A method and system for diagnosing faults of an electric discharge machine

Abstract

The application relates to the technical field of machine tool fault diagnosis, and discloses an electric spark machine tool fault diagnosis method and system, an electric spark machine tool fault diagnosis method, which comprises the following steps: step S101, obtaining an electric conductivity characteristic quantity and a machining liquid characteristic quantity through pretreatment; step S102, obtaining a balance coefficient by executing a micro-process stimulation on the machine tool; step S103, obtaining an equivalent coupling driving vector through calculation; step S104, obtaining a hysteresis characteristic quantity through calculation; and step S105, obtaining the running state of the electric spark machine tool through a fault diagnosis model. According to the application, the electric conductivity and the machining liquid characteristic quantity are obtained through orthogonal decoupling, the balance coefficient is obtained by combining the micro-process stimulation, the equivalent coupling driving vector is constructed, and the hysteresis characteristic quantity is extracted, so that the running state of the electric spark machine tool can be accurately diagnosed, targeted maintenance strategies can be provided for different fault types, the blindness of maintenance can be avoided, the fault disposal efficiency can be greatly improved, and the risk of equipment damage can be reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine tool fault diagnosis, more particularly, it relates to a spark machine tool fault diagnosis method and system. BACKGROUND

[0002] The spark machine tool realizes machining by discharging pulse to erode the workpiece, and its running stability directly depends on the machining liquid state and the discharge environment. The conductivity of the machining liquid and the medium vaporization of the machining area are the core factors affecting the discharge state. The conductivity is affected by the ion concentration of the machining liquid and the gap temperature, and high concentration or temperature rise will exacerbate the arc risk; the medium vaporization of the machining area is affected by the flow field pressure and the nozzle state, and bubble accumulation will destroy the discharge uniformity and easily cause abnormalities.

[0003] In actual machining process, the conductivity and the medium vaporization of the machining area exist natural coupling, for example, temperature rise not only changes the conductivity, but also affects the vaporization rate of the machining liquid, and indirectly changes the medium vaporization; and bubble accumulation may hinder heat diffusion, further push up the temperature, forming a vicious cycle. This coupling makes the fault performance complex, and the existing spark machine tool fault diagnosis method mostly depends on a single index, such as arc proportion or conductivity absolute value, without considering the hysteresis characteristics of the coupling of the two, that is, when an abnormality occurs, it is difficult to judge whether the electric conduction temperature rise dominates the fault (for example, the ion concentration is too high) or the machining liquid dominates the fault (for example, the bubble is not discharged smoothly), leading to blind repair, for example, mistakenly considering the arc caused by bubble accumulation as a conductivity problem, and blindly replacing the resin will exacerbate the flow field disturbance, and instead, the fault is worsened.

[0004] Therefore, there is an urgent need for a spark machine tool fault diagnosis method to solve the above technical problems. SUMMARY

[0005] The present application provides a spark machine tool fault diagnosis method and system to solve the technical problems in the background art.

[0006] The present application provides a spark machine tool fault diagnosis method, comprising the following steps:

[0007] Step S101, in a preset time period, the running parameters of the spark machine tool are collected at a preset time interval, and the running parameters are preprocessed to obtain the conductivity feature quantity and the machining liquid feature quantity;

[0008] The running parameters include: display conductivity, gap temperature, acoustic emission original signal and nozzle front pressure;

[0009] Step S102, first orthogonal decoupling of the conductivity characteristic quantity and the processing liquid characteristic quantity, obtaining the conductivity channel characteristic vector and the processing liquid channel characteristic vector, then obtaining the balance coefficient by performing micro-process stimulation on the machine tool;

[0010] Step S103, adjusting the machine tool process parameters, calculating the equivalent coupling driving vector according to the conductivity channel characteristic vector, the processing liquid channel characteristic vector and the balance coefficient, and marking the up and down of each element value;

[0011] Step S104, extracting the uplink switching point and downlink switching point of the equivalent coupling driving vector, and calculating the hysteresis characteristic quantity according to the uplink switching point and downlink switching point;

[0012] The hysteresis characteristic quantity includes: main loop width, hysteresis loop area, critical slope ratio and recovery hysteresis time;

[0013] Step S105, inputting the hysteresis characteristic quantity into the fault diagnosis model, and outputting the running state of the electric spark machine tool;

[0014] The running state includes: normal state, electric conduction temperature rise dominant state and processing liquid dominant state.

