Bad material prediction method and device, electronic equipment and computer storage medium

Abstract

The invention relates to the technical field of production scheduling, and discloses a bad material prediction method and device, electronic equipment and a computer storage medium, and the method comprises the steps: inputting target material feature data into a bad material proportion prediction model, and obtaining a target bad part proportion value; inputting the target material feature data into a dichotomy model to obtain a bad piece generation probability; performing parameter optimization based on the target bad part proportion value and the bad part generation probability to obtain a Pareto optimal solution set; and obtaining an optimal bad piece number based on the satisfaction rate and the dull number of each data point in the Pareto optimal solution set. According to the method, parameter optimization adjustment is carried out through the bad part proportion value and the bad part generation probability, the Pareto optimal solution set of the satisfaction rate and the dull number is obtained, the minimum dull number can be obtained under different satisfaction rates, then the optimal bad part number is obtained through the Pareto optimal solution set, the bad part number is accurately predicted, and the accuracy of bad part prediction is improved.

Description

Technical Field

[0001] This application relates to the technical field of production scheduling, and in particular, to a method, device, electronic device, and computer storage medium for predicting defective materials. Background Art

[0002] Material management is an important part of supply chain management. Manufacturers need to prepare the materials required for production in advance to ensure that products can be produced and delivered within the specified time. Since some materials may be damaged during the production process, manufacturers need to predict the damage ratio of materials in the next production cycle when purchasing, so as to reduce the situation of insufficient material preparation.

[0003] During the prediction process of the damage ratio, considering the influence of the fulfillment rate and the stagnant quantity, it is necessary to obtain the fulfillment rate and the stagnant quantity. The existing method is to obtain the fulfillment rate and the stagnant quantity through manual adjustment or grid search. Therefore, the obtained fulfillment rate and stagnant quantity cannot take into account the optimization of multiple objectives at the same time, resulting in the inability to accurately predict the number of defective parts, which is likely to cause overstocking of materials by manufacturers or insufficient material preparation, resulting in the inability to produce on schedule, and the accuracy of defective material prediction is low. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the related art. To this end, embodiments of this application provide a method, device, electronic device, and computer storage medium for predicting defective materials, which can accurately predict the number of defective parts and improve the accuracy of defective material prediction.

[0005] In a first aspect, embodiments of this application provide a method for predicting defective materials, including:

[0006] Inputting the target material characteristic data to be processed into a defective material ratio prediction model to obtain a target defective part ratio value to be optimized output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on sample material characteristic data and its corresponding defective part ratio value;

[0007] Inputting the target material characteristic data into a binary classification model to obtain a probability of generating defective parts output by the binary classification model; the binary classification model is trained based on the sample material characteristic data and its corresponding label data of whether defective parts are generated;

[0008] Performing parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a fulfillment rate and its corresponding stagnant quantity;

[0009] Based on the fulfillment rate and stagnant quantity of each data point in the Pareto optimal solution set, obtaining the optimal number of defective parts.

[0010] In a second aspect, an embodiment of the present application provides a defective material prediction device, including:

[0011] A defective parts ratio prediction module, configured to input target material feature data into a defective material ratio prediction model, and obtain a target defective parts ratio value output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on sample material feature data and their corresponding defective parts ratio values;

[0012] A defective parts probability prediction module, configured to input the target material feature data into a binary classification model, and obtain a probability of generating defective parts output by the binary classification model; the binary classification model is trained based on the sample material feature data and their corresponding labels indicating whether defective parts are generated;

[0013] A parameter optimization module, configured to perform parameter optimization based on the target defective parts ratio value and the probability of generating defective parts, and obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding stagnant number;

[0014] A defective parts number determination module, configured to obtain an optimal defective parts number based on the satisfaction rates and stagnant numbers of the respective data points in the Pareto optimal solution set.

[0015] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory storing multiple instructions; a processor loads instructions from the memory to execute any one of the defective material prediction methods provided by the embodiments of the present application.

[0016] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing multiple instructions, and the instructions are suitable for being loaded by a processor to execute any one of the defective material prediction methods provided by the embodiments of the present application.

[0017] In a fifth aspect, an embodiment of the present application further provides a computer program product, including a computer program or instructions, and when the computer program or instructions are executed by a processor, any one of the defective material prediction methods provided by the embodiments of the present application is implemented.

[0018] The embodiments of the present application perform optimization and adjustment through the defective parts ratio value and the probability of generating defective parts, obtain a Pareto optimal solution set of the satisfaction rate and the stagnant number, so that the minimum stagnant number can be obtained under different satisfaction rates, and the optimal defective parts number is obtained through the Pareto optimal solution set, realizing accurate prediction of the defective parts number and improving the accuracy of defective material prediction. Description of the Drawings

[0019] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0020] Figure 1 It is one of the schematic flowcharts of the blank prediction method provided in the embodiments of the present application;

[0021] Figure 2 It is the schematic diagram of the Pareto optimal solution set provided in the embodiments of the present application;

[0022] Figure 3 It is the schematic diagram of the residual connection layer provided in the embodiments of the present application;

[0023] Figure 4 It is the schematic flowchart of the defective parts ratio value provided in the embodiments of the present application;

[0024] Figure 5 It is the second schematic flowchart of the blank prediction method provided in the embodiments of the present application;

[0025] Figure 6 It is the schematic structural diagram of the blank prediction device provided in the embodiments of the present application;

[0026] Figure 7 It is the schematic structural diagram of the electronic device provided in the embodiments of the present application. Specific Embodiments

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application. At the same time, in the description of the embodiments of the present application, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of the embodiments of the present application, "a plurality of" means two or more, unless otherwise clearly and specifically defined.

