Real-time detection method and system for finishing amount of micro electrode based on voltage signal

Abstract

The invention belongs to the technical field of non-traditional machining, and relates to a voltage signal-based micro electrode dressing amount real-time detection method and system, which are suitable for micron-level precision electrode dressing process detection. The voltage transient characteristics in the initial stage of discharge plasma generation are used as detection input, and real-time, non-contact and high-precision monitoring is achieved. According to the method, an electrode and a finishing block are fixed to a machine tool working area and connected with a positive electrode and a negative electrode correspondingly, an initial interelectrode gap is set through a machine tool touch function, then electric spark discharge machining is started, and gap classification is conducted on collected discharge signals in real time through a deep learning model in the process; and identifying the inter-electrode gap variation and setting an auxiliary electrode-auxiliary workpiece discharge group as a physical verification group. According to the method, the real-time, non-contact and high-precision monitoring and control of the electrode finishing process are realized, and the online correction efficiency and precision of the micro electrode are improved.

Description

Technical Field

[0001] This invention belongs to the field of special processing technology, specifically relating to a method and system for real-time detection of micro-electrode trimming amount based on voltage signals. It is suitable for detecting electrode trimming processes with micron-level precision. Through the collaborative design of deep learning and dedicated hardware, it achieves real-time, non-contact, and high-precision monitoring. Background Technology

[0002] Microelectrode trimming technology is a key process for the online fabrication of high-precision microelectrodes in the field of electrical discharge machining (EDM), and it is indispensable, especially in micron-level precision machining. The core principle of EDM electrode dressing is the same as that of EDM. A pulse voltage (usually 60-300V) is applied between the electrode and the workpiece. When the gap is reduced to the micrometer level (5-50μm) and the electric field strength exceeds the dielectric strength threshold of the working fluid, the working fluid is ionized and broken down, forming a plasma discharge channel. The discharge channel has extremely low resistance, which causes the pulse energy to be released in a concentrated manner in a very short time, which is converted into high heat energy. This causes the temperature of the discharge area to rise sharply to 10,000-15,000℃, exceeding the melting and boiling points of the material. This causes the material between the electrode and the dressing block (or workpiece) to be eroded. By controlling the discharge parameters (current, pulse width, etc.), the material on the electrode surface can be eroded in a controllable and precise manner, thereby correcting the shape, size or surface roughness of the electrode. However, the difference is that the electrode is connected to the positive electrode and the workpiece is connected to the negative electrode during the electrode dressing process. The purpose is to promote a relatively higher erosion rate of the electrode material (positive electrode) and reduce the wear of the dressing block (negative electrode).

[0003] Currently common contact measurement methods require pausing processing to detect electrode trimming, which disrupts continuous discharge and significantly reduces processing efficiency. Another common non-contact laser measurement method suffers from inaccuracies due to the refraction of the working fluid, resulting in large errors that fail to meet accuracy requirements. In signal processing after collecting discharge signals, some traditional methods rely on manually designed features, such as spectral peaks extracted by Fourier transform and pulse width features from wavelet analysis. However, the microsecond-level transient process of discharge signals involves three nonlinear stages: breakdown, sustaining, and extinction. Manual features struggle to quantify the subtle differences in minute gap variations. Furthermore, working fluid jet noise easily interferes with peak voltage characteristics, leading to significant errors. In contrast, the hardware-software co-processing method used in this invention, targeting the effective voltage signal input, enables online, real-time, high-precision detection. Summary of the Invention

[0004] This invention provides a method and system for real-time detection of micro-electrode trimming amount based on voltage signal. Through the coordinated use of physical detection and deep learning detection hardware and software, the accuracy of the trimmed amount during the electrode trimming process is dynamically monitored in real time, improving the trimming accuracy to within the range of ±2μm. Furthermore, it proposes to use the initial stage of discharge plasma generation in the discharge pulse waveform as an effective input signal for the detection model.

[0005] A method and system for real-time detection of micro-electrode trimming amount based on voltage signal, comprising the following steps:

[0006] S.1. Fix the trimming block and electrode, the auxiliary workpiece and the auxiliary electrode, connect the trimming block and the auxiliary workpiece to the negative terminal of the power supply, and connect the electrode and the auxiliary electrode to the positive terminal of the power supply.

