An abnormal discharge detection system for electric discharge machining

Abstract

This invention discloses an abnormal discharge detection system for electrical discharge machining (EDM), relating to the field of special machining technology. It includes an extraction module for wavelet decomposition of the discharge pulse signal during EDM to calculate the energy spectrum ratio; a monitoring module for constructing a sliding time window and linearly fitting the energy spectrum ratio within the sliding time window to calculate the real-time slope; a decision module for continuously verifying the real-time slope using a counter to determine the abnormal arcing state. This invention introduces a real-time discrimination mechanism combining dynamic trend analysis and multi-dimensional feature fusion to capture the instantaneous amplitude of the discharge signal, continuously track the dynamic change trajectory and statistical distribution law of the instantaneous amplitude, and continuously verify the slope using a counter and dynamic trend threshold. This significantly improves the timeliness and accuracy of fault detection, providing reliable technical support for rapid fault response and closed-loop control in high-precision machining processes.

Description

Technical Field

[0001] This invention relates to the field of special processing technology, and in particular to an abnormal discharge detection system for electrical discharge machining. Background Technology

[0002] In recent years, abnormal discharge phenomena, especially arcing discharge, are prone to occur in actual electrical discharge machining (EDM) due to the accumulation of etched products and changes in the chip removal environment. If these are not detected and suppressed in time, they can lead to workpiece burns, increased electrode wear, and even damage to the machining equipment, seriously affecting machining quality and production efficiency. Therefore, real-time and accurate monitoring and identification of the discharge state is the key to ensuring the stability and safety of EDM.

[0003] Currently, Chinese invention patent application number CN201910590900.9 discloses an abnormal discharge state detection device and method for insulating ceramic coating-metal electrical discharge machining, including: a voltage waveform sampling circuit, a current signal sampling circuit, and a statistical judgment module. The voltage waveform sampling circuit is used to provide a gap voltage sampling signal for the statistical judgment module; the current signal sampling circuit is used to provide a gap current sampling signal for the statistical judgment module; the statistical judgment module judges the abnormal discharge state by statistically analyzing the collected gap current and voltage signals. This invention identifies abnormal discharge states such as short-circuit pulses, non-short-circuit to short-circuit pulses, short-circuit to non-short-circuit pulses, and arcs by statistically analyzing and distinguishing the breakdown delay time and falling edge time of a single pulse waveform and the current waveform. This enables the detection of abnormal discharge states in the EDM (Electrical Discharge Machining) of insulating ceramic coatings with metal, thereby improving the machining quality of EDM. However, existing technologies rely on setting fixed voltage or current thresholds to distinguish between normal discharge and abnormal arcing. This single threshold judgment method is difficult to adapt to the complex changes in working conditions during machining. When there is a fuzzy transition region between normal discharge and abnormal arcing, it is easy to produce misjudgments or omissions. It lacks the ability to dynamically track the trend of signal changes and cannot capture the early characteristics of the discharge state transitioning from normal to abnormal in a timely manner, resulting in detection lag and untimely response, which cannot meet the needs of high-precision machining for rapid identification and processing of abnormal states. Summary of the Invention

[0004] The technical problem solved by this invention is that relying on setting fixed voltage or current thresholds to distinguish between normal discharge and abnormal arcing is difficult to adapt to the complex working conditions during processing. When there is a fuzzy transition area between normal discharge and abnormal arcing, it is easy to make misjudgments or omissions. It lacks the ability to dynamically track the trend of signal changes and cannot capture the early characteristics of the discharge state transitioning from normal to abnormal in time, resulting in detection lag and untimely response, which cannot meet the needs of high-precision processing for rapid identification and processing of abnormal states.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an abnormal discharge detection system for electrical discharge machining, comprising an extraction module, a monitoring module, and a decision module;

[0006] The extraction module is used to perform wavelet decomposition on the discharge pulse signal of electrical discharge machining, and calculate the high-frequency energy and low-frequency energy based on the wavelet decomposition results. The high-frequency energy is divided by the low-frequency energy to obtain the energy spectrum ratio.

[0007] The monitoring module is used to construct a sliding time window, perform linear fitting on the energy spectrum ratio within the sliding time window, calculate the real-time slope, identify the fuzzy transition state based on the energy spectrum ratio, and determine the state of the energy spectrum ratio based on the real-time slope.

[0008] The decision module is used to continuously verify the real-time slope using a counter when the fuzzy transition state is identified and the energy spectrum ratio is in a rapidly decreasing state. When the counter count value is greater than or equal to a preset confirmation threshold and the cumulative time of the continuous verification reaches the preset confirmation duration, the discharge state is determined to be an abnormal arcing state.

[0009] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining according to the present invention, the extraction module includes a decomposition unit, a statistical unit, and a calculation unit.

[0010] The decomposition unit is used to acquire the time-domain gap voltage signal of electrical discharge machining, and detect the amplitude of the time-domain gap voltage signal, and take the voltage waveform data corresponding to the time segment with an amplitude greater than a preset breakdown threshold as the discharge pulse signal.

[0011] The discharge pulse signal is decomposed into five levels using a preset wavelet basis function to obtain the wavelet decomposition result.

