Training method of industrial prediction model, parameter prediction method and device of industrial equipment, equipment and medium

Abstract

The invention belongs to the technical field of industrial equipment, and particularly relates to a training method of an industrial prediction model, a parameter prediction method and device of industrial equipment, equipment and a medium, and the method comprises the steps: obtaining first historical production data of a plurality of pieces of industrial equipment and corresponding molds; based on the first historical production data, training an industrial prediction model to obtain a trained candidate industrial prediction model, the candidate industrial prediction model being a regression model based on an OmniPred framework; and obtaining second historical production data of to-be-predicted industrial equipment and a corresponding mold, and performing fine adjustment processing on the model parameters of the candidate industrial prediction model based on the second historical production data to obtain an industrial prediction model subjected to fine adjustment. According to the method, the dependence on a large amount of new data is remarkably reduced, the training efficiency is improved, and the production process can be optimized, the production efficiency is improved and the cost is reduced by accurately predicting the process and product parameters.

Description

Technical Field

[0001] The present application belongs to the field of industrial equipment technology, and specifically relates to a training method for an industrial prediction model, a parameter prediction method, device, equipment and medium for industrial equipment. Background Art

[0002] In industries such as injection molding, predictive models are often used to optimize production processes, improve product quality, reduce costs and increase efficiency. By analyzing relevant industrial data, predictive models can help predict quality defects, optimize process parameters, improve production stability and reduce energy consumption.

[0003] However, traditional prediction models usually face the problem of high data dependence during training. Every time industrial equipment or mold combination is replaced, a large amount of data needs to be collected to retrain the model. This method not only seriously affects production efficiency, but also increases the complexity of maintenance and updating, consuming a lot of time and resources. Summary of the invention

[0004] In response to the above problems, the present application provides a method for training an industrial prediction model, a method, device, equipment and medium for predicting parameters of industrial equipment.

[0005] In a first aspect, the present application provides a method for training an industrial prediction model, the method comprising:

[0006] Acquire first historical production data of a plurality of industrial equipment and corresponding molds;

[0007] Based on the first historical production data, the industrial prediction model is trained to obtain a candidate industrial prediction model that has been trained, wherein the candidate industrial prediction model is a regression model based on the OmniPred framework;

[0008] Acquire the second historical production data of the industrial equipment to be predicted and the corresponding mold, and based on the second historical production data, fine-tune the model parameters of the candidate industrial prediction model to obtain a fine-tuned industrial prediction model, which is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0009] In a possible implementation, the training of the industrial prediction model based on the first historical production data to obtain a trained candidate industrial prediction model includes:

[0010] Determine a first sample data set according to the first historical production data;

[0011] Dividing the first sample data set into a training set and a validation set;

[0012] Based on the training set and the validation set, the industrial prediction model is trained and optimized to obtain a trained candidate industrial prediction model.

[0013] In a possible implementation manner, determining a first sample data set according to the first historical production data includes:

[0014] Determine initial characteristic parameters and corresponding result parameters according to the first historical production data, wherein the initial characteristic parameters include at least one of equipment parameters, mold parameters and production process parameters of industrial equipment, and the result parameters represent process parameters and / or product parameters corresponding to the initial characteristic parameters;

[0015] For each characteristic parameter in the initial characteristic parameters, determining a correlation coefficient between the characteristic parameter and the result parameter, wherein the correlation between the characteristic parameter and the result parameter is proportional to the correlation coefficient;

[0016] The feature parameter whose correlation coefficient is greater than a preset threshold is used as the first target feature parameter, and a first sample data set is obtained according to the target feature parameter and the corresponding result parameter.

[0017] In a possible implementation, fine-tuning the model parameters of the candidate industrial forecasting model based on the second historical production data to obtain a fine-tuned industrial forecasting model includes:

[0018] Determine a second target characteristic parameter and a corresponding result parameter according to the second historical production data, and obtain a second sample data set according to the second target characteristic parameter and the corresponding result parameter;

[0019] Acquire a fine-tuning target preset by the user end, where the fine-tuning target includes at least one of accuracy, recall, and mean absolute percentage error;

[0020] The trained industrial model is loaded, and based on the second sample data set and the fine-tuning target, the trained industrial model is fine-tuned to obtain a fine-tuned industrial model.