[0007] S.2. After touching the trimming block with the electrode, retract 10 micrometers to form the initial gap value. After the auxiliary electrode touches the auxiliary workpiece, retract 12 micrometers.

[0008] S.3. Set the electrical parameters, start applying voltage, and simultaneously open the working fluid nozzle to start jetting. The purpose is to promptly remove the molten metal particles generated by the etching of the electrode and the trimming block surface, effectively avoid abnormal discharge or short circuit caused by the accumulation of electro-erosion products in the gap, and maintain the stability of the working fluid dielectric strength and gap state.

[0009] S.4. When processing starts, the electrodes remain stationary, maintaining an initial gap of x micrometers. The system uses a high-speed synchronous acquisition module to acquire the voltage signal U(t) and current signal I(t) of the discharge circuit in real time at an ultra-high sampling rate of ≥1GSa / s. The built-in deep learning classification model extracts and classifies the real-time electrical signals, classifying them into the preset gap size category (the initial state is classified as the "10 micrometers" category).

[0010] When the amount of erosion on the electrode sidewall reaches a certain amount, the deep learning model automatically classifies the received electrical signal into the next category (the classification interval used in this invention is 2 micrometers), and performs physical verification using the discharge situation between the auxiliary electrode and the auxiliary workpiece. If the verification is successful, the electrode is automatically moved 2 micrometers toward the trimming block to return to the 10-micrometer category.

[0011] Repeat the above operation until the electrode trimming amount reaches the target trimming amount δ, then terminate the processing.

[0012] A method and system for real-time detection of micro-electrode trimming amount based on voltage signal is provided to implement the above method. The system includes a power supply system, an implementation system, a working fluid system, and a prediction system.

[0013] The power supply system consists of an RC pulse power supply composed of four capacitors of different sizes and four resistors. The positive terminal of the power supply is connected to the electrode, and the negative terminal is connected to the workpiece.

[0014] The implementation system includes a dressing block, auxiliary workpiece, workpiece base, workpiece fixing block, electrode, auxiliary electrode, electric spindle, auxiliary axis, electric spindle fixing device, worktable, and CNC Z-axis.

[0015] The working fluid system includes a working fluid nozzle, a solution tank, and the working fluid.

[0016] The prediction system consists of a high-speed signal acquisition module (≥1GSa / s) and a deep learning model. Through an analysis of the signal contribution values ​​at each stage of the discharge waveform, it was found that the classification model primarily relies on the voltage transient characteristics of the initial stage of discharge plasma generation for classification. Figure 2 The waveform of a single discharge pulse signal is shown in the figure. There are three nodes: T1, T2, and T3. T1 represents the starting point of the discharge triggering stage; before T1, the charging stage is underway. T2 represents the starting point of the spark discharge sustaining stage. The signal segment from T1 to T2 represents the spark discharge triggering stage, where the discharge has just begun, the interelectrode medium rapidly ionizes to form a conductive channel, the voltage drops sharply, and the current begins to rise slowly. T3 is the cutoff point of the initial stage of discharge plasma generation. Figure 3 The graph showing the relationship between the contribution value of the discharge pulse signal and time clearly shows that the contribution value of the T1-T3 stage is significantly higher than that of other stages. Here, the T1-T3 stage is named the "initial stage of discharge plasma generation".

[0017] Furthermore, the electrode can be any one of a cylindrical electrode or a chamfered electrode.

[0018] Furthermore, the electrode trimming method can be applied to both stationary and rotating electrode spindles.

[0019] Furthermore, the implementation system also includes an electric spindle and a clamping device. The electric spindle is fixed by an electric spindle fixing device and connected to the electrodes. The clamping device ensures that the electric spindle will not deviate during rotation.

[0020] Furthermore, the electrode is fixed to the electric spindle by a Φ3.4mm chuck, and the electric spindle is bolted to a support plate.

[0021] Furthermore, the implementation system also includes a CNC display screen and a CNC Z-axis, the CNC display screen being mounted on the CNC Z-axis, and the CNC Z-axis being connected to the electrode.