[0012] The wavelet decomposition results include low-frequency approximation coefficients and high-frequency detail coefficients;

[0013] The statistical unit is used to extract high-frequency and low-frequency components from the wavelet decomposition results, and to calculate the cumulative square of the coefficient values ​​of the high-frequency and low-frequency components in the corresponding preset analysis time period to obtain the high-frequency energy and low-frequency energy.

[0014] The arithmetic unit is used to divide the high-frequency energy by the low-frequency energy to obtain the energy spectrum ratio, and associate the energy spectrum ratio with the corresponding discharge pulse signal. At the same time, it assigns time sequence tags to each discharge pulse signal according to the order in which the discharge pulse signals are generated, and generates discharge pulse data.

[0015] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the logic for extracting high-frequency and low-frequency components from the wavelet decomposition results includes:

[0016] The high-frequency detail coefficients from the first to the second layer of wavelet decomposition are taken as high-frequency components;

[0017] The high-frequency detail coefficients of the third to fifth layers of wavelet decomposition and the low-frequency approximation coefficients of the fifth layer of wavelet decomposition are taken as the low-frequency components.

[0018] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining according to the present invention, the monitoring module includes a construction unit, a fitting unit, a judgment unit, a definition unit, and a modification unit.

[0019] The construction unit is used to construct a sliding time window whose width covers a continuously preset number of discharge pulse data;

[0020] The sliding time window is updated in steps as the discharge pulse data is generated. Each update removes the earliest discharge pulse data and adds the latest generated discharge pulse signal.

[0021] The fitting unit is used to take the time label of each discharge pulse data in the sliding time window as the independent variable and the energy spectrum ratio corresponding to each discharge pulse data in the sliding time window as the dependent variable, and perform linear fitting using the least squares method to obtain the fitting line. The slope of the fitting line is then calculated as the current real-time slope.

[0022] The judgment unit is used to determine the state of the energy spectrum ratio value based on the real-time slope.

[0023] If the real-time slope is negative, and the absolute value of the real-time slope is greater than or equal to the absolute value of the slope of the previous sliding time window, then the energy spectrum ratio is determined to be in a state of rapid decline.

[0024] If the real-time slope is negative and the absolute value of the real-time slope is less than the absolute value of the slope of the previous sliding time window, then it is determined that the energy spectrum ratio is in a slow decreasing state.

[0025] If the real-time slope is positive, it is determined that the energy spectrum ratio is in an upward state;

[0026] If the real-time slope is 0, it is determined that the energy spectrum ratio is in a stable state.

[0027] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the defining unit is used to define the discharge state of the discharge pulse signal, and the discharge state includes normal discharge state, abnormal arcing state and fuzzy transition state.

[0028] The logic for defining this includes:

[0029] The statistical upper limit of the normal discharge energy spectrum ratio is set as the upper limit threshold, and the statistical lower limit of the abnormal arcing energy spectrum ratio is set as the lower limit threshold.

[0030] If the currently detected energy spectrum ratio is greater than or equal to the upper limit threshold, the current discharge state is identified as a normal discharge state.

[0031] If the currently detected energy spectrum ratio is between the upper and lower threshold values, the current discharge state is identified as a fuzzy transition state.

[0032] If the detected energy spectrum ratio is less than or equal to the lower threshold, the current discharge state is identified as an abnormal arcing state.

[0033] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the modification unit is used to control the step update cycle of the sliding time window to be shortened to half of the current step update cycle setting value when a fuzzy transition state is identified.

[0034] When the discharge state is detected to switch from an ambiguous transition state to a normal discharge state or an abnormal arcing state, the step update cycle of the sliding time window is controlled to gradually increase according to the preset recovery increment until it is restored to the initial set value of the step update cycle.

[0035] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining according to the present invention, the decision module includes a parameter acquisition unit, a threshold calculation unit, an anomaly determination unit, and an output control unit.

[0036] The parameter acquisition unit is used to acquire the discharge current of electrical discharge machining in real time at a preset sampling frequency, as the original sample value, and to perform smoothing filtering on the original sample value to obtain the average discharge current.

[0037] The threshold calculation unit is used to calculate the corresponding dynamic trend threshold based on the magnitude of the average discharge current.

[0038] In a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the calculation formula for the dynamic trend threshold is as follows:

[0039] ;

[0040] in, Indicates the dynamic trend threshold. This represents the preset basic threshold coefficient. Indicates the average discharge current. Indicates the preset reference current. Indicates the adjustment coefficient and >0.

[0041] As a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the abnormality determination unit is used to continuously verify the real-time slope based on the dynamic trend threshold when the current discharge state is identified as a fuzzy transition state and the energy spectrum ratio is in a rapidly decreasing state.

[0042] The logic for continuous verification includes:

[0043] When the real-time slope is less than the negative value of the dynamic trend threshold, the control counter is incremented.

[0044] When the real-time slope is greater than or equal to the negative value of the dynamic trend threshold, the control counter is reset to zero.

[0045] When the counter value is greater than or equal to the preset confirmation threshold, and the cumulative time of the continuous verification reaches the preset confirmation duration, the discharge state is determined to be an abnormal arcing state.

[0046] In a preferred embodiment of the abnormal discharge detection system for electrical discharge machining described in this invention, the output control unit is used to output an interrupt control signal to perform a pulse power supply pause action when the discharge state is identified as an abnormal arcing state.