[0021] In a possible implementation, the step of obtaining the second historical production data of the industrial equipment to be predicted and the corresponding mold includes:

[0022] Obtaining a debugging instruction from a user terminal, wherein the debugging instruction is used to instruct to fine-tune the candidate industrial prediction model for the industrial equipment to be predicted and the corresponding mold;

[0023] According to the debugging instruction, determining identification information of the industrial equipment to be predicted and the corresponding mold;

[0024] According to the identification information, second historical production data of the to-be-predicted industrial equipment and the corresponding mold are acquired.

[0025] In a second aspect, the present application provides a method for predicting parameters of industrial equipment, the method comprising:

[0026] Acquire production data of the industrial equipment to be predicted and the corresponding mold, and perform data processing on the production data to obtain characteristic parameters;

[0027] The characteristic parameters are input into an industrial prediction model to obtain a prediction result output by the industrial prediction model, wherein the industrial prediction model is obtained by the training method of the industrial prediction model described in any one of the first aspects, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction result is used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0028] In a third aspect, the present application provides a training device for an industrial prediction model, the device comprising:

[0029] A first acquisition module, used to acquire first historical production data of a plurality of industrial equipment and corresponding molds;

[0030] A training module, used for training an industrial prediction model based on the first historical production data to obtain a trained candidate industrial prediction model, wherein the candidate industrial prediction model is a regression model based on an OmniPred framework;

[0031] In a possible implementation, the training module is specifically used to:

[0032] Determine a first sample data set according to the first historical production data;

[0033] Dividing the first sample data set into a training set and a validation set;

[0034] Based on the training set and the validation set, the industrial prediction model is trained and optimized to obtain a trained candidate industrial prediction model.

[0035] In a possible implementation, the training module is specifically used to:

[0036] Determine initial characteristic parameters and corresponding result parameters according to the first historical production data, wherein the initial characteristic parameters include at least one of equipment parameters, mold parameters and production process parameters of industrial equipment, and the result parameters represent process parameters and / or product parameters corresponding to the initial characteristic parameters;

[0037] For each characteristic parameter in the initial characteristic parameters, determining a correlation coefficient between the characteristic parameter and the result parameter, wherein the correlation between the characteristic parameter and the result parameter is proportional to the correlation coefficient;

[0038] The feature parameter whose correlation coefficient is greater than a preset threshold is used as the first target feature parameter, and a first sample data set is obtained according to the target feature parameter and the corresponding result parameter.

[0039] The second acquisition module is used to acquire second historical production data of the industrial equipment to be predicted and the corresponding mold.

[0040] In a possible implementation, the second acquisition module is specifically used to:

[0041] Obtaining a debugging instruction from a user terminal, wherein the debugging instruction is used to instruct to fine-tune the candidate industrial prediction model for the industrial equipment to be predicted and the corresponding mold;

[0042] According to the debugging instruction, determining identification information of the industrial equipment to be predicted and the corresponding mold;

[0043] According to the identification information, second historical production data of the to-be-predicted industrial equipment and the corresponding mold are acquired.

[0044] A fine-tuning module is used to fine-tune the model parameters of the candidate industrial prediction model based on the second historical production data to obtain a fine-tuned industrial prediction model, wherein the industrial prediction model is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0045] In a possible implementation, the fine-tuning module is specifically used to:

[0046] Determine a second target characteristic parameter and a corresponding result parameter according to the second historical production data, and obtain a second sample data set according to the second target characteristic parameter and the corresponding result parameter;

[0047] Acquire a fine-tuning target preset by the user end, where the fine-tuning target includes at least one of accuracy, recall, and mean absolute percentage error;

[0048] The trained industrial model is loaded, and based on the second sample data set and the fine-tuning target, the trained industrial model is fine-tuned to obtain a fine-tuned industrial model.