[0028] Embodiments of the present application provide a defective material prediction method, apparatus, electronic device, and computer storage medium. Specifically, embodiments of the present application will be described from the perspective of a defective material prediction apparatus, which can be specifically integrated in an electronic device, that is, the defective material prediction method in embodiments of the present application can be executed by the electronic device. Optionally, the electronic device includes a terminal device. The terminal device can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop computer, a game console, or a personal computer (PC), etc. Optionally, the electronic device includes a server, which can be an independent server or a server network or server cluster composed of servers, including but not limited to a computer, a network host, a single network server, a set of network servers, or a cloud server composed of servers. Among them, the cloud server is composed of a large number of computers or network servers based on cloud computing.

[0029] It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments. Although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from that shown in the drawings.

[0030] In the first aspect, the existing method obtains the satisfaction rate and the number of slow-moving items by manual adjustment or grid search. Therefore, the obtained satisfaction rate and the number of slow-moving items cannot simultaneously take into account the optimization of multiple objectives, resulting in the inability to accurately predict the number of defective parts, which is likely to cause overstocking of materials by manufacturers due to over-preparation or inability to produce on schedule due to insufficient material preparation, resulting in low accuracy of defective material prediction.

[0031] In order to improve the prediction accuracy of defective materials and take into account the optimal conditions of the satisfaction rate and the number of slow-moving items, it is necessary to combine a defective material ratio prediction model and a binary classification model to output an accurate target defective part ratio value and the probability of generating defective parts. Through the target defective part ratio value and the probability of generating defective parts, a Pareto optimal solution set of the satisfaction rate and the number of slow-moving items is optimized, and then the optimal number of defective parts is obtained through the Pareto optimal solution set. Therefore, the minimum number of slow-moving items can be obtained under different satisfaction rates. By obtaining the optimal number of defective parts through the Pareto optimal solution set, the number of defective parts can be accurately predicted, and the accuracy of defective material prediction is improved.

[0032] Optionally, embodiments of the present application take a defective material prediction apparatus as the execution subject for illustration. The following will be described in detail with reference to the accompanying drawings respectively. Refer to Figure 1 , Figure 1 is one of the flowcharts of the defective material prediction method provided in embodiments of the present application. The specific process of the defective material prediction method provided in embodiments of the present application can be as follows: steps 10 to 40, including:

[0033] Step 10: Input the target material feature data to be processed into the defective material ratio prediction model, and obtain the target defective part ratio value to be optimized output by the defective material ratio prediction model.

[0034] The defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value. Specifically, it is described in Steps 101 to 104.

[0035] Optionally, the defective material prediction device inputs the target material feature data to be processed into the defective material ratio prediction model, and obtains the target defective part ratio value to be optimized output by the defective material ratio prediction model. Among them, the target material feature data is obtained after processing the metadata required for defective part prediction, and the defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value.

[0036] Optionally, the target material feature data obtained by the defective material prediction device after processing the metadata required for defective part prediction is specifically: the defective material prediction device collects the metadata required for defective part prediction, and the metadata includes material feeding history records, material defective part history records, factory information, material basic attribute information, etc.

[0037] Further, the defective material prediction device cleans and filters the abnormal samples in the collected metadata, such as samples with a defective material ratio greater than 0.05, samples belonging to trial production, etc.

[0038] Further, the defective material prediction device constructs the features required by the model according to the defective material ratio prediction model to obtain the target material feature data. Specifically, in one embodiment, the features required by the model include historical monthly defective material ratio features, static features, and monthly demand features.

[0039] Among them, the historical monthly defective material ratio feature y T-h:T-1 , T is the current month, h is the defective material ratio of the input historical h months. Since the production process of the factory will improve over time, h can be set to 3 to 6 months.

[0040] The static feature a at least includes material basic features, factory basic features, factory historical window features, material historical window features, and category historical window features. The material basic features include category, supplier, price, specification, etc. The factory basic features include factory id and number of workers, etc. The factory historical window features include the defective material ratio of the factory in the previous quarter, the average number of defective materials in the previous quarter of the factory, and the defective material ratio of the factory in the same period last year, etc. The material historical window features include the defective material ratio of the material in the previous quarter, the average number of defective materials in the previous quarter of the material, and the defective material ratio of the material in the same period last year, etc. The category historical window features include the defective material ratio of the category in the previous quarter, the average number of defective materials in the previous quarter of the category, and the defective material ratio of the category in the same period last year, etc.

[0041] The monthly demand feature xT-h:T For the feeding requirements of each month.

[0042] Step 20: Input the target material characteristic data into the binary classification model to obtain the probability of defective parts generated by the binary classification model.