[0022] Furthermore, the auxiliary workpiece and trimming block are stainless steel workpieces measuring 50mm × 50mm × 5mm.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention utilizes a deep learning classification model to monitor the electrode dressing process in real time during electrical discharge machining (EDM). Physical verification is achieved by setting up an auxiliary electrode-auxiliary workpiece group for discharge, effectively preventing the misjudgment risks that may exist with a single deep learning model and improving the reliability and stability of the monitoring system. It overcomes the drawback of traditional contact-based detection methods that require pausing processing, while maintaining dressing accuracy within ±2μm, thus improving electrode dressing efficiency and precision. Furthermore, it proposes prioritizing the initial stage of discharge plasma generation in the discharge pulse waveform as the classification model's selection phase, ensuring processing quality and improving processing efficiency during electrode dressing. Simultaneously, it enhances the service performance and lifespan of the electrodes. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the device structure of the present invention;

[0026] Figure 2 Waveform of discharge pulse signal

[0027] Figure 3 Graph showing the relationship between the contribution value of the discharge pulse signal and time

[0028] Illustration number:

[0029] 1. RC power supply; 2. Electrode; 3. Auxiliary electrode; 4. Trimming block; 5. Auxiliary workpiece;

[0030] 6. Workpiece fixing block; 7. CNC Z-axis; 8. Electric spindle; 9. Auxiliary axis; 10. Electrode fixture; 11. Working fluid nozzle; 12. Electric spindle fixing device; 13. Worktable; 14. Workpiece base;

[0031] 15. Working fluid; 16. Solution tank. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] A method and system for real-time detection of micro-electrode trimming amount based on voltage signal, comprising the following steps:

[0034] S.1. Prepare working fluid 15, a portion of which is sprayed out by the working fluid nozzle during the processing, and the other portion is placed in the solution tank 16;

[0035] S.2 Place the trimming block 4 on the base and fix it with the workpiece fixing block 6. Similarly, place the auxiliary workpiece 5 on the base and fix it with the fixing block, and place it in the solution tank 16. Fix the electrode 2 and the auxiliary electrode 3 respectively with the electrode clamp.

[0036] S.3. Extend electrode 2 and auxiliary electrode 3 to the sides of trimming block 4 and auxiliary workpiece 5 respectively. Move the electric spindle 8 towards trimming block 4 until it touches it. The buzzer will sound an alarm. At this time, the electric spindle 8 will retract 10 micrometers and be positioned as the starting position. The auxiliary spindle 9 will reach a distance of 12 micrometers from the auxiliary workpiece 5 in the same way (the accuracy can be selected by the user, such as 14 micrometers, 16 micrometers, etc.). After each retraction, the auxiliary electrode will be controlled to rotate a certain angle to avoid the influence of fine wear.

[0037] S.4. RC power supply 1 is powered on (voltage is 100-250V, pulse width is 15-50μs, working fluid nozzle 11 starts spraying deionized water, electrode trimming begins, and the deep learning classification model will automatically classify the gap size into the 10-micrometer category based on the electrical signal detected in real time at this time (the discharge waveform voltage U(t) is collected at a sampling rate of ≥1GSa / s).

[0038] As the discharge proceeds, the insulating working fluid is broken down and generates an electric spark. At this time, the material in the side wall area of ​​the electrode 2 facing the trimming block 4 is eroded. As the trimming continues, the gap between the electrode 2 and the trimming block 4 begins to increase. The deep learning classification model will automatically detect the gap change until the gap size reaches the next classification label (12 micrometers in this method). The model will automatically classify it into the 12-micrometer category. At this time, the auxiliary electrode 3 and the auxiliary workpiece 5 will play a physical verification role. The physical verification refers to: monitoring the discharge state of the auxiliary group composed of the auxiliary electrode and the auxiliary workpiece; when the deep learning model classifies the electrical signal of the main discharge group composed of the electrode and the trimming block into the next gap size category, if the auxiliary group discharge circuit is detected to start continuous discharge, the physical verification is determined to be successful; if the auxiliary group discharge circuit is not formed, the verification is determined to be unsuccessful; if the verification is successful, the electrode (2) will be automatically controlled to move d micrometers towards the trimming block and return to the x-micrometer category. At this time, the accuracy of the model classification can be verified, and the function of physical and model dual verification can be played.

