High-speed electric spark small hole machining penetration detection method based on sound signals

Abstract

A high-speed electric spark small hole machining penetration detection method based on sound signals comprises the steps that the sound signals in the high-speed electric spark small hole machining process are collected, noise reduction and filtering are conducted, and then the penetration occurrence effective rate is calculated through sliding counting; and multi-source data fusion is carried out according to the penetration occurrence effective value sequence, the machining position and the voltage signal, the starting time point and the ending time point of the penetration process are comprehensively judged, a machining stop instruction and an electrode rollback movement instruction are generated, and penetration detection and hole site machining are achieved. According to the method, the sound signals generated in the high-speed electric spark small hole machining process are collected and processed in real time, the occurrence moment of penetration is judged based on the change of the characteristics of the sound signals before and after penetration, comprehensive judgment is conducted in combination with a multi-source data fusion method, and therefore penetration detection is achieved; the problem that an original method is poor in reliability and redundancy is solved.

Description

Technical Field

[0001] The present invention relates to a technology in the field of mechanical processing, in particular to a high-speed electric spark small hole processing penetration detection method based on acoustic signals. Background Art

[0002] Electrode wear during high-speed electrospark small-hole machining is high and uncertain, making it difficult for control systems to determine whether penetration has occurred during the machining process. Currently, two common approaches are to set a fixed machining depth, but this method relies on experience and prior experiments, resulting in low reliability and efficiency. Another approach is to process and determine penetration based on the electrical signals generated during the machining process. However, this electrical signal is highly random and susceptible to interference, and the signal source is single and lacks redundancy, making it less reliable. New signal sources and methods are needed to enhance the reliability of penetration detection. Summary of the Invention

[0003] The present invention aims to address the shortcomings of the existing technology that are unable to identify the different stages of high-speed electrospark small hole machining, namely high-energy discharge before penetration, energy drop during penetration, and stable discharge after penetration, and cannot perform feedback control on the machining process. A high-speed electrospark small hole machining penetration detection method based on acoustic signals is proposed. The method collects the acoustic signals generated during the high-speed electrospark small hole machining process and processes them in real time. The moment of penetration is determined based on the changes in the characteristics of the acoustic signals before and after penetration. A comprehensive judgment is performed in combination with a multi-source data fusion method, thereby realizing penetration detection and solving the problems of lack of reliability and redundancy of the original method.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a method for detecting penetration during high-speed electric spark small hole machining based on acoustic signals. The method collects audio signals during the high-speed electric spark small hole machining process, calculates the penetration efficiency by sliding counting after noise reduction filtering, and then performs multi-source data fusion based on the penetration efficiency value sequence, machining position and voltage signal to comprehensively determine the start and end time points of the penetration process, generate a machining stop instruction and an electrode retraction motion instruction, thereby realizing penetration detection and hole machining.

[0006] The penetration occurrence efficiency value sequence is obtained by the following method:

[0007] A) In each acoustic signal processing cycle, the smoothed RMS value of the acoustic signal in each acoustic signal processing cycle is calculated by the RMS smoothing filter algorithm, and the smoothed RMS value is stored in the smoothed RMS value buffer of the signal acquisition and processing module to obtain the smoothed RMS value sequence, specifically: smoothed RMS value Where: i is the sequence number in the smoothed RMS value buffer, M is the smoothing window size used by the RMS smoothing filter algorithm, and the unsmoothed RMS value x[n] is the amplitude of the sound signal after filtering, and N is the RMS window size for calculating the unsmoothed RMS value.

