Disconnection Anomaly Prediction Method, Device and Computer Equipment

Abstract

The present application provides a method, device, and computer equipment for predicting abnormal wire breakage, particularly related to the field of wire cutting. The method includes: obtaining cutting data of the previous cutting cycle; dividing the previous cutting cycle into multiple cutting conditions according to the main roller linear velocity and the gradient of the main roller linear velocity, decomposing the cutting process data into multiple sub-data sequences, and using a convolutional neural network to extract features from the sub-data sequences corresponding to each cutting condition to obtain the temporal features of each cutting condition; performing feature fusion on the temporal features of multiple cutting conditions to obtain fused features; obtaining a pre-trained reliability prediction model, and inputting at least the fused features and the wear amount of the cutting wire into the reliability prediction model to obtain the reliability value of the cutting wire in the next cutting cycle. By dividing the cutting cycle into multiple conditions and performing feature extraction and fusion on the cutting data of multiple conditions, the method improves the accuracy of wire breakage prediction.

Description

Technical Field

[0001] The present application relates to the technical field of wire cutting, and particularly to a method, apparatus, and computer device for predicting abnormal wire breakage. Background Art

[0002] In modern manufacturing, wire cutting technology, as a precise machining method, is widely used in the machining processes of various materials. This technology cuts workpieces by means of a high-speed thin cutting wire (such as a diamond wire) in cooperation with cutting fluid. However, during the cutting process, the cutting wire is prone to being affected by various factors, such as the installation accuracy of the workpiece, clamping force, cutting speed, tension fluctuation, and the wear of the cutting wire itself. These factors may cause the cutting wire to break.

[0003] Once an abnormal wire breakage occurs, it will not only interrupt the machining process, resulting in a decrease in production efficiency, but also may cause damage to the workpiece, increasing production costs. Currently, most of the wire breakage monitoring methods in related technologies rely on single-sensor monitoring or manual monitoring. These methods cannot effectively achieve real-time early warning of abnormal wire breakage. Usually, they can only monitor and intervene after the wire breakage occurs and cannot prevent the occurrence of wire breakage in advance. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, apparatus, and electronic device for predicting abnormal wire breakage to solve the above technical problems, and this method can improve the accuracy and timeliness of predicting abnormal wire breakage.

[0005] In a first aspect, the present application provides a method for predicting abnormal wire breakage, which includes:

[0006] Obtain the cutting data of the previous cutting cycle, where the cutting data at least includes cutting process data and the wear amount of the cutting wire, and the cutting process data at least includes the tension of the cutting wire and the main roller wire speed;

[0007] According to the main roller wire speed and the gradient of the main roller wire speed, divide the previous cutting cycle into multiple cutting working conditions, and decompose the cutting process data into multiple sub-data sequences according to the main roller wire speed and the gradient of the main roller wire speed of adjacent cutting working conditions, where each sub-data sequence corresponds to a cutting working condition;

[0008] Use a convolutional neural network to extract features from the sub-data sequence corresponding to each cutting working condition to obtain the time-series features of each cutting working condition;

[0009] Fuse the time-series features of multiple cutting working conditions to obtain the fused features;

[0010] Obtain a pre-trained reliability prediction model, and input at least the fused features and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting cycle; among them, the smaller the reliability value, the greater the possibility of wire breakage abnormality of the cutting line.

[0011] In one embodiment, a convolutional neural network is used to extract features from the sub-data sequences corresponding to each cutting condition to obtain the time-series features of each cutting condition, including:

[0012] Perform padding on the sub-data sequences shorter than the set sequence length to complete them;

[0013] Use the parameters in the mask to mark whether the elements in the padded sub-data sequence are valid data. The parameters in the mask include a first parameter and a second parameter of different sizes. The first parameter is used to indicate that the element in the sub-data sequence corresponding to this position is valid data, and the second parameter indicates that the element in the sub-data sequence corresponding to this position is padded data;

[0014] When using a convolutional neural network to extract features from the sub-data sequences corresponding to each cutting condition, multiply the padded sub-data sequence by the corresponding parameter in the mask to filter out the padded data so that the padded data does not participate in the convolution calculation.

[0015] In one embodiment, perform feature fusion on the time-series features of multiple cutting conditions to obtain the fused features, including:

[0016] For the time-series features of each cutting condition, map the time-series features to a query matrix Q i 、a key matrix K i and a value matrix V i , and obtain a new feature matrix Q i 、a key matrix K i and a value matrix V i for each cutting condition; Z i ;

[0017] Connect the new feature matrices end to end according to the time order of each condition in the previous cutting cycle to obtain the fused features;

[0018] Among them, the query matrix Q i = W Q y i , and the key matrix Ki = W K y i , the value matrix V i = W V y i , W Q , W K and W V are all parameter matrices, y i represents the timing feature corresponding to the i-th cutting condition.

[0019] In one embodiment, according to the mapped query matrix Q i , the key matrix K i and the value matrix V i , the new feature matrix of each cutting condition is obtained through the following calculation method Z i :

[0020] ;

[0021] wherein, is the -th value of the key matrix of the j-th cutting condition, softmax is exponential normalization, and the formula is:

[0022] ;

[0023] wherein, X represents the product result sequence of the query matrix of the condition and the key matrix of the 1, 2,... l ..., L , L indicates that the length of the product result sequence X is L .

[0024] In one embodiment, training a reliability prediction model includes:

[0025] Constructing a reliability curve of the cutting line based on the cumulative distribution function of the Weibull distribution;

[0026] Generate a piecewise reliability function based on the reliability curve of the cutting line;

[0027] Determine the label in the feature label pair based on the piecewise reliability function; the feature label pair includes a label and a feature, the feature in the feature label pair is the time series feature of each working condition during the historical cutting period, and the label in the feature label pair is the reliability value corresponding to the historical cutting period;

[0028] Train a reliability prediction model using the feature label pair, the cutting line wear amount, and the actual position of the workbench;

[0029] Among them, at least input the fused feature and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting period, including: input the fused feature, the wear amount of the cutting line, and the current actual position of the workbench into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting period.

[0030] In one embodiment, generating a piecewise reliability function based on the reliability curve of the cutting line includes:

[0031] Determine the piecewise reliability function according to the first time and the second time; wherein, if the usage time of the cutting line is less than the first time, the reliability value of the cutting line is 1, and if the usage time of the cutting line is greater than the second time, the reliability value of the cutting line is 0;

[0032] Piecewise reliability function is: ;

[0033] Among them, represents the first time, represents the second time, represents the usage time of the cutting line, k represents the shape parameter, and λ represents the scale parameter.

[0034] In one embodiment, determine the wear amount of the cutting line through the following formula;

[0035] ;

[0036] Among them, represents the wear amount of the cutting line, idx represents the cycle number of the previous cutting period, cir_len represents the cutting line length for one revolution of the main roller, len_n represents the net cutting line length in the previous cutting period, v b represents the instantaneous speed of the workbench at the end of the previous cutting period, v l represents the instantaneous linear speed of the main roller at the end of the previous cutting period.