[0039] Since the deep learning model can only receive the electrical signals generated by the discharge between the trimming block 4 and the electrode 2, the electrical signal detection is interrupted when the auxiliary workpiece 5 and the auxiliary electrode 3 discharge. At this time, the control system controls the electrode 2 to move 2 micrometers towards the trimming block, and the processing gap becomes 10 micrometers. The discharge between the electrode 2 and the trimming block 4 starts again. The above operation is repeated until the processing is completed. The trimming amount of the electrode is calculated according to the trimming amount per time and the number of trimming times.

[0040] Figure 1The diagram shows a method and system for real-time detection of micro-electrode trimming amount based on voltage signal, including an implementation system, a power supply system, a working fluid system, and a prediction system.

[0041] The implementation system includes a trimming block 4, an auxiliary workpiece 5, a workpiece base 14, a workpiece fixing block 6, an electrode 2, an auxiliary electrode 3, and an electrode fixture 10. The trimming block 4 and the auxiliary workpiece 5 can be made of any conductive and wear-resistant material. They are fixed to the workpiece base 14 and the workpiece fixing block 6. The former is used to trim the electrode, and the latter is used for physical verification. The electrode 2 and the auxiliary electrode 3 can be either a chamfered electrode or a cylindrical electrode.

[0042] The power supply system is an RC pulse power supply 1. The negative terminal of the RC power supply 1 is connected in parallel with the trimmer 4 and the auxiliary workpiece 5, and the positive terminal is connected in parallel with the electrode 2 and the auxiliary electrode 3.

[0043] The working fluid system includes a working fluid nozzle 11, a solution tank 16, and a working fluid 15. The solution tank 16 is used to hold the working fluid, and the working fluid nozzle 11 is used to spray the working fluid during the machining process to remove the erosion material in the machining gap in a timely manner, promote the renewal of the machining gap and the working fluid, and provide machining stability.

[0044] The prediction system includes a high-speed signal acquisition module (≥1GSa / s) and a deep learning model. The model achieves real-time classification using an industrial-grade GPU.

[0045] In this embodiment, both the trimming block 4 and the auxiliary workpiece 5 are made of stainless steel.

[0046] In this embodiment, both electrode 2 and auxiliary electrode 3 are cemented carbide cylindrical electrodes.

[0047] In this embodiment, the system further includes an electric spindle 8, an electric spindle fixing device 12, and an auxiliary shaft 9; the electrode 2 is fixed on the electric spindle 8 by an electrode clamp 10; the electric spindle 8 is fixed by the electric spindle fixing device 12.

[0048] In this embodiment, the system also includes a CNC display screen and a CNC Z-axis 7. The CNC display screen displays the classification category of the deep learning model in real time, and the CNC Z-axis 7 controls the electrode to move toward the trimming block.

[0049] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method and system for real-time detection of micro-electrode trimming amount based on voltage signal, characterized in that, Includes the following steps: S.

1. Fix the trimming block (4) and electrode (2), auxiliary workpiece (5) and auxiliary electrode (3), connect the trimming block (4) and auxiliary workpiece (5) to the negative terminal of the power supply (1), and connect the electrode (2) and auxiliary electrode (3) to the positive terminal; S.

2. After the electrode (2) touches the trimming block (4) by the electric spindle (8), it retracts by x micrometers to form an initial gap. The initial gap x is 10-20 μm and is adaptively selected according to the conductivity of the electrode material. After the auxiliary electrode (3) touches the auxiliary workpiece (5), it retracts by x+d micrometers to form an initial gap. d is the classification interval value, which is 2-10 μm and is adaptively selected according to the trimming accuracy required by the electrode. δ is the target trimming amount of the electrode. S.

3. Set electrical parameters, start applying voltage, and at the same time open the working fluid nozzle (11) to spray working fluid (15) into the working area. The purpose is to remove the molten metal particles generated by the etching of the electrode and the trimming block surface in time, and effectively avoid abnormal discharge caused by the accumulation of electro-erosion products in the gap. S.