[0008] The non-smoothed RMS value is obtained by saving a mean square sum value sq(i), and adding x[i] to sq(i) each time a new sound signal data x[i] is added to the buffer. 2 , and subtract the oldest x[i-N+1] in the buffer 2 Then, in each cycle, the latest value of the unsmoothed RMS value rms(i) can be obtained based on the latest sq(i) value. This method does not depend on the size of the window width and can support the calculation of the unsmoothed RMS value rms(i) of any length window width;

[0009] The smoothed RMS value is obtained by saving an arithmetic sum value sum(i), adding rms(i) to sum(i) each time a new unsmoothed RMS value rms(i) is added to the buffer, and subtracting the oldest rms(i-M+1) in the buffer. Then, in each cycle, the smoothed RMS value rms can be obtained based on the latest sum(i) value. m The latest value of (i).

[0010] This method does not depend on the size of the window width and can support the smoothed RMS value rms of any length window width. m (i) Calculation.

[0011] B) In each acoustic signal processing cycle, the normalized RMS value of the smoothed RMS value sequence is calculated by the RMS process normalization algorithm, and the normalized RMS value is stored in the normalized RMS value buffer of the signal acquisition and processing module to generate a normalized smoothed RMS value sequence, specifically: normalized RMS value Where: i is the sequence number in the normalized RMS value buffer, i s is the buffer data sequence number after the start of EDM, t b is the reference time offset, fs is the sampling rate, rms m (i) is the data value of the corresponding sequence number in the previously calculated smoothed RMS value buffer, and the process normalized reference average RMS value N mean The sliding average window size for calculating the normalized reference average RMS value of the process, i bo The buffer sequence number when penetration occurs.

[0012] C) In each acoustic signal processing cycle, the penetration occurrence efficiency of the normalized smoothed RMS value sequence in each acoustic signal processing cycle is calculated and stored in the local penetration occurrence efficiency value buffer of the signal acquisition and processing module to generate a penetration occurrence efficiency value sequence, specifically: the efficiency of the sliding counter component Where: N sr is the sliding window size of the sliding counter, the effective value function The value of the sliding counter data sequence i is the number of the data in the sequence, rms n (i) is the value of the normalized RMS value sequence calculated in the previous step at the corresponding data number, rms nt Normalized RMS threshold set for penetration detection, i s The data sequence number to start processing.

[0013] The multi-source data fusion means that in each acoustic signal processing cycle, the acoustic signal, processing speed change signal and gap voltage change signal of the processing control system are collected simultaneously, and stored in the local acoustic signal, processing speed and gap voltage buffer of the signal acquisition and processing module to perform a comprehensive judgment of the three sound sources.

[0014] The three-source comprehensive judgment means that only when at least two groups of signal sources determine penetration can the occurrence and end of penetration be determined, thus completing the penetration detection, and having better redundancy and reliability, wherein:

[0015] i) For the acoustic signal, in each cycle, when the acoustic signal penetration efficiency sr(i) is greater than the set penetration efficiency threshold sr t , then the acoustic signal source is determined to be penetrating;

[0016] ii) For the processing speed change signal, read the processing speed change signal v(i) in the buffer in each cycle and calculate the short-term processing speed fluctuation rate Characterizes whether there are short-term fluctuations, stagnation or abnormalities in the processing speed caused by the penetration process, is the sliding mean of processing speed, N v is the size of the sliding window for processing speed statistics. When the short-term processing speed fluctuation rate e v (i) Greater than the set threshold e ut , then the processing speed change signal source is determined to be penetrating;

[0017] iii) For the gap voltage change signal, in each cycle, read the gap voltage signal u(i) in the buffer and calculate the short-time gap voltage fluctuation rate Characterizes whether there is a short-term fluctuation in the gap voltage caused by the penetration process, where is the sliding mean of the gap voltage, N u is the size of the gap voltage statistical sliding window. When the short-time gap voltage fluctuation rate e u (i) Greater than the set threshold e ut , then the gap voltage change signal source determines penetration. Technical Effects