[0037] In one embodiment, the cutting conditions include a wire feeding condition and a wire retracting condition. The main roller linear velocity in the wire feeding condition is greater than zero, and the main roller linear velocity in the wire retracting condition is less than zero. Among them, the wire feeding condition includes:

[0038] Wire feeding acceleration condition: v > 0 and f(v, t) > 0;

[0039] Wire feeding constant speed cutting condition: v > 0 and f(v, t) = 0;

[0040] Wire feeding deceleration condition: v > 0 and f(v, t) < 0;

[0041] The wire retracting condition includes:

[0042] Wire retracting acceleration condition: v < 0 and f(v, t) < 0;

[0043] Wire retracting constant speed cutting condition: v < 0 and f(v, t) = 0;

[0044] Wire retracting deceleration condition: v < 0 and f(v, t) > 0;

[0045] v represents the main roller linear velocity, f(v, t) represents the gradient of the main roller linear velocity.

[0046] In a second aspect, the present application also provides a wire breakage anomaly prediction device, which includes:

[0047] An acquisition module, configured to acquire cutting data of the previous cutting cycle. The cutting data includes at least cutting process data and the wear amount of the cutting wire. The cutting process data includes at least the tension of the cutting wire and the main roller linear velocity;

[0048] A segmentation module, configured to segment the previous cutting cycle into multiple cutting conditions according to the main roller linear velocity and the gradient of the main roller linear velocity, and decompose the cutting process data into multiple sub-data sequences according to the main roller linear velocity and the gradient of the main roller linear velocity of adjacent cutting conditions. Among them, each sub-data sequence corresponds to a cutting condition;

[0049] A feature extraction module, configured to extract features of each sub-data sequence corresponding to each cutting condition by using a convolutional neural network to obtain the time series features of each cutting condition;

[0050] A feature fusion module, configured to perform feature fusion on the time series features of multiple cutting conditions to obtain the fused features;

[0051] A prediction module, configured to obtain a pre-trained reliability prediction model, and input at least the fused features and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle; wherein, the smaller the reliability value, the greater the possibility of abnormal wire breakage of the cutting line.

[0052] In a third aspect, the present application further provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the wire breakage abnormal prediction method in the first aspect is implemented.

[0053] For the above-mentioned wire breakage abnormal prediction method, the method obtains the cutting data of the previous cutting cycle, and the cutting data at least includes the tension of the cutting line, the main roller line speed, and the wear amount of the cutting line. According to the changes in the main roller line speed and its gradient, the previous cutting cycle is divided into multiple cutting working conditions, and the cutting process data is decomposed into multiple sub-data sequences, and each sub-sequence corresponds to a specific cutting working condition. A convolutional neural network is used to extract features from the sub-data sequences of each cutting working condition to obtain the time-series features of each working condition. The time-series features are fused. The fused features and the wear amount of the cutting line are input into a pre-trained reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle. The smaller the reliability value, the greater the possibility of abnormal wire breakage of the cutting line. By dividing the cutting cycle into multiple working conditions and extracting and fusing the cutting data of multiple working conditions, the method realizes the prediction of abnormal wire breakage and improves the accuracy of wire breakage prediction. Description of the Drawings

[0054] Figure 1 It is a flowchart of the wire breakage abnormal prediction method in an embodiment;

[0055] Figure 2 It is a schematic diagram of the wire speed and working condition segmentation in the cutting process in an embodiment;

[0056] Figure 3 It is a schematic diagram of the trend of the output torque of the bilateral spindle motors in an embodiment;

[0057] Figure 4 It is a flowchart of obtaining the time-series features of each cutting working condition in an embodiment;

[0058] Figure 5 It is a flowchart of fusing the time-series features of multiple cutting working conditions to obtain the fused features in an embodiment;

[0059] Figure 6 It is a flowchart of training the reliability prediction model in an embodiment;

[0060] Figure 7 It is a reliability curve diagram in an embodiment;

[0061] Figure 8 A diagram of a disconnection abnormality prediction device in an embodiment;

[0062] Figure 9 A structural diagram of a computer device in one embodiment. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0064] In one embodiment, Figure 1 As shown, a disconnection anomaly prediction method is provided, the method comprising the following steps:

[0065] Step 101: obtaining cutting data of the last cutting cycle, wherein the cutting data at least includes cutting process data and the wear amount of the cutting line, and the cutting process data at least includes the tension of the cutting line and the linear speed of the main roller;

[0066] It should be noted that a complete cutting operation includes thousands of cutting cycles, and the previous cutting cycle is one of the thousands of cutting cycles. The cutting data of the previous cutting cycle at least includes the cutting process data and the wear of the cutting wire, wherein the cutting wire can be a diamond wire, which is a wire with diamond particles fixed on the metal wire. The cutting process data mainly records the changes in various key parameters during the cutting process, including at least the tension of the cutting wire and the linear speed of the main roller. The cutting process data can also include the output torque of the double-sided spindle motor, the position of the worktable, the tension torque, etc.

[0067] It should be noted that the tension of the cutting line can refer to the tension on the left and right sides of the cutting line, that is, the pulling force on the left and right sides of the cutting line during the cutting process; the tension torque refers to the torque value used to control and maintain the tension of the cutting line during the cutting process; the output torque of the double-sided spindle motor refers to the torque output by the two-sided spindle motors when they are running in the cutting equipment. Furthermore, the main roller linear speed is the linear speed of the main roller driving the cutting line to perform reciprocating cutting motion, and its size and stability directly affect the cutting efficiency and quality. Among them, the main roller can be a roller wheel installed on the main shaft, which can rotate at high speed to drive the groove wheel and the wire mesh to rotate, contact with the silicon carbide crystal rod, and the crystal rod is cut into wafers through the grinding action of the cutting wire and the crystal rod.

[0068] Step 102: dividing the previous cutting cycle into a plurality of cutting conditions according to the main roller linear velocity and the gradient of the main roller linear velocity, and decomposing the cutting process data into a plurality of sub-data sequences according to the main roller linear velocity and the gradient of the main roller linear velocity of adjacent cutting conditions, wherein each sub-data sequence corresponds to a cutting condition;

[0069] According to the changes in the main roller linear velocity and its gradient, the working condition switching points in the cutting cycle can be determined. For example, when the main roller linear velocity starts to increase from a positive value, it indicates entering the wire feeding acceleration working condition; when the main roller linear velocity reaches a stable value, it switches to the wire feeding constant velocity working condition; when the main roller linear velocity starts to decrease, it enters the wire feeding deceleration working condition. When the main roller linear velocity becomes negative and starts to increase the absolute value of the negative value, it enters the wire retracting acceleration working condition; when the main roller linear velocity is a stable negative value, it is the wire retracting constant velocity working condition; when the main roller linear velocity decreases the absolute value of the negative value, it is the wire retracting deceleration working condition.