[0015] Further, the pre-processing of the conductivity characteristic quantity and the processing liquid characteristic quantity includes the following steps:

[0016] Step S201, obtaining the median of the running parameters of the electric spark machine tool within the baseline window;

[0017] The baseline window represents the time period when the electric spark machine tool is in a stable machining state, and the size of the baseline window is a self-defined parameter;

[0018] Step S202, the temperature correction coefficient is an exponential growth factor with a natural constant as the base, and its index is equal to the product of the gap temperature minus the standard temperature and the temperature coefficient;

[0019] The standard temperature is a self-defined parameter; the temperature coefficient is obtained by fitting the mapping relationship between the gap temperature and the conductivity characteristic quantity through the least squares method;

[0020] Step S203, the conductivity characteristic quantity is equal to the ratio of the display conductivity multiplied by the temperature correction coefficient to the median of the conductivity within the baseline window;

[0021] Step S204, the processing liquid characteristic quantity is equal to the weighted sum of the acoustic emission original signal characteristic quantity and the nozzle front pressure characteristic quantity;

[0022] The acoustic emission original signal characteristic quantity is equal to the square of the amplitude of the acoustic emission original signal at each time within a preset time period, and then the acoustic emission energy is obtained by integrating the time, and then divided by the median of the acoustic emission energy within the baseline window;

[0023] The nozzle front pressure characteristic quantity is equal to the square of the nozzle front pressure at each time point in the preset time period, and the square root of the integral of the time is obtained after being divided by the length of the preset time period. The effective value of the nozzle front pressure is obtained, and then divided by the median value of the effective value of the nozzle front pressure in the baseline window.

[0024] The weighting coefficients of the acoustic emission original signal characteristic quantity and the nozzle front pressure characteristic quantity are self-defined parameters with a total value of 1.

[0025] Further, the conductivity characteristic quantity in the decoupling window is obtained as the conductivity channel characteristic vector, the machining liquid characteristic quantity in the decoupling window is obtained as the machining liquid characteristic vector, and the machining liquid channel characteristic vector is calculated in combination with the conductivity channel characteristic vector. The size of the decoupling window is a self-defined parameter.

[0026] The machining liquid channel characteristic vector The calculation formula is as follows:

[0027] , wherein represents the machining liquid characteristic vector, represents the conductivity channel characteristic vector, represents the vector inner product operation.

[0028] Further, the micro-process stimulation represents that two independent disturbance operations are applied under the premise of keeping the discharge energy and wire trajectory constant, which are the conductivity channel micro-stimulation and the machining liquid channel micro-stimulation. The conductivity channel micro-stimulation adjusts the bypass ratio of deionized water or the resin regeneration flow by a small amount, and the arc ratio sequence and the conductivity channel characteristic vector are collected in the pre-adjustment time period and the post-adjustment time period. The conductivity effective gain is calculated according to the arc ratio sequence and the conductivity channel characteristic vector.

[0029] The conductivity effective gain The calculation formula is as follows:

[0030] , wherein and represent the mean value of the arc ratio sequence in the pre-adjustment time period and the post-adjustment time period, respectively, and represent the mean value of the conductivity channel characteristic vector in the pre-adjustment time period and the post-adjustment time period, respectively.

[0031] The machining liquid channel micro-stimulation adjusts the flushing pump frequency or the nozzle pressure command by a small amount, and the arc ratio sequence and the machining liquid channel characteristic vector are collected in the pre-adjustment time period and the post-adjustment time period. The machining liquid effective gain is calculated according to the arc ratio sequence and the machining liquid channel characteristic vector.

[0032] The machining liquid effective gain The calculation formula is as follows:

[0033] wherein and respectively represent the mean value of the arc ratio sequence of the pre-adjustment time period and the post-adjustment time period, and respectively represent the mean value of the machining fluid channel feature vector of the pre-adjustment time period and the post-adjustment time period;

[0034] The balance coefficient is equal to the ratio of the conductivity effective gain and the machining fluid effective gain.

[0035] Further, the upper and lower limit values of the closed-loop triangular excitation amplitude are determined according to the current working condition, under the premise of keeping the discharge energy and the wire trajectory constant, the regenerative current proportion is reduced or the bypass is increased, the flushing intensity is simultaneously reduced, until the element value of the equivalent coupling driving vector is equal to the upper limit value of the closed-loop triangular excitation amplitude, then the time period is marked as an uplink segment, and then the reverse adjustment is performed until the element value of the equivalent coupling driving vector is equal to the lower limit value of the closed-loop triangular excitation amplitude, then the time period is marked as a downlink segment, and the arc ratio of the uplink segment and the downlink segment is recorded synchronously; the i-th element value of the equivalent coupling driving vector is The calculation formula is as follows:

[0036] wherein and respectively represent the i-th element value of the conductivity channel feature vector and the machining fluid channel feature vector, represents the balance coefficient.

[0037] Further, if the change rate of the element value of the equivalent coupling driving vector in the uplink segment is greater than 0, an uplink mark is added to the element value, if the change rate of the element value of the equivalent coupling driving vector in the downlink segment is less than 0, a downlink mark is added to the element value, and when the change rate of the element value is 0, the mark of the previous element value is used.

[0038] The upper and lower limit values of the closed-loop triangular excitation amplitude are symmetric boundaries, the upper limit value of the closed-loop triangular excitation amplitude is The calculation formula is as follows: wherein represents the balance coefficient, and respectively represent the maximum value of the absolute value of the element value change rate in the conductivity channel feature vector and the machining fluid channel feature vector.

[0039] Further, the element value marked by the uplink is added to the uplink element set, the element value marked by the downlink is added to the downlink element set, the element index in the uplink element set is synchronized with the moment when the corresponding pull arc proportion is greater than the proportion threshold for the first time, and the element index in the downlink element set is synchronized with the moment when the corresponding pull arc proportion is less than the proportion threshold for the last time, wherein the proportion threshold is a self-defined parameter.