[0018] The present invention utilizes an RMS smoothing filter algorithm, employs a dynamic sliding signal window to perform sliding RMS calculation processing on the acoustic signal, and performs sliding average filtering, thereby obtaining a more continuous and smoother RMS value curve. Furthermore, due to the characteristics of the sliding algorithm, the algorithm can achieve the same computational effort for any window width, thus ensuring the real-time stability of the algorithm. Secondly, the present invention utilizes an RMS process normalization algorithm to simultaneously eliminate differences in signal characteristics under different parameter conditions and interference on acoustic signal characteristics caused by changes in machining depth, thereby ensuring the stability and consistency of the normalized results. Thirdly, the present invention utilizes the above RMS process normalization processing of the acoustic signal and a sliding counter statistical component to calculate the penetration occurrence efficiency, corresponding to the energy drop stage during the penetration process. This allows accurate distinction between the three different stages of high energy discharge before the penetration process, energy drop during the penetration process, and stable energy discharge after the penetration process, thereby enabling penetration detection and determination. Fourthly, the present invention utilizes multi-source signal fusion to comprehensively determine penetration detection, integrating multiple signal sources including machining speed change signals, gap voltage change signals, and acoustic signals. The machining control system processes and comprehensively determines penetration detection in real time, thereby improving the redundancy and reliability of penetration detection in small-hole EDM machining. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Flowchart of the present invention;

[0020] Figure 2 Schematic diagram of the embodiment system;

[0021] Figure 3 For the embodiment, the waveform diagram and frequency energy distribution diagram of the original sound signal during the processing are collected;

[0022] Figure 4 This is a diagram showing the effects of noise reduction and filtering processing on acoustic signals in the embodiment;

[0023] Figure 5 This is a diagram showing the effect of applying the RMS smoothing filter algorithm to the acoustic signal in the embodiment;

[0024] Figure 6 This is a diagram showing the effect of applying the RMS process normalization algorithm and the sliding counter component statistics to the acoustic signal in the embodiment;

[0025] Figure 7This is a diagram showing the effect of the embodiment on the fusion processing of multi-source signals of machining speed change, gap voltage change signal and acoustic signal and comprehensive penetration detection judgment;

[0026] In the figure: device table 1, workpiece 2, servo processing axis 3, rotating spindle 4, voltage probe 5, pulse power supply 6, sampling and signal processing module 7, acoustic signal sensor 8, electrode 9, processing automatic control system 10. DETAILED DESCRIPTION

[0027] like Figure 2 As shown, a high-speed electric spark small hole machining system involved in this embodiment includes: a pulse power supply 6, a sampling and signal processing module 7, a device table 1, and a servo machining axis 3, a rotating spindle 4, an electrode 9 and a workpiece 2 arranged opposite to the electrode in sequence on the device table 1, wherein: the two poles of the pulse power supply 6 are respectively connected to the rotating spindle 4 and the workpiece 2, and the sampling and signal processing module 7 performs penetration monitoring and generates a feedback signal by collecting the voltage difference between the electrode 9 and the workpiece 2 and the audio signal during the machining process, and outputs it to the machining automatic control system 10 of the servo machining axis 3.

[0028] The device platform 1 serves as a machine tool body for carrying a complete set of processing equipment.

[0029] The audio signal is realized by the acoustic signal sensor 8 arranged on the side of the electrode 9, and the acoustic signal sensor 8 is connected to the sampling and signal processing module 7.

[0030] The workpiece 2 is preferably connected to the positive terminal of the pulse power supply 6; the rotating spindle 4 is connected to the negative terminal of the pulse power supply 6.

[0031] like Figure 1 As shown in FIG. 1 , a high-speed electrospark small hole machining penetration detection method based on acoustic signals in this embodiment of the system is shown, which specifically includes:

[0032] Step 1: Acoustic signal collection, specifically including:

[0033] 1.1 Set the voltage sampling period of the sampling and signal processing module to 1 millisecond, the acoustic signal sampling frequency to 192,000 Hz, the servo control period of the machining automatic control system to 1 millisecond, and the acoustic signal processing period to 1 millisecond;

[0034] 1.2 Use an acoustic signal sensor placed near the processing position to collect the acoustic signal of high-speed EDM small hole processing at a sampling frequency of 192,000 Hz;

[0035] 1.3 The data obtained by the sensor is transmitted to the sampling and signal processing module, and then converted into digital signals and stored in the local ring buffer of the signal acquisition and processing module.