[0049] In a fourth aspect, the present application provides a parameter prediction device for industrial equipment, the device comprising:

[0050] An acquisition module is used to acquire production data of the industrial equipment to be predicted and the corresponding mold, and perform data processing on the production data to obtain characteristic parameters;

[0051] A prediction module, used to input the characteristic parameters into an industrial prediction model to obtain a prediction result output by the industrial prediction model, wherein the industrial prediction model is obtained by the training method of the industrial prediction model described in any one of the first aspects, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction result is used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0052] In a fifth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in any one of the first aspect and / or the second aspect.

[0053] In a sixth aspect, the present application provides an electronic device, comprising: at least one processor and a memory; wherein:

[0054] The memory stores computer-executable instructions;

[0055] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method as described in any one of the first aspect and / or the second aspect.

[0056] In a seventh aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, can implement the steps of the method described in any one of the first aspect and / or the second aspect.

[0057] The training method of the industrial prediction model, the parameter prediction method, device, equipment and medium of industrial equipment provided in this application obtain the historical production data of multiple devices and molds, perform pre-training under the OmniPred framework, and fine-tune with the latest data of specific devices and molds. The model can quickly adapt to new production scenarios and avoid repeated data collection and training from scratch. This method not only significantly reduces the dependence on a large amount of new data and improves training efficiency, but also optimizes the production process, improves production efficiency and reduces costs by accurately predicting process and product parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0059] Figure 1Process of training the industrial prediction model provided in the embodiment of the present application Figure 1 ;

[0060] Figure 2 Process of training the industrial prediction model provided in the embodiment of the present application Figure 2 ;

[0061] Figure 3 A flow chart of a method for predicting parameters of industrial equipment provided in an embodiment of the present application;

[0062] Figure 4 A training device for an industrial prediction model provided by an embodiment of the present invention;

[0063] Figure 5 A diagram of a parameter prediction device for industrial equipment provided by an embodiment of the present invention;

[0064] Figure 6 A hardware schematic diagram of an electronic device provided by an embodiment of the present invention.

[0065] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0066] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0067] The terms "first", "second", "third", "fourth", etc. (if any) in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein, for example.

[0068] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0069] In industrial industries such as injection molding, traditional model prediction methods require data collection and training for each specific set of injection molding machines and mold combinations, which not only leads to a large demand for data volume, but also inefficiency, because data collection and model training need to be re-performed every time a mold or injection molding machine is replaced. Existing methods are usually traditional machine learning methods such as linear regression, tree models, and DenseNet. Based on the incremental learning of traditional learning methods, the self-updating method of remodeling training combines the initial sample set with the newly added samples for training, which is not only very time-consuming, but also cannot guarantee the generalization of the current samples. In addition, there are studies that propose a method for predicting the quality of injection molding products based on width learning. This method improves the model by adding the minimum p-norm to the ordinary width learning system (BLS) to obtain the p-norm width learning system (pN-BLS), solves the problems of small samples and unbalanced data, and improves the model's detection performance for outliers.

[0070] However, the above methods all have limitations: linear regression and support vector machine models are classic methods in the field of injection molding machine quality prediction. However, linear regression has problems of overfitting and multicollinearity, and support vector machines have high computational complexity in solving quadratic programming when the data scale is large. The most important problem is that training faces the problem of high dependence on data and poor generalization ability on unseen data. Based on the incremental learning of traditional learning methods, the self-updating method of remodeling training combines the initial sample set with the newly added samples for training. This not only wastes a lot of time and server resources, but also cannot guarantee generalization on the current samples.

[0071] Considering that the OmniPred framework can process (x, y) evaluation data from a variety of real-world experiments by training a language model as a universal end-to-end regressor. The data used by this framework comes from Google Vizier, one of the world's largest black-box optimization databases. The innovation of OmniPred is that it is a scalable and simple metric prediction framework based on constraint-independent text representations and is applicable to general input spaces. By performing multi-task learning on different input spaces and targets, OmniPred demonstrates the ability to surpass traditional regression models in many cases. The benefits of this transfer learning persist even after fine-tuning OmniPred locally on unseen tasks.

