 291 is attracted to the actuator 215, the work 291 also rises as the actuator 215 rises, and is separated from the belt conveyor 225.

Next, by driving the motor 212 on the positive side in the Y direction, the work 291 moves directly above the belt conveyor 226. After that, the work 291 is placed on the belt conveyor 226 by rotating the motor 211 again. The actuator 215 puts the work 291 on the belt conveyor 226 by releasing the work 291 at the timing when the held work 291 comes into contact with the belt conveyor 226. After releasing the actuator 215, the motor 211 and the motor 212 are driven to return the actuator 215 to the starting position. Finally, by rotating the motor 214, the work 291 is carried out toward the negative side in the Y direction of the belt conveyor 226.

In the process of the operation described above, the diagnosis target 200 emits a driving sound when it is driven. The drive sound generated by the diagnosis target 200 and the rotation angles of the

motors

211, 212, 213, and 214 are recorded and acquired by the drive sound detection unit 11 and the operating state detection unit 12, and are recorded and acquired by the server device 250 by wireless communication. Will be sent.

The server device 250 communicates with the drive sound detection unit 11 and the operation state detection unit 12 through the wireless network device 240 at any time to acquire various information about the diagnosis target 200 including the drive sound and the operation state, and the device data storage unit 252. Store information in. As preprocessing stored in the device data storage unit 252, the time synchronization unit 13, the operation mode extraction unit 14, the sound vibration time series spectrum acquisition unit 15 and the feature point extraction unit 16 are used to obtain sound vibration data and operation of the time series data. Process the data and calculate the operation mode and feature point data. The device data storage unit 252 includes the feature point data calculated by the time synchronization unit 13, the operation mode extraction unit 14, the sound vibration time series spectrum acquisition unit 15, and the feature point extraction unit 16, and the time-synchronized operation data of the feature points. And other necessary information about the diagnosis target 200 is stored.

At the request of the user, the user terminal 260 acquires various information of the system required by the user, such as the presence or absence of an abnormality in the system, through the server device 250, and displays it on the display unit 263 of the user terminal 260. In addition, the user can perform the analysis necessary for the maintenance and operation of the diagnosis target 200 and the drive sound diagnosis system 10B by using the database stored in the device data storage unit 252 of the server device 250.

Next, the drive sound diagnosis procedure in the drive sound diagnosis system 10B in the second embodiment will be described. The basic procedure for diagnosing the driving sound is almost the same as that of the first embodiment, but the differences from the first embodiment will be mainly described below.

In the second embodiment, the time synchronization unit 13 of the server device 250 synchronizes the sound vibration data with the operation data by using the time stamps set by the drive sound detection unit 11 and the operation state detection unit 12. This makes it possible to synchronize the sound vibration data and the operation data even when the real-time performance of the communication cannot be ensured due to poor communication quality or the like.

Further, when the operation mode extraction unit 14 of the server device 250 or the like requires data between specific times, it is assumed that the data having the time stamps between the times specified by the time stamps is extracted. By extracting by this method, it is possible to perform analysis as synchronized data even if the acquisition cycles of sound vibration data and operation data are different.

As another method when the detection cycles of sound vibration data and operation data are different, the data is thinned out after filtering so that the cycles are the same, or conversely, the data between them is linearly interpolated. A method of interpolation can be considered.

Further, the operation mode extraction unit 14 sets the operation mode by dividing the process according to which of the plurality of motors 211,212,213,214 and the actuator 215 is mainly operated. Specifically, the operation mode extraction unit 14 is based on the operation data, and based on the speed command of the motors 211,212,213,214 and the data of the crimping state of the actuator 215, which motor 211,212,213,214 or It is determined whether the actuator 215 is operating, and the operation mode is determined. In one example, the motor 213 mainly operates while the work is being carried in, the motors 211 and 212 mainly operate while the work is being lifted, the motor 214 is mainly operated while the work is being carried out, and the actuator 215 is mainly operated during the pump operation. Works on. Therefore, in one example, the operation mode extraction unit 14 divides the operation mode during the work loading, the work lifting, the work unloading, and the pump operation.

The factor determination unit 18 of the user terminal 260 compares the feature point data with the numerical range generated by the server device 250, which is the factor determination condition, and determines the cause of the driving sound. As a method of generating a numerical range in the server device 250, a numerical range may be generated based on a parameter estimated to be a factor based on knowledge, or a numerical range may be generated by machine learning based on operation data and sound vibration data. It may be generated. Further, the server device 250 may dynamically generate the factor determination condition immediately before the factor determination unit 18 determines the condition. In this case, since factor diagnosis based on a more urgent case can be performed, more accurate diagnosis can be performed.