[0040] Further, the main loop width is equal to the equivalent driving amount corresponding to the uplink switching point minus the equivalent driving amount corresponding to the downlink switching point; the hysteresis loop area is equal to the closed area formed by all elements in the uplink element set and the downlink element set, and the calculation formula of the closed area S is as follows:

[0041] , wherein total represents the total set of the uplink element set and the downlink element set, and respectively represent the pull arc proportion corresponding to the jth and j+1th element index of the total set, and respectively represent the jth and j+1th element value of the total set;

[0042] The critical slope ratio is equal to the ratio of the pull arc proportion change rate of the uplink switching point neighborhood to the equivalent driving amount change rate divided by the ratio of the pull arc proportion change rate of the downlink switching point neighborhood to the equivalent driving amount change rate, wherein the neighborhood is composed of the switching point and one moment before and after the switching point;

[0043] The recovery hysteresis time is equal to the time interval between the downlink switching point and the next recovery to the uplink switching point.

[0044] Further, the fault diagnosis model is constructed based on a multilayer perceptron, the sample labels of the training samples used for training the fault diagnosis model are obtained by expert annotation; the maintenance strategy corresponding to the conductive temperature rise dominant state is to increase the deionized water regeneration ratio or replace the resin and filter core; the maintenance strategy corresponding to the machining liquid dominant state is to improve the flushing pressure and flow, correct the nozzle fitting gap, and clean the nozzle and pipeline.

[0045] The present application provides a kind of electric spark machine tool fault diagnosis system, comprising:

[0046] The preprocessing module is used to collect the operating parameters of the electric spark machine tool at a preset time interval within a preset time period, and to obtain the conductivity characteristic quantity and the machining liquid characteristic quantity by preprocessing;

[0047] The balance coefficient calculation module is used to first orthogonally decouple the conductivity characteristic quantity and the machining liquid characteristic quantity to obtain the conductivity channel feature vector and the machining liquid channel feature vector, and then to obtain the balance coefficient by executing the micro-amplitude process stimulation on the machine tool.

[0048] an up-down marking module for adjusting the machine tool process parameters, calculating the equivalent coupling driving vector according to the conductivity channel feature vector, the machining fluid channel feature vector and the balance coefficient, and marking each element value of the equivalent coupling driving vector up or down;

[0049] a hysteresis feature quantity calculation module for extracting the up-switching point and the down-switching point of the equivalent coupling driving vector, and calculating the hysteresis feature quantity according to the up-switching point and the down-switching point;

[0050] a fault diagnosis module for inputting the hysteresis feature quantity into the fault diagnosis model, and outputting the running state of the electric spark machine tool.

[0051] The present application has the beneficial effects that: the present application obtains the conductivity and machining fluid feature quantities through orthogonal decoupling, obtains the balance coefficient by combining with the micro-amplitude process stimulation, constructs the equivalent coupling driving vector and extracts the hysteresis feature quantities such as the main loop width and the hysteresis loop area, realizes the accurate diagnosis of the running state of the electric spark machine tool, clearly distinguishes the normal state, the electric conduction temperature rise dominant state and the machining fluid dominant state, provides the targeted maintenance strategy for different fault types, avoids the blindness of maintenance, greatly improves the fault disposal efficiency, reduces the risk of equipment damage, and protects the machining stability of the electric spark machine tool. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a flowchart of an electric spark machine tool fault diagnosis method of the present application;

[0053] Figure 2 is a flowchart of the pre-processing to obtain the conductivity feature quantity and the machining fluid feature quantity of the present application;

[0054] Figure 3 is a schematic diagram of an electric spark machine tool fault diagnosis system of the present application;

[0055] Figure 4 is a schematic diagram of a confusion matrix tested by the fault diagnosis model of the present application.

[0056] In the figure: the pre-processing module 301, the balance coefficient calculation module 302, the up-down marking module 303, the hysteresis feature quantity calculation module 304, and the fault diagnosis module 305. DETAILED DESCRIPTION

[0057] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is merely meant to provide a better understanding of the subject matter described herein and can be changed in the function and arrangement of the elements discussed without departing from the scope of the content of the present specification. Various processes or components can be omitted, replaced or added according to needs of various examples. In addition, the features described with respect to some examples can also be combined in other examples.

[0058] It should be noted that unless otherwise defined, technical terms or scientific terms used in one or more embodiments of the present application shall have the ordinary meanings as understood by one of ordinary skill in the art to which the present application pertains. The terms "first", "second", and similar terms used in one or more embodiments of the present application do not denote any order, quantity, or importance, but are used to distinguish different components. The terms "include", "comprise", and similar terms mean that the elements or objects before the term encompass the elements or objects listed after the term and equivalents thereof, and do not exclude other elements or objects. The terms "connected" or "coupled" or similar terms do not limit to physical or mechanical connections or couplings, but can include electrical connections or couplings, whether direct or indirect. The terms "upper", "lower", "left", "right", and the like are used only to indicate relative positions, and when the absolute positions of the described objects are changed, the relative positions can also be changed accordingly.