[0043] Optionally, the defective material prediction device inputs the target material characteristic data into the binary classification model to obtain the probability of defective parts generated by the binary classification model. The binary classification model is trained based on the sample material characteristic data and its corresponding label data indicating whether defective parts are generated. Specifically:

[0044] The binary classification model uses a machine learning model for binary classification prediction of the probability of defective materials. The machine learning model includes but is not limited to linear regression, random forest, gradient boosting tree, LightGBM (Light Gradient Boosting Machine), deep learning, etc. The constructed data features are divided into a training set, a validation set, and a test set in a ratio of 7:2:1 according to consecutive months. Then, they are input into machine learning for training and prediction to obtain the probability of defective materials.

[0045] Step 30: Optimize the parameters based on the target defective parts ratio value and the probability of defective parts generated to obtain the Pareto optimal solution set.

[0046] Optionally, the defective material prediction device optimizes the parameters of the target defective parts ratio value and the probability of defective parts generated by setting spare part thresholds or non-spare part thresholds through a multi-objective optimization algorithm, obtains a set of data points including the satisfaction rate and its corresponding stagnant number, and determines this set of data points as the Pareto optimal solution set. Therefore, it can be understood that the Pareto optimal solution set is the Pareto optimal solution set of the satisfaction rate and the stagnant number. The Pareto optimal solution set includes at least two data points, and each data point includes the satisfaction rate and its corresponding stagnant number, as specifically described in Steps 301 to 303.

[0047] In one embodiment, for the case where the probability of defective parts generated is less than the threshold low_thes, no materials are prepared, that is, no materials are prepared for the case with a low probability of defects. For the case where the probability of defective parts generated is greater than high_thes, one more material is prepared. For the case where the target defective parts ratio value is less than the threshold bad_thes, no materials are prepared, and other threshold parameters are used for multi-objective tuning to obtain the Pareto optimal solution set. The Pareto optimal solution set is as Figure 2 shown Figure 2 It is a schematic diagram of the Pareto optimal solution set provided in the embodiment of the present application.

[0048] Step 40: Obtain the optimal number of defective parts based on the satisfaction rate and the stagnant number of each data point in the Pareto optimal solution set.

[0049] Optionally, the Pareto optimal solution set consists of multiple groups of different approximate optimal parameters. Therefore, it is necessary to select the optimal group of parameters according to the actual scenario requirements. Thus, the defective material prediction device predicts the optimal data point in the Pareto optimal solution set based on the satisfaction rate and the number of stagnant items of each data point in the Pareto optimal solution set. Further, the defective material prediction device calculates the number of defective items according to the optimal threshold parameter corresponding to the optimal data point, and determines the number of defective items obtained after calculating the optimal threshold parameter as the optimal number of defective items, as specifically described in steps 401 to 403.

[0050] In the embodiment of the present application, optimization and adjustment are performed through the defective item ratio value and the probability of generating defective items to obtain the Pareto optimal solution set of the satisfaction rate and the number of stagnant items, so that the minimum number of stagnant items can be obtained under different satisfaction rates. The optimal number of defective items is obtained through the Pareto optimal solution set, realizing the accurate prediction of the number of defective items and improving the accuracy of defective material prediction.

[0051] In the second aspect, most of the existing defective material ratio prediction models are time series prediction structures based on LSTM, GRU, or attention mechanism, and the accuracy of the defective item ratio output by the defective material ratio prediction model with the time series prediction structure of LSTM, GRU, or attention mechanism is low.

[0052] The defective material ratio prediction model in the embodiment of the present application is composed of a feature fusion layer, a first fully connected layer, multiple residual connection layers, and a second fully connected layer. Therefore, the defective material ratio prediction model based on residual connection in the embodiment of the present application has fewer model parameters and can accurately output the target defective item ratio value.

[0053] In an optional embodiment, the descriptions of steps 101 to 104 are as follows:

[0054] Step 101: Input the target material feature data into the feature fusion layer for feature splicing to obtain material feature spliced data;

[0055] Step 102: Input the material feature spliced data into the first fully connected layer for data extraction to obtain hidden layer data;

[0056] Step 103: Input the hidden layer data and the material feature spliced data into the residual connection layer for data extraction to obtain residual layer data;

[0057] Step 104: Input the residual layer data into the second fully connected layer for result prediction to obtain the target defective item ratio value output by the defective material ratio prediction model.

[0058] It should be noted that the scrap ratio prediction model is a scrap ratio prediction model based on residual connection. In one embodiment, the scrap ratio prediction model includes a feature fusion layer, a first fully connected layer, a plurality of residual connection layers, and a second fully connected layer. The output of the feature fusion layer is connected to the input of the first fully connected layer, the output of the first fully connected layer is connected to the input of the residual connection layer, and the output of the residual connection layer is connected to the input of the second fully connected layer. Each residual connection layer is as Figure 3 shown, Figure 3 which is a schematic diagram of the residual connection layer provided in the embodiment of the present application.

[0059] Optionally, referring to Figure 4 , Figure 4 which is a schematic diagram of the defective parts ratio value in the embodiment of the present application. The scrap prediction device inputs the target material feature data into the feature fusion layer, and the feature fusion layer performs feature splicing to obtain the spliced material feature data. Continuing with the above embodiment, the target material feature data includes the historical monthly scrap ratio feature y T-h:T-1 , static feature a, and monthly demand feature x T-h:T . Therefore, the spliced material feature data can be expressed as (y T-h:T-1 , a, x T-h:t ).