[0047] The beneficial effects of this invention are as follows: By introducing a real-time discrimination mechanism that combines dynamic trend analysis and multi-dimensional feature fusion, this invention effectively overcomes the inherent defects of traditional fixed threshold methods, such as poor adaptability to changes in operating conditions and fuzzy identification of transition zones. It captures the instantaneous amplitude of the discharge signal, continuously tracks the dynamic change trajectory and statistical distribution law of the instantaneous amplitude, and continuously verifies the slope by combining a counter and a dynamic trend threshold. This enables the identification of characteristic trends in the early stage of abnormal arc formation, achieving accurate perception of the gradual change process of the discharge state. It significantly improves the timeliness and accuracy of fault detection, providing reliable technical support for rapid response and closed-loop control of faults in high-precision machining processes. Attached Figure Description

[0048] Figure 1 This is a basic flowchart of an abnormal discharge detection system for electrical discharge machining provided in one embodiment of the present invention. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0050] Reference Figure 1 As an embodiment of the present invention, an abnormal discharge detection system for electrical discharge machining is provided, including an extraction module, a monitoring module and a decision module;

[0051] The extraction module is used to perform wavelet decomposition on the discharge pulse signal of electrical discharge machining, and calculate the high-frequency energy and low-frequency energy based on the wavelet decomposition results. The high-frequency energy is divided by the low-frequency energy to obtain the energy spectrum ratio.

[0052] The monitoring module is used to construct a sliding time window, perform linear fitting on the energy spectrum ratio within the sliding time window, calculate the real-time slope, identify the fuzzy transition state based on the energy spectrum ratio, and determine the state of the energy spectrum ratio based on the real-time slope.

[0053] The decision module is used to continuously verify the real-time slope using a counter when the state is identified as an ambiguous transition state and the energy spectrum ratio is in a rapidly decreasing state. When the counter value is greater than or equal to the preset confirmation threshold and the cumulative time of continuous verification reaches the preset confirmation duration, the discharge state is determined to be an abnormal arcing state.

[0054] By setting up extraction, monitoring, and decision modules, a dual detection architecture based on energy spectrum ratio and real-time slope was constructed. The high-frequency to low-frequency energy ratio extracted by wavelet decomposition is used as a feature parameter, which can effectively reflect the physical state of the discharge gap. Combined with the linear fitting slope within the sliding time window to determine the trend of state change, this method overcomes the limitations of single threshold judgment. Even under conditions of fluctuating processing parameters or complex operating conditions, it can accurately identify the key node of the transition from normal discharge to abnormal arcing, and realize early warning and accurate judgment of abnormal discharge state.

[0055] The extraction module includes a decomposition unit, a statistical unit, and a calculation unit;

[0056] The decomposition unit is used to acquire the time-domain gap voltage signal of electrical discharge machining and detect the amplitude of the time-domain gap voltage signal. The voltage waveform data corresponding to the time segment with an amplitude greater than the preset breakdown threshold is used as the discharge pulse signal.

[0057] The preset breakdown threshold is set to 80% of the open circuit voltage. The basis for this setting is based on the breakdown characteristics of the EDM gap. 80% of the open circuit voltage is used as the critical point for determining breakdown. This value is higher than the holding voltage and lower than the upper limit of the breakdown voltage, which can effectively identify the spark establishment time and avoid false triggering caused by gap fluctuations.

[0058] The discharge pulse signal is decomposed into five levels using a preset wavelet basis function to obtain the wavelet decomposition results.

[0059] The preset wavelet basis function is set to the Db4 wavelet basis. The basis for this setting is that the waveform shape of the Db4 wavelet has a very high similarity to the transient pulse waveform of electric spark discharge, and it has good localization characteristics in both the time and frequency domains, which can match and extract the non-stationary features in the discharge signal to the greatest extent.

[0060] Starting from the breakdown moment, a data segment containing the complete discharge process and the set buffer margin is extracted as an analysis sample to ensure the integrity of the boundary data required for wavelet decomposition and avoid the edge effect caused by truncation affecting the accuracy of coefficient calculation, thereby realizing the real-time extraction of high-frequency energy and low-frequency energy.

[0061] The wavelet decomposition results include low-frequency approximation coefficients containing signal contour information and high-frequency detail coefficients containing signal detail information;

[0062] The statistical unit is used to extract high-frequency and low-frequency components from the wavelet decomposition results, and to calculate the cumulative square of the coefficient values ​​of the high-frequency and low-frequency components in the corresponding preset analysis time period to obtain the high-frequency energy and low-frequency energy.

[0063] The preset analysis period is set to the duration of a single discharge pulse. The basis for this setting is to lock the analysis period within a single pulse cycle, which ensures that the energy calculation only includes information on the current complete discharge process, eliminates interference between consecutive pulses, and thus accurately reflects the energy distribution characteristics of a single discharge.

[0064] The preset analysis period is configured to be synchronized with the sliding time window in the monitoring module, specifically referring to the time span corresponding to the preset number of consecutive discharge pulses contained within the sliding time window.

[0065] The arithmetic unit is used to divide the high-frequency energy by the low-frequency energy to obtain the energy spectrum ratio used to characterize the discharge state, and associate the energy spectrum ratio with the corresponding discharge pulse signal. At the same time, it assigns time sequence labels to each discharge pulse signal according to the order in which the discharge pulse signals are generated, and generates discharge pulse data.