[0059] As shown in Figures 1-4 A method for diagnosing a fault of an EDM machine tool, comprising the following steps:

[0060] In step S101, operating parameters of the EDM machine tool are collected at preset time intervals within a preset time period, and preprocessed to obtain conductivity features and machining fluid features.

[0061] The operating parameters include display conductivity, gap temperature, acoustic emission raw signal, and nozzle front pressure.

[0062] In step S102, the conductivity features and machining fluid features are first orthogonally decoupled to obtain conductivity channel feature vectors and machining fluid channel feature vectors, and then a micro-amplitude process stimulus is performed on the machine tool to obtain a balance coefficient.

[0063] In step S103, the process parameters of the machine tool are adjusted, the equivalent coupling driving vectors are calculated based on the conductivity channel feature vectors, the machining fluid channel feature vectors, and the balance coefficient, and each element value of the equivalent coupling driving vectors is marked as uplink or downlink.

[0064] In step S104, the uplink switching points and the downlink switching points of the equivalent coupling driving vectors are extracted, and the hysteresis features are calculated based thereon.

[0065] The hysteresis features include main loop width, hysteresis loop area, critical slope ratio, and recovery hysteresis time.

[0066] In step S105, the hysteresis features are input into a fault diagnosis model, and the operating state of the EDM machine tool is output.

[0067] The operating state includes a normal state, a conductive temperature rise dominant state, and a machining fluid dominant state.

[0068] It should be noted that the operating parameters of the electric spark machine tool also include discharge pulse statistical data, the discharge pulse statistical data including: total pulse number, arc pulling pulse number, arc pulling proportion, each pulse can be distinguished through the pulse distinguishing circuit built in the machine tool power supply or controller, including: normal spark, arc pulling, short circuit, etc., the discharge pulse statistical data can be obtained by reading the cumulative count once every 1 second interval of the PLC or numerical control system; the display conductivity represents the display value of the online conductivity instrument, which can be obtained through the online conductivity instrument of the medium circulating system of the electric spark machine tool; the gap temperature represents the actual temperature of the discharge gap region, which can be obtained by collecting the small thermosensitive element embedded in the nozzle seat; the acoustic emission original signal can be obtained by collecting the piezoelectric AE (acoustic emission) sensor pasted or magnetically attracted on the machine head or workpiece clamp; the nozzle front pressure can be obtained by collecting the pressure sensor connected in series or embedded on the nozzle water inlet pipeline; before obtaining the operating parameters of the electric spark machine tool, the machine tool is first placed in a stable machining state, and the operating parameters of 2 to 3 minutes are recorded as baseline data, and then a data collection stage of 3 to 5 minutes is entered, that is, the preset time period is between 3 to 5 minutes of self-defined parameters, preferably, the preset time period is set to 3 minutes, all channels are synchronized by the same clock, and each 1 second is taken as a time window for statistical and calculation once, that is, the preset time interval is set to 1 second.

[0069] In an embodiment of the present application, as shown in Figure 2 The pre-processing obtains the conductivity characteristic quantity and the machining liquid characteristic quantity, including the following steps:

[0070] Step S201, obtaining the median of the operating parameters of the electric spark machine tool in the baseline window;

[0071] The baseline window represents the time period when the electric spark machine tool is in a stable machining state, and the size of the baseline window is a self-defined parameter, preferably, the size of the baseline window is set to 2 minutes;

[0072] Step S202, the temperature correction coefficient is an exponential growth factor with a natural constant as the base, and the index is equal to the product of the gap temperature minus the standard temperature and the temperature coefficient;

[0073] The standard temperature is a self-defined parameter, and preferably, the standard temperature is set to 25℃;

[0074] The temperature coefficient is obtained by fitting the mapping relationship between the gap temperature and the conductivity characteristic quantity through the least square method, specifically, under the condition that the machine tool is in a stable machining state, the display conductivity is collected while the gap temperature is changed, for example, between 20℃ to 30℃, the conductivity is collected every 1℃ interval, then the conductivity characteristic quantity is calculated and obtained according to the display conductivity, and finally the temperature coefficient can be obtained by fitting the mapping relationship between the gap temperature and the conductivity characteristic quantity through the least square method.

[0075] The conductivity characteristic quantity is equal to the ratio of the conductivity multiplied by the temperature correction coefficient to the median conductivity in the baseline window.

[0076] The machining fluid characteristic quantity is equal to the weighted sum of the acoustic emission raw signal characteristic quantity and the nozzle front pressure characteristic quantity.

[0077] The acoustic emission raw signal characteristic quantity is equal to the acoustic emission energy obtained by integrating the square of the amplitude of the acoustic emission raw signal at each time in a preset time period, and then divided by the median acoustic emission energy in the baseline window.

[0078] The nozzle front pressure characteristic quantity is equal to the effective value of the nozzle front pressure obtained by integrating the square of the nozzle front pressure at each time in a preset time period, dividing by the length of the preset time period, and then taking the square root, and then divided by the median effective value of the nozzle front pressure in the baseline window.

[0079] The weighting coefficients of the acoustic emission raw signal characteristic quantity and the nozzle front pressure characteristic quantity are self-defined parameters with a total value of 1.

