RU2696121C1 - Method for 3d printing on cnc equipment with intelligent optimization of modes - Google Patents

Abstract

FIELD: metallurgy.

SUBSTANCE: method includes billet formation by electric arc surfacing in medium of protective gases from layers consisting of metal surfaced beads contacting each other, and finish machining of formed workpiece by removal of excess metal using blade tool. All motions of equipment units are coordinated by CNC system. CNC includes control program containing specified surfacing modes for formation of workpiece. Data of the three-dimensional printing process when forming the workpiece is fixed by means of a fixing equipment complex, comprising a visual control system of the deposition zone of the workpiece, which is configured to illuminate the deposition zone with ultraviolet radiation, radiation filtration, coming from welding-up zone, and obtaining image of welded-up area, system for fixing acoustic signals from formed workpiece, laser system for measuring temperature of welding bath, monitoring system of protective gas supply and its concentration in the printing zone, welding torch movement speed measurement system and surfacing speed of surfacing wire and current measurement system, value and frequency of welding pulses, which is made in form of a single system connected through analogue-to-digital converters with CNC system by means of built-in high-performance computing modules using nVidia CUDA technology or its analogues and neural network of deep training, providing feedback for optimization of modes of electric arc-welding facing in process of three-dimensional printing. Said neural network is configured to autonomously make a decision on optimization of welding surfacing and generation of commands to change parameters of equipment with numerical control.

EFFECT: invention relates to a method of making an article by three-dimensional printing of an electric arc surfaced on CNC equipment.

1 cl, 2 dwg, 1 tbl

Description

The invention relates to a method for manufacturing volumetric parts and structures.

The prior art method for creating three-dimensional objects using electric arc surfacing in a protective gas environment, patent: EP 1354658 A2. The formation of the object according to the specified patent is due to the creation of layers consisting of welded metal rollers in contact with each other. Each subsequent layer is formed on the surface of the previous layer, except for the first layer, which is deposited on the substrate or part of the product mounted on the table. The movement of the welding torch and / or the movement of the table are controlled by a numerical control system (CNC).

The disadvantages of the prior art invention include: low dimensional accuracy and low surface roughness of the formed objects, due to the lack of fine machining, as well as the lack of high-performance computing modules of nVidia CUDA technology or its analogues and deep training of neural networks for monitoring and analysis deposition parameters in order to adjust the modes in the deposition process to improve the quality of the metal structure of the formed object.

Also known from the prior art is a device for creating three-dimensional objects using electric arc surfacing, cutting processing, laser processing and others on a CNC metal cutting machine according to patent WO 2014013247 A1. This device forms a workpiece in the working area of the machine by means of electric arc surfacing, and then performs machining of the printed object by removing excess metal using a blade tool. All movements of the machine nodes are coordinated and controlled using the CNC system.

Among the disadvantages of the prior art invention are the following: the device’s lack of diagnostic and adjustment systems during the formation and further processing of the object, in particular, there is no hardware and software to perform these functions. Also, the device does not include high-performance computing modules of nVidia CUDA technology or its analogues and deep training of neural networks for data analysis in order to find the optimal operating mode.

Moreover, methods of diagnosing welded joints and weld objects are known from the prior art:

According to patent RU2312745C2, a method is described for “monitoring the welding zone of a product during a welding process, during which the welding zone is illuminated with ultraviolet radiation to obtain an image of the welding zone, while the radiation emanating from the welding zone in the direction of the image acquisition apparatus is filtered, characterized in that that they use a band-pass filter for filtering, which transmits radiation near a certain wavelength lying in the ultraviolet wavelength range. " The images of the weld pool obtained during surfacing are processed and analyzed and compared with reference images. Based on the information received, a decision is made to adjust one or more parameters.

Among the disadvantages of the known from the prior art invention should be attributed to: the clarity of the mechanism for adjusting surfacing modes, namely, by means of which the adjustment of modes occurs. Also, the method does not include high-performance computing modules of nVidia CUDA technology or its analogues and deep training of neural networks for image analysis in order to find the optimal operating mode.

In addition to the method described above, a method is known for detecting defects in welds during welding and determining their location from acoustic signals according to patent RU2424510C2. The method includes: “receiving acoustic signals arising in the welding and cooling zone placed on the welded structure along the weld by broadband acoustic transducers, filtering them according to the value of the specified peak amplitude, analog-to-digital conversion, recording the times of arrival of acoustic emission signals to acoustic transducers, calculating the coordinates of the sources of acoustic signals, according to the results of acoustic emission control, they build a picture of localization in the zone of welding and cooling, according to le analysis which judging the quality of the weld and of the degree of danger of defects found in it. "

The disadvantages of the prior art invention include: the inability of the proposed system to make changes to the modes of welding or surfacing in the process of performing operations on welding or surfacing. Also, the method does not include high-performance computing modules of nVidia CUDA technology or its analogues and deep training of neural networks for analyzing acoustic emission data in order to find the optimal operating mode.