[0070] According to the working condition switching points, the main roller linear velocity data sequence of the cutting cycle can be divided into multiple sub-data sequences. Each sub-data sequence corresponds to a specific cutting working condition. For example, the sub-data sequence of the wire feeding acceleration working condition includes all data points from when the main roller linear velocity starts to increase to when the main roller linear velocity reaches a stable value; the sub-data sequence of the wire feeding constant velocity working condition includes all data points during which the main roller linear velocity remains stable; the sub-data sequence of the wire feeding deceleration working condition includes all data points from when the main roller linear velocity starts to decrease to when the main roller linear velocity drops to zero. Similarly, the sub-data sequences of the wire retracting acceleration working condition, the wire retracting constant velocity working condition, and the wire retracting deceleration working condition are also divided in a similar manner.

[0071] Step 103: Use a convolutional neural network to extract features from the sub-data sequences corresponding to each cutting working condition to obtain the time series features of each cutting working condition;

[0072] The sub-data sequence of each cutting working condition can be regarded as a time series data. The convolutional neural network can perform a convolution operation on the sub-data sequence of each cutting working condition through the convolution kernels in the convolutional layer to extract features such as edges, trends, and periodicity. Among them, the size and number of the convolution kernels can be set according to the characteristics of the sub-data sequence.

[0073] In the convolution operation, the parameters (weights and biases) of the convolution kernels are learned and optimized through training, enabling the convolutional neural network to automatically learn the features most useful for wire break prediction. After the processing of the convolutional layer, the sub-data sequence corresponding to each cutting working condition is converted into the time series features of each cutting working condition, and the time series features can more effectively reflect the dynamic changes in the cutting cycle.

[0074] It should be noted that in a convolutional neural network, the parameters of the convolutional kernel mainly include weights and biases. The weights are the numerical matrices in the convolutional kernel used for element-wise multiplication operations with the input data, and their size and shape are determined by the size of the convolutional kernel. The bias is a scalar used to shift the result after the convolution operation, increasing the flexibility of the model. The parameters of the convolutional kernel can be learned and adjusted during the training process through optimization algorithms (such as gradient descent) so that the convolutional kernel can effectively extract the features in the sub-data sequence corresponding to each cutting condition.

[0075] Step 104: Perform feature fusion on the temporal features of multiple cutting conditions to obtain the fused features.

[0076] There are various methods for feature fusion, and common ones include feature concatenation, weighted average, and attention mechanism, etc.

[0077] Exemplarily, assume that feature concatenation is performed on the temporal features of two cutting conditions, which are the feature vector F 1 = f 11, f 12 ,..., f 1n of the wire feeding acceleration condition and the feature vector F 2 = f 21 , f 22 ,..., f 2m of the wire feeding constant speed condition. For feature fusion, these two feature vectors are fused in sequence into a longer feature vector, that is, F = f 11 , f 12 ,..., f 1n , f 21 , f 22 ,..., f 2m .

[0078] Step 105: Obtain a pre-trained reliability prediction model, and input at least the fused features and the wear amount of the cutting wire into the reliability prediction model to obtain the reliability value of the cutting wire for the next cutting cycle; among them, the smaller the reliability value, the greater the possibility of wire breakage abnormality of the cutting wire.

[0079] The reliability prediction model performs complex non - linear transformations and calculations on the input data based on an internal deep - learning structure, such as a long short - term memory network (LSTM), etc., and outputs a reliability value. The reliability value reflects the probability that the cutting line can remain in normal operation in the next cutting cycle. The smaller the reliability value, the more obvious the abnormal feature of wire breakage during the cutting process, and thus the greater the likelihood of wire breakage abnormality.

[0080] Exemplarily, an alarm threshold for the reliability value can be set in advance. For example, the alarm threshold is that the reliability value is 0.1. If the reliability value output by the reliability prediction model is less than 0.1, a wire breakage warning is issued.

[0081] In this embodiment, the method obtains the cutting data of the previous cutting cycle. The cutting data at least includes the tension of the cutting line, the main roller linear speed, and the wear amount of the cutting line. According to the change of the main roller linear speed and its gradient, the previous cutting cycle is divided into multiple cutting working conditions, and the cutting process data is decomposed into multiple sub - data sequences, and each sub - sequence corresponds to a specific cutting working condition. A convolutional neural network is used to extract features from the sub - data sequences of each cutting working condition to obtain the time - series features of each working condition. The time - series features are fused. The fused features and the wear amount of the cutting line are input into a pre - trained reliability prediction model to obtain the reliability value of the cutting line in the next cutting cycle. The smaller the reliability value, the greater the likelihood of wire breakage abnormality of the cutting line. By dividing the cutting cycle into multiple working conditions and extracting and fusing the cutting data of multiple working conditions, this method realizes the prediction of wire breakage abnormality and improves the accuracy of wire breakage prediction.

[0082] In one embodiment, the cutting working conditions include a wire - feeding working condition and a wire - taking - up working condition. The main roller linear speed in the wire - feeding working condition is greater than zero, and the main roller linear speed in the wire - taking - up working condition is less than zero; among them, the wire - feeding working condition includes: wire - feeding acceleration working condition: v > 0 and f(v,t)>0; wire - feeding uniform cutting working condition: v > 0 and f(v,t)=0; wire - feeding deceleration working condition: v > 0 and f(v,t)<0;

[0083] The wire - taking - up working condition includes: wire - taking - up acceleration working condition: v < 0 and f(v,t)<0; wire - taking - up uniform cutting working condition: v < 0 and f(v,t)=0; wire - taking - up deceleration working condition: v < 0 and f(v,t)>0;

[0084] Among them, v represents the main roller linear speed, f(v,t) represents the gradient of the main roller linear speed.

[0085] As Figure 2 shown, the main roller linear speed can be based onv and its gradient According to the differences of , the cutting process of a single cutting cycle can be divided into multiple cutting conditions. Among them, the abscissa is time t, and the ordinate is the main roller linear velocity v. Each cutting condition has its specific judgment logic. For example, the wire feeding acceleration condition corresponds to v >0 and f(v,t)>0, while the wire taking-up deceleration condition corresponds to v <0 and f(v,t)>0.

[0086] Furthermore, by determining the segmentation nodes of different cutting conditions, the original continuous process data sequence can be segmented into several sub-data sequences according to the segmentation nodes. The trend of the cutting process data under different cutting conditions is as Figure 3 shown. It should be noted that Figure 3 Taking the output torque of the bilateral main shaft motor as an example, different colors in the figure represent data points under different cutting conditions, where the abscissa is time and the ordinate is the magnitude of the data passing through the output torque of the bilateral main shaft motor. Furthermore, under different cutting conditions, the trend and volatility characteristics of the cutting process data are very different, and the dynamic characteristics between different cutting conditions are relatively strong. Therefore, for a complete cutting cycle X n,L , n is the quantity of the cutting process data, L is the time span. According to the segmentation nodes of each condition, the cutting cycle is decomposed into 6 sub-data sequences, which are respectively X n,L1 , X n,L2 , X n,L3 , X n,L4 , X n,L5 , X n,L6 , among which L 1~ L 6 is the length of the sub-data sequence of each cutting condition.