[0072] In response to the above problems, this application proposes a method for training an industrial prediction model, which uses the OmniPred framework for pre-training and uses historically accumulated data on other industrial equipment and mold combinations to obtain a model with certain generalization capabilities. When a new injection molding machine and mold combination needs to be predicted, only a small amount of current data needs to be used to fine-tune the pre-trained model to quickly adapt to the new situation without having to collect and train a large amount of data from scratch. This method can significantly reduce dependence on data, improve the efficiency of model training, and help improve production efficiency and reduce costs.

[0073] On the one hand, pre-training is performed using historically accumulated data on other industrial equipment and mold combinations to obtain a model with a certain generalization ability. Then, for new injection molding machines and mold combinations, only a small amount of current data needs to be used to fine-tune the pre-trained model to quickly adapt to new situations without the need to collect and train a large amount of data from scratch. On the other hand, traditional prediction methods such as industrial equipment quality face the problem of high dependence on data. Through pre-training and transfer learning, the present invention reduces the dependence on a large amount of specific data, improves the efficiency of model training and the accuracy of prediction. In addition, by using the OmniPred framework, the generalization ability of the model on new tasks is improved. This method avoids collecting and training a large amount of data from scratch, and solves the problems of overfitting, multicollinearity, and high computational complexity.

[0074] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments may exist independently or in combination with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0075] This embodiment provides a method for training an industrial prediction model. Figure 1 Process of training the industrial prediction model provided in the embodiment of the present application Figure 1 The method comprises:

[0076] S101. Acquire first historical production data of a plurality of industrial equipment and corresponding molds.

[0077] In this step, industrial equipment and molds refer to the combination of machines and molds used in industrial production processes such as injection molding, which determine the process parameters (such as temperature, pressure, time) and the final product quality. The first historical production data contains production data under multiple combinations of industrial equipment and molds, such as process parameters, product parameters (size, weight, defect rate, etc.), environmental conditions (temperature, humidity, etc.), and equipment and mold parameters. For example, in the scenario of injection molding machines and molds, the first historical production data may include injection molding machine tonnage, mold capacity, mold tonnage, nozzle temperature, barrel temperature, cooling time, plasticizing speed, injection speed, production cycle, energy consumption, etc., as well as whether the quality of injection molded products is qualified.

[0078] Through this step, a broad basic data set can be constructed, covering the production conditions of various equipment and mold combinations, providing rich sample support for pre-training industrial prediction models. By collecting diverse historical data, the model can learn to capture the common characteristics of different equipment and mold combinations and improve the generalization ability of the model. For example, real-time or historical data can be obtained from controllers (PLCs), sensor networks, and SCADA systems in industrial equipment. Combined with records exported from production management systems (such as MES), parameters related to equipment and molds are extracted to ensure the comprehensiveness of data collection, including samples during normal and abnormal production.

[0079] S102. Based on the first historical production data, train an industrial prediction model to obtain a trained candidate industrial prediction model, wherein the candidate industrial prediction model is a regression model based on an OmniPred framework.

[0080] In this step, the industrial prediction model is a regression model used to predict process parameters and / or product parameters. The OmniPred framework represents a pre-trained framework designed specifically for industrial prediction scenarios, which can use multi-source heterogeneous data to extract common features and form a basic model with generalization capabilities. The candidate industrial prediction model represents a model obtained through pre-training that has not yet been fine-tuned for a specific scenario.

[0081] Through pre-training, a basic model that can adapt to a variety of industrial equipment and mold combinations is built, reducing dependence on a single data set, improving the model's generalization ability for unseen data, and shortening the training time under new tasks. During the pre-training process, you can first perform feature engineering to extract meaningful features from the original data, and then define the objective function (such as mean square error or mean square logarithmic error) and optimization algorithm (such as Adam or SGD). Then perform batch training to adjust the model weights to adapt it to the data distribution of multiple devices and multiple molds. For model verification and storage, you can test the model performance on multiple independent verification data sets to evaluate its generalization ability. Save the trained candidate model in a portable format.