[0080] In an embodiment of the present application, the conductivity characteristic quantity in the decoupling window is obtained as the conductivity channel characteristic vector, the machining fluid characteristic quantity in the decoupling window is obtained as the machining fluid characteristic vector, and the machining fluid channel characteristic vector is calculated in combination with the conductivity channel characteristic vector, wherein the size of the decoupling window is a self-defined parameter, and preferably, the size of the decoupling window is set to 2 minutes. The calculation formula of the machining fluid channel characteristic vector is as follows:

[0081] , wherein represents the machining fluid characteristic vector, represents the conductivity channel characteristic vector, represents the vector inner product operation.

[0082] It should be noted that since the conductivity is mainly affected by the ion concentration and temperature of the processing liquid, and the processing liquid is mainly affected by the flow field pressure and nozzle state, the two are naturally weakly coupled, therefore, the conductivity channel feature vector directly reflects the real change of the conductivity itself, which is composed of the conductivity feature quantity at each time in the decoupling window in order, and is not disturbed by the processing liquid; and for the processing liquid channel feature vector, by calculating the inner product relationship between the processing liquid feature vector and the conductivity channel feature vector, the part coupled with the conductivity is stripped from the processing liquid feature vector, and the processing liquid pure change vector reflecting only the influence of the flow field pressure and the nozzle state is obtained, the correlation between the processing liquid feature quantity and the conductivity feature quantity is removed, and the two respectively accurately characterize the state change caused by the corresponding driving factor, thereby providing independent and clear feature quantity basis for subsequent process control and state analysis.

[0083] In an embodiment of the application, the micro-process stimulation represents that under the premise of keeping the discharge energy and wire trajectory constant, two independent disturbance operations are performed, which are the conductivity channel micro-stimulation and the processing liquid channel micro-stimulation; the conductivity channel micro-stimulation adjusts the bypass ratio of the deionized water or the resin regeneration flow by a small amount, collects the arc ratio sequence and the conductivity channel feature vector in the pre-adjustment time period and the post-adjustment time period, and calculates the conductivity effective gain according to the arc ratio sequence and the conductivity channel feature vector;

[0084] The conductivity effective gain The calculation formula is as follows:

[0085] , wherein and respectively represent the mean values of the arc ratio sequences in the pre-adjustment time period and the post-adjustment time period, and respectively represent the mean values of the conductivity channel feature vectors in the pre-adjustment time period and the post-adjustment time period;

[0086] The processing liquid channel micro-stimulation adjusts the flushing pump frequency or the nozzle pressure command by a small amount, collects the arc ratio sequence and the processing liquid channel feature vector in the pre-adjustment time period and the post-adjustment time period, and calculates the processing liquid effective gain according to the arc ratio sequence and the processing liquid channel feature vector;

[0087] The processing liquid effective gain The calculation formula is as follows:

[0088] , wherein and respectively represent the mean values of the arc ratio sequences in the pre-adjustment time period and the post-adjustment time period, and respectively represent the mean values of the processing liquid channel feature vectors in the pre-adjustment time period and the post-adjustment time period;

[0089] The balance coefficient is equal to the ratio of the conductivity effective gain and the machining liquid effective gain.

[0090] It should be noted that the micro-process stimulation keeps the discharge energy (for example, the discharge power determined by the pulse parameters) and the wire trace (for example, the electrode wire movement path) constant, the purpose is to isolate the interference of the two on the machining state, and to separately explore the influence of the conductivity and the machining liquid channel change on the discharge; the length of the time period before adjustment and the time period after adjustment are both self-defined parameters, preferably, both are set to 20 seconds when the machine tool is in a stable machining state; the conductivity effective gain and the machining liquid effective gain respectively represent the influence degree of the conductivity channel feature vector change and the machining liquid channel feature vector change on the arc striking ratio; the balance coefficient represents the influence weight of the conductivity and the machining liquid channel on the arc striking ratio; by comparing the mean value changes of the arc striking ratio and the conductivity channel feature vector and the machining liquid channel feature vector before and after adjustment, the change trend of the discharge arc state when the conductivity and the machining liquid change is quantified, thereby providing a measurement basis for subsequent process control.

[0091] In an embodiment of the present application, the upper and lower limit values of the closed-loop triangular excitation amplitude are determined according to the current working condition, under the premise of keeping the discharge energy and the wire trace constant, the regenerative flow proportion is reduced or the bypass is increased, the flushing intensity is simultaneously reduced, until the element value of the equivalent coupling driving vector is equal to the upper limit value of the closed-loop triangular excitation amplitude, then the time period is marked as an uplink segment, and then the reverse adjustment is performed until the element value of the equivalent coupling driving vector is equal to the lower limit value of the closed-loop triangular excitation amplitude, then the time period is marked as a downlink segment, and the arc striking ratios of the uplink segment and the downlink segment are recorded synchronously.

[0092] The i-th element value of the equivalent coupling driving vector The calculation formula is as follows:

[0093] , wherein and respectively represent the i-th element value of the conductivity channel feature vector and the machining liquid channel feature vector, represents the balance coefficient.

