Abnormal prediction method, system and equipment for processing parameters of plastic plate and storage medium

Abstract

The invention discloses an anomaly prediction method, system and device for plastic plate machining parameters and a storage medium, and belongs to the technical field of plastic machining, the method comprises the steps that sample data of a machine tool in the machining process is acquired, and the sample data comprises a key factor data sample and a machining result data sample; constructing an initial processing prediction model, inputting the key factor data sample into the initial processing prediction model, and performing iterative training by taking the processing result data sample as output until a completely trained target processing prediction model is obtained; real-time key factor data of the machine tool in the machining process are obtained, and machining result abnormity prediction is conducted on the real-time key factor data according to the target machining prediction model. And integrally checking the key factor data and the processing result data through the processing prediction model, thereby determining the relationship between the processing result data and the key factor data, and further predicting the processing result of the plastic plate in real time according to the real-time key factor data.

Description

Technical Field

[0001] The present invention relates to the technical field of plastic processing, and particularly to a method, system, device and storage medium for predicting anomalies in plastic plate processing parameters. Background Art

[0002] Engineering plastics have become an ideal processing material for profile plates due to their characteristics such as light weight, corrosion resistance, good insulation, and easy processing and forming. However, during numerical control processing, the surface of plastic plates often faces problems of hot melt adhesion and hot melt distortion caused by work hardening and thermal softening phenomena, which directly affect the scrap rate of workpieces and the wear of cutting tools.

[0003] Currently, with the continuous breakthroughs and improvements in materials and cutting tools, optimizing processing parameters has become the main direction for solving the problems of hot melt adhesion and hot melt distortion caused by work hardening and thermal softening during plastic processing. However, the processing parameters in plastic processing are diverse, making it difficult to accurately determine the parameters that cause anomalies in workpieces or cutting tools, and thus it is also difficult to accurately determine the quality of workpieces obtained by the machine tool according to the current parameters.

[0004] Therefore, in the prior art, during the process of processing plastic plates, there is a problem that it is difficult to determine the processing results of plastic plates in real time. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, system, device and storage medium for predicting anomalies in plastic plate processing parameters to solve the problem in the prior art that it is difficult to determine the processing results of plastic plates in real time during the process of processing plastic plates.

[0006] To solve the above problems, the present invention provides a method for predicting anomalies in plastic plate processing parameters, including:

[0007] Obtain sample data during the machining process of the machine tool, where the sample data includes key factor data samples and machining result data samples;

[0008] Construct an initial machining prediction model, input the key factor data samples into the initial machining prediction model, and use the machining result data samples as the output for iterative training until a trained complete target machining prediction model is obtained;

[0009] Obtain real-time key factor data during the machining process of the machine tool, and predict anomalies in machining results based on the target machining prediction model.

[0010] Further, obtaining sample data during the machining process of the machine tool includes:

[0011] Obtain possible factors that cause anomalies in plastic plate processing, and initial sample data during the machining process of the machine tool;

[0012] Perform a correlation analysis on the initial sample data according to the grey relational analysis method to determine the key factors among the possible factors;

[0013] Screen the initial sample data according to the key factors to determine the sample data.

[0014] Furthermore, obtain the possible factors that cause abnormal plastic plate processing, including:

[0015] Analyze the abnormal conditions in the plastic plate processing process through the TRIZ theory to determine the abnormal processing phenomena;

[0016] Propose corresponding solutions for the abnormal processing phenomena through the invention principle and the substance-field model, and determine the possible factors based on the feedback of the solutions.

[0017] Furthermore, the possible factors include temperature, tool material, plastic plate material, spindle speed, feed rate, depth of cut, and humidity. The initial sample data includes the possible factor data samples and the initial processing result data samples; perform a correlation analysis on the initial sample data according to the grey relational analysis method to determine the key factors among the possible factors, including:

[0018] Take the possible factor data samples as the comparison sequences;

[0019] Take the initial processing result data samples as the reference sequences;

[0020] Calculate the correlation coefficients between the comparison sequences and the reference sequences respectively, and determine the possible factors whose correlation coefficients exceed the correlation coefficient threshold as the key factors.