[0082] S103: Acquire second historical production data of the industrial equipment to be predicted and the corresponding mold.

[0083] In this step, the industrial equipment and molds to be predicted refer to the combination of equipment and molds for which process parameters or product parameters need to be predicted. Due to the generalization ability of the model for unseen data, the industrial equipment and molds to be predicted can be a group of industrial equipment and molds among the multiple industrial equipment and corresponding molds in step S101, or a group of industrial equipment and molds other than the multiple industrial equipment and corresponding molds in step S101, and there is no restriction here. The second historical production data represents a small-scale data set for this specific combination, containing a small amount of historical records, but the data types it contains are the same as the first historical data.

[0084] This step provides scenario-specific calibration data for candidate industrial prediction models to ensure the prediction accuracy of the model under the current combination, reduce dependence on large-scale data collection, and quickly adapt to new tasks.

[0085] Exemplarily, the step of obtaining the second historical production data of the industrial equipment to be predicted and the corresponding mold includes:

[0086] Obtaining a debugging instruction from a user terminal, wherein the debugging instruction is used to instruct to fine-tune the candidate industrial prediction model for the industrial equipment to be predicted and the corresponding mold;

[0087] According to the debugging instruction, determining identification information of the industrial equipment to be predicted and the corresponding mold;

[0088] According to the identification information, second historical production data of the to-be-predicted industrial equipment and the corresponding mold are acquired.

[0089] It should be noted that obtaining the debugging instructions from the user side here is the trigger condition for the entire fine-tuning process. Through the debugging instructions, the user explicitly indicates the need to fine-tune the model for a specific industrial equipment and mold combination (i.e., the combination to be predicted). These instructions usually contain specific parameters or operating requirements, such as equipment number, mold type, etc. According to the debugging instructions, the identification information (such as a unique serial number or product number) of the industrial equipment and mold to be predicted is extracted to ensure that the targeted equipment and mold combination is correctly identified. This method reduces the complexity of data search by clarifying the target equipment and mold, and provides accurate input for subsequent steps. Through the identification information, the second historical production data related to the equipment and mold combination can be quickly located from the production database or log.

[0090] S104. Based on the second historical production data, fine-tune the model parameters of the candidate industrial prediction model to obtain a fine-tuned industrial prediction model, where the industrial prediction model is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0091] In this step, fine-tuning refers to the process of adjusting only some parameters for specific tasks or data based on the existing pre-trained model. Model parameters include weights, biases, and other optimizer-related hyperparameters of the neural network. This step uses the second historical production data to quickly adapt the candidate model to ensure that the model can accurately predict the process parameters and product parameters of the current equipment and mold combination, avoiding training the model from scratch and saving time and computing resources.

[0092] For example, the early layers that capture common features in the pre-trained model can be frozen, and only the parameters of the later layers can be optimized to adapt to the characteristics of specific equipment and molds. This process incrementally trains the model by defining a loss function (such as mean square error) and a small learning rate (such as 0.001) for a specific task, ensuring accuracy improvement on new data while avoiding overfitting. After fine-tuning, the model can more accurately predict the process parameters or product performance of the current equipment and mold.

[0093] The training method of the industrial prediction model provided in this embodiment obtains historical production data of multiple devices and molds, performs pre-training under the OmniPred framework, and fine-tunes the model in combination with the latest data of specific devices and molds. The model can quickly adapt to new production scenarios and avoid repeated data collection and training from scratch. This method not only significantly reduces the dependence on a large amount of new data and improves training efficiency, but also optimizes the production process, improves production efficiency, and reduces costs by accurately predicting process and product parameters.