Note that, in the second embodiment, as shown in FIG. 1, the driving vibration extraction unit 17 is not provided. Therefore, the factor determination unit 18 determines the factor by using the feature point data without using the vibration data.

The drive sound diagnosis system 10B according to the second embodiment is driven by comparing the feature point data obtained by time-frequency analysis of the sound vibration data and the operation data with the numerical range generated by the server device 250 for each operation mode. Diagnose the cause of sound. Therefore, it is not necessary to prepare operation data in a normal state or an abnormal state in advance, and even if the configuration of the drive device or the machine is changed, the general-purpose and immediate diagnosis of the drive sound can be performed.

Further, according to the drive sound diagnosis system 10B according to the second embodiment, it is possible to immediately determine the cause of the drive sound generated by the driven machine 220 driven by the plurality of motors 211,212,213,214 and the actuator 215. it can. In particular, it becomes easy to determine which motor drives the cause. As a result, when there is a problem in the driving sound of the driven machine 220, the user can take an appropriate countermeasure from the factor and can shorten the time required for the countermeasure.

Embodiment 3.
FIG. 18 is a block diagram showing an example of the functional configuration of the factor determination unit in the drive sound diagnosis system according to the third embodiment. The drive sound diagnosis system according to the third embodiment uses the factor determination unit 18 of the arithmetic processing unit 140 in the first embodiment to determine the cause of the drive sound by using a learner that has already learned about the factor determination of the drive sound. It is replaced with the part 18A. In the following, the same components as those in the first embodiment are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those in the first embodiment will be described.

The factor determination unit 18A includes a learning result storage unit 31 and a factor inference unit 32. The learning result storage unit 31 stores the learning result obtained by performing machine learning in advance for estimating the factor of the driving sound with respect to the feature point data extracted by the feature point extraction unit 16. The factor inference unit 32 executes arithmetic processing based on the learning result of the learning result storage unit 31.

The factor determination unit 18A receives the feature point data extracted by the feature point extraction unit 16 as an input, and performs arithmetic processing based on the learning result of the learning result storage unit 31 in the factor inference unit 32 to cause the driving sound. Make a judgment.

At this time, the learning result saved in the learning result storage unit 31 may be the result of machine learning using all of the feature point data, or the result of machine learning using a part of the feature point data. You may. Further, the factor inference unit 32 extracts the input data according to the feature point data used at the time of learning of the learning result storage unit 31, and performs arithmetic processing.

Further, examples of the machine learning learning model for deriving the learning result stored in the learning result storage unit 31 include the K-nearest neighbor method, a decision tree, a support vector machine, and a kernel approximation. Further, deep learning may be used as a learning model.

The factor determination unit 18A in the third embodiment determines the factor determination of the driving sound using a learned learner. Thereby, it is possible to provide more accurate factor determination of the driving sound. Further, since the factor determination unit 18A uses a learner that performs machine learning using the same data as the factor determination unit 18, it is possible to perform determination that does not depend on the drive pattern. From this, the learning result stored in the learning result storage unit 31 can be applied to various driving devices.

Embodiment 4.
FIG. 19 is a block diagram showing an example of the functional configuration of the machine learning device of the drive sound diagnosis system according to the fourth embodiment. The machine learning device 50 is a server device 250 and a user terminal 260 having a function processing unit that discriminates factors using sound vibration data and operation data according to the second embodiment, and is provided with a machine learning function. In the following, the same components as those of the first and second embodiments are designated by the same reference numerals, the description thereof will be omitted, and the parts different from those of the first and second embodiments will be described.

The machine learning device 50 learns from a drive sound detection unit 11, an operating state detection unit 12, a sound vibration time series spectrum acquisition unit 15, a feature point extraction unit 16, a factor acquisition unit 51, a factor learning unit 52, and so on. A result storage unit 53 is provided.

The factor acquisition unit 51 acquires data related to the factors that generate the driving sound. The factor acquisition unit 51 may, for example, acquire data relating to the factors that generate the measured driving sound input by the operation of the designer or the user. Alternatively, the factor acquisition unit 51 may acquire the diagnosis result of the driving sound generation factor according to the first embodiment or the second embodiment, or the diagnosis result of the driving sound generation factor in another driving sound diagnosis system. It may be the one to acquire.

The factor learning unit 52 follows a training data set created based on a combination of the feature point data extracted by the feature point extraction unit 16 and the data related to the generation factor of the driving sound acquired by the factor acquisition unit 51. Learn the factors that generate driving noise. As the learning model used in the factor learning unit 52, the learning model mentioned in the third embodiment can be used. As the diagnostic data used in the training data set, those of a plurality of driven machines 220 may be used. In one example, more accurate diagnosis can be made by connecting to a plurality of driven machines 220 via a communication line 280 or the like and collecting diagnostic data. The learning result storage unit 53 stores the learning result by the factor learning unit 52.