[0066] The five-layer wavelet decomposition determines the frequency band of each layer based on a preset sampling frequency. When the sampling frequency is 2MHz, the detail coefficients of the first to fifth layers after decomposition correspond to the frequency ranges of 500kHz-1MHz, 250kHz-500kHz, 125kHz-250kHz, 62.5kHz-125kHz, and 31.25kHz-62.5kHz, respectively. This layering method accurately covers the high-frequency components of spark breakdown to the low-frequency components of sustaining discharge.

[0067] The time-domain gap voltage signal is obtained by the decomposition unit and a preset wavelet basis function is selected for five-level wavelet decomposition. The coefficient components representing high-frequency noise and low-frequency trends are accurately separated. The statistical unit performs cumulative square calculation on the coefficient values ​​within a specific time period to quantify the energy characteristics. The calculation unit divides the high-frequency energy by the low-frequency energy to obtain the energy spectrum ratio and associates it with the time sequence label. This wavelet transform-based energy ratio calculation method can more sensitively capture the weak physical changes in the discharge gap, significantly improving the signal-to-noise ratio and anti-interference ability of feature extraction.

[0068] The logic for extracting high-frequency and low-frequency components from wavelet decomposition results includes:

[0069] The high-frequency detail coefficients from the first to the second layer of wavelet decomposition are taken as high-frequency components;

[0070] Used to capture high-frequency burrs generated by normal spark discharge.

[0071] The high-frequency detail coefficients of the third to fifth layers of wavelet decomposition and the low-frequency approximation coefficients of the fifth layer of wavelet decomposition are taken as the low-frequency components.

[0072] The high-frequency detail coefficients of the first to second layers are defined as high-frequency components, which correspond to the extremely short breakdown delay and high-frequency electromagnetic radiation and noise generated at the moment of channel establishment in electrical discharge machining. Their energy reflects the intensity of the discharge. The high-frequency detail coefficients of the third to fifth layers and the low-frequency approximation coefficient of the fifth layer are defined as low-frequency components, which correspond to the relatively stable oscillation of the plasma channel and the ejection of erosion products during the discharge maintenance period. Their energy reflects the stability of the discharge channel. In the calculation process of the statistical unit, the coefficient values ​​of the above high-frequency and low-frequency components within the preset analysis period are squared and summed. That is, the energy is equal to the sum of the squares of the coefficient values ​​at each moment. Through this cumulative square calculation, the complex waveform data is transformed into intuitive numerical features, so that the ratio of high-frequency energy to low-frequency energy can directly characterize the degree of shift of the discharge state from normal spark to abnormal arcing.

[0073] Used to capture the low-frequency band where energy is concentrated when abnormal arcing occurs.

[0074] By using the high-frequency detail coefficients of the first to second layers as high-frequency components and the high-frequency detail coefficients of the third to fifth layers and the low-frequency approximation coefficients of the fifth layer as low-frequency components, this specific frequency band division strategy optimizes the spectral characteristics of electrical discharge machining signals. The high-frequency components can reflect the transient intensity of the spark discharge, while the low-frequency components reflect the stability of the discharge channel. By reasonably configuring the frequency band range, invalid background noise is effectively filtered out, making the calculated energy spectrum ratio more realistically reflect the essence of the discharge, thereby improving the accuracy of anomaly identification.

[0075] The monitoring module includes a construction unit, a fitting unit, a judgment unit, a definition unit, and a modification unit;

[0076] The construction unit is used to construct a sliding time window whose width covers a preset number of consecutive discharge pulse data;

[0077] The preset number is set to 128, which is determined based on the law of large numbers in statistics and the real-time requirements of the system. 128 data points are sufficient to eliminate the influence of random noise from a single pulse to ensure fitting accuracy. At the same time, the window width is small, which can ensure that the response speed to state changes is in the millisecond range.

[0078] The sliding time window is updated in steps as discharge pulse data is generated. Each update removes the earliest discharge pulse data and adds the latest generated discharge pulse signal.

[0079] The fitting unit uses the time-series label of each discharge pulse data within the sliding time window as the independent variable and the energy spectrum ratio corresponding to each discharge pulse data within the sliding time window as the dependent variable. It uses the least squares method to perform linear fitting to obtain the fitted line and calculates the slope of the fitted line as the current real-time slope.

[0080] The judgment unit is used to determine the state of the energy spectrum ratio based on the real-time slope;

[0081] If the real-time slope is negative, and the absolute value of the real-time slope is greater than or equal to the absolute value of the slope of the previous sliding time window, then the energy spectrum ratio is determined to be in a state of rapid decline.

[0082] If the real-time slope is negative and the absolute value of the real-time slope is less than the absolute value of the slope of the previous sliding time window, then the energy spectrum ratio is determined to be in a slow decreasing state.

[0083] If the real-time slope is positive, it is determined that the energy spectrum ratio is in an upward trend;

[0084] If the real-time slope is 0, it is determined that the energy spectrum ratio is in a stable state.

[0085] The width of the sliding time window is set to cover 128 consecutive discharge pulse data. This window width can effectively smooth out random interference from a single pulse and ensure that the response time to state changes is in the millisecond range.

[0086] In the fitting unit, the time series labels 1 to 128 of 128 discharge pulse data are used as independent variables X, and the corresponding energy spectrum ratios are used as dependent variables Y. The slope K of the regression line is calculated using the least squares formula.