[0094] This embodiment provides a method for training an industrial prediction model. Figure 2 Process of training the industrial prediction model provided in the embodiment of the present application Figure 2 .like Figure 2 As shown, in this embodiment Figure 1 Based on the embodiment, the model training process is described in detail. The method includes:

[0095] S201. Determine a first sample data set according to the first historical production data.

[0096] In this step, a sample data set that can be used to train the industrial prediction model is determined by analyzing the first historical production data. Historical production data refers to relevant information collected during past production processes, including but not limited to equipment operating status, production parameters, output, quality indicators, etc. The sample data set represents a portion of data extracted from historical data, which is usually used to train and test models. Exemplarily, valuable samples can be cleaned and filtered out from historical production data through data preprocessing. These samples should include production data under different states of equipment and mold combinations, and should cover various production conditions to ensure the diversity and breadth of the data set. Feature selection methods (such as correlation analysis and principal component analysis) can also be used to determine the features most relevant to the prediction target, thereby constructing an effective sample data set.

[0097] Exemplarily, initial characteristic parameters and corresponding result parameters are determined according to the first historical production data, wherein the initial characteristic parameters include at least one of equipment parameters, mold parameters and production process parameters of industrial equipment, and the result parameters represent process parameters and / or product parameters corresponding to the initial characteristic parameters;

[0098] For each characteristic parameter in the initial characteristic parameters, determining a correlation coefficient between the characteristic parameter and the result parameter, wherein the correlation between the characteristic parameter and the result parameter is proportional to the correlation coefficient;

[0099] The feature parameter whose correlation coefficient is greater than a preset threshold is used as the first target feature parameter, and a first sample data set is obtained according to the target feature parameter and the corresponding result parameter.

[0100] It should be noted that the initial characteristic parameters are various measurable parameters in the first historical production data, such as equipment parameters (equipment speed, pressure, temperature, etc.), mold parameters (mold material, shape, wear degree, etc.) and production process parameters (such as injection cycle, cooling time, raw material ratio, etc.). Result parameters refer to process parameters and / or product parameters determined by the initial characteristic parameters, such as process parameters such as product qualification rate, production efficiency, etc., and product parameters such as product strength, surface finish, dimensional accuracy, or directly indicate whether the quality is qualified, etc.

[0101] Exemplarily, for each initial feature parameter, a statistical method (such as Pearson correlation coefficient, Spearman correlation coefficient) can be used to calculate the correlation coefficient between it and the result parameter. The range of the correlation coefficient is usually between -1 and 1, where 1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no correlation. The higher the coefficient value, the stronger the relationship between the feature parameter and the result parameter. By calculating the correlation coefficient of all feature parameters, the features that have a significant impact on the result parameters can be identified. Feature parameters with a correlation coefficient greater than a threshold are regarded as important target feature parameters. This ensures that there is a significant positive correlation between the selected feature parameters and the result parameters. Finally, the first sample data set is constructed by the selected target feature parameters and their corresponding result parameters. This data set will be used for subsequent model training to provide the model with the necessary input features and output labels so that the model can learn the mapping relationship between features and results.

[0102] For example, before constructing a sample data set, the initial data can be cleaned up in advance to remove missing values, outliers, and illogical data to ensure data quality. The sample data can also be standardized or normalized to make the dimensions of different feature parameters consistent, which is more convenient for subsequent model training and analysis.

[0103] S202: Divide the first sample data set into a training set and a validation set.

[0104] In this step, the training set is used to train the dataset of the machine learning model. The model will learn parameters on this dataset, and the validation set is used to tune the model and select the best hyperparameter dataset. The validation set is used to check the performance of the model and decide whether the model structure or training method needs to be adjusted. For example, the dataset can be divided by a preset ratio; or the sample dataset can be divided into different subsets by random methods to ensure that the data in each subset is representative. Stratified sampling methods can also be used to ensure that the category distribution in each subset is consistent with the overall dataset, thereby avoiding overfitting or underfitting of certain categories during training.

[0105] S203: Based on the training set and the validation set, the industrial prediction model is trained and optimized to obtain a trained candidate industrial prediction model.