[0094] In an embodiment of the present application, when the element value change rate of the equivalent coupling driving vector in the uplink segment is greater than 0, an uplink mark is added to it, when the element value change rate of the equivalent coupling driving vector in the downlink segment is less than 0, a downlink mark is added to it, and when the element value change rate is 0, the mark of the previous element value is used.

[0095] The upper and lower limit values of the closed-loop triangular excitation amplitude are symmetric boundaries, the upper limit value of the closed-loop triangular excitation amplitude The calculation formula is as follows:

[0096] wherein represents the balance coefficient, and respectively represent the maximum value of the absolute value of the rate of change of the element value in the conductivity channel feature vector and the processing liquid channel feature vector.

[0097] In an embodiment of the present application, the element value marked as uplink is added to the uplink element set, the element value marked as downlink is added to the downlink element set, the element index in the uplink element set is synchronized with the time when the corresponding arc ratio is greater than the ratio threshold for the first time, and the element index in the downlink element set is synchronized with the time when the corresponding arc ratio is less than the ratio threshold for the last time, wherein the ratio threshold is a self-defined parameter, and preferably, the ratio threshold is set to 30%.

[0098] In an embodiment of the present application, the main loop width is equal to the equivalent driving amount (element value) corresponding to the uplink switching point minus the equivalent driving amount (element value) corresponding to the downlink switching point.

[0099] The hysteresis loop area is equal to the area of the closed region formed by all elements in the uplink element set and the downlink element set, and the calculation formula of the closed region area S is as follows:

[0100] wherein total represents the total set of the uplink element set and the downlink element set, and respectively represent the arc ratio corresponding to the jth and j+1th element index in the total set, and respectively represent the jth and j+1th element value (equivalent driving amount) in the total set.

[0101] The critical slope ratio is equal to the ratio of the arc ratio change rate and the equivalent driving amount change rate in the neighborhood of the uplink switching point divided by the ratio of the arc ratio change rate and the equivalent driving amount change rate in the neighborhood of the downlink switching point, wherein the neighborhood is composed of the switching point and one time before and after the switching point.

[0102] The recovery hysteresis time is equal to the time interval between the downlink switching point and the next recovery to the uplink switching point.

[0103] It should be noted that the main loop width represents the difference between the equivalent driving force of the first uplink boundary position and the equivalent driving force of the last downlink non-boundary position, and the greater the value, the more obvious the difference between the required conditions and the exit conditions for entering harmful discharge; the hysteresis loop area is used to reflect the cumulative effect and irreversibility of path memory in the whole process, and is used to determine the instability dominance under the coupling of multiple factors; the critical slope ratio is used to measure the sensitivity difference in different directions near the threshold, and is used to identify whether the entry and exit stages are asymmetric and assist in distinguishing the dominant factors; the recovery hysteresis time is used to measure the speed of the system out of the critical state, and is used to evaluate the continuous influence of the processing liquid and the conductive temperature rise and the shutdown risk.

[0104] In an embodiment of the present application, the fault diagnosis model is constructed based on a multilayer perceptron, and the sample labels of the training samples for training the fault diagnosis model are obtained by expert annotation; the maintenance strategy corresponding to the conductive temperature rise dominant state is to increase the deionized water regeneration ratio or replace the resin and filter core; and the maintenance strategy corresponding to the processing liquid dominant state is to improve the flushing pressure and flow, correct the nozzle fitting gap, and clean the nozzle and pipeline.

[0105] It should be noted that in the training process of the fault diagnosis model, the difference between the predicted category and the expert annotation label (the sample label of the training sample) is measured by a cross-entropy loss function, and then the error calculated by the loss function is propagated to each layer of the network by a gradient descent algorithm or a gradient optimizer, so as to update the parameters such as weights and biases, gradually reduce the prediction error, and complete the training of the fault diagnosis model until the model converges, which will not be repeated here.

[0106] It should be noted that when the collected multi-source operating parameters are directly used for analysis, it is difficult to intuitively reflect the fault nature due to coupling interference and large amount of data, and by extracting hysteresis characteristic quantities such as the main loop width and the hysteresis loop area, redundant information can be stripped and the core differences in different states can be highlighted; these characteristic quantities have clear physical meaning, which facilitates expert accurate labeling of sample labels and provides high-quality training data for the fault diagnosis model. At the same time, the simplified characteristic quantities reduce the input dimension of the model, reduce the calculation burden, speed up the convergence speed of the model, and improve the diagnosis efficiency.

[0107] In an embodiment of the present application, as shown in Figure 3 A fault diagnosis system of an electric spark machine tool, comprising:

[0108] A preprocessing module 301 is configured to collect operating parameters of the electric spark machine tool at a preset time interval within a preset time period, and to obtain conductivity characteristic quantities and processing liquid characteristic quantities by preprocessing the operating parameters;

[0109] a balance coefficient calculation module 302, configured to obtain a conductivity channel feature vector and a machining fluid channel feature vector by orthogonal decoupling the conductivity feature and the machining fluid feature, and then obtain a balance coefficient by performing a micro-process stimulation on the machine tool;

[0110] an up-down marking module 303, configured to adjust the machine tool process parameters, obtain an equivalent coupling driving vector according to the conductivity channel feature vector, the machining fluid channel feature vector and the balance coefficient, and mark each element value of the equivalent coupling driving vector as up or down;

[0111] a hysteresis feature calculation module 304, configured to extract up-switching points and down-switching points of the equivalent coupling driving vector, and calculate a hysteresis feature according to the up-switching points and the down-switching points;

[0112] a fault diagnosis module 305, configured to input the hysteresis feature into a fault diagnosis model, and output a running state of the electric spark machine tool.