[0021] Furthermore, the initial processing prediction model includes an input layer, a convolutional feature extraction layer, a first fully connected layer, a gated recurrent network layer, an attention unit, a second fully connected layer, and an output layer; construct the initial processing prediction model, input the key factor data samples into the initial processing prediction model, and take the processing result data samples as the output, and perform iterative training until a trained and complete target processing prediction model is obtained, including:

[0022] Input the key factor data samples from the input layer into the convolutional feature extraction layer to extract the spatial features of the key factor data samples;

[0023] Perform deep fusion and dimensional transformation on the spatial features through the first fully connected layer to obtain a preliminary feature integration vector;

[0024] Perform temporal modeling and capture long-term dependence relationships on the preliminary feature integration vector through the gated recurrent network layer to obtain features containing time series;

[0025] Perform dynamic weight assignment and importance screening on the features containing time series through the attention unit to obtain a weighted fusion vector;

[0026] The weighted fusion vector is subjected to feature compression and classification through a second fully connected layer to obtain a decision vector;

[0027] The output layer outputs the processed result data sample according to the decision vector, and performs iterative training until a trained and complete target processing prediction model is obtained.

[0028] Further, before inputting the key factor data sample from the input layer to the convolutional feature extraction layer, it further includes:

[0029] The hyperparameters of the initial processing prediction model are optimized according to the chaotic optimization algorithm to obtain an initial processing prediction model with optimized hyperparameters.

[0030] Further, real-time key factor data during the machining process of the machine tool is obtained, and an abnormal prediction of the machining result is performed on the real-time key factor data according to the target machining prediction model, including:

[0031] An abnormal prediction of the machining result is performed on the real-time key factor data according to the target machining prediction model to determine the machining result data of the machine tool;

[0032] The machining prediction result of the machine tool is determined according to the machining result data, and the machining prediction result includes normal machining result and abnormal machining result.

[0033] Further, after determining the machining prediction result of the machine tool according to the machining result data, it further includes:

[0034] When the machining prediction result is an abnormal machining result, the machining parameters of the machine tool are adjusted according to the corresponding abnormal machining result data.

[0035] To solve the above problems, the present invention further provides an abnormal prediction system for plastic plate processing parameters, including:

[0036] A sample data acquisition module, configured to acquire sample data during the machining process of the machine tool, where the sample data includes key factor data samples and machining result data samples;

[0037] A target machining prediction model determination module, configured to construct an initial machining prediction model, input the key factor data sample into the initial machining prediction model, and use the machining result data sample as the output, and perform iterative training until a trained and complete target machining prediction model is obtained;

[0038] A machining result prediction module, configured to acquire real-time key factor data during the machining process of the machine tool, and perform an abnormal prediction of the machining result on the real-time key factor data according to the target machining prediction model.

[0039] To solve the above problems, the present invention also provides an abnormal prediction device for plastic plate processing parameters, including a processor and a memory. A computer program is stored on the memory. When the computer program is executed by the processor, the abnormal prediction method for plastic plate processing parameters as described above is implemented.

[0040] The beneficial effects of adopting the present invention are as follows: By constructing a processing prediction model, the present invention comprehensively examines the key factor data and processing result data during the machining process of the machine tool, and uses the powerful data processing ability of the processing prediction model itself to perform data learning, so as to determine the direct data relationship between the processing result data of the machine tool and the key factor data during the machining process of the machine tool, facilitating subsequent real-time analysis of the key factor data in the processing parameters during the machining process of the machine tool, and then determining the processing result data corresponding to the real-time key factor data, realizing real-time prediction of the plastic plate processing result. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic flowchart of an embodiment of the abnormal prediction method for plastic plate processing parameters provided by the present invention;

[0042] Figure 2 It is a schematic flowchart of an embodiment of obtaining sample data during the machining process of the machine tool provided by the present invention;

[0043] Figure 3 It is a schematic structural diagram of an embodiment of the initial processing prediction model provided by the present invention;

[0044] Figure 4 It is a schematic flowchart of an embodiment of training the processing prediction model provided by the present invention;

[0045] Figure 5 It is a schematic structural diagram of an embodiment of the gated recurrent network layer provided by the present invention;

[0046] Figure 6 It is a schematic block diagram of an embodiment of the abnormal prediction system for plastic plate processing parameters provided by the present invention;

[0047] Figure 7 It is a schematic block diagram of an embodiment of the abnormal prediction device for plastic plate processing parameters provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will specifically describe the preferred embodiments of the present invention in conjunction with the drawings. The drawings constitute a part of this application and are used together with the embodiments of the present invention to explain the principles of the present invention, rather than to limit the scope of the present invention.