[0060] Further, the scrap prediction device inputs the spliced material feature data into the first fully connected layer (Dense fully connected layer), and the first fully connected layer performs data extraction to obtain the hidden layer data, which can be specifically expressed as:

[0061]

[0062] wherein, is the hidden layer data.

[0063] Further, the scrap prediction device inputs the spliced material feature data and the hidden layer data into the residual connection layer, and the residual connection layer performs effective information extraction to obtain the residual layer data, which can be specifically expressed as:

[0064]

[0065]

[0066] wherein, is the residual layer data.

[0067] Further, the scrap prediction device inputs the residual layer data into the second fully connected layer, and the second fully connected layer performs result prediction to obtain the target defective parts ratio value output by the scrap ratio prediction model. The target defective parts ratio value can be expressed as

[0068] It should be noted that the loss function of the defective material ratio prediction model adopts the MSE loss of the prediction task:

[0069]

[0070] where S represents all the constructed training samples, T represents the current month, and y T represents the true defective material ratio, and represents the target defective part ratio value.

[0071] The embodiment of the present application adopts a defective material ratio prediction model based on residual connection. This model splices the input features into each layer of the residual connection layer, enabling each layer of the residual connection layer to obtain the input features, avoiding the problem of network degradation caused by too many network layers, and improving the prediction accuracy of the model. Therefore, the target defective part ratio value can be accurately output through the defective material ratio prediction model.

[0072] In the third aspect, the existing multi-objective optimization methods are mainly genetic algorithms, and the Pareto optimal solution set obtained by the genetic algorithm cannot simultaneously take into account the tuning of multiple objectives.

[0073] The embodiment of the present application can simultaneously take into account the tuning of multiple objectives through non-dominated sorting and crowding distance sorting, obtain the Pareto optimal solution set of the satisfaction rate and the number of slow-moving items, so that the minimum number of slow-moving items can be obtained under different satisfaction rates, and the optimal number of defective parts can be obtained through the Pareto optimal solution set, realizing the accurate prediction of the number of defective parts and improving the accuracy of defective material prediction.

[0074] In an optional embodiment, the descriptions of steps 301 to 303 are as follows:

[0075] Step 301, determine the objective function of maximizing the satisfaction rate and the objective function of minimizing the number of slow-moving items;

[0076] Step 302, taking the target defective part ratio value and the defective part generation probability as optimization variables, and taking the first threshold of the target defective part ratio value and the second threshold of the defective part generation probability as constraint conditions, perform parameter optimization on the objective function of maximizing the satisfaction rate and the objective function of minimizing the number of slow-moving items to obtain candidate satisfaction rate solutions and their corresponding candidate slow-moving item numbers;

[0077] Step 303, sort the candidate satisfaction rate solutions and their corresponding candidate slow-moving item numbers to obtain the Pareto optimal solution set.

[0078] Optionally, the defective material prediction device determines the maximization satisfaction rate objective function and the minimization stagnant number objective function, and uses the target defective part ratio value and the defective part generation probability as optimization variables, and the first threshold of the target defective part ratio value and the second threshold of the defective part generation probability as constraint conditions to perform parameter optimization on the maximization satisfaction rate objective function and the minimization stagnant number objective function, and obtains the satisfaction rate candidate solution and its corresponding stagnant number candidate solution, where the first threshold and the second threshold are continuously adjusted according to the actual situation.

[0079] Further, the defective material prediction device sorts the satisfaction rate candidate solution and its corresponding stagnant number candidate solution to obtain the Pareto optimal solution set, specifically:

[0080] The defective material prediction device performs non-dominated sorting on the satisfaction rate candidate solution and its corresponding stagnant number candidate solution to obtain the non-dominated sorting result. The principle of non-dominated sorting is: for individuals x0 and x1, if x0 is smaller than x1 in both objective function values, the solution obtained by x0 is better than that of x1, that is, x0 dominates x1, and at this time x0 is retained. If one objective function value of x0 is smaller than x1, and x1 is smaller than x0 in the other objective function value, then x0 cannot dominate x1, and at this time x0 and x1 are retained.

[0081] Further, the defective material prediction device performs crowding distance sorting on the non-dominated sorting result to obtain the Pareto optimal solution set, specifically: in the non-dominated sorting result, calculate the average distance of each sample from its two nearest neighbors to obtain the crowding degree. Among them, the greater the crowding degree, the farther the current sample is from other samples, and the lower the similarity. Retaining it will make the diversity of the solution set better.

[0082] In the embodiment of the present application, iterative mutation update is performed through non-dominated sorting and crowding distance sorting, and the tuning of multiple objectives is considered at the same time, and the Pareto optimal solution set of the satisfaction rate and the stagnant number is obtained, so that the minimum stagnant number can be obtained under different satisfaction rates.

[0083] Fourthly, the preparation of defective materials should not only meet the condition that the number of prepared materials is greater than the defective materials (satisfaction rate), but also meet the condition that the fewer the number of prepared materials is better (stagnant number). The prior art obtains the optimal solution of the parameters through manual adjustment or grid search. Therefore, the prior art cannot consider the tuning of multiple objectives at the same time.