[0113] As shown in Figure 4 the confusion matrix is obtained by testing the completed fault diagnosis model, the number of test samples of each state is 100, 96 samples are correctly identified in the identification of the normal state, the identification accuracy is 96.0%; 95 samples are correctly identified in the identification of the conductive temperature rise dominant state, the accuracy is 95.0%; 97 samples are correctly identified in the identification of the machining fluid dominant state, the accuracy is 97.0%.

[0114] It should be noted that the setting of the interval and the threshold size is for easy comparison, and the size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data, as long as it does not affect the proportional relationship between the parameters and the quantized values. And the above formula is a dimensionless calculation of the value, and the formula is obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation, and the preset parameters in the formula are set by the person skilled in the art according to the actual situation.

[0115] The above describes the embodiments of the present embodiment, but the present embodiment is not limited to the specific embodiments described above, and the specific embodiments described above are only illustrative and not limiting, and a person of ordinary skill in the art can make many forms under the inspiration of the present embodiment, which all belong to the protection of the present embodiment.

Claims

1. An electric discharge machine failure diagnosis method characterized by, The method comprises the following steps: Step S101, within a preset time period, collecting the running parameters of the EDM machine according to a preset time interval, and pre-processing the running parameters to obtain the conductivity characteristic quantity and the machining liquid characteristic quantity; The running parameters include: display conductivity, gap temperature, acoustic emission original signal and nozzle front pressure; Step S102, first, orthogonal decoupling is performed on the conductivity characteristic quantity and the machining liquid characteristic quantity to obtain the conductivity channel characteristic vector and the machining liquid channel characteristic vector, and then a micro-amplitude process stimulus is performed on the machine tool to obtain the balance coefficient; Step S103, adjusting the process parameters of the machine tool, calculating the equivalent coupling driving vector according to the conductivity channel characteristic vector, the machining liquid channel characteristic vector and the balance coefficient, and marking each element value of the equivalent coupling driving vector as uplink or downlink; Step S104, extracting the uplink switching point and the downlink switching point of the equivalent coupling driving vector, and calculating the hysteresis characteristic quantity according to the uplink switching point and the downlink switching point; The hysteresis characteristic quantity includes: main loop width, hysteresis loop area, critical slope ratio and recovery hysteresis time; Step S105, inputting the hysteresis characteristic quantity into the fault diagnosis model to output the running state of the EDM machine; The running state includes: normal state, electric conduction temperature rise dominant state and machining liquid dominant state; According to the current working condition, the upper and lower limit values of the closed-loop triangular excitation amplitude are determined, under the premise of keeping the discharge energy and wire track constant, the regenerative current is reduced or the bypass is increased, the flushing intensity is simultaneously reduced, until the element value of the equivalent coupling driving vector is equal to the upper limit value of the closed-loop triangular excitation amplitude, then the time period is marked as an uplink section, and then the reverse adjustment is performed until the element value of the equivalent coupling driving vector is equal to the lower limit value of the closed-loop triangular excitation amplitude, then the time period is marked as a downlink section, and the arc draw ratio of the uplink section and the downlink section is recorded simultaneously; the i th element value of the equivalent coupling driving vector The calculation formula is as follows: wherein and respectively represent the i-th element value of the conductivity channel feature vector and the machining fluid channel feature vector, represents the balance coefficient; If the element value change rate of the equivalent coupling driving vector in the uplink segment is greater than 0, an uplink mark is added to the element value; if the element value change rate of the equivalent coupling driving vector in the downlink segment is less than 0, a downlink mark is added to the element value; when the element value change rate is 0, the mark of the previous element value is used; The upper and lower limit values of the closed-loop triangular excitation amplitude are symmetrical boundaries, and the upper limit value of the closed-loop triangular excitation amplitude The calculation formula is as follows: Wherein Balancing coefficient is represented by And The maximum value of the absolute value of the element value change rate of the conductivity channel characteristic vector and the machining fluid channel characteristic vector is represented by The element value of the uplink mark is added to the uplink element set, and the element value of the downlink mark is added to the downlink element set. The time when the element index in the uplink element set corresponds to the first time when the arc ratio is greater than the ratio threshold is taken as the uplink switching point, and the time when the element index in the downlink element set corresponds to the last time when the arc ratio is less than the ratio threshold is taken as the downlink switching point. The ratio threshold is a self-defined parameter.