[0087] In one embodiment, as Figure 4 shown, a convolutional neural network is used to extract features from the sub-data sequences corresponding to each cutting condition, and the time series features of each cutting condition are obtained, including the following steps:

[0088] Step 401: Pad the sub-data sequences with a length less than the set sequence length through padding operation;

[0089] It should be noted that each cutting cycle has a complete 6 cutting conditions, but the data lengths of the same-named conditions in each cutting cycle are not exactly the same, there are sampling errors, and length padding is required to make the data lengths of the same-named conditions consistent.

[0090] During the cutting process of the cutting line, in order to eliminate the differences in dimension and magnitude between the data of different cutting processes and improve the comparability of the data and the efficiency of model training, it is necessary to perform min-max normalization operations on the data of each process. Further, set the maximum number of elements in the sub-data sequence, that is, tokens. For sub-data sequences with data lengths less than the set value, fill -1 values at their tails to ensure that all sub-data sequences input into the model have the same length.

[0091] Exemplarily, assume that the maximum element length of the set sub-data sequence is 10. If a certain sub-data sequence has only 7 data points: [0.2, 0.5, 0.7, 0.9, 0.1, 0.3, 0.6], then add 3 -1s at the end of this sub-data sequence to make its length reach 10, becoming: [0.2, 0.5, 0.7, 0.9, 0.1, 0.3, 0.6, -1, -1, -1].

[0092] Step 402: Mark whether the elements in the padded sub-data sequence are valid data through the parameters in the mask. The parameters in the mask include a first parameter and a second parameter of different sizes. The first parameter is used to indicate that the element in the sub-data sequence corresponding to this position is valid data, and the second parameter indicates that the element in the sub-data sequence corresponding to this position is padding data;

[0093] The mask is a sequence with the same length as the padded sequence, which contains two types of parameters: the first parameter and the second parameter. The first parameter can be 1, which is used to indicate that the element at this position is the valid data in the original sub-data sequence, and the second parameter can be 0, which can indicate that the element at this position is the padding data.

[0094] Exemplarily, assume that the sub-data sequence becomes [0.2, 0.5, 0.7, 0.9, 0.1, 0.3, 0.6, -1, -1, -1] after normalization and padding, then the corresponding mask should be [1, 1, 1, 1, 1, 1, 1, 0, 0, 0]. In this mask, the 1s in the first 7 positions indicate that the corresponding data is valid data, and the 0s in the last 3 positions indicate that the corresponding data is the -1 of the padding data.

[0095] Step 403: When using a convolutional neural network to extract features from the sub-data sequence corresponding to each cutting condition, multiply the padded sub-data sequence by the corresponding parameters in the mask to filter out the padding data so that the padding data does not participate in the convolution calculation.

[0096] Exemplarily, assume that the padded sub - data sequence is: [0.2, 0.5, 0.7, 0.9, 0.1, 0.3, 0.6, - 1, - 1, - 1], and the corresponding mask is: [1, 1, 1, 1, 1, 1, 1, 0, 0, 0]. By multiplying these two sequences element - by - element, the resulting sequence is: [0.2, 0.5, 0.7, 0.9, 0.1, 0.3, 0.6, 0, 0, 0].

[0097] Furthermore, the filled -1 is multiplied by 0 in the mask, and the resulting data is 0, that is, the filled data does not participate in the convolutional kernel calculation. The valid data remains unchanged. During the convolutional calculation process, the filled positions corresponding to 0 values do not have a substantial impact on the calculation of the convolutional kernel, thus avoiding the interference of the filled data on feature extraction.

[0098] In this embodiment, through the padding operation and mask marking, sub - data sequences of different lengths can be effectively unified. By filtering out the filled data before convolutional calculation, the accuracy and efficiency of feature extraction are guaranteed, and the interference of invalid data on model training and prediction is avoided.

[0099] In one embodiment, in order to effectively extract temporal features from sub - data sequences of different cutting conditions, 6 parallel feature extractors are constructed according to the types of cutting conditions. Each extractor consists of 2 Temporal Convolutional Network (TCN) layers. Since the data characteristics under different cutting conditions are different, for example, the duration of the uniform cutting condition is long and the temporal feature fluctuation is small, while the temporal feature fluctuation of the acceleration - deceleration condition is intense, different convolutional kernel sizes can be set for the temporal features. In the uniform cutting condition, a larger convolutional kernel size such as 5 is used to capture a wider range of time dependencies; while in the acceleration - deceleration cutting condition, a smaller convolutional kernel size such as 3 is used to more carefully extract local detail information. The temporal convolutional kernel slides one step along the cutting time direction, and then parallelly extracts the temporal features under 6 different cutting conditions. The feature extractor for each cutting condition can focus on its specific type of temporal features, thus improving the accuracy and efficiency of temporal feature extraction.

[0100] In one embodiment, as Figure 5 shown, fusing the temporal features of multiple cutting conditions to obtain the fused features includes the following steps:

[0101] Step 501: For the temporal features of each cutting condition, map the temporal features to a query matrix Q i , key matrix K i and value matrix V i , and according to the mapped query matrix Q i, key matrix K i and the value matrix V i Get the new feature matrix for each cutting condition Z i ;

[0102] For the time series features extracted for the i-th cutting condition y 1, y 2, ..., y 6, each y i The shape is n × L i , n Indicates the number of cutting process data, L i represents the length of the time series feature of the i-th cutting condition in the time dimension, that is, the number of time steps contained in the cutting condition. yi Mapped to query ( Q i = W Q y i ),key( K i = W K y i ) and value ( V i = W V y i ) three parts, among which W Q , W K , W V is the parameter matrix learned during training. W Q The matrix transforms the time series features into query representation. W K The matrix converts the time series features into key representations, W V The matrix converts the time series features into a value representation.

[0103] It should be noted that the time series characteristics of each cutting condition are mapped into three matrices, namely the query matrix Q i , key matrix K i Sum Matrix V iThe mappings of these three matrices are implemented through a fully-connected network, and the dimensions and contents of each matrix are related to the original time-series features.

[0104] Query matrix Q i It is used to query the current time-series features for finding correlations with other time-series features, which can help determine which other time-series features are most relevant to the current time-series features. Key matrix K i It is used to match with the queries of other time-series features to determine the correlations between time-series features. Value matrix V i It contains the value information of the original time-series features and is used to generate a new time-series feature representation after correlation calculation.

[0105] By performing matrix multiplication on the query matrix Q i and the key matrix K i and applying the softmax function for normalization, an attention score matrix can be obtained, which represents the correlation weights between various time-series features. This weight matrix is multiplied by the value matrix V i to generate a new feature matrix for each cutting condition Z i . The new feature matrix Z i retains the information of the original time-series features and the correlations between the fused time-series features.