[0049] Engineering plastics have become an ideal processing material for footprint plates due to their characteristics such as light weight, corrosion resistance, good insulation, and easy processing and forming. However, during the CNC machining process, the plastic plate surface often faces problems of hot melt adhesion and hot melt distortion caused by work hardening and thermal softening phenomena, which directly affect the scrap rate of workpieces and the wear of cutting tools.

[0050] Currently, with the continuous breakthroughs and improvements in materials and cutting tools, optimizing processing parameters has become the main direction to solve the problems of hot melt adhesion and hot melt distortion caused by work hardening and thermal softening during plastic processing. However, the processing parameters in plastic processing are diverse, making it difficult to accurately determine the parameters that cause abnormalities in workpieces or cutting tools, and thus it is also difficult to accurately determine the quality of workpieces obtained by the machine tool according to the current parameters.

[0051] Therefore, in the prior art, there is a problem that it is difficult to determine the processing result of the plastic plate in real time during the processing of the plastic plate.

[0052] To solve the above problems, the present invention provides an abnormal prediction method, system, device, and storage medium for plastic plate processing parameters, which will be described in detail below.

[0053] Figure 1 It is a schematic flowchart of an embodiment of the abnormal prediction method for plastic plate processing parameters provided by the present invention. As Figure 1 shown, the abnormal prediction method for plastic plate processing parameters includes:

[0054] S101: Obtain sample data during the machining process of the machine tool, where the sample data includes key factor data samples and processing result data samples;

[0055] S102: Construct an initial machining prediction model, input the key factor data samples into the initial machining prediction model, and use the processing result data samples as the output for iterative training until a trained complete target machining prediction model is obtained;

[0056] S103: Obtain real-time key factor data during the machining process of the machine tool, and perform abnormal prediction of the processing result based on the target machining prediction model for the real-time key factor data.

[0057] In this embodiment, by specifically constructing a machining prediction model to overall control the key factor data and processing result data during the machining process of the machine tool and perform data learning, the direct data relationship between the processing result data of the machine tool and the key factor data during the machining process of the machine tool is determined, so as to facilitate subsequent real-time analysis of the key factor data in the processing parameters during the machining process of the machine tool, and then determine the processing result data corresponding to the real-time key factor data, realizing real-time prediction of the plastic plate processing result.

[0058] As a preferred embodiment, in S101, since the data of the machine tool is diverse and very large in quantity, in order to obtain reliable sample data, as Figure 2 shown, Figure 2 FIG. is a schematic flowchart of an embodiment for obtaining sample data of a machine tool during machining provided by the present invention, including:

[0059] S201: Obtain the possible factors that cause abnormal plastic plate machining and the initial sample data of the machine tool during machining;

[0060] S202: Perform a correlation degree analysis on the initial sample data according to the grey correlation analysis method to determine the key factors among the possible factors;

[0061] S203: Screen the initial sample data according to the key factors to determine the sample data.

[0062] In this embodiment, first, obtain the possible factors that cause abnormal plastic plate machining and the initial sample data of the machine tool during machining; then, perform a correlation degree analysis on the initial sample data according to the grey correlation analysis method to determine the key factors among the possible factors; finally, screen the initial sample data according to the key factors to determine the sample data.

[0063] In this embodiment, by first preliminarily diagnosing the data that causes abnormal plastic plate machining to determine the possible factors that cause abnormal plastic plate machining, then performing targeted analysis on the initial sample data according to the grey correlation analysis method to determine the key factors, and finally screening the initial sample data according to the key factors to determine the sample data, the correlation degree between the sample data and the machining result can be effectively improved, the quantity of the sample data can be reduced, and the reliability of the data can be improved.

[0064] As a preferred embodiment, in S201, in order to obtain the possible factors that cause abnormal plastic plate machining, first, analyze the abnormal conditions during the plastic plate machining process through the TRIZ theory to determine the abnormal machining phenomena; then, propose corresponding solutions for the abnormal machining phenomena through the inventive principles and the substance-field model, and determine the possible factors based on the feedback of the solutions.