[0106] In this step, you can use the training set to train the model and adjust the model parameters so that the model can fit the rules in the data. For example, the regression model based on the OmniPred framework is trained through standard cross entropy loss to implicitly learn the numerical distance from the training data. At the same time, during the training process, the model will be evaluated on the validation set after each round of training, and the model will be adjusted according to the performance indicators of the validation set (such as accuracy, recall, etc.). Common optimization methods include adjusting the learning rate, modifying the network structure, adding regularization terms, etc.

[0107] S204. Determine a second target characteristic parameter and a corresponding result parameter according to the second historical production data, and obtain a second sample data set according to the second target characteristic parameter and the corresponding result parameter.

[0108] In this step, new features and result parameters related to the target prediction are extracted from the second historical production data to form a new sample data set to provide data support for the fine-tuning model. The types of these features are similar or identical to the feature parameters extracted from the first historical production data. The types of the result parameters are the same as the result parameters determined in the first historical production data. That is, the data type of the second sample data set is similar or identical to the first sample data set, but the data values ​​therein are different due to differentiation factors such as the corresponding industrial equipment and molds.

[0109] S205: Obtain a fine-tuning target preset by the user end, where the fine-tuning target includes at least one of accuracy, recall, and mean absolute percentage error.

[0110] In this step, we make sure that the fine-tuned model can achieve the expected results by clarifying the performance indicators that need to be optimized during the fine-tuning process. Accuracy is a measure of the proportion of correct predictions made by the model; recall is a measure of the model's ability to correctly identify positive examples (such as qualified products, successful production, etc.); mean absolute percentage error is a measure of the average percentage of error between the predicted value and the actual value. Users can choose different optimization goals based on actual needs. For example, if the accuracy of the prediction results is critical, you may choose accuracy as the optimization goal; if you want to reduce the error, you may choose mean absolute percentage error. According to different performance indicators, set specific optimization goals (such as increasing the accuracy to 95%, reducing the mean absolute percentage error to less than 5%, etc.).

[0111] S206: Load the trained industrial model, and fine-tune the trained industrial model based on the second sample data set and the fine-tuning target to obtain a fine-tuned industrial model.

[0112] In this step, the previously trained candidate model is loaded into the training framework, that is, the weights and configuration parameters of the candidate industrial prediction model are extracted from the storage (such as disk or cloud) to the current training environment. The second sample data set is used, and the model is trained a small amount according to the fine-tuning goal set by the user (such as improving accuracy or reducing error). The fine-tuning in this step is the same as that in S103, and will not be repeated here.

[0113] The training method of the industrial prediction model provided in this embodiment can obtain an initial model with good generalization ability based on historical data and the combination of different equipment by using the OmniPred framework for pre-training. When a new equipment or mold combination needs to be predicted, only a small amount of current data needs to be used for fine-tuning, thus avoiding the complex training process from scratch. Not only does it significantly shorten the training time of the model, it also improves the accuracy and adaptability of the prediction, helps to improve production efficiency, reduce costs, and enhance the application ability of the model in different scenarios.

[0114] This embodiment provides a method for predicting parameters of industrial equipment. Figure 3 Flow chart of the parameter prediction method of industrial equipment provided in the embodiment of the present application. Figure 3 As shown, the method includes:

[0115] S301, obtaining production data of the industrial equipment to be predicted and the corresponding mold, and performing data processing on the production data to obtain characteristic parameters.

[0116] In this step, the acquired production data usually needs to be preprocessed. This process may include denoising, standardization, normalization, feature extraction, etc., so as to convert the data into feature parameters suitable for input into the industrial prediction model. Feature parameters are usually quantitative representations of production processes, equipment performance, mold characteristics, etc. The types of production data and feature parameters are similar to those in the above-mentioned training process embodiment and will not be described in detail.

[0117] S302: Input the characteristic parameters into an industrial prediction model to obtain a prediction result output by the industrial prediction model.