[0084] In the embodiment of the present application, the Pareto optimal solution set of the satisfaction rate and the stagnant number is obtained through parameter optimization of the target defective part ratio value and the defective part generation probability, and then the optimal number of defective parts is obtained through the Pareto optimal solution set. Therefore, the minimum stagnant number can be obtained under different satisfaction rates. By obtaining the optimal number of defective parts through the Pareto optimal solution set, the accurate prediction of the number of defective parts is realized, and the accuracy of defective material prediction is improved.

[0085] In an optional embodiment, the descriptions of steps 401 to 403 are as follows:

[0086] Step 401: Obtain a set of target data points in the Pareto optimal solution set where both the satisfaction rate and the number of stagnant items meet the preset satisfaction rate and the preset number of stagnant items.

[0087] Step 402: Based on the set of target data points and the number of data points in the set of target data points, determine the optimal sample data points in the Pareto optimal solution set.

[0088] Step 403: Determine the optimal number of defective parts according to the threshold parameter corresponding to the optimal sample data point.

[0089] It should be noted that in the embodiments of the present application, intervals for the satisfaction rate and the number of stagnant items are set according to the actual situation. In one embodiment, the satisfaction rate of the interval needs to be greater than or equal to 0.98, and the number of stagnant items is less than or equal to 10,000.

[0090] Optionally, the defective material prediction device obtains a set of target data points in the Pareto optimal solution set where both the satisfaction rate and the number of stagnant items meet the preset satisfaction rate and the preset number of stagnant items, that is, obtains a set of target data points in the Pareto optimal solution set where the satisfaction rate and the number of stagnant items are within the above intervals.

[0091] Further, the defective material prediction device determines the number of data points in the set of target data points. Among them, the number can be greater than the preset threshold, or can be less than or equal to the preset threshold. The preset threshold is set according to the actual situation. For example, the preset threshold is 1.

[0092] Further, the defective material prediction device determines the optimal sample data points in the Pareto optimal solution set according to the set of target data points and the number of data points in the set of target data points, as specifically described in Steps 4021 to 4027. Further, the defective material prediction device calculates the number of defective parts according to the optimal threshold parameter corresponding to the optimal data point, and determines the number of defective parts obtained after calculating the optimal threshold parameter as the optimal number of defective parts.

[0093] In the embodiments of the present application, the optimal number of defective parts is obtained through the Pareto optimal solution set. Therefore, the minimum number of stagnant items can be obtained under different satisfaction rates. By obtaining the optimal number of defective parts through the Pareto optimal solution set, the number of defective parts can be accurately predicted, and the accuracy of defective material prediction is improved.

[0094] In an optional embodiment, the descriptions of Steps 4021 to 4024 are as follows:

[0095] Step 4021: If the number is greater than the preset threshold, obtain the first data point and the second data point in the set of target data points.

[0096] Step 4022: Calculate the target satisfaction rate of each data point in the target data point set based on the satisfaction rates of the data points in the target data point set, the satisfaction rate of the first data point, and the satisfaction rate of the second data point.

[0097] Step 4023: Calculate the target stagnation number of each data point in the target data point set based on the stagnation numbers of the data points in the target data point set, the stagnation number of the first data point, and the stagnation number of the second data point.

[0098] Step 4024: Determine the optimal sample data point based on the target satisfaction rate and target stagnation number of each data point in the target data point set.

[0099] Optionally, if the number of data points in the target data point set is greater than a preset threshold, the defective material prediction device acquires the first data point x corresponding to the minimum satisfaction rate in the target data point set min and the second data point x corresponding to the maximum stagnation number max .

[0100] Furthermore, the defective material prediction device calculates the target satisfaction rate of each data point in the target data point set according to the satisfaction rates of the data points in the target data point set, the satisfaction rate of the first data point, and the satisfaction rate of the second data point. The specific calculation formula is:

[0101]

[0102] where is the target satisfaction rate of each data point, x 满足率 is the satisfaction rate of each data point, x min(满足率) is the satisfaction rate of the first data point x min and x max(满足率) is the satisfaction rate of the second data point x max .

[0103] Furthermore, the defective material prediction device calculates the target stagnation number of each data point in the target data point set according to the stagnation numbers of the data points in the target data point set, the stagnation number of the first data point, and the stagnation number of the second data point. The specific calculation formula is:

[0104]

[0105] where is the target stagnation number of each data point, x 呆滞数 is the stagnation number of each data point, x min(呆滞数) is the stagnation number of the first data point x min and x max(呆滞数) is the stagnation number of the second data point x max .

[0106] Further, the defective material prediction device calculates based on the target satisfaction rate and target stagnant number of each data point in the target data set, obtains a calculation result, and determines the data point corresponding to the smallest value in the calculation result as the optimal sample data point, specifically:

[0107]

[0108] In the embodiment of the present application, the optimal number of defective parts is obtained through the Pareto optimal solution set, so that the minimum stagnant number can be obtained under different satisfaction rates. By obtaining the optimal number of defective parts through the Pareto optimal solution set, the number of defective parts can be accurately predicted, and the accuracy of defective material prediction is improved.