[0065] In a specific embodiment, when conducting a component function analysis on a plastic sheet processing system, the tracing plate, chips, and sheet are classified as supersystem components, while the tool, nozzle, spindle, and compressed air are classified as system components. Through component function analysis, the executors and receivers of the functions can be known, and it can be determined whether the nature of the function is a useful function or a harmful function. If it is a useful function, further analysis will be carried out to determine whether its performance level is normal, insufficient, or excessive. Through research, the functional relationships between every two components mainly include insufficient functions and harmful functions. Among them, the harmful functions mainly include the spindle being heated by the tool, friction between the tracing plate and the tool, the tool heating the tracing plate, chips accumulating on the tool due to chip adhesion, and the movement of the sheet affecting the tracing plate. The insufficient functions mainly include the insufficient cooling effect of compressed air on the tool and the plastic sheet.

[0066] Furthermore, by classifying and analyzing the abnormal problems of the plastic sheet, the following results are obtained:

[0067]

[0068] That is to say, during the machining process of the machine tool, the problems that may lead to abnormal machining results, the solutions, and the specific implementation methods for implementing the solutions are shown in the above table.

[0069] As a preferred embodiment, in S202, the possible factors include temperature, tool material, plastic sheet material, spindle speed, feed rate, depth of cut, and humidity. The initial sample data includes the possible factor data samples and the initial machining result data samples.

[0070] In order to conduct a correlation analysis on the initial sample data according to the grey relational analysis method and determine the key factors among the possible factors, first, the possible factor data samples are used as the comparison sequences; then, the initial machining result data samples are used as the reference sequences; finally, the correlation coefficients between the comparison sequences and the reference sequences are calculated respectively, and the possible factors whose correlation coefficients exceed the correlation coefficient threshold are determined as the key factors.

[0071] In this embodiment, the correlation coefficients between the possible factor data samples and the initial machining result data samples are quantitatively determined through the grey relational analysis method, so as to reliably determine the key factors related to the machining results.

[0072] In a specific embodiment, during the milling process of a plastic sheet by a numerically controlled machine tool, the occurrence of hot melt adhesion and hot melt distortion problems involves multiple complex factors. In order to deeply analyze the internal relationship between the machine tool parameters and the phenomena of hot melt adhesion and hot melt distortion, and accordingly extract several key influencing factors that are strongly related to the hot melt adhesion and hot melt distortion problems, first, collect the data for analyzing the factors affecting tool sticking as the comparison sequences. Suppose there are n groups of data to form the following matrix X:

[0073]

[0074] Among them, X' i represents the i-th column vector of matrix X, and x' i (m) represents the i-th data of the m-th influencing factor, m represents the number of plastic processing influencing factors, and n represents the number of data groups.

[0075] For these n groups of data, the corresponding processing result sequence is defined as Y' = (y'(1), y'(2)…, y'(m)).

[0076] Then, during the plastic processing, since the dimensions of the influencing factors and the corresponding results are different, it is necessary to perform dimensionless processing on the data. Specifically, the matrix after dimensionless processing is:

[0077]

[0078] Next, calculate the correlation coefficients of the corresponding elements of each comparison sequence and the reference sequence. Among them, the calculation formula of the correlation coefficient is:

[0079]

[0080] Among them, i = 1, 2,…, n, is the index of the comparison sequence, ρ is the resolution coefficient, where ρ = 0.5, and h = 1, 2,…, m.

[0081] To facilitate data comparison, by calculating the correlation order r i value to determine the corresponding possible factor as the key factor, the calculation formula of the correlation order r i is:

[0082]

[0083] Among them, m represents the number of possible factors.

[0084] It should be noted that the larger the correlation order r i value, the greater the influence of this factor on the result.

[0085] In a specific embodiment, select the 5 possible factors with the largest correlation order value as the key factors.

[0086] In other embodiments, the number of key factors can also be adjusted according to needs.

[0087] As a preferred embodiment, in S102, the initial processing prediction model includes an input layer, a convolutional feature extraction layer, a first fully connected layer, a gated recurrent network layer, an attention unit, a second fully connected layer, and an output layer, as Figure 3 shown, Figure 3Schematic structural diagram of an embodiment of the initial processing prediction model provided by the present invention.