[0087] The decision unit determines the state by comparing the current window slope with the previous window slope. If the current slope K is negative and If K is negative and... If K > 0, it is determined to be a slow descent state; if K > 0, it is determined to be an ascending state; if K = 0, it is determined to be a stable state. This logic based on the rate of change of slope can accurately capture the acceleration of the deterioration of the discharge state.

[0088] By using building units to create sliding time windows and updating them step by step as data is generated, the fitting unit uses the least squares method to calculate the slope of the fitted line as the real-time slope. The judgment unit defines four states—rapid decline, slow decline, rise, and stability—based on the sign and absolute value of the slope. This slope calculation method based on least squares linear fitting can extract core indicators reflecting the changing trend from the data sequence, smooth out random interference, and clearly depict the evolution trajectory of the energy spectrum ratio, providing a reliable quantitative basis for judging the deterioration trend of the discharge state.

[0089] The defining unit is used to define the discharge state of the discharge pulse signal, which includes normal discharge state, abnormal arcing state, and ambiguous transition state.

[0090] The logic for defining this includes:

[0091] The statistical upper limit of the normal discharge energy spectrum ratio is set as the upper limit threshold, and the statistical lower limit of the abnormal arcing energy spectrum ratio is set as the lower limit threshold.

[0092] If the detected energy spectrum ratio is greater than or equal to the upper limit threshold, the current discharge state is identified as a normal discharge state.

[0093] If the currently detected energy spectrum ratio is between the upper and lower thresholds, the current discharge state is identified as a fuzzy transition state.

[0094] If the detected energy spectrum ratio is less than or equal to the lower threshold, the current discharge state is identified as an abnormal arcing state.

[0095] The upper and lower thresholds were obtained by conducting a large number of normal discharge and arcing experiments under standard processing parameters: thousands of normal discharge samples were collected, the average value of their energy spectrum ratios was calculated and then added to three times the standard deviation as the upper threshold. Similarly, abnormal arcing samples were collected, their average value was calculated and then subtracted from three times the standard deviation as the lower threshold.

[0096] The region where the energy spectrum ratio is between these two thresholds is defined as a fuzzy transition state. In actual processing, this state usually corresponds to the initial stage where chip removal is not smooth and the discharge gap becomes smaller due to the accumulation of electro-erosion products. By identifying this black-and-white intermediate region separately, we can intervene in monitoring in advance and avoid frequent false alarms of critical states caused by fixed threshold cutting.

[0097] By defining the upper limit of the normal discharge energy spectrum ratio and the lower limit of the abnormal arcing energy spectrum ratio, the current energy spectrum ratio is compared with these two thresholds to define the discharge state. This hierarchical threshold judgment logic clearly distinguishes between normal discharge, fuzzy transition state and abnormal arcing state. In particular, the independent identification of fuzzy transition state enables the system to remain vigilant in the critical region of state transition, providing a clear state entry point for subsequent fine judgment.

[0098] The modification unit is used to control the step update cycle of the sliding time window to be shortened to half of the current step update cycle setting value when a fuzzy transition state is detected.

[0099] When the discharge state is detected to switch from an ambiguous transition state to a normal discharge state or an abnormal arcing state, the step update cycle of the control sliding time window is gradually increased according to the preset recovery increment until it is restored to the initial set value of the step update cycle.

[0100] The preset recovery increment is set to 0.1 pulse cycles. The basis for this setting is to adopt a small step gradual recovery strategy to prevent data jumps caused by sudden changes in window width during state switching. Each increase of 0.1 pulse cycles can make the monitoring frequency smoothly transition to the normal level, balancing the steady decrease of computing load with the continuity of monitoring for abnormal recurrence.

[0101] The default step update cycle of the sliding time window is once for every discharge pulse data generated. When the modification unit detects that it has entered a fuzzy transition state, the step logic is adjusted to trigger a window shift once every 0.5 pulse data generation time interval, i.e. half a pulse cycle. This is equivalent to speeding up the data update rate through interpolation or high-frequency sampling, so as to monitor state changes more intensively.

[0102] Once the discharge state returns to normal discharge or is confirmed as abnormal arcing, the modification unit controls the step update cycle to gradually recover at a rate of 0.1 pulse cycles per cycle until it returns to the initial 1 pulse cycle. This mechanism ensures the highest time resolution during the most risky fuzzy transition period, while reducing computational load to save system resources during the stable period.

[0103] When the modification unit detects a fuzzy transition state, it shortens the step update cycle of the sliding time window. After the state transition is restored, the cycle is gradually increased according to the preset recovery increment until the initial value is restored. This adaptive adjustment mechanism of the time window step cycle enables the system to collect data with higher time resolution and capture rapidly changing details in the fuzzy stage of unstable state. After the state stabilizes, the normal cycle is restored to reduce the computational load. This dynamic adjustment strategy optimizes the utilization efficiency of system resources while ensuring detection sensitivity, and achieves the best balance between response speed and computational overhead.

[0104] The decision module and the monitoring module form a serial verification mechanism. The monitoring module, as a front-end processor, is responsible for outputting the current discharge state and the energy spectrum ratio in real time as intermediate criteria. The decision module, as a back-end arbitrator, receives the intermediate criteria output by the monitoring module and calculates the dynamic trend threshold by combining it with the current parameters it collects. Only when the monitoring module determines that the state meets the preset conditions, that is, it is in a fuzzy transition and rapidly decreasing, will the decision module start the counter for final verification. This hierarchical interaction logic effectively reduces the false trigger rate of the system and avoids unnecessary complex calculations during stable processing.