[0036] like Figure 3 The figure shows the original acoustic signal, normalized to an amplitude range of -1 to 1, and its frequency and energy distribution. The upper portion of the figure shows the curve of the original acoustic signal collected in this embodiment over time. Based on the data characteristics, it can be categorized into three stages: 1 - before penetration, 2 - during penetration, and 3 - after penetration. The lower portion shows the energy frequency distribution of the original acoustic signal over time.

[0037] Step 2: Noise reduction and filtering: During each sound signal processing cycle, a 5th-order 7000 Hz to 20000 Hz bandpass filter and a spectral subtraction module are configured through the signal acquisition and processing module to perform noise reduction and filtering on the sound signal in the local ring buffer. After processing, a filtered sound signal is obtained, and then the filtered sound signal is stored in the filtered sound signal buffer of the signal acquisition and processing module.

[0038] like Figure 4 As shown, the comparison between the filtered sound signal after noise reduction and filtering and the unfiltered signal is shown. The left side of the figure is the amplitude and energy distribution data curve of the original sound signal collected in this embodiment, and the right side is the amplitude and energy distribution data curve of the filtered sound signal after the filtering method. It can be seen that low-frequency noise such as environmental noise, air pump noise, flushing noise, and spindle rotation is effectively filtered out.

[0039] Step 3: Calculate the penetration efficiency by sliding count, which includes:

[0040] 3.1 In each acoustic signal processing cycle, the signal acquisition and processing module uses the RMS smoothing filter algorithm to calculate the smoothed RMS value of the acoustic signal in each acoustic signal processing cycle, and stores the smoothed RMS value in the local smoothed RMS value buffer of the signal acquisition and processing module to obtain the smoothed RMS value sequence, specifically: smoothed RMS value Where: i is the sequence number in the smoothed RMS value buffer, M is the smoothing window size used by the RMS smoothing filter algorithm, and the unsmoothed RMS value x[n] is the amplitude of the sound signal after filtering, and N is the RMS window size for calculating the unsmoothed RMS value.

[0041] like Figure 5 As shown, the smoothing window is 100 milliseconds, the RMS window size for calculating the non-smoothed RMS value is 10 milliseconds, and the smoothed RMS value sequence is calculated. The lower part of the figure is the filtered sound signal data in this embodiment. After the RMS smoothing filtering algorithm, the smoothed RMS value data curve is obtained. It can be seen that a relatively continuous and smooth data result curve is obtained.

[0042] 3.2 In each acoustic signal processing cycle, the signal acquisition and processing module uses the RMS process normalization algorithm to calculate the normalized RMS value of the smoothed RMS value sequence, and stores the normalized RMS value in the normalized RMS value buffer of the signal acquisition and processing module to generate a normalized smoothed RMS value sequence, specifically: normalized RMS value Where: i is the sequence number in the normalized RMS value buffer, i s is the buffer data sequence number after the start of EDM, t b is the reference time offset, fs is the sampling rate, rms m (i) is the data value of the corresponding sequence number in the previously calculated smoothed RMS value buffer, and the process normalized reference average RMS value N mean The sliding average window size for calculating the normalized reference average RMS value of the process, i bo The buffer sequence number when penetration occurs.

[0043] like Figure 6 As shown, the reference time offset is 2 seconds, the sliding average window size of the normalized reference average RMS value of the calculation process is 1 second, and the normalized RMS value sequence is calculated. The upper part of the figure is the data curve of the unnormalized smoothed RMS value in this embodiment. It can be seen that it decays with time (increases in processing depth) during the processing; the middle part is the normalized RMS value data curve obtained after processing by the RMS process normalization algorithm. It can be seen that this curve eliminates the interference factor of the unnormalized smoothed RMS value data decaying with time.