[0106] Exemplarily, assume that the dimension of the time-series features of a certain cutting condition is n × m (where n is the number of features, m is the time step). After being mapped through the fully-connected network, the dimensions of the query matrix Q i , the key matrix K i and the value matrix V i are respectively n × d , n × d and n × d (where d is the dimension after mapping). Through the attention calculation process, the dimension of the obtained new feature matrix Z i remains n × m , and the new feature matrix Z iIt contains feature information adjusted by the attention mechanism, which can better reflect the importance and mutual relationship of features in wire break prediction.

[0107] Step 502: Connect the new feature matrices head to tail according to the time order of each working condition in the previous cutting cycle to obtain the fused features. Among them, the query matrix Q i = W Q y i , the key matrix K i = W K y i , the value matrix V i = W V y i , W Q , W K and W V are all parameter matrices, y i represents the time series features corresponding to the i-th cutting working condition.

[0108] Each cutting working condition corresponds to a new feature matrix Z i , and these new feature matrices are obtained by processing the time series features of their respective working conditions through the attention mechanism. Exemplarily, assuming there are 6 working conditions, corresponding to 6 new feature matrices Z 1, Z 2, Z 3, Z 4, Z 5, Z 6. In the previous cutting cycle, the order of the cutting working conditions is wire feeding acceleration, wire feeding at a constant speed, wire feeding deceleration, wire retracting acceleration, wire retracting at a constant speed, and wire retracting deceleration. Connect the new feature matrices head to tail according to this time order, that is, connect the end of Z 1 to the beginning of Z 2, Z the end of 2 to the beginning of Z 3, and so on, until all new feature matrices are connected in order to obtain a fused feature matrix Z n,L .

[0109] In this embodiment, by dynamically fusing the temporal features under different cutting conditions through the attention mechanism, the dynamic changes and the correlations between features during the cutting process can be captured more effectively. The fused features not only retain the information of the original temporal features but also highlight the feature parts that are more important for the wire break prediction task.

[0110] In one embodiment, according to the mapped query matrix Q i , key matrix K i , and value matrix V i , a new feature matrix for each cutting condition is obtained through the following calculation method Z i :

[0111] ;

[0112] where is the -th value of the key matrix of the j-th cutting condition, and softmax is exponential normalization, with the formula:

[0113] ;

[0114] where X represents the product result sequence of the query matrix of the condition and the key matrix of the l condition. The product result sequence X is [x1, x2, … x L …, x L , and x L represents the length of the product result sequence X as

[0115] Multiply the query matrix Q i of one cutting condition with the K j -th values of the key matrices of all other cutting conditions l , where l represents the K j -th value in the key matrix l . The purpose of the multiplication operation is to measure the correlation between the query matrix Q i and the key matrix K j at a specific position. Here, it is to measure the correlation between the query matrix Q i and the K j -th value in the key matrix l . Further, the query matrixQ i The key matrix for all other cutting conditions K j The l product results obtained by multiplying with the X -th values can form a product result sequence

[0116] Furthermore, perform normalization processing on the product result sequence X using the softmax function. The softmax function can convert each value in the product result sequence X into a weight value ranging from 0 to 1. Furthermore, multiply each normalized weight value obtained using the softmax function with the corresponding value matrix V j and perform weighted summation to obtain the new feature matrix corresponding to this cutting condition Z i

[0117] It should be noted that performing exponential normalization on each value in the product result sequence X means adding all exponential terms to obtain the denominator to ensure that the output of the softmax function is a probability distribution. The weight l at the -th position is calculated as: . It represents the relative importance of the l -th position in the product result sequence X. The larger the weight value, the more important this position is

[0118] Exemplarily, assume there are 6 cutting conditions, and the query matrix Q i , key matrix K i and value matrix V i for each condition have been mapped through a fully connected network. For the 1st cutting condition, multiply its query matrix Q1 with the l -th values of the key matrices K1 to K6 of all 6 conditions respectively to obtain 6 values, and these 6 values can form a product result sequence X. Normalize the 6 values in the product result sequence X using the softmax function to obtain 6 weight values. Among them, the 6 weight values range from 0 to 1. Multiply these 6 weight values with the value matrices V1 to V6 under the corresponding cutting conditions respectively and perform weighted summation to obtain the new feature matrix Z1 of the 1st cutting condition. The same process is also applied to the other 5 cutting conditions, and finally, the new feature matrices Z1 to Z6 of all 6 cutting conditions are obtained

[0119] ​In this embodiment, the new feature matrix of each cutting condition can dynamically fuse the feature information of other conditions, and adjust the weights according to the relevance in the wire break prediction task, so as to improve the richness of feature representation and the accuracy of the prediction model.

[0120] In one embodiment, as Figure 6 shown, training a reliability prediction model includes the following steps:

[0121] Step 601: Construct a reliability curve of the cutting wire based on the cumulative distribution function of the Weibull distribution;

[0122] The reliability curve constructed by the Weibull distribution is as Figure 7 shown. The abscissa is the cutting time in seconds (s), and the ordinate is the reliability value. The Weibull distribution is a commonly used life distribution model. The life degradation curve of the cutting wire is based on the cumulative distribution function (CDF) of the Weibull distribution. Taking the cutting data with wire breaks in the historical records as the research object, the reliability during the normal cutting period is set to 1. As the cutting wire wears and deteriorates, the life gradually declines until the reliability is set to 0 when a wire break anomaly occurs.

[0123] When constructing the reliability curve of the cutting wire, based on the cumulative distribution function of the Weibull distribution, the reliability change trend of the cutting wire during use can be effectively characterized. The expression of the cumulative distribution function (CDF) of the Weibull distribution is:

[0124] ,

[0125] where x is a non - negative random variable, usually representing time or life data; k is the shape parameter. For the wire break failure caused by the wear of the cutting wire, in the case where the device failure rate and failure rate gradually increase over time, it generally takes a value of 1.5; λ is the scale parameter, which affects the scale of the distribution and represents the characteristic life of the product.

[0126] It should be noted that by recording the time T e when each cutting wire break occurs and the time T s when the process parameters start to show anomalies, the time difference Δt = T e - T s is calculated. Δt can reflect the time span from the appearance of the anomaly to the final wire break. Further, after collecting the Δt data of multiple cutting cycles, the maximum likelihood estimation method can be used to fit the Δt data of multiple cutting cycles into the Weibull distribution, so as to solve the scale parameter λ.

[0127] Step 602: Generate a segmented reliability function based on the reliability curve of the cutting wire;

[0128] Collect data on the cutting line during actual use, including records of normal working hours and the time of wire breakage. Statistically analyze this data and use life distribution models such as the Weibull distribution to determine the shape parameter k and scale parameter λ for each stage. Further, through the shape parameter k and scale parameter λ, a piecewise reliability function can be constructed, and the function form for each stage is R(t)= , where t is time, λ is the scale parameter, and k is the shape parameter. The piecewise function can more accurately describe the reliability change trend of the cutting line at different stages.