FIG. 20 is a block diagram schematically showing an example of the functional configuration of the server device according to the fourth embodiment. The server device 250A is an information processing device installed on the network by the communication line 280. The server device 250A includes a time synchronization unit 13, an operation mode extraction unit 14, a sound vibration time series spectrum acquisition unit 15, a feature point extraction unit 16, a factor learning unit 52, a learning result storage unit 53, and a communication unit. 251 and.

The server device 250A processes the device data including the drive sound and the operation state sent from the drive sound detection unit 11 and the operation state detection unit 12 to the communication line 280 as necessary, and then learns and stores the device data. As an example, first, as in the second embodiment, all or part of the feature point data, which is a set of operation data at the frequency, time, power, waveform, and time of the feature point, is input to the factor learning unit 52. Will be done. Next, the factors that generate the driving sound are acquired from the user terminal 260 via the communication line 280 and the communication unit 251 and combined to form a training data set. Finally, the factor learning unit 52 learns the factors that generate the driving sound using the training data set, and stores the result in the learning result storage unit 53.

FIG. 21 is a block diagram schematically showing an example of the functional configuration of the user terminal according to the fourth embodiment. The user terminal 260A includes a communication unit 261, an input unit 262A, a display unit 263, a factor determination unit 18, and a factor acquisition unit 51.

The factor acquisition unit 51 acquires data on the factors that generate the driving sound from, for example, the user. The acquired data regarding the cause of the driving sound is transmitted to the server device 250A via the communication unit 261.

The input unit 262A is an input interface with the user. The input unit 262A is an interface when the user inputs an instruction to execute a factor determination process of the driven machine 220, and when the user inputs data related to a driving sound generation factor acquired by the factor acquisition unit 51. It becomes.

The drive sound diagnosis system according to the fourth embodiment has both a machine learning function for the factors of the drive sound and a factor discrimination function using the result of the machine learning function. As a result, it is possible to improve the accuracy of the factor diagnosis by advancing the learning while using the factor diagnosis of the driving sound.

Further, by connecting to a plurality of driven machines 220 or user terminals 260A via a communication line 280, more accurate factor diagnosis can be used at a plurality of locations regardless of the installation location of the driven machine 220. .. As a result, it becomes possible to learn a highly versatile diagnosis.

The drive sound diagnosis system according to the fourth embodiment has both a machine learning function for the factors of the drive sound and a factor discrimination function using the result of the machine learning function, but the machine learning device 50 of the drive sound diagnosis system is independently configured. Therefore, factor determination using the learning result may be a function of another device. An example of the factor discriminating device using the learning result has the configuration shown in FIG. 18 of the third embodiment.

The configuration shown in the above-described embodiment shows an example of the content of the present invention, can be combined with another known technique, and is one of the configurations without departing from the gist of the present invention. It is also possible to omit or change the part.

10, 10A, 10B drive sound diagnosis system, 11 drive sound detection unit, 12 operation state detection unit, 13 time synchronization unit, 14 operation mode extraction unit, 15 sound vibration time series spectrum acquisition unit, 16 feature point extraction unit, 17 operation Vibration extraction unit, 18,18A factor determination unit, 31,53 learning result storage unit, 32 factor reasoning unit, 50 machine learning device, 51 factor acquisition unit, 52 factor learning unit, 100,200 diagnosis target, 110, 211,212 , 213,214 motor, 120,220 driven machine, 130,230 drive device, 131,231,232,233 motor drive, 132,234 motor control equipment, 133 display, 140 arithmetic processing unit, 215 actuator, 250, 250A server device, 251,261 communication unit, 252 device data storage unit, 260, 260A user terminal, 270 network equipment, 280 communication line.