[0088] Further, in order to construct the initial processing prediction model, the key factor data samples are input into the initial processing prediction model, and the processing result data samples are used as the output for iterative training until a trained and complete target processing prediction model is obtained. As Figure 4 shown, Figure 4 Schematic flow diagram of an embodiment of the training processing prediction model provided by the present invention, including:

[0089] S401: Input the key factor data samples from the input layer into the convolutional feature extraction layer to extract the spatial features of the key factor data samples;

[0090] S402: Perform deep fusion and dimensional transformation on the spatial features through the first fully connected layer to obtain a preliminary feature integration vector;

[0091] S403: Perform temporal modeling and capture long-term dependence relationships on the preliminary feature integration vector through the gated recurrent network layer to obtain time series features;

[0092] S404: Perform dynamic weight allocation and importance screening on the time series features through the attention unit to obtain a weighted fusion vector;

[0093] S405: Perform feature compression and classification on the weighted fusion vector through the second fully connected layer to obtain a decision vector;

[0094] S406: The output layer outputs the processing result data samples according to the decision vector for iterative training until a trained and complete target processing prediction model is obtained.

[0095] In this embodiment, the spatial features of the key factor data samples are extracted through the convolutional feature extraction layer to realize the extraction of local features of the plastic sheet processing data; the temporal modeling and long-term dependence relationship capture are performed on the preliminary feature integration vector through the gated recurrent network layer to obtain time series features, realizing the capture of the temporal relationship of the plastic sheet processing data, which can better ensure the reliability of the feature analysis results of the data.

[0096] In a specific embodiment, the gated recurrent network layer consists of the current memory unit H t , update gate Z t and reset gate R t as Figure 5 shown, Figure 5 Schematic structural diagram of an embodiment of the gated recurrent network layer provided by the present invention.

[0097] Among them, the calculation steps of different states in the gated recurrent network are as follows:

[0098] Rt = σ(W R · [H t-1, x t + b R )

[0099] R t is the parameter of the reset gate. The larger R t is, the less the previous state information is retained; x t is the t-th component of the input sequence X; H t-1 represents the information at the previous time step t - 1; + represents the addition operation; W R represents the weight matrix; b R represents the offset; σ represents the sigmoid function.

[0100] Z t = σ(W z · [H t-1 , x t + b z )

[0101] Z t is the parameter of the update gate, and x t is the input vector at the t-th time step, that is, the t-th component of the input sequence X. The update gate will add x z after a linear transformation (multiplied by the weight matrix W t ) to the information H t-1 at the previous time step t - 1, and compress the output result to between 0 and 1 through the Sigmoid activation function. Z t represents the output vector of the sigmoid neural layer. The larger its value, the more information from the previous state is brought in. x t is the t-th component of the input sequence X; H t-1 represents the information at the previous time step t - 1; + represents the addition operation; W z represents the weight matrix; b z represents the offset; σ represents the sigmoid function.

[0102] Specifically, the update gate helps the model determine how much past information needs to be passed to the future, or how much information from the previous time step and the current time step needs to be passed on.

[0103] Furthermore, the calculation formula for generating the candidate activation value is as follows:

[0104]

[0105] Represents a candidate state, and tanh represents an activation function whose output range is between [-1, 1], which enables the gated recurrent network to effectively handle the problems of vanishing gradients and exploding gradients while maintaining the potential of information; x t Is the t-th component of the input sequence X; H t-1 Represents the information at the previous time step t - 1; + represents the superposition operation; W h Represents the weight matrix; b H Represents the offset; the symbol * represents the Hadamard product.

[0106] The calculation formula of the output value is as follows:

[0107]

[0108] Z t Represents the activation result of the Update gate, and also controls the inflow of information in the form of a gate. (1 - Z t ) * H t-1 Represents the information retained from the previous time step to the final memory; Represents the information retained from the current memory to the final memory; + represents the superposition operation, and the symbol * represents the Hadamard product.

[0109] In this embodiment, by using the gated recurrent network to capture the temporal relationship of sequence data, it can better fit the training data to improve the effect of model training.

[0110] As a preferred embodiment, in S401, in order to improve the training efficiency of the initial processing prediction model, before inputting the key factor data samples from the input layer to the convolutional feature extraction layer, it is necessary to optimize the hyperparameters of the initial processing prediction model according to the chaos optimization algorithm to obtain an initial processing prediction model with optimized hyperparameters.