[0129] Step 603: Determine the labels in the feature-label pairs based on the piecewise reliability function; the feature-label pairs include labels and features, the features in the feature-label pairs are the temporal features of each working condition during the historical cutting cycle, and the labels in the feature-label pairs are the reliability values corresponding to the historical cutting cycle;

[0130] The temporal features can reflect the dynamic changes in the cutting process and the wear condition of the cutting line. The labels can be determined based on the piecewise reliability function. The piecewise reliability function is generated according to the reliability data of the cutting line at different stages and can more accurately describe the reliability change of the cutting line at different usage times or conditions. Substituting the time data of each historical cutting cycle into the corresponding piecewise reliability function, the corresponding reliability value can be obtained as the label in the feature-label pair.

[0131] Step 604: Use the feature-label pairs, the wear amount of the cutting line, and the actual position of the workbench to train the reliability prediction model; among them, at least the fused features and the wear amount of the cutting line are input into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle, including: inputting the fused features, the wear amount of the cutting line, and the current actual position of the workbench into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle.

[0132] The feature-label pairs can include the temporal features of each cutting working condition and the corresponding reliability values during the historical cutting cycle. By learning the relationship between the temporal features and reliability values of each cutting working condition during the historical cutting cycle, the reliability prediction model can understand the patterns that affect the reliability of the cutting line under different cutting working conditions.

[0133] The wear amount of the cutting line can directly reflect the loss degree of the cutting line during the cutting process. As the cutting progresses, the cutting line will gradually wear, its diameter decreases, and its strength decreases, thus increasing the risk of wire breakage. By using the wear amount of the cutting line as an input to the reliability prediction model, the reliability prediction model can more accurately evaluate the remaining life and reliability of the cutting line. The calculation of the wear amount is based on parameters such as the usage length of the cutting line, the length of one revolution of the main roller, the workbench speed, and the main roller line speed during each cutting cycle, and these parameters can accurately quantify the loss situation of the cutting line.

[0134] The workbench is a component that fixes the ingot and feeds it downward slowly. During the cutting process, the stability of the current actual position of the workbench also has a great impact on the cutting effect. The actual position of the workbench can be monitored in real time by sensors or encoders, reflecting the precise coordinates or position information of the workbench during the cutting process. When the workbench feeds to the position where the contact length between the ingot and the wire mesh is the longest, the cutting resistance is the greatest. At this time, if the workbench vibrates unstably, it is very easy to occur abnormal wire breakage.

[0135] In the training stage of the reliability prediction model, the fused features, the wear amount of the cutting wire, and the current actual position of the workbench are used as inputs. By selecting a suitable algorithm, such as the long short-term memory network (LSTM), and using the time-series features and reliability values of each cutting condition in the historical cutting cycle for training, the reliability prediction model can learn the relationship between the time-series features and the reliability value of the cutting wire. During the training process, by continuously sliding the sliding window along the cutting time direction, a mapping combination of feature labels is constructed, and the parameters of the model are continuously adjusted to minimize the error between the predicted value and the actual value until the reliability prediction model shows good prediction performance.

[0136] The trained reliability prediction model can predict the reliability of the cutting wire in the next cutting cycle based on the time-series features of different cutting conditions in the current cutting cycle during the actual cutting process.

[0137] In one embodiment, based on the reliability curve of the cutting wire, a piecewise reliability function is generated, including:

[0138] According to the first time and the second time, the piecewise reliability function is determined; wherein, if the usage time of the cutting wire is less than the first time, the reliability value of the cutting wire is 1, and if the usage time of the cutting wire is greater than the second time, the reliability value of the cutting wire is 0;

[0139] Piecewise reliability function is:

[0140] ;

[0141] wherein, represents the first time, represents the second time, represents the usage time of the cutting wire, k represents the shape parameter, and λ represents the scale parameter.

[0142] Suppose there is a set of data on the failure time of the cutting wire, denoted as t 1, t 2,..., t n ,The following formula can be used to estimate the value of λ: λ = .

[0143] Among them, k is the shape parameter, which can be estimated by the maximum likelihood estimation method or other methods.

[0144] Determine two time points, namely the first time T s and the second time T e . When the usage time of the cutting line t does not exceed the first time T s , it can be considered that the cutting line is in the initial usage stage, and the reliability is maintained at the highest level. Therefore, its reliability value is set to 1; while when the usage time t exceeds T e , the cutting line has entered the stage of excessive wear, and the reliability has decreased significantly. At this time, its reliability value is set to 0. During the stage between T s and T e , the reliability of the cutting line shows a gradually decreasing trend. The reliability value of this stage can be calculated by the cumulative distribution function of the Weibull distribution, and the formula is R ( t ) = , where k is the shape parameter, reflecting the trend of the failure rate changing with time, λ is the scale parameter, representing the characteristic life. The entire piecewise reliability function R can be expressed as:

[0145] ;

[0146] In practical applications, T s and T e need to be determined based on a large amount of cutting data and the actual service life of the cutting line. By collecting the time data of the cutting line from the start of use to wire breakage in multiple cutting cycles, as well as the time data when the process parameters start to show abnormalities, statistical analysis methods can be used to determine the T s and T e specific values. Further, using the maximum likelihood estimation method, fitting the Weibull distribution according to these data, the best estimated values of the shape parameter k and the scale parameter λ can be obtained.

[0147] In one embodiment, the wear amount of the cutting line is determined by the following formula;

[0148] ;

[0149] Wherein, represents the wear amount of the cutting line, idx represents the cycle number of the previous cutting cycle, cir_len represents the length of the cutting line around the main roller for one week, len_n represents the net length of the cutting line in the previous cutting cycle, v b represents the instantaneous speed of the workbench at the end of the previous cutting cycle, v l represents the instantaneous speed of the main roller line at the end of the previous cutting cycle.

[0150] wear The wear amount of the cutting line can be obtained by accumulating the wear conditions of each cutting cycle. idx represents the cycle number of the previous cutting cycle, which is used to determine the accumulation range of the wear condition. cir _ len can represent the length of the cutting line around the main roller for one week, and can calculate the wear condition of the cutting line in each cycle. lenn can represent the net length of the cutting line in the previous cutting cycle, which can reflect the actual used length of the cutting line in this cycle. V b can represent the instantaneous speed of the workbench at the end of the previous cutting cycle, v l can represent the instantaneous speed of the main roller line at the end of the previous cutting cycle. These two speed parameters can jointly affect the wear rate of the cutting line during the cutting process. The workbench moves slowly downward during the cutting process, thus pressing the ingot against the wire mesh and giving the ingot a downward feeding force. Therefore, there is an instantaneous speed of the workbench during the cutting process.

[0151] Furthermore, by substituting these parameters into the formula and accumulating for all cutting cycles, the total wear amount of the cutting line can be obtained wear , which further provides important data support for subsequent reliability prediction and wire break warning.

[0152] In this embodiment, this division method helps to more accurately analyze and control different stages during the cutting process.