[0105] The decision module includes a parameter acquisition unit, a threshold calculation unit, an anomaly detection unit, and an output control unit;

[0106] The parameter acquisition unit is used to acquire the discharge current of electrical discharge machining in real time at a preset sampling frequency, which is used as the raw sample value. The raw sample value is then smoothed and filtered to obtain the average discharge current.

[0107] The threshold calculation unit is used to calculate the corresponding dynamic trend threshold based on the magnitude of the average discharge current.

[0108] The preset sampling frequency is set to 2MHz, based on the Nyquist sampling theorem and the bandwidth of the discharge signal. This frequency is at least 4 times the effective high-frequency component to ensure that the steep leading edge and high-frequency details of the discharge pulse can be fully reproduced and to avoid spectral aliasing.

[0109] The smoothing filtering algorithm used for the original sampled values ​​is the five-point moving average method, that is, the average current value at the current moment is the arithmetic mean of the sampled values ​​at the current moment and the previous four moments. Through this weighted averaging process, high-frequency spike interference is filtered out, and an average discharge current that can truly reflect the discharge energy level is obtained. The average discharge current calculated in real time is used as the input variable for subsequent dynamic trend threshold calculation, so that the threshold adjustment can closely follow the actual fluctuation of the processing current and ensure the real-time tracking of the detection.

[0110] The discharge current is collected by the parameter acquisition unit at a preset sampling frequency and smoothed and filtered to obtain the average discharge current. The threshold calculation unit calculates the dynamic trend threshold based on this. This method of introducing the real-time discharge current into the threshold calculation process establishes a dynamic correlation between the processing parameters and the detection standard, so that the judgment standard can be automatically adjusted with the change of the actual processing current. This avoids the problem of fixed thresholds failing under different processing specifications and ensures the applicability and adaptability of the detection strategy under different process conditions.

[0111] The formula for calculating the dynamic trend threshold is:

[0112] ;

[0113] in, Indicates the dynamic trend threshold. This represents the preset basic threshold coefficient. Indicates the average discharge current. Indicates the preset reference current. Indicates the adjustment coefficient and >0.

[0114] The parameters in the formula for calculating the dynamic trend threshold have all been normalized in terms of dimensions.

[0115] The preset basic threshold coefficient is set to -0.05. The setting is based on the critical value that the system can tolerate for the rate of decrease of the energy spectrum ratio under the standard reference processing current. The decrease rate corresponding to -0.05 distinguishes between normal chip removal fluctuations and abnormal arcing trends, and is the benchmark sensitivity for judging the deterioration of the discharge state.

[0116] The preset reference current is set to 15A. The setting is based on the optimal current value of this model of machining power supply when transitioning from roughing to intermediate machining. This serves as the zero point of the working condition for calibrating the detection sensitivity, ensuring that the dynamic threshold calculation has an accurate reference within the normal machining range.

[0117] The adjustment coefficient is set to 0.002. The setting is based on the weight of the influence of current change on the threshold. This value is calibrated through processing experiments so that the threshold is appropriately relaxed during high current processing to adapt to energy fluctuations, and the threshold is strictly tightened during low current fine processing, so as to achieve adaptive matching between detection sensitivity and processing conditions.

[0118] The adjustment coefficient α is greater than 0, indicating that there is a positive correlation between the dynamic trend threshold K and the average discharge current Ia. When the average discharge current Ia increases by 1A, the dynamic trend threshold K increases by 0.002 in absolute value. That is, the slope change range is allowed to be slightly larger when processing with high current, while the slope change is more strictly controlled when processing with low current. Through this linear adjustment model, the adaptive matching between the detection threshold and the processing conditions is realized.

[0119] This linear weighted mathematical model directly establishes a proportional relationship between the dynamic trend threshold and the average discharge current. When the average discharge current increases, the trend threshold increases accordingly, and vice versa. This adaptive threshold algorithm based on the correlation of physical quantities is more in line with the actual physical laws of the processing process than empirical value settings. It effectively eliminates misjudgments caused by current fluctuations and keeps the boundary of anomaly judgment at the optimal position.

[0120] The anomaly determination unit is used to identify the current discharge state as an ambiguous transition state and the energy spectrum ratio is in a rapidly decreasing state. Based on the dynamic trend threshold, it uses a counter to continuously verify the real-time slope.

[0121] The logic for continuous validation includes:

[0122] When the real-time slope is less than the negative value of the dynamic trend threshold, the control counter will increment the count.

[0123] When the real-time slope is greater than or equal to the negative value of the dynamic trend threshold, the control counter is reset to zero.

[0124] When the counter value is greater than or equal to the preset confirmation threshold, and the cumulative time of continuous verification reaches the preset confirmation duration, the discharge state is determined to be an abnormal arcing state.

[0125] The preset confirmation threshold is set to 3. The setting is based on the time constant of arc formation and anti-interference requirements. It requires that the conditions be met in three consecutive tests to ensure that the abnormal state has time continuity, effectively filter out misjudgments caused by random electromagnetic interference, and at the same time take into account the system's rapid response capability.