The prior art method for automatically controlling the depth of penetration in automatic arc welding according to patent RU 2613255 C1. In which a method is described where reference values of the welding parameter from the group of welding current, welding speed and welding voltage are set, calculation of the calculated temperature value of a given point on the surface of the product and measurement of the temperature of a given point on the surface of the product during welding, after which the difference between the current and given parameters on the basis of which the process is regulated by the differences obtained.

The disadvantages of the prior art invention include the fact that the method does not include other factors capable of influencing the surfacing process such as: the effect of the atmosphere or protective gas on the surfacing zone, the distance of the welding torch from the weld pool. Also, the method does not include high-performance computing modules of nVidia CUDA technology or its analogues and deep training of neural networks for data analysis in order to find the optimal operating mode.

The objective of the invention is to improve the quality of the metal structure of metal products printed according to the technology of electric arc surfacing in a protective gas environment, as well as improving the control system for 3D printing and optimization of modes.

The technical result of the invention is to expand the technological capabilities of the method of 3D printing by electric arc surfacing on CNC equipment.

The technical result is achieved by integrating a system consisting of a set of sensors and other fixing equipment into the CNC equipment that implements 3D printing method for electric arc surfacing: this is a system of visual control of the surfacing zone of the product during the formation of the object, during which the surfacing zone is illuminated with ultraviolet radiation to obtain an image deposition zone, the radiation emanating from the deposition zone in the direction of the image acquisition apparatus is filtered; system for fixing acoustic signals during surfacing; laser system for measuring the temperature of the weld pool; measuring system used in the process of arc welding of shielding gas; a system that collects data about the CNC equipment itself, in particular, the speed of movement of the welding torch, the speed of rewinding of the wire; A system that collects data on the current, magnitude and frequency of welding pulses. All of the above fixing equipment is combined into a single system, which, through analog-to-digital converters, is connected to the high-performance computing modules of nVidia CUDA technology or its analogs and deep training of neural networks with providing feedback for optimizing the electric arc surfacing during 3D printing. In turn, the described system is associated with a CNC system of equipment.

 This system allows you to comprehensively optimize surfacing during 3D printing on CNC equipment, a schematic diagram of its operation is presented in FIG. one.

The proposed method is implemented as follows: The mechanical parts of the CNC equipment bring the print head - 1 to the required distance from the substrate - 2, with the print start command that comes from the CNC system - 3 equipment, the wire - 4 and protective wire begin to flow through the print head into the surfacing zone gas - 5, the welding pulse generator - 6 also turns on, which ignites the electric arc between the substrate and the wire, through which the wire melts and forms the product - 7 on the substrate. At the same time, throughout the entire process of object formation, a system for optimizing 3D printing modes operates, which works as follows: video sensor - 8 collects information about the distance between the layer being formed and the end of the welding torch, data on the width and length of the weld pool; AE-9 sensor registers signals both when printing a part and when it cools; temperature measurement system - 10, by processing the laser radiation supplied to the weld pool, registers the temperature of surfacing; shielding gas supply sensor - 11 collects data on the supply of shielding gas to the print zone and its concentration; speed sensors - 12 transmit data on the wire feed speed and the speed of movement of the print head along the formed layer; wire presence monitoring sensor - 13 transmits data on the presence of wire in the surfacing zone and its remaining quantity; current and voltage monitoring sensor - 14 collects data on the magnitude of the welding current and voltage, the magnitude and duration of the welding pulses.

All signals from the sensors pass through analog-to-digital converters - 15, where they are converted into a digital signal and transmitted to the neuromodule - 16. In the neuromodule, using the control microcontroller - 17 and the neuromorphic processor - 18, it processes all the information received and calculates to make a decision for optimization of the 3D printing process and transmission of commands to the executive bodies through the CNC system - 3 equipment. All data is stored in the cloud - 19 or a separate server storage - 20. The following data are input parameters for the system:

- the distance between the formed layer and the end of the print head;

- data on the width and length of the weld pool;

- data on the magnitude and duration of the welding pulses;

- data on the flow of protective gas into the print zone;

- the current position of the print head along the Z axis;

- data on the presence of wire and its remaining quantity;

- information about the current speed of movement of the print head;

- information about the current wire feed speed;

- data on the temperature in the surfacing zone;

- acoustic signals of the surfacing process

Based on these data, a self-learning neural network can decide on the adjustment of surfacing modes autonomously without human intervention to issue commands to change the parameters to the following executive bodies:

- pulse generator of welding current;

- wire feed mechanism;

- mechanical parts of CNC equipment responsible for the position of the welding torch along the Z axis;

- mechanical parts of CNC equipment responsible for the speed of movement;

- An electric circuit of CNC equipment responsible for switching on and adjusting the magnitude of the welding current and voltage, the supply of wire and protective gas.