[0044] 3.3 In each acoustic signal processing cycle, the signal acquisition and processing module calculates the penetration efficiency of the normalized smoothed RMS value sequence in each acoustic signal processing cycle and stores it in the local penetration efficiency value buffer of the signal acquisition and processing module to generate a penetration efficiency value sequence, specifically: the efficiency of the sliding counter component Where: N sr is the sliding window size of the sliding counter, the effective value function The value of the sliding counter data sequence i is the number of the data in the sequence, rms n (i) is the value of the normalized RMS value sequence calculated in the previous step at the corresponding data number, rms nt Normalized RMS threshold set for penetration detection, i s The data sequence number to start processing.

[0045] Preferably, when the current normalized RMS value is less than the normalized RMS threshold, the sliding counter component is set to a valid operation, that is, the value at the corresponding data number is assigned to 1; otherwise, the sliding counter component is set to an invalid operation, that is, the value at the corresponding data number is assigned to 0.

[0046] The normalized RMS threshold of the penetration detection setting is 0.4.

[0047] like Figure 6 As shown, the sliding window size of the sliding counter is 200 milliseconds, and the resulting penetration efficiency value sequence is obtained. The lower part of the figure is the penetration efficiency value curve obtained by applying the normalized RMS value obtained in the previous step to the sliding counter component statistical method.

[0048] Step 4. Penetration determination based on acoustic signals: In each acoustic signal processing cycle, the machining automatic control system determines the start and end time points of the penetration determination process according to the penetration occurrence efficiency value sequence and generates a machining stop instruction and an electrode retraction movement instruction to complete the machining of a hole.

[0049] The start and end time points of the penetration process are as follows: in the pre-penetration stage, the normalized RMS value sequence before penetration is at a relatively stable and high level, and the corresponding penetration efficiency rate also remains near 0. When the penetration efficiency rate begins to rise until it exceeds the penetration efficiency threshold of 0.6 set in this embodiment, the normalized RMS value shows a continuous downward trend, and the surface acoustic signal intensity weakens. At this time, it is determined that penetration has occurred, the judgment state is switched to penetration, and the discharge machining is continued; in the penetration stage, the normalized RMS value is at a low level, and the penetration efficiency rate is above the set threshold. When the penetration efficiency rate begins to decrease until it falls below the penetration end threshold of 0.6 set in this embodiment, the normalized RMS value shows a continuous upward trend again, indicating an increase in acoustic signal intensity. At this time, it is determined that penetration has ended, the judgment state is switched to penetration completion, and the discharge machining is terminated. The electrode is then controlled to withdraw from the hole position, completing the machining of this hole position.

[0050] like Figure 6 As shown, the peak generated in the penetration efficiency curve during the period of 10 to 11 seconds can be determined, and the acoustic signal characteristics at this time mark the occurrence and end of penetration.

[0051] Step 5. Multi-source signal fusion to determine penetration detection: In each acoustic signal processing cycle, by simultaneously collecting and analyzing the gap voltage change signal, acoustic signal change data, processing position and processing speed change signals, penetration detection is determined through a comprehensive analysis of multiple signal sources. That is, when at least two of the three signal sources determine penetration, the occurrence and end of penetration are determined, and the penetration detection is completed with better redundancy and reliability.

[0052] like Figure 7 As shown, the waveform curve of the electrode feed curve and the gap voltage waveform change curve during the discharge process, the filtered acoustic signal, the normalized RMS value, the penetration efficiency, the servo axis position, the gap voltage, and the gap voltage standard deviation over time are recorded simultaneously, demonstrating the synchronous comprehensive judgment process of multi-source signal fusion.

[0053] like Figure 7 (B) shows the penetration detection data based on the acoustic signal data of this method, where t bo1 The time point is the moment when the penetration ends after the acoustic signal data collected in this embodiment is processed by the method of the present invention.