[0153] Based on the same concept, as Figure 8 shown, the present application also provides a wire break abnormal prediction device, which includes:

[0154] An acquisition module 801, configured to acquire cutting data of the previous cutting cycle, where the cutting data at least includes cutting process data and the wear amount of the cutting line, and the cutting process data at least includes the tension of the cutting line and the main roller linear speed;

[0155] A segmentation module 802, configured to segment the previous cutting cycle into multiple cutting working conditions according to the main roller linear speed and the gradient of the main roller linear speed, and decompose the cutting process data into multiple sub-data sequences according to the main roller linear speed and the gradient of the main roller linear speed of adjacent cutting working conditions, where each sub-data sequence corresponds to one cutting working condition;

[0156] A feature extraction module 803, configured to perform feature extraction on the sub-data sequence corresponding to each cutting working condition by using a convolutional neural network to obtain the temporal features of each cutting working condition;

[0157] A feature fusion module 804, configured to perform feature fusion on the temporal features of multiple cutting working conditions to obtain fused features;

[0158] A prediction module 805, configured to obtain a pre-trained reliability prediction model, and input at least the fused features and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting cycle; where the smaller the reliability value, the greater the possibility of wire breakage abnormality of the cutting line.

[0159] In one embodiment, the feature extraction module 803 performs feature extraction on the sub-data sequence corresponding to each cutting working condition by using a convolutional neural network to obtain the temporal features of each cutting working condition, and specifically is configured to: perform padding on the sub-data sequence with a length less than the set sequence length; mark whether the elements in the padded sub-data sequence are valid data through the parameters in the mask, and the parameters in the mask include a first parameter and a second parameter with different sizes, the first parameter is used to indicate that the element in the sub-data sequence corresponding to this position is valid data, and the second parameter indicates that the element in the sub-data sequence corresponding to this position is padding data; when performing feature extraction on the sub-data sequence corresponding to each cutting working condition by using a convolutional neural network, multiply the padded sub-data sequence by the corresponding parameters in the mask to filter out the padding data so that the padding data does not participate in the convolution calculation.

[0160] In one embodiment, the feature fusion module 804 performs feature fusion on the temporal features of multiple cutting working conditions to obtain fused features, and specifically is configured to: for the temporal features of each cutting working condition, map the temporal features to a query matrix Q i a key matrix K i and a value matrix V i and, based on the mapped query matrix Qi , the key matrix K i and the value matrix V i to obtain a new feature matrix for each cutting condition Z i ; according to the time sequence of each condition in the previous cutting cycle, connect the new feature matrices end to end to obtain the fused features; among them, the query matrix Q i = W Q y i , the key matrix K i = W K y i , the value matrix V i = W V y i , W Q , W K and W V are all parameter matrices, y i represents the timing feature corresponding to the i-th cutting condition.

[0161] In one embodiment, the feature fusion module 804, based on the mapped query matrix Q i , the key matrix K i and the value matrix V i , obtains a new feature matrix for each cutting condition through the following calculation method Z i : ; among them, is the -th value of the key matrix of the j-th cutting condition, softmax is exponential normalization, and the formula is: ;

[0162] where x represents the product result sequence of the query matrix and the key matrix of one of the cutting conditions, and the product result sequence x is [x1, x2,... x l ..., x L , x L represents that the length of the product result sequence x is L .

[0163] In one embodiment, the prediction module 805 trains a reliability prediction model, specifically for: constructing a reliability curve of the cutting line based on the cumulative distribution function of the Weibull distribution; generating a segmented reliability function based on the reliability curve of the cutting line; determining the label in the feature-label pair based on the segmented reliability function; the feature-label pair includes a label and a feature, the feature in the feature-label pair is the time series feature of each working condition during the historical cutting period, and the label in the feature-label pair is the reliability value corresponding to the historical cutting period; training the reliability prediction model using the feature-label pair, the wear amount of the cutting line, and the actual position of the workbench; wherein, at least inputting the fused feature and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting period, including: inputting the fused feature, the wear amount of the cutting line, and the current actual position of the workbench into the reliability prediction model to obtain the reliability value of the cutting line in the next cutting period.

[0164] In one embodiment, the prediction module 805 generates a segmented reliability function based on the reliability curve of the cutting line, specifically for: determining the segmented reliability function according to the first time and the second time; wherein, if the usage time of the cutting line is less than the first time, the reliability value of the cutting line is 1, and if the usage time of the cutting line is greater than the second time, the reliability value of the cutting line is 0; the segmented reliability function is:

[0165] ;

[0166] wherein, represents the first time, represents the second time, represents the usage time of the cutting line, k represents the shape parameter, and λ represents the scale parameter.

[0167] In one embodiment, the prediction module 805 determines the wear amount of the cutting line through the following formula;

[0168] ;

[0169] wherein, represents the wear amount of the cutting line, idx represents the cycle number of the previous cutting period, cir_len represents the length of the cutting line around the main roller for one week, len_n represents the net length of the cutting line in the previous cutting period, v b represents the instantaneous speed of the workbench at the end of the previous cutting period, v l represents the instantaneous speed of the main roller line at the end of the previous cutting period.

[0170] In one embodiment, the cutting conditions of the splitting module 802 include a wire feeding condition and a wire retracting condition. The main roller linear speed in the wire feeding condition is greater than zero, and the main roller linear speed in the wire retracting condition is less than zero. Among them, the wire feeding condition includes: wire feeding acceleration condition: v > 0 and f(v, t) > 0; wire feeding constant speed cutting condition: v > 0 and f(v, t) = 0; wire feeding deceleration condition: v > 0 and f(v, t) < 0; the wire retracting condition includes: wire retracting acceleration condition: v < 0 and f(v, t) < 0; wire retracting constant speed cutting condition: v < 0 and f(v, t) = 0; wire retracting deceleration condition: v < 0 and f(v, t) > 0; v represents the main roller linear speed, f(v, t) represents the gradient of the main roller linear speed.

[0171] Based on the same concept, the present application also provides a computer device, including a memory and a processor. This computer device can be a terminal, and its internal structure diagram can be as Figure 9 shown. This computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. When the computer program is executed by the processor, it is used to implement the disconnection anomaly prediction method. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of this computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad set on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0172] Those skilled in the art can understand that Figure 9 the structure shown in

[0173] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0174] The above embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for predicting abnormal wire breakage, characterized in that, The method includes: Obtaining cutting data of the previous cutting cycle, where the cutting data at least includes cutting process data and the wear amount of the cutting line, and the cutting process data at least includes the tension of the cutting line and the main roller linear speed; Dividing the previous cutting cycle into multiple cutting conditions according to the main roller linear speed and the gradient of the main roller linear speed, and decomposing the cutting process data into multiple sub-data sequences according to the main roller linear speed and the gradient of the main roller linear speed of adjacent cutting conditions, where each sub-data sequence corresponds to one of the cutting conditions; Using a convolutional neural network to extract features from the sub-data sequence corresponding to each cutting condition to obtain the time-series features of each cutting condition; Performing feature fusion on the time-series features of multiple cutting conditions to obtain fused features; Obtaining a pre-trained reliability prediction model, and inputting at least the fused features and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle; the smaller the reliability value, the greater the possibility of wire breakage abnormality of the cutting line; Training the reliability prediction model includes: constructing a reliability curve of the cutting line based on the cumulative distribution function of the Weibull distribution; generating a segmented reliability function based on the reliability curve of the cutting line; determining the label in the feature-label pair based on the segmented reliability function; the feature-label pair includes a label and a feature, the feature in the feature-label pair is the time-series feature of each condition in the historical cutting cycle, and the label in the feature-label pair is the reliability value corresponding to the historical cutting cycle; training the reliability prediction model using the feature-label pair, the wear amount of the cutting line, and the actual position of the workbench.