[0111] In a specific embodiment, the chaos optimization algorithm (COA, Chaos Optimization Algorithm) is a new type of meta-heuristic optimization algorithm, which has the characteristics of strong evolutionary ability, fast search speed, and strong optimization ability.

[0112] The COA algorithm simulates the hunting behavior of raccoons through the processes of population initialization, the hunting and attacking strategies of iguanas (exploration stage), and escaping from predators (exploitation stage), and selects the optimal solution as the result.

[0113] In this embodiment, the COA algorithm is improved by using the Bernoulli chaos mapping, variable spiral search strategy, non-linear decreasing weight, and Wright flight mechanism to obtain better search efficiency and robustness. Among them, the process of randomly initializing the population by the Bernoulli chaos mapping is:

[0114]

[0115] Among them, is a random number between (0, 1); X k represents the K - th iteration value of the chaotic mapping sequence.

[0116]

[0117]

[0118] is the new position of the i - th raccoon in the j - th dimension; V is a random number between [0, 1]; G j is the position of the iguana in the j - th dimension, actually representing the position of the best member; I is a number randomly selected from the set {1, 2}; N represents the number of raccoons, and [N / 2] represents the largest integer not exceeding [N / 2]; m represents the number of decision variables.

[0119] The calculation formula of the variable spiral search strategy is:

[0120]

[0121] The z parameter changes with the number of operations; k is the change coefficient, k = 5; I represents a random number uniformly distributed between [-1, 1].

[0122] The calculation formula of the exploration stage based on non - linear decreasing weight update is:

[0123] W(t) = (M - n + 1) / M n

[0124]

[0125]

[0126]

[0127] W is the non - linear decreasing weight; M is the maximum number of iterations; n is the current number of iterations; V is a random number between [0, 1]; is the new position of the i - th raccoon in the j - th dimension; F i P1 is the objective function value of the i - th raccoon at the new position; F i is the objective function value of the i - th raccoon at the previous position; X i (t) represents the position of the i - th raccoon in the t - th iteration, is the landing position of the iguana in the j - th dimension; represents the objective function value after the iguana lands in the j - th dimension; Fi,j represents the objective function value of the $i$-th raccoon under the $j$-th dimension; $F$ i P1 is the objective function value of the $i$-th raccoon at the new position; $F$ i is the objective function value of the $i$-th raccoon at the previous position.

[0128]

[0129] is the local lower bound of the $j$-th decision variable; is the local upper bound of the $j$-th decision variable; $t$ is the number of iterations; $T$ is the maximum number of iterations; is the lower bound of the $j$-th decision variable; is the upper bound of the $j$-th decision variable.

[0130] By combining the Levy flight mechanism and the non-linear decreasing weight, the search method is balanced and the search ability of the algorithm is enhanced, so that the quality of each solution is improved to a certain extent. Specifically, the calculation process is as follows:

[0131]

[0132]

[0133]

[0134]

[0135] $V$ is a random number between $[0,1]$; is the new position of the $i$-th raccoon in the $j$-th dimension; $F$ i P2 is the objective function value of the $i$-th raccoon at the new position; $F$ i is the objective function value of the $i$-th raccoon at the previous position; $X$ i $(t)$ represents the position of the $i$-th raccoon in the $t$-th iteration; $X$ i '$(t)$ represents the position after the Levy flight mechanism update; represents the dot product; $Z$ represents the step size control parameter; $\Gamma$ is the gamma function; $levy(\lambda)$ represents the path following the Levy distribution; $\sigma$ ν is 1; $\gamma$ is 1.5.

[0136] In this embodiment, by tuning the hyperparameters of the initial processing prediction model, the reliability of the hyperparameters at the beginning of model training is improved, so as to reduce the difficulty of model training.

[0137] As a preferred embodiment, in S103, in order to predict abnormal processing results based on the target processing prediction model for real-time key factor data, first, predict abnormal processing results based on the target processing prediction model for real-time key factor data to determine the processing result data of the machine tool; then, determine the processing prediction result of the machine tool according to the processing result data, where the processing prediction result includes normal processing results and abnormal processing results.