[0054] like Figure 7 (C) shows the electrode feed curve in this embodiment, describing the control process of the servo axis by the automatic machining system throughout the entire process. The automatic machining control system first touches the electrode edge, then retracts a certain distance after the edge touch is completed, and then begins machining. After machining to a certain depth, the automatic machining control system integrates the signal processing results of the voltage signal and the acoustic signal, determines that penetration has occurred, and continues to feed a fixed margin (to ensure complete penetration of the inclined hole). Finally, the machining is terminated and the electrode is retracted out of the hole position, completing the entire machining process. Around time t5, the machining speed fluctuates and changes abnormally. This is the machining penetration moment determined by the machining position and machining speed change signals.

[0055] like Figure 7 (D) shows the data curve of the voltage signal and the standard deviation of the voltage signal collected in this embodiment, where bo2 At this time point, the gap voltage shows abnormal fluctuations, which is the machining penetration moment determined by the gap voltage signal. bo1 , t5 and t bo2 The three moments are close to each other and all have been judged, so this embodiment judges that penetration occurs near this moment, and the judgment of penetration detection is completed.

[0056] Compared with the existing technology, this method uses the acoustic signal generated during the machining process as the signal source, and realizes penetration detection based on acoustic signals based on technical methods such as RMS smoothing filtering algorithm, RMS process normalization algorithm, and sliding counter component statistics; through penetration detection based on multi-source signal fusion, combined with analysis of changes in machining position, speed, and gap voltage signals, the judgment of penetration detection is comprehensively completed, thereby improving the redundancy, stability, and reliability of penetration detection judgment in EDM small hole machining.

[0057] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principles and purpose of the present invention. The scope of protection of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. All implementation schemes within its scope shall be subject to the constraints of the present invention.

Claims

1. A high-speed electrospark small hole machining penetration detection method based on acoustic signals, characterized in that: By collecting audio signals during high-speed EDM small hole machining, the penetration efficiency is calculated by sliding counting after noise reduction filtering. Multi-source data fusion is then performed based on the penetration efficiency value sequence, machining position and voltage signal to comprehensively determine the start and end time points of the penetration process, generate machining stop instructions and electrode retraction movement instructions, and realize penetration detection and hole machining.

2. The method for detecting penetration of a high-speed electric spark hole machining based on acoustic signals according to claim 1 is characterized in that: The penetration occurrence efficiency value sequence is obtained by the following method: A) In each acoustic signal processing cycle, the smoothed RMS value of the acoustic signal in each acoustic signal processing cycle is calculated by the RMS smoothing filter algorithm, and the smoothed RMS value is stored in the smoothed RMS value buffer of the signal acquisition and processing module to obtain the smoothed RMS value sequence, specifically: smoothed RMS value Where: i is the sequence number in the smoothed RMS value buffer, M is the smoothing window size used by the RMS smoothing filter algorithm, and the unsmoothed RMS value x[n] is the amplitude of the acoustic signal after filtering, and N is the RMS window size for calculating the unsmoothed RMS value; B) In each acoustic signal processing cycle, the normalized RMS value of the smoothed RMS value sequence is calculated by the RMS process normalization algorithm, and the normalized RMS value is stored in the normalized RMS value buffer of the signal acquisition and processing module to generate a normalized smoothed RMS value sequence, specifically: normalized RMS value Where: i is the sequence number in the normalized RMS value buffer, i s is the buffer data sequence number after the start of EDM, t b is the reference time offset, fs is the sampling rate, rms m (i) is the data value of the corresponding sequence number in the previously calculated smoothed RMS value buffer, and the process normalized reference average RMS value rms mean (i)= N mean The sliding average window size for calculating the normalized reference average RMS value of the process, i bo The buffer sequence number when penetration occurs; C) In each acoustic signal processing cycle, the penetration occurrence efficiency of the normalized smoothed RMS value sequence in each acoustic signal processing cycle is calculated and stored in the local penetration occurrence efficiency value buffer of the signal acquisition and processing module to generate a penetration occurrence efficiency value sequence, specifically: the efficiency of the sliding counter component Where: N sr is the sliding window size of the sliding counter, the effective value function The value of the sliding counter data sequence i is the number of the data in the sequence, rms n (i) is the value of the normalized RMS value sequence calculated in the previous step at the corresponding data number, rms nt Normalized RMS threshold set for penetration detection, i s The data sequence number to start processing.