2. The disconnection anomaly prediction method according to claim 1, wherein Using a convolutional neural network to extract features from the sub-data sequence corresponding to each cutting condition to obtain the time-series features of each cutting condition, including: Padding the sub-data sequence shorter than the set sequence length through a padding operation; Marking whether the elements in the padded sub-data sequence are valid data through the parameters in the mask, where the parameters in the mask include a first parameter and a second parameter with different sizes, the first parameter is used to indicate that the element in the sub-data sequence corresponding to this position is valid data, and the second parameter indicates that the element in the sub-data sequence corresponding to this position is padding data; When using a convolutional neural network to extract features from the sub-data sequence corresponding to each cutting condition, multiplying the padded sub-data sequence by the corresponding parameter in the mask to filter out the padding data so that the padding data does not participate in the convolution calculation.

3. The wire breakage abnormality prediction method according to claim 1, wherein Performing feature fusion on the time-series features of multiple cutting conditions Obtaining fused features, including: Map the timing features for each cutting condition to a query matrix Q i , a key matrix K i and a value matrix V i . Based on the mapped query matrix Q i , the key matrix K i and the value matrix V i , obtain a new feature matrix for each cutting condition Z i ; Connecting the new feature matrices end to end in the time order of each condition appearing in the previous cutting cycle to obtain fused features; Among them, the query matrix Q i = W Q y i , the key matrix K i = W K y i , the value matrix V i = W V y i , the W Q , W K and W V are all parameter matrices, and the y i represents the timing characteristics corresponding to the i-th cutting condition.

4. The method for predicting abnormal wire breakage according to claim 3, wherein, According to the queried matrix of the mapping Q i , the key matrix K i and the value matrix V i , a new feature matrix for each cutting condition is obtained through the following calculation method Z i : ; Among them, is the l -th value of the key matrix for the j-th cutting condition, and softmax is exponential normalization, and the formula is: ； Among them, X represents the product result sequence of the query matrix of the working condition and the key matrix of the working condition. The product result sequence X is 1, 2, … l …, L , L indicating that the length of the product result sequence X is L .

5. The method for predicting wire breakage abnormality according to claim 1, wherein Among them, Input at least the wear amount of the fused feature and the cutting line into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle, including: input the fused feature, the wear amount of the cutting line, and the current actual position of the workbench into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle.

6. The method for predicting abnormal wire breakage according to claim 5, wherein Generate a piecewise reliability function based on the reliability curve of the cutting line, including: Determine the piecewise reliability function according to the first time and the second time; wherein, if the usage time of the cutting line is less than the first time, the reliability value of the cutting line is 1, and if the usage time of the cutting line is greater than the second time, the reliability value of the cutting line is 0. The piecewise reliability function is as follows: ; Among them, represents the first time, represents the second time, represents the usage time of the cutting line, k represents the shape parameter, and λ represents the scale parameter.

7. The wire break anomaly prediction method according to claim 1, wherein Determine the wear amount of the cutting line through the following formula; ； Among them, represents the wear amount of the cutting line, idx represents the cycle number of the previous cutting cycle, cir_len represents the length of the cutting line around the main roller for one week, len_n represents the net length of the cutting line in the previous cutting cycle, v b represents the instantaneous speed of the workbench at the end of the previous cutting cycle, v l represents the instantaneous linear speed of the main roller at the end of the previous cutting cycle.

8. The method for predicting abnormal wire breakage according to any one of claims 1 to 7, characterized in that The cutting working conditions include a wire feeding working condition and a wire rewinding working condition. The main roller wire speed in the wire feeding working condition is greater than zero, and the main roller wire speed in the wire rewinding working condition is less than zero; wherein, the wire feeding working condition includes: Wire feeding acceleration condition: v > 0 and f(v, t) > 0; Wire feeding and uniform cutting condition: v > 0 and f(v, t) = 0; Wire feeding deceleration condition: v > 0 and f(v, t) < 0; The wire rewinding working condition includes: Wire take-up acceleration condition: v < 0 and f(v, t) < 0; Wire take-up uniform cutting condition: v < 0 and f(v, t) = 0; Wire take-up deceleration condition: v < 0 and f(v, t) > 0; v represents the linear velocity of the main roller, f(v,t) represents the gradient of the linear velocity of the main roller.

9. A wire break anomaly prediction device, characterized in that, The device includes: An acquisition module, configured to acquire the cutting data of the previous cutting cycle. The cutting data at least includes cutting process data and the wear amount of the cutting line. The cutting process data at least includes the tension of the cutting line and the main roller wire speed. A segmentation module, configured to segment the previous cutting cycle into multiple cutting working conditions according to the main roller wire speed and the gradient of the main roller wire speed, and decompose the cutting process data into multiple sub-data sequences according to the main roller wire speed and the gradient of the main roller wire speed of adjacent cutting working conditions. Each sub-data sequence corresponds to one of the cutting working conditions. A feature extraction module, configured to extract features from the sub-data sequence corresponding to each cutting working condition by using a convolutional neural network to obtain the time series features of each cutting working condition. A feature fusion module, configured to fuse the time series features of multiple cutting working conditions to obtain a fused feature. A prediction module, configured to obtain a pre-trained reliability prediction model, and input at least the fused feature and the wear amount of the cutting line into the reliability prediction model to obtain the reliability value of the cutting line for the next cutting cycle; wherein, the smaller the reliability value, the greater the possibility of the cutting line breaking abnormally. Training the reliability prediction model includes: constructing the reliability curve of the cutting line based on the cumulative distribution function of the Weibull distribution; generating a piecewise reliability function based on the reliability curve of the cutting line; determining the label in the feature-label pair based on the piecewise reliability function. The feature-label pair includes a label and a feature. The feature in the feature-label pair is the time series feature of each working condition in the historical cutting cycle, and the label in the feature-label pair is the reliability value corresponding to the historical cutting cycle; training the reliability prediction model by using the feature-label pair, the wear amount of the cutting line, and the actual position of the workbench.

10. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the wire breakage abnormal prediction method according to any one of claims 1 to 8.

Images