[0138] In a specific embodiment, by presetting the threshold range of the processing result data, compare the directly predicted processing result data with the threshold range of the processing result data. When the processing result data is within the threshold range of the processing result data, determine that the processing prediction result is a normal processing result; when the processing result data is outside the threshold range of the processing result data, determine that the processing prediction result is an abnormal processing result.

[0139] Further, in order to adaptively adjust the machine tool according to the processing prediction result to avoid unqualified workpieces, when the processing prediction result is an abnormal processing result, adjust the processing parameters of the machine tool according to the corresponding abnormal processing result data.

[0140] In a specific embodiment, when the processing prediction result is an abnormal processing result, feedback-adjust the processing parameters of the numerically controlled machine tool, that is, when hot melt adhesion and hot melt distortion occur, the system adjusts the spindle speed, feed speed, and depth of cut to avoid abnormal conditions; if there is no abnormal condition, process according to the original processing parameters.

[0141] In the above manner, by specially constructing a processing prediction model to overall check the key factor data and processing result data during the processing of the machine tool, and perform data learning, so as to determine the direct data relationship between the processing result data of the machine tool and the key factor data during the processing of the machine tool, so as to facilitate subsequent real-time analysis of the key factor data in the processing parameters during the processing of the machine tool, and then determine the processing result data corresponding to the real-time key factor data, and realize real-time prediction of the plastic plate processing result.

[0142] To solve the above problems, the present invention also provides an abnormal prediction system for plastic plate processing parameters, as Figure 6 shown, Figure 6 is a structural block diagram of an embodiment of the abnormal prediction system for plastic plate processing parameters provided by the present invention. The abnormal prediction system 600 for plastic plate processing parameters includes:

[0143] A sample data acquisition module 601, configured to acquire sample data during the processing of the machine tool, where the sample data includes key factor data samples and processing result data samples;

[0144] The target processing prediction model determination module 602 is configured to construct an initial processing prediction model, input the key factor data samples into the initial processing prediction model, and take the processing result data samples as the output, and perform iterative training until a trained complete target processing prediction model is obtained;

[0145] The processing result prediction module 603 is configured to obtain the real-time key factor data during the machining process of the machine tool, and perform machining result anomaly prediction on the real-time key factor data according to the target machining prediction model.

[0146] The present invention also correspondingly provides an abnormal prediction device for plastic plate processing parameters, as Figure 7 shown Figure 7 is a structural block diagram of an embodiment of the abnormal prediction device for plastic plate processing parameters provided by the present invention. The abnormal prediction device 700 for plastic plate processing parameters may be a computing device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The abnormal prediction device 700 for plastic plate processing parameters includes a processor 701 and a memory 702, wherein, an abnormal prediction program 703 for plastic plate processing parameters is stored on the memory 702.

[0147] In some embodiments, the memory 702 may be an internal storage unit of the computer device, such as the hard disk or memory of the computer device. In other embodiments, the memory 702 may also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. Further, the memory 702 may also include both the internal storage unit and the external storage device of the computer device. The memory 702 is used to store the application software installed on the computer device and various types of data, such as the program code installed on the computer device. The memory 702 may also be used to temporarily store the data that has been output or will be output. In one embodiment, the abnormal prediction program 703 for plastic plate processing parameters can be executed by the processor 701, so as to implement the abnormal prediction methods, systems, devices, and storage media for plastic plate processing parameters in various embodiments of the present invention.

[0148] In some embodiments, the processor 701 may be a Central Processing Unit (CPU), a microprocessor, or other data processing chips, and is used to run the program code stored in the memory 702 or process data, such as executing the abnormal prediction program for plastic plate processing parameters, etc.

[0149] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0150] As mentioned above, the above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. An abnormal prediction method for plastic sheet processing parameters, characterized in that Including: Obtain sample data during the machining process of the machine tool, where the sample data includes key factor data samples and machining result data samples; Construct an initial machining prediction model, input the key factor data samples into the initial machining prediction model, and use the machining result data samples as the output for iterative training until a trained complete target machining prediction model is obtained; Obtain real-time key factor data during the machining process of the machine tool, and predict machining result anomalies based on the target machining prediction model for the real-time key factor data.

2. The abnormal prediction method for plastic sheet processing parameters according to claim 1, wherein The obtaining of the sample data during the machining process of the machine tool includes: Obtain the possible factors causing abnormal machining of the plastic plate and the initial sample data during the machining process of the machine tool; Conduct correlation analysis on the initial sample data according to the grey relational analysis method to determine the key factors among the possible factors; Screen the initial sample data according to the key factors to determine the sample data.