Claims

A drive sound detector that detects an actuator or a drive sound that is a mechanical vibration or a sound generated by a driven machine driven by the actuator.
An operating state detection unit that acquires the driving position, driving speed, or force generated by driving the actuator in time series, and
A frequency spectrum corresponding to each time of the sound vibration data, which is the time-series data of the detected driving sound, is calculated, and the calculated power of the frequency spectrum is associated with the frequency and the time to form a set of time-series spectra. The output sound vibration time series spectrum acquisition unit and
A point in which the frequency of the power of the time series spectrum and the waveform with respect to the time satisfy a predetermined condition is extracted as a feature point, and the frequency of the feature point, the time, the waveform of the feature point, and the feature point are extracted. A feature point extraction unit that outputs feature point data that is a set of operation data that is the drive position, drive speed, or force generated by the actuator at the time mentioned above.
For each phenomenon that occurs in the actuator or the driven machine that is the cause of the driving sound, at least one of the frequency and the time including the feature point generated in association with the phenomenon, and the actuator at the time. Comparing the factor determination condition that defines the first numerical range when the combination of the driving position, the driving speed, or the driving force generated by the driving is taken as multidimensional data, and the numerical value of the feature point data. A factor determination unit that determines the cause of the driving sound detected by the operation,
A driving sound diagnostic system characterized by being equipped with.
Further provided with an operation mode extraction unit that divides the operating state of the driven machine into two or more sections separated by time based on the operating data.
The first aspect of claim 1, wherein the sound vibration time series spectrum acquisition unit calculates the frequency spectrum using at least one point of the sound vibration data detected between the start time and the end time of the section. Driving sound diagnostic system.
The driving sound diagnosis system according to claim 2, wherein the operation mode extraction unit extracts a section in which the speed of the actuator is within a certain width for a time longer than a predetermined time.
Further equipped with an operation vibration extraction unit that extracts vibration components from the operation data,
In addition to comparing the numerical value of the feature point data with the first numerical value range, the factor determination unit includes the vibration component and the vibration component generated in association with the phenomenon for each phenomenon. The driving sound diagnostic system according to any one of claims 1 to 3, wherein the cause of the driving sound detected is determined by comparing the range with the range.
The feature point to be inspected by the feature point extraction unit is any one of claims 1 to 4, characterized in that the waveform of the power of the time series spectrum with respect to the frequency and the time is the apex. The drive sound diagnostic system described.
The factor determination unit
A learning result storage unit that stores learning results obtained by performing machine learning for estimating the feature point data and sound factors with respect to the driving data.
A factor inference unit that receives the multidimensional data as an input, executes arithmetic processing based on the learning result stored in the learning result storage unit, and determines the cause of the driving sound.
The driving sound diagnostic system according to any one of claims 1 to 5, wherein the driving sound diagnostic system is characterized by having.
The process of detecting the actuator or the driving sound that is the mechanical vibration or the sound generated by the driven machine driven by the actuator, and
The process of acquiring the drive position, drive speed, or force generated by the drive of the actuator in time series, and
A frequency spectrum corresponding to each time of the sound vibration data, which is the time-series data of the detected driving sound, is calculated, and the calculated power of the frequency spectrum is associated with the frequency and the time to form a set of time-series spectra. Output process and
A point in which the frequency of the power of the time series spectrum and the waveform with respect to the time satisfy a predetermined condition is extracted as a feature point, and the frequency of the feature point, the time, the waveform of the feature point, and the feature point are extracted. A step of outputting feature point data that is a set of operation data that is a drive position, a drive speed, or a force generated by the drive at the time mentioned above.
For each phenomenon that occurs in the actuator or the driven machine that is the cause of the driving sound, at least one of the frequency and the time including the feature point generated in association with the phenomenon, and the actuator at the time. Comparing the factor determination condition that defines the numerical range when the combination of the driving position, the driving speed, or the driving data, which is the force generated by the driving, is regarded as multidimensional data, and the numerical value of the feature point data. The step of determining the cause of the driving sound detected in
A driving sound diagnostic method characterized by including.
A drive sound detector that detects an actuator or a drive sound that is a mechanical vibration or a sound generated by a driven machine driven by the actuator.
An operating state detection unit that acquires the drive position, drive speed, or force generated by the actuator in chronological order.
A frequency spectrum corresponding to each time of the sound vibration data, which is the time-series data of the detected driving sound, is calculated, and the calculated power of the frequency spectrum is associated with the frequency and the time to form a set of time-series spectra. The output sound vibration time series spectrum acquisition unit and
A point in which the frequency of the power of the time series spectrum and the waveform with respect to the time satisfy a predetermined condition is extracted as a feature point, and the frequency of the feature point, the time, the waveform of the feature point, and the feature point are extracted. A feature point extraction unit that outputs feature point data that is a set of operation data that is the drive position, drive speed, or force generated by the actuator at the time mentioned above.
For each phenomenon that occurs in the actuator or the driven machine that is the cause of the driving sound, at least one of the frequency and the time including the feature point generated in association with the phenomenon, and the actuator at the time. A factor learning unit that learns the factors that generate the driving sound based on the combination of the driving position, the driving speed, or the driving data, which is the force generated by the driving, as multidimensional data.
A learning result storage unit that stores the learning results learned in the factor learning unit,
A machine learning device for a driving sound diagnosis system, which comprises a factor acquisition unit for acquiring a factor for generating the driving sound given to the factor learning unit.