[0109] In an alternative embodiment, the descriptions of steps 4025 to 4027 are as follows:

[0110] Step 4025, if the quantity is less than or equal to a preset threshold, central normalization is performed on the Pareto optimal solution set to obtain the central data point of the Pareto optimal solution set;

[0111] Step 4026, obtain the central satisfaction rate and central stagnant number of the central data point;

[0112] Step 4027, based on the satisfaction rate and stagnant number of each data point in the Pareto optimal solution set, and the central satisfaction rate and the central stagnant number, determine the optimal sample data point.

[0113] Optionally, if the number of data points in the target data set is less than or equal to the preset threshold, the defective material prediction device performs central normalization on the Pareto optimal solution set to obtain the central data point of the Pareto optimal solution set. Specifically, it can be understood that the preset threshold is 0, that is, if there are no data points in the interval, central normalization is performed on all data points in the Pareto optimal solution set to obtain the central data point of the Pareto optimal solution set. The central data point can be expressed as x ideal , where the specific formula for central normalization (Z-score normalization) is as follows:

[0114]

[0115] Among them, x is the data point, mean(x) represents the mean of x, and std(x) represents the standard deviation of x.

[0116] Further, the defective material prediction device obtains the central satisfaction rate x ideal(满足率) and the central stagnant number x ideal(呆滞数) .

[0117] Further, the defective material prediction device calculates the Euclidean distance between each data point in the Pareto optimal solution set and the central data point according to the satisfaction rate and the number of sluggish items of each data point, as well as the central satisfaction rate and the central number of sluggish items, and takes the data point with the closest Euclidean distance as the optimal sample data point. The specific formula is:

[0118] Optimal sample point data point = min[(x 满足率 - x ideal(满足率) ) 2 +(x 呆滞数 - x ideal(呆滞数) ) 2 )

[0119] In the embodiment of the present application, the optimal number of defective parts is obtained through the Pareto optimal solution set. Therefore, the minimum number of sluggish items can be obtained under different satisfaction rates. By obtaining the optimal number of defective parts through the Pareto optimal solution set, the number of defective parts can be accurately predicted, and the accuracy of defective material prediction is improved.

[0120] Further, referring to Figure 5 , Figure 5 is the second schematic flowchart of the defective material prediction method provided in the embodiment of the present application. As can be seen from the figure, the defective material prediction device collects the metadata required for defective part prediction, cleans and filters the abnormal samples in the collected metadata, and then constructs the features required for the model according to the defective material ratio prediction model to obtain the target material feature data. The target material feature data is input into the defective material ratio prediction model to obtain the target defective part ratio value, and the target material feature data is input into the binary classification model to obtain the probability of generating defective parts. The target defective part ratio value and the probability of generating defective parts are processed through a multi-objective optimization algorithm to obtain a Pareto optimal solution set, and then the optimal number of defective parts is obtained according to the Pareto optimal solution set.

[0121] Next, the defective material prediction device provided in the embodiment of the present application will be described. The defective material prediction device described below can be correspondingly referred to the defective material prediction method described above.

[0122] Referring to Figure 6 shown, Figure 6 is the structural schematic diagram of the defective material prediction device provided in the embodiment of the present application. The defective material prediction device may include:

[0123] A defective part ratio prediction module 601, configured to input the target material feature data into the defective material ratio prediction model to obtain the target defective part ratio value output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value;

[0124] A defective part probability prediction module 602 for inputting the target material feature data into a binary classification model to obtain the defective part generation probability output by the binary classification model; the binary classification model is trained based on the sample material feature data and its corresponding defective part generation label data;

[0125] A parameter optimization module 603 for performing parameter optimization based on the target defective part ratio value and the defective part generation probability to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding stagnant number;

[0126] A defective part number determination module 604 for obtaining the optimal defective part number based on the satisfaction rate and the stagnant number of each data point in the Pareto optimal solution set.

[0127] In the embodiment of the present application, optimization adjustment is performed through the defective part ratio value and the defective part generation probability to obtain a Pareto optimal solution set of the satisfaction rate and the stagnant number, so that the minimum stagnant number can be obtained under different satisfaction rates, and the optimal defective part number is obtained through the Pareto optimal solution set, realizing accurate prediction of the defective part number and improving the accuracy of defective material prediction.

[0128] In an optional example, the defective part ratio prediction module 601 is further configured to:

[0129] Input the target material feature data into the feature fusion layer for feature splicing to obtain material feature splicing data;

[0130] Input the material feature splicing data into the first fully connected layer for data extraction to obtain hidden layer data;

[0131] Input the hidden layer data and the material feature splicing data into the residual connection layer for data extraction to obtain residual layer data;

[0132] Input the residual layer data into the second fully connected layer for result prediction to obtain the target defective part ratio value output by the defective material ratio prediction model.

[0133] In an optional example, the parameter optimization module 603 is further configured to:

[0134] Determine a maximum satisfaction rate objective function and a minimum stagnant number objective function;

[0135] Taking the target defective part ratio value and the defective part generation probability as optimization variables, and taking the first threshold of the target defective part ratio value and the second threshold of the defective part generation probability as constraint conditions, perform parameter optimization on the maximum satisfaction rate objective function and the minimum stagnant number objective function to obtain a satisfaction rate candidate solution and its corresponding stagnant number candidate solution;

[0136] Sort the candidate solutions for the satisfaction rate and their corresponding candidate solutions for the stagnant number to obtain the Pareto optimal solution set.