3. The abnormal prediction method for plastic sheet processing parameters according to claim 2, characterized in that The obtaining of the possible factors causing abnormal machining of the plastic plate includes: Analyze the abnormal conditions during the plastic plate machining process through the TRIZ theory to determine the machining abnormal phenomena; Propose corresponding solutions for the machining abnormal phenomena through the invention principle and the substance-field model, and determine the possible factors based on the feedback of the solutions.

4. The abnormal prediction method for plastic sheet processing parameters according to claim 2, characterized in that, The possible factors include temperature, tool material, plastic plate material, spindle speed, feed rate, depth of cut, and humidity. The initial sample data includes possible factor data samples and initial machining result data samples. The conducting of the correlation analysis on the initial sample data according to the grey relational analysis method to determine the key factors among the possible factors includes: Use the possible factor data samples as the comparison sequences; Use the initial machining result data samples as the reference sequences; Calculate the correlation coefficients between the comparison sequences and the reference sequences respectively, and determine the possible factors with the correlation coefficients exceeding the correlation coefficient threshold as the key factors.

5. The abnormal prediction method for plastic sheet processing parameters according to claim 1, characterized in that, The initial machining prediction model includes an input layer, a convolutional feature extraction layer, a first fully connected layer, a gated recurrent network layer, an attention unit, a second fully connected layer, and an output layer. The constructing of the initial machining prediction model, inputting the key factor data samples into the initial machining prediction model, and using the machining result data samples as the output for iterative training until a trained complete target machining prediction model is obtained includes: Input the key factor data samples from the input layer into the convolutional feature extraction layer to extract the spatial features of the key factor data samples; Conduct deep fusion and dimensional transformation on the spatial features through the first fully connected layer to obtain a preliminary feature integration vector; Conduct temporal modeling and capture long-term dependence relationships on the preliminary feature integration vector through the gated recurrent network layer to obtain time series features; Conduct dynamic weight assignment and importance screening on the time series features through the attention unit to obtain a weighted fusion vector; Conduct feature compression and classification on the weighted fusion vector through the second fully connected layer to obtain a decision vector; The output layer outputs the processed result data sample according to the decision vector, and performs iterative training until a trained complete target processing prediction model is obtained.

6. The abnormal prediction method for plastic sheet processing parameters according to claim 5, characterized in that, Before inputting the key factor data sample from the input layer to the convolutional feature extraction layer, it further includes: Optimizing the hyperparameters of the initial processing prediction model according to the chaotic optimization algorithm to obtain an initial processing prediction model with optimized hyperparameters.

7. The abnormal prediction method for plastic sheet processing parameters according to claim 1, characterized in that Obtaining the real-time key factor data during the machining process of the machine tool, and predicting the abnormal machining result of the real-time key factor data according to the target processing prediction model, including: Predicting the abnormal machining result of the real-time key factor data according to the target processing prediction model, and determining the machining result data of the machine tool; Determining the machining prediction result of the machine tool according to the machining result data, where the machining prediction result includes normal machining result and abnormal machining result.

8. The abnormal prediction method for plastic sheet processing parameters according to claim 1, characterized in that, After determining the machining prediction result of the machine tool according to the machining result data, it further includes: When the machining prediction result is an abnormal machining result, adjusting the machining parameters of the machine tool according to the corresponding abnormal machining result data.

9. An abnormal prediction system for plastic sheet processing parameters, characterized in that, It includes: A sample data acquisition module, configured to acquire sample data during the machining process of the machine tool, where the sample data includes key factor data samples and processed result data samples; A target processing prediction model determination module, configured to construct an initial processing prediction model, input the key factor data sample into the initial processing prediction model, and use the processed result data sample as the output, and perform iterative training until a trained complete target processing prediction model is obtained; A machining result prediction module, configured to acquire the real-time key factor data during the machining process of the machine tool, and predict the abnormal machining result of the real-time key factor data according to the target processing prediction model.

10. An abnormal prediction device for plastic plate processing parameters, characterized in that, It includes a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, it implements the abnormal prediction method for plastic plate machining parameters according to any one of claims 1-8.

Images