2. The method according to claim 1, wherein The pre-processing to obtain the conductivity characteristic quantity and the machining liquid characteristic quantity comprises the following steps: Step S201, obtaining the median of the running parameters of the EDM machine within the baseline window; The baseline window represents the time period when the EDM machine is in a stable machining state. The size of the baseline window is a self-defined parameter; Step S202, the temperature correction coefficient is an exponential growth factor with a natural constant as the base number, and the index is equal to the product of the gap temperature minus the standard temperature and the temperature coefficient; The standard temperature is a self-defined parameter; the temperature coefficient is obtained by fitting the mapping relationship between the gap temperature and the conductivity characteristic quantity through the least squares method; Step S203, the conductivity characteristic quantity is equal to the ratio of the display conductivity multiplied by the temperature correction coefficient to the median of the conductivity within the baseline window; Step S204, the machining liquid characteristic quantity is equal to the weighted sum of the acoustic emission original signal characteristic quantity and the nozzle front pressure characteristic quantity; The acoustic emission original signal characteristic quantity is equal to the integral of the square of the amplitude of the acoustic emission original signal at each time within the preset time period to obtain the acoustic emission energy, and then divided by the median of the acoustic emission energy within the baseline window; The nozzle front pressure characteristic quantity is equal to the square of the nozzle front pressure at each time point in a preset time period, and the square of the integral of the nozzle front pressure with respect to time is obtained by dividing the length of the preset time period; the effective value of the nozzle front pressure is obtained by dividing the median value of the nozzle front pressure effective values in the baseline window; The weighting coefficients of the acoustic emission original signal characteristic quantity and the nozzle front pressure characteristic quantity are self-defined parameters with a total value of 1.

3. The method according to claim 1, wherein The conductivity channel characteristic vector is obtained by acquiring the conductivity characteristic quantity in the decoupling window, the machining liquid characteristic vector is obtained by acquiring the machining liquid characteristic quantity in the decoupling window, and the machining liquid channel characteristic vector is obtained by combining the conductivity channel characteristic vector; the size of the decoupling window is a self-defined parameter.

4. The method according to claim 1, wherein The micro-process stimulation represents that two independent disturbance operations are applied under the premise of keeping the discharge energy and the wire trace constant, which are the conductivity channel micro-stimulation and the machining liquid channel micro-stimulation, respectively. The conductivity channel micro-stimulation adjusts the bypass ratio of the deionized water or the resin regeneration flow by a small amount, and the arc ratio sequence and the conductivity channel characteristic vector are collected in the pre-adjustment time period and the post-adjustment time period, and the conductivity effective gain is calculated accordingly; The machining liquid channel micro-stimulation adjusts the flushing pump frequency or the nozzle pressure command by a small amount, and the arc ratio sequence and the machining liquid channel characteristic vector are collected in the pre-adjustment time period and the post-adjustment time period, and the machining liquid effective gain is calculated accordingly; The balance coefficient is equal to the ratio of the conductivity effective gain to the machining liquid effective gain.

5. The method according to claim 1, wherein The main loop width is equal to the equivalent driving amount corresponding to the up-switching point minus the equivalent driving amount corresponding to the down-switching point; the hysteresis loop area is equal to the area of the closed region formed by all elements in the up-element set and the down-element set, and the calculation formula of the closed region area S is as follows: where total denotes the total set of the uplink element set and the uplink element set, and respectively denote the jth and j+1th element index of the total set corresponding to the pull arc proportion, and respectively denote the jth and j+1th element value of the total set. The critical slope ratio is equal to the ratio of the arc ratio change rate of the up-switching point neighborhood to the equivalent driving amount change rate divided by the ratio of the arc ratio change rate of the down-switching point neighborhood to the equivalent driving amount change rate, wherein the neighborhood is composed of the switching point and one time point before and after the switching point; The recovery hysteresis time is equal to the time interval between the down-switching point and the next recovery to the up-switching point.

6. The method according to claim 1, wherein The fault diagnosis model is constructed based on a multilayer perceptron, and the sample labels of the training samples for training the fault diagnosis model are obtained by expert annotation; the maintenance strategy corresponding to the conductive temperature rise dominant state is to increase the deionized water regeneration ratio or replace the resin and filter core; the maintenance strategy corresponding to the machining liquid dominant state is to improve the flushing pressure and flow, correct the nozzle fitting gap, and clean the nozzle and pipeline.

7. An electric discharge machine failure diagnosis system characterized by comprising: The method comprises: A preprocessing module is used to collect the operating parameters of the EDM machine tool at preset time intervals within a preset time period, and the operating parameters are preprocessed to obtain the conductivity characteristic quantity and the machining liquid characteristic quantity; A balance coefficient calculation module is used to orthogonally decouple the conductivity characteristic quantity and the machining liquid characteristic quantity to obtain the conductivity channel characteristic vector and the machining liquid channel characteristic vector, and then perform micro-process stimulation on the machine tool to obtain the balance coefficient. An up-down mark module is configured to adjust the machine tool process parameters, calculate an equivalent coupling driving vector according to the conductivity channel feature vector, the machining fluid channel feature vector and the balance coefficient, and mark each element value of the equivalent coupling driving vector as up or down; A hysteresis feature quantity calculation module is configured to extract the up-switching point and the down-switching point of the equivalent coupling driving vector, and calculate a hysteresis feature quantity according to the up-switching point and the down-switching point; A fault diagnosis module is configured to input the hysteresis feature quantity into a fault diagnosis model, and output the running state of the electric spark machine tool.

Images