3. The method for detecting penetration of a high-speed electric spark hole machining based on acoustic signals according to claim 2 is characterized in that: The non-smoothed RMS value is obtained by saving a mean square sum value sq(i), and adding x[i] to sq(i) each time a new sound signal data x[i] is added to the buffer. 2 , and subtract the oldest x[i-N+1] in the buffer 2 Then, in each cycle, the latest value of the unsmoothed RMS value rms(i) can be obtained based on the latest sq(i) value. This method does not depend on the size of the window width and can support the calculation of the unsmoothed RMS value rms(i) of any length window width; The smoothed RMS value is obtained by saving an arithmetic sum value sum(i), adding rms(i) to sum(i) each time a new unsmoothed RMS value rms(i) is added to the buffer, and subtracting the oldest rms(i-M+1) in the buffer. Then, in each cycle, the smoothed RMS value rms can be obtained based on the latest sum(i) value. m The latest value of (i).

4. The method for detecting penetration of a high-speed electric spark hole machining based on acoustic signals according to claim 1 is characterized in that: The multi-source data fusion means that in each acoustic signal processing cycle, the acoustic signal, processing speed change signal and gap voltage change signal of the processing control system are collected simultaneously, and stored in the local acoustic signal, processing speed and gap voltage buffer of the signal acquisition and processing module to perform a comprehensive judgment of the three sound sources.

5. The method for detecting penetration of a high-speed electric spark hole machining based on acoustic signals according to claim 4 is characterized in that: The three-source comprehensive judgment means that only when at least two groups of signal sources determine penetration can the occurrence and end of penetration be determined, thus completing the penetration detection, and having better redundancy and reliability, wherein: i) For the acoustic signal, in each cycle, when the acoustic signal penetration efficiency sr(i) is greater than the set penetration efficiency threshold sr t , then the acoustic signal source is determined to be penetrating; ii) For the processing speed change signal, read the processing speed change signal v(i) in the buffer in each cycle and calculate the short-term processing speed fluctuation rate Characterizes whether there are short-term fluctuations, stagnation or abnormalities in the processing speed caused by the penetration process, is the sliding mean of processing speed, N v is the size of the sliding window for processing speed statistics. When the short-term processing speed fluctuation rate e v (i) Greater than the set threshold e ut , then the processing speed change signal source is determined to be penetrating; iii) For the gap voltage change signal, in each cycle, read the gap voltage signal u(i) in the buffer and calculate the short-time gap voltage fluctuation rate Characterizes whether there is a short-term fluctuation in the gap voltage caused by the penetration process, where is the sliding mean of the gap voltage, N u is the size of the gap voltage statistical sliding window. When the short-time gap voltage fluctuation rate e u (i) Greater than the set threshold e ut , then the gap voltage change signal source determines penetration.

6. The method for detecting penetration of a high-speed electric spark hole machining based on acoustic signals according to claim 1 is characterized in that: The hole processing is achieved by a high-speed electric spark small hole processing system, which includes: a pulse power supply, a sampling and signal processing module, a device table, and a servo processing axis, a rotating spindle, an electrode, and a workpiece arranged on the device table in sequence, wherein: the two poles of the pulse power supply are respectively connected to the rotating spindle and the workpiece, and the sampling and signal processing module collects the voltage difference between the electrode and the workpiece and the audio signal during the processing process to perform penetration monitoring and generate a feedback signal and output it to the servo processing axis processing automatic control system.

Images