All the executive bodies of the equipment are directly connected to the CNC system of the machine (network interaction). Based on the obtained data, the neuromodule can instruct the CNC system to raise or lower the print head to the surfacing zone, increase or decrease the wire feed speed and the speed of movement of the print head along the layer being formed, adjust the magnitude and frequency of welding pulses and the magnitude of current and voltage, change the volume and the concentration of the supplied protective gas, thereby achieving optimal 3D printing modes with stable dimensions of the formed workpiece and metal structure. Moreover, if the equipment does not have enough consumables and energy for high-quality printing, then the system, based on the sensors, will give a command to stop printing with a message indicating that it is necessary to add the missing component to the equipment.

The main neuromodule 16 is a bunch of a control microcontroller 17 and a neuromorphic processor 18 and provides the formation of a setpoint for the control system of CNC equipment. The input layer of a neuromorphic processor receives a digitized signal from a complex of sensors and other fixing equipment described above, in particular, the current value of the fractal dimension of the AE signal. Based on the input data of the neuromorphic processor, it forms a certain state of the output neurons, which determine the state vector for the setting action. Because Since a neuromorphic processor is built into the feedback loop with CNC equipment, the system has the opportunity to adapt to optimal printing modes. One of the optimality criteria is the value of the fractal dimension of the attartor of the AE signal. During operation, all signals from the AE sensor are digitized and sent to the cloud for storage and subsequent retraining of the neural network. Thus, such systems can accumulate experience and exchange it between similar systems. The composition also includes an additional neurochip - 21, trained to identify typical defects in surfacing. To increase the accuracy of determining the fractal dimension DF of the AE signal attractor, special wavelet filters are used to remove the noise component based on the decomposition of the signal at different levels into a useful and harmful component.

In FIG. Figure 2 shows the attractors of AE signals for various surfacing modes, i.e. with optimal (stable) (Fig. 2A) and unstable surfacing process (Fig. 2b). These attractors were obtained by surfacing with 08G2S steel in a protective gas environment. When training the neural network, a series of test surfacing was carried out in various modes and under different conditions, which are shown in table 1.

Table 1

No. Current, A Voltage Clearance burner - substrate, mm one 161 21 eleven 2 152 22 eleven 3 158 24 eleven four 160 24 eleven five 160 24 eleven 6 120 20 eleven 7 72 nineteen eleven eight 102 24 20 9 20 35 20 ten 180 20 five eleven thirty 34 five

At the same time, when setting any initial mode, the system subsequently independently selected the optimal surfacing mode based on current conditions. Through feedback, acting on the executive bodies listed above. In particular, on an example of 08G2S steel surfacing, after training, the system independently found the optimal mode, it became a mode with parameters: 161A, 21B, a gap of 11 mm (see table 1).

Thus, the technological capabilities of the 3D printing method by electric arc welding on CNC equipment have been expanded, the process of determining and assigning optimal 3D printing modes to obtain the best quality of printed blanks and minimize production defects has been implemented.

Claims (1)

A method of manufacturing an article by three-dimensional printing by electric arc welding on CNC equipment, comprising forming an electric arc surfacing blank in a shielding gas medium from layers consisting of welded metal rollers in contact with each other, and performing machining of the formed blank by removing excess metal using a blade tool , while all the movements of the equipment nodes are coordinated by means of the CNC system, including the control program comprising specified deposition modes for forming a preform, characterized in that the data of the three-dimensional printing process during the formation of the preform is recorded using a set of fixing equipment containing a visual control system for the surfacing zone of the workpiece, configured to illuminate the surfacing zone with ultraviolet radiation, filtering radiation emanating from the zone deposition, and obtaining images of the deposition zone, a system for fixing acoustic signals from the workpiece being formed, a laser system for measuring weld pool temperature, a control system for the supply of shielding gas and its concentration in the print zone, a system for measuring the speed of movement of the welding torch and the speed of rewinding of the welding wire and a system for measuring current, magnitude and frequency of welding pulses, which is made in the form of a single system connected via analog-digital converters with a CNC system through built-in high-performance computing modules using nVidia CUDA technology or its analogues and a deep learning neural network providing feedback to optimize the modes of electric arc surfacing in the process of three-dimensional printing, and the aforementioned neural network is performed with the possibility of an autonomous decision on the optimization of surfacing modes and the formation of commands for changing the parameters of CNC equipment.

Images