[0137] In an alternative example, the parameter optimization module 603 is further configured to:

[0138] Perform non-dominated sorting on the candidate solutions for the satisfaction rate and their corresponding candidate solutions for the stagnant number to obtain a non-dominated sorting result;

[0139] Perform crowding distance sorting on the non-dominated sorting result to obtain the Pareto optimal solution set.

[0140] In an alternative example, the defective parts number determination module 604 is further configured to:

[0141] Obtain a set of target data points in the Pareto optimal solution set where both the satisfaction rate and the stagnant number meet the preset satisfaction rate and the preset stagnant number;

[0142] Based on the set of target data points and the number of data points in the set of target data points, determine the optimal sample data point in the Pareto optimal solution set;

[0143] Determine the optimal defective parts number according to the threshold parameter corresponding to the optimal sample data point.

[0144] In an alternative example, the defective parts number determination module 604 is further configured to:

[0145] If the quantity is greater than a preset threshold, obtain a first data point and a second data point in the set of target data points; the first data point is the data point corresponding to the minimum satisfaction rate, and the second data point is the data point corresponding to the maximum stagnant number;

[0146] Based on the satisfaction rate of each data point in the set of target data points, the satisfaction rate of the first data point, and the satisfaction rate of the second data point, calculate the target satisfaction rate of each data point in the set of target data points;

[0147] Based on the stagnant number of each data point in the set of target data points, the stagnant number of the first data point, and the stagnant number of the second data point, calculate the target stagnant number of each data point in the set of target data points;

[0148] Based on the target satisfaction rate and the target stagnant number of each data point in the set of target data points, determine the optimal sample data point.

[0149] In an alternative example, the defective parts number determination module 604 is further configured to:

[0150] If the quantity is less than or equal to the preset threshold, perform central normalization on the Pareto optimal solution set to obtain the central data point of the Pareto optimal solution set;

[0151] Obtain the center satisfaction rate and the center stagnant number of the center data point;

[0152] Based on the satisfaction rate and the stagnant number of each data point in the Pareto optimal solution set, as well as the center satisfaction rate and the center stagnant number, determine the optimal sample data point.

[0153] The specific embodiments of the defective material prediction device provided in this application are basically the same as those of the defective material prediction method embodiments, and will not be elaborated here.

[0154] Optionally, as Figure 7 shown, Figure 7 is a schematic structural diagram of an electronic device provided by an embodiment of this application. The electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communication interface 720, and the memory 730 complete mutual communication through the communication bus 740. The processor 710 may call a computer program in the memory 730 to execute the steps of the defective material prediction method, for example, including:

[0155] Input the target material feature data to be processed into the defective material ratio prediction model to obtain the target defective part ratio value to be optimized output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value;

[0156] Input the target material feature data into the binary classification model to obtain the probability of generating defective parts output by the binary classification model; the binary classification model is trained based on the sample material feature data and its corresponding label data of whether defective parts are generated;

[0157] Based on the target defective part ratio value and the probability of generating defective parts, perform parameter optimization to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding stagnant number;

[0158] Based on the satisfaction rate and the stagnant number of each data point in the Pareto optimal solution set, obtain the optimal number of defective parts.

[0159] In addition, when the logical instructions in the above-mentioned memory 730 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0160] On the other hand, an embodiment of this application also provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium includes a computer program. The computer program can be stored on the non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the steps of the defective material prediction method provided in the above-mentioned various embodiments, for example, including:

[0161] Input the target material feature data to be processed into the defective part ratio prediction model to obtain the target defective part ratio value to be optimized output by the defective part ratio prediction model; the defective part ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value;

[0162] Input the target material feature data into the binary classification model to obtain the probability of generating defective parts output by the binary classification model; the binary classification model is trained based on the sample material feature data and its corresponding label data indicating whether defective parts are generated;

[0163] Perform parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding sluggish number;

[0164] Obtain the optimal number of defective parts based on the satisfaction rate and sluggish number of each data point in the Pareto optimal solution set.

[0165] On another aspect, an embodiment of this application also provides a computer product. The computer product includes a computer program. The computer program can be stored on the computer product. When the computer program is executed by a processor, the computer can execute the steps of the defective material prediction method provided in the above-mentioned various embodiments, for example, including:

[0166] Input the target material feature data to be processed into the defective material ratio prediction model, and obtain the target defective part ratio value to be optimized output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value;

[0167] Input the target material feature data into the binary classification model, and obtain the probability of generating defective parts output by the binary classification model; the binary classification model is trained based on the sample material feature data and its corresponding label data indicating whether defective parts are generated;

[0168] Perform parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain the Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes the satisfaction rate and its corresponding stagnant number;

[0169] Obtain the optimal number of defective parts based on the satisfaction rate and stagnant number of each data point in the Pareto optimal solution set.