[0126] The preset confirmation time is set to 20ms. The setting is based on the critical time window required for abnormal arcing to cause burns on the workpiece surface. 20ms can ensure that the arcing state has fully formed and reached the level of irreversible damage, while minimizing the detection delay and achieving rapid protection.

[0127] When the counter continuously verifies the real-time slope, it adopts an anti-bouncing logic of accumulation and clearing: once the real-time slope is less than the negative value of the dynamic trend threshold K, the counter is incremented by 1. If the real-time slope rises back to above the negative value of the threshold, the counter is immediately cleared and reset. Only when the counter value is greater than or equal to the preset confirmation threshold, and the cumulative time experienced by these verifications reaches the preset confirmation duration, is it finally determined to be an abnormal arcing state. This dual confirmation mechanism is equivalent to performing integral judgment on abnormal features in the time dimension, effectively eliminating false alarms caused by electromagnetic interference or extremely short unstable processes.

[0128] When the anomaly detection unit is in a fuzzy transition state and the energy spectrum ratio drops rapidly, it uses a counter to continuously verify the real-time slope. When the slope is less than the negative value of the dynamic trend threshold, the count is incremented; otherwise, it is reset to zero. An abnormal arc is detected when both the count value and the accumulated time meet the standard. This dual confirmation mechanism, which combines the counter and the duration, is equivalent to continuously integrating the abnormal features on the time axis, effectively filtering out false positive signals caused by instantaneous interference. An alarm is only triggered when the downward trend is continuous and stable. This anti-jitter design greatly improves the robustness of the system and ensures the rigor and reliability of the anomaly judgment.

[0129] The output control unit is used to output an interrupt control signal to perform a pulse power supply pause when the discharge state is identified as an abnormal arcing state.

[0130] The output control unit is directly connected to the pulse power controller of the EDM machine tool. It uses a high-priority control signal at the hardware interrupt level. Once the judgment module determines that there is an abnormal arcing state, the output control unit immediately sends a level transition signal to the control logic terminal of the pulse power supply, forcing the drive circuit to stop outputting pulse energy and executing the pulse power supply pause action. This action is usually accompanied by the tool retraction command of the servo control system, which quickly widens the gap between the electrode and the workpiece. Through this millisecond-level rapid cut-off and tool retraction linkage, it can effectively avoid continuous arcing current from burning the workpiece surface and electrode, protect machining accuracy and shorten fault recovery time.

[0131] When an abnormal arcing state is detected by the output control unit, an interrupt control signal is output to execute a pulse power supply pause action. This closed-loop design of detection and control realizes rapid linkage from signal monitoring to the actuator. Once the system determines that an abnormal arcing has occurred, it immediately cuts off the power supply, which can minimize the duration of abnormal discharge, prevent workpiece surface burns or electrode damage, control the processing risk at the bud stage, and provide a solid safety guarantee for high-precision electrical discharge machining.

[0132] The output control unit is equipped with automatic recovery logic. After the output interruption control signal executes the pulse power supply pause action, the output control unit starts an exhaust waiting timer. During this period, it controls the servo system to perform a tool lifting action to remove electrolytic corrosion products. When the exhaust waiting timer ends and the average discharge current drops back to a safe range, the output control unit outputs a recovery pulse signal to control the pulse power supply to resume discharge attempts. If an abnormal arcing state is detected again after the recovery attempt, the exhaust waiting time is extended and the tool lifting height is increased until the machining process is stable.

[0133] This invention effectively overcomes the inherent defects of traditional fixed threshold methods, such as poor adaptability to changes in operating conditions and fuzzy identification of transition zones, by introducing a real-time discrimination mechanism that combines dynamic trend analysis and multi-dimensional feature fusion. It captures the instantaneous amplitude of the discharge signal, continuously tracks the dynamic change trajectory and statistical distribution law of the instantaneous amplitude, and continuously verifies the slope by combining a counter and a dynamic trend threshold. It can identify characteristic trends in the early stage of abnormal arc formation, achieve accurate perception of the gradual change process of discharge state, significantly improve the timeliness and accuracy of fault detection, and provide reliable technical support for rapid response and closed-loop control of faults in high-precision machining processes.