4. When the processing starts, the electrode (2) remains stationary, maintaining an initial gap of x micrometers. The built-in deep learning classification model extracts and classifies the real-time electrical signal, classifying it into the preset gap size category. The initial state is classified as the "x micrometers" category. A deep learning model with a transient capture module is used to classify the real-time collected discharge signals, initially classifying them into the "x micrometers" category. When the model determines that the gap increases to x+d micrometers, the discharge of the auxiliary electrode (3) and the auxiliary workpiece (5) is triggered as a physical verification. The physical verification refers to monitoring the discharge state of the auxiliary group composed of the auxiliary electrode and the auxiliary workpiece. When the deep learning model classifies the electrical signal of the main discharge group composed of the electrode and the trimming block into the next gap size category, if an effective discharge pulse is detected in the auxiliary group discharge circuit, the physical verification is determined to be successful. If the auxiliary group discharge circuit is not formed, the verification is determined to be unsuccessful and processing needs to continue. If the verification is successful, the electrode (2) is automatically controlled to move d micrometers towards the trimming block to return to the x micrometers category. The above operation is repeated until the electrode trimming amount reaches the target trimming amount δ and the processing is terminated.

2. The method and system for real-time detection of micro-electrode trimming amount based on voltage signal according to claim 1, characterized in that: In S.2, there are two discharge regions, namely the electrode (2) and the trimming block (4), and the auxiliary electrode (3) and the auxiliary workpiece (5). The former is used to discharge and trim the electrode, and the generated discharge signal is transmitted to the deep learning model. The latter is used to provide physical verification methods to corroborate the model classification.

3. The method for real-time detection of micro-electrode trimming amount based on voltage signal according to claim 1, characterized in that: The deep learning model in S.4 innovatively uses a method of analyzing the contribution value of different stages of the discharge signal before classification to determine the signal stage with the greatest impact on classification in each discharge pulse signal, which is named "initial stage of discharge plasma generation"; and in subsequent deep learning classification tasks, the model prioritizes to identify the electrical signal of the initial stage of discharge plasma generation, thereby speeding up the model classification speed and reducing the real-time monitoring delay time.

4. A method for implementing the method of claim 1, characterized in that: This includes the implementation system, power system, working fluid system, and prediction system; The implementation system includes a trimming block (4), an auxiliary workpiece (5), a workpiece base (14), a workpiece fixing block (6), an electrode (2), an auxiliary electrode (3), and an electrode fixture (10). The trimming block (4) and the auxiliary workpiece (5) are fixed together by the workpiece base (14) and the workpiece fixing block (6). The former is used to trim the electrode, and the latter is used for physical verification. The power supply system is a power supply (1), the negative terminal of which is connected to the trimming block (4) and the auxiliary workpiece (5), and the positive terminal is connected to the electrode (2) and the auxiliary electrode (3); The working fluid system includes a working fluid nozzle (11), a solution tank (16), and a working fluid (15). The solution tank (16) is used to hold the working fluid, and the working fluid nozzle (11) is used to spray the working fluid (15) during the processing. The prediction system includes a high-speed signal acquisition module and the deep learning model as described in claim 3. The model achieves real-time classification with a time of ≤1ms using an industrial-grade GPU.

5. The method and system for real-time detection of micro-electrode trimming amount based on voltage signal according to claim 4, characterized in that: The real-time detection system for the micro-electrode adopts a dual discharge device, wherein the electrode (2) and the trimming block (4) are used for discharge trimming of the electrode, and the auxiliary electrode (3) and the auxiliary workpiece (5) are used for physical detection. The electrode (2) and the auxiliary electrode (3) can be cylindrical electrodes, edge-trimmed electrodes, etc., and the trimming block (4) and the auxiliary workpiece (5) can be any conductive and high-temperature-resistant metal material.

6. The real-time detection system for micro-electrode trimming amount based on voltage signal according to claim 4, characterized in that: The implementation system also includes an electric spindle (8), an electric spindle fixing device (12), and an auxiliary shaft (9); the electrode (2) is fixed on the electric spindle (8) by an electrode clamp (10), and the electric spindle (8) is connected to the electric spindle fixing device (12) by bolts.

7. The method and system for real-time detection of micro-electrode trimming amount based on voltage signal according to claim 4, characterized in that: The implementation system also includes a CNC display screen and a CNC Z-axis (7), the CNC Z-axis being connected to the electrode (2).

8. The method and system for real-time detection of micro-electrode trimming amount based on voltage signal according to claim 3, characterized in that: The working fluid (15) ejected from the working fluid nozzle (11) and the working fluid in the solution tank (16) can be deionized water or EDM oil.

Images