[0170] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0171] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A defective material prediction method, characterized in that, Including: Inputting the target material feature data to be processed into the defective material ratio prediction model to obtain the target defective part ratio value to be optimized output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on the sample material feature data and its corresponding defective part ratio value; Inputting the target material feature data into the binary classification model to obtain the probability of generating defective parts output by the binary classification model; The binary classification model is trained based on the sample material feature data and its corresponding label data of whether defective parts are generated; Performing parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding stagnant number; Obtaining the optimal number of defective parts based on the satisfaction rate and stagnant number of each data point in the Pareto optimal solution set.

2. The blank prediction method according to claim 1, characterized in that The obtaining the optimal number of defective parts based on the satisfaction rate and stagnant number of each data point in the Pareto optimal solution set includes: Obtaining a target data point set in which the satisfaction rate and stagnant number in the Pareto optimal solution set both meet the preset satisfaction rate and preset stagnant number; Determining the optimal sample data point in the Pareto optimal solution set based on the target data point set and the number of data points in the target data point set; Determining the optimal number of defective parts according to the threshold parameter corresponding to the optimal sample data point.

3. The blank prediction method according to claim 2, wherein The determining the optimal sample data point in the Pareto optimal solution set based on the target data point set and the number of data points in the target data point set includes: If the number is greater than the preset threshold, obtaining a first data point and a second data point in the target data point set; the first data point is the data point corresponding to the minimum satisfaction rate, and the second data point is the data point corresponding to the maximum stagnant number; Calculating the target satisfaction rate of each data point in the target data point set based on the satisfaction rate of each data point in the target data point set, the satisfaction rate of the first data point, and the satisfaction rate of the second data point; Calculating the target stagnant number of each data point in the target data point set based on the stagnant number of each data point in the target data point set, the stagnant number of the first data point, and the stagnant number of the second data point; Determining the optimal sample data point based on the target satisfaction rate and target stagnant number of each data point in the target data point set.

4. The blank prediction method according to claim 2, wherein The determining the optimal sample data point in the Pareto optimal solution set based on the target data point set and the number of data points in the target data point set includes: If the number is less than or equal to the preset threshold, performing central normalization on the Pareto optimal solution set to obtain the central data point of the Pareto optimal solution set; Obtaining the central satisfaction rate and central stagnant number of the central data point; Determining the optimal sample data point based on the satisfaction rate and stagnant number of each data point in the Pareto optimal solution set, and the central satisfaction rate and central stagnant number.

5. The blank prediction method according to claim 1, wherein The performing parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain a Pareto optimal solution set includes: Determining the objective function of maximizing the satisfaction rate and the objective function of minimizing the stagnant number; Taking the target defective part ratio value and the probability of generating defective parts as optimization variables, and taking the first threshold of the target defective part ratio value and the second threshold of the probability of generating defective parts as constraint conditions, perform parameter optimization on the maximized satisfaction rate objective function and the minimized stagnant number objective function to obtain a satisfaction rate candidate solution and its corresponding stagnant number candidate solution; Sort the satisfaction rate candidate solution and its corresponding stagnant number candidate solution to obtain the Pareto optimal solution set.

6. The blank prediction method according to claim 5, wherein The sorting of the satisfaction rate candidate solution and its corresponding stagnant number candidate solution to obtain the Pareto optimal solution set includes: Perform non-dominated sorting on the satisfaction rate candidate solution and its corresponding stagnant number candidate solution to obtain a non-dominated sorting result; Perform crowding distance sorting on the non-dominated sorting result to obtain the Pareto optimal solution set.

7. The blank prediction method according to any one of claims 1 to 6, characterized in that, The defective material ratio prediction model includes a feature fusion layer, a first fully connected layer, a residual connection layer, and a second fully connected layer; The inputting the target material feature data into the defective material ratio prediction model to obtain the target defective part ratio value output by the defective material ratio prediction model includes: Input the target material feature data into the feature fusion layer for feature splicing to obtain material feature splicing data; Input the material feature splicing data into the first fully connected layer for data extraction to obtain hidden layer data; Input the hidden layer data and the material feature splicing data into the residual connection layer for data extraction to obtain residual layer data; Input the residual layer data into the second fully connected layer for result prediction to obtain the target defective part ratio value output by the defective material ratio prediction model.

8. A defective material prediction device, characterized in that, Including: A defective part ratio prediction module, configured to input target material feature data into a defective material ratio prediction model to obtain the target defective part ratio value output by the defective material ratio prediction model; the defective material ratio prediction model is trained based on sample material feature data and its corresponding defective part ratio value; A defective part probability prediction module, configured to input the target material feature data into a binary classification model to obtain the probability of generating defective parts output by the binary classification model; The binary classification model is trained based on the sample material feature data and its corresponding label data indicating whether defective parts are generated; A parameter optimization module, configured to perform parameter optimization based on the target defective part ratio value and the probability of generating defective parts to obtain a Pareto optimal solution set; the Pareto optimal solution set includes at least two data points, and each data point includes a satisfaction rate and its corresponding stagnant number; A defective part number determination module, configured to obtain the optimal defective part number based on the satisfaction rate and the stagnant number of each data point in the Pareto optimal solution set.

9. An electronic device, characterized in that, Including a processor and a memory, the memory stores multiple instructions; the processor loads the instructions from the memory to execute the defective material prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores multiple instructions, and the instructions are suitable for being loaded by a processor to execute the defective material prediction method according to any one of claims 1 to 7.

Images