[0134] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0135] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A system for detecting abnormal discharges in electrical discharge machining, characterized in that, It includes an extraction module, a monitoring module, and a decision module; The extraction module is used to perform wavelet decomposition on the discharge pulse signal of electrical discharge machining, and calculate the high-frequency energy and low-frequency energy based on the wavelet decomposition results. The high-frequency energy is divided by the low-frequency energy to obtain the energy spectrum ratio. The monitoring module is used to construct a sliding time window, perform linear fitting on the energy spectrum ratio within the sliding time window, calculate the real-time slope, identify the fuzzy transition state based on the energy spectrum ratio, and determine the state of the energy spectrum ratio based on the real-time slope. The decision module is used to continuously verify the real-time slope using a counter when the fuzzy transition state is identified and the energy spectrum ratio is in a rapidly decreasing state. When the counter value is greater than or equal to a preset confirmation threshold and the cumulative time of the continuous verification reaches a preset confirmation duration, the discharge state is determined to be an abnormal arcing state. The monitoring module includes a construction unit, a fitting unit, a judgment unit, and a definition unit; The construction unit is used to construct a sliding time window whose width covers a continuously preset number of discharge pulse data; The sliding time window is updated in steps as the discharge pulse data is generated. Each update removes the earliest discharge pulse data and adds the latest generated discharge pulse signal. The fitting unit is used to take the time label of each discharge pulse data in the sliding time window as the independent variable and the energy spectrum ratio corresponding to each discharge pulse data in the sliding time window as the dependent variable, and perform linear fitting using the least squares method to obtain the fitting line. The slope of the fitting line is then calculated as the current real-time slope. The judgment unit is used to determine the state of the energy spectrum ratio value based on the real-time slope. If the real-time slope is negative, and the absolute value of the real-time slope is greater than or equal to the absolute value of the slope of the previous sliding time window, then the energy spectrum ratio is determined to be in a state of rapid decline. If the real-time slope is negative and the absolute value of the real-time slope is less than the absolute value of the slope of the previous sliding time window, then it is determined that the energy spectrum ratio is in a slow decreasing state. If the real-time slope is positive, it is determined that the energy spectrum ratio is in an upward state; If the real-time slope is 0, it is determined that the energy spectrum ratio is in a stable state; The defining unit is used to define the discharge state of the discharge pulse signal, which includes normal discharge state, abnormal arcing state, and ambiguous transition state. The logic for defining this includes: The statistical upper limit of the normal discharge energy spectrum ratio is set as the upper limit threshold, and the statistical lower limit of the abnormal arcing energy spectrum ratio is set as the lower limit threshold. If the currently detected energy spectrum ratio is greater than or equal to the upper limit threshold, the current discharge state is identified as a normal discharge state. If the currently detected energy spectrum ratio is between the upper and lower threshold values, the current discharge state is identified as a fuzzy transition state. If the detected energy spectrum ratio is less than or equal to the lower threshold, the current discharge state is identified as an abnormal arcing state.

2. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 1, characterized in that, The extraction module includes a decomposition unit, a statistical unit, and a calculation unit; The decomposition unit is used to acquire the time-domain gap voltage signal of electrical discharge machining, and detect the amplitude of the time-domain gap voltage signal, and take the voltage waveform data corresponding to the time segment with an amplitude greater than a preset breakdown threshold as the discharge pulse signal. The discharge pulse signal is decomposed into five levels using a preset wavelet basis function to obtain the wavelet decomposition result. The wavelet decomposition results include low-frequency approximation coefficients and high-frequency detail coefficients; The statistical unit is used to extract high-frequency and low-frequency components from the wavelet decomposition results, and to calculate the cumulative square of the coefficient values ​​of the high-frequency and low-frequency components in the corresponding preset analysis time period to obtain the high-frequency energy and low-frequency energy. The arithmetic unit is used to divide the high-frequency energy by the low-frequency energy to obtain the energy spectrum ratio, and associate the energy spectrum ratio with the corresponding discharge pulse signal. At the same time, it assigns time sequence tags to each discharge pulse signal according to the order in which the discharge pulse signals are generated, and generates discharge pulse data.

3. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 2, characterized in that, The logic for extracting high-frequency and low-frequency components from the wavelet decomposition results includes: The high-frequency detail coefficients from the first to the second layer of wavelet decomposition are taken as high-frequency components; The high-frequency detail coefficients of the third to fifth layers of wavelet decomposition and the low-frequency approximation coefficients of the fifth layer of wavelet decomposition are taken as the low-frequency components.

4. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 3, characterized in that, The monitoring module also includes a modification unit; The modification unit is used to control the step update cycle of the sliding time window to be shortened to half of the current step update cycle setting value when a fuzzy transition state is detected. When the discharge state is detected to switch from an ambiguous transition state to a normal discharge state or an abnormal arcing state, the step update cycle of the sliding time window is controlled to gradually increase according to the preset recovery increment until it is restored to the initial set value of the step update cycle.

5. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 4, characterized in that, The decision module includes a parameter acquisition unit, a threshold calculation unit, an anomaly determination unit, and an output control unit; The parameter acquisition unit is used to acquire the discharge current of electrical discharge machining in real time at a preset sampling frequency, as the original sample value, and to perform smoothing filtering on the original sample value to obtain the average discharge current. The threshold calculation unit is used to calculate the corresponding dynamic trend threshold based on the magnitude of the average discharge current.

6. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 5, characterized in that, The formula for calculating the dynamic trend threshold is: ； in, Indicates the dynamic trend threshold. This represents the preset basic threshold coefficient. Indicates the average discharge current. Indicates the preset reference current. Indicates the adjustment coefficient and >

0.

7. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 6, characterized in that, The anomaly determination unit is used to continuously verify the real-time slope based on the dynamic trend threshold when the current discharge state is identified as a fuzzy transition state and the energy spectrum ratio is in a rapidly decreasing state. The logic for continuous verification includes: When the real-time slope is less than the negative value of the dynamic trend threshold, the control counter is incremented. When the real-time slope is greater than or equal to the negative value of the dynamic trend threshold, the control counter is reset to zero. When the counter value is greater than or equal to the preset confirmation threshold, and the cumulative time of the continuous verification reaches the preset confirmation duration, the discharge state is determined to be an abnormal arcing state.

8. The electrical discharge detection system for abnormal discharge in electrical discharge machining as described in claim 7, characterized in that, The output control unit is used to output an interrupt control signal to perform a pulse power supply pause action when the discharge state is identified as an abnormal arcing state.

Images