[0118] In which, the industrial prediction model is obtained through the training method of the industrial prediction model described in any one of the above embodiments, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction result is used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0119] In this step, the industrial prediction model calculates and outputs the prediction results based on the input characteristic parameters. The prediction results usually include the process parameters (such as temperature, pressure, time, etc.) and / or product parameters (such as product quality, size, surface treatment, etc.) of the industrial equipment and mold to be predicted.

[0120] For example, in the injection molding process, predicting product quality is a common industrial application scenario. Through the industrial prediction model, a prediction model can be trained based on historical data to infer the performance of product quality under given process conditions. Collect real-time production data of the current injection molding machine and mold combination, including the pressure, temperature, injection speed, cooling time, mold temperature, etc. of the injection molding machine. Obtain the corresponding production result data, such as the actual size of the plastic part, surface finish, product strength, etc. The above production data is preprocessed to obtain feature parameters, and the processed feature parameters are input into the trained industrial prediction model. The model predicts the product quality under the current process conditions based on historical data and the rules learned during the training process. For example, predict the dimensional error range of the product, whether the surface has defects, and the tensile strength of the part under this set of process parameters. If the prediction results show that the product quality does not meet the standards (for example, the dimensional error is too large), the process parameters can be adjusted immediately, such as adjusting the temperature or pressure of the injection molding machine, to optimize the production process and reduce the scrap rate. If the prediction results show that the product quality meets the standards, production can continue without stopping for quality inspection, thereby improving production efficiency.

[0121] The parameter prediction method of industrial equipment provided in this embodiment can quickly obtain accurate predictions of equipment and mold combinations by inputting real-time production data, thereby greatly improving the intelligent level of the production process, reducing manual intervention, and optimizing the production process and product quality in real time, thereby improving production efficiency and reducing costs.

[0122] This embodiment also provides a training device for an industrial prediction model. Figure 4 A training device for an industrial prediction model provided by an embodiment of the present invention, such as Figure 4 As shown, the 40 includes:

[0123] A first acquisition module 401 is used to acquire first historical production data of a plurality of industrial equipment and corresponding molds;

[0124] A training module 402 is used to train an industrial prediction model based on the first historical production data to obtain a trained candidate industrial prediction model, wherein the candidate industrial prediction model is a regression model based on an OmniPred framework;

[0125] In a possible implementation, the training module 402 is specifically used for:

[0126] Determine a first sample data set according to the first historical production data;

[0127] Dividing the first sample data set into a training set and a validation set;

[0128] Based on the training set and the validation set, the industrial prediction model is trained and optimized to obtain a trained candidate industrial prediction model.

[0129] In a possible implementation, the training module 402 is specifically used for:

[0130] Determine initial characteristic parameters and corresponding result parameters according to the first historical production data, wherein the initial characteristic parameters include at least one of equipment parameters, mold parameters and production process parameters of industrial equipment, and the result parameters represent process parameters and / or product parameters corresponding to the initial characteristic parameters;

[0131] For each characteristic parameter in the initial characteristic parameters, determining a correlation coefficient between the characteristic parameter and the result parameter, wherein the correlation between the characteristic parameter and the result parameter is proportional to the correlation coefficient;

[0132] The feature parameter whose correlation coefficient is greater than a preset threshold is used as the first target feature parameter, and a first sample data set is obtained according to the target feature parameter and the corresponding result parameter.

[0133] The second acquisition module 403 is used to acquire second historical production data of the industrial equipment to be predicted and the corresponding mold.

[0134] In a possible implementation, the second acquisition module 403 is specifically configured to:

[0135] Obtaining a debugging instruction from a user terminal, wherein the debugging instruction is used to instruct to fine-tune the candidate industrial prediction model for the industrial equipment to be predicted and the corresponding mold;

[0136] According to the debugging instruction, determining identification information of the industrial equipment to be predicted and the corresponding mold;

[0137] According to the identification information, second historical production data of the to-be-predicted industrial equipment and the corresponding mold are acquired.

[0138] The fine-tuning module 404 is used to fine-tune the model parameters of the candidate industrial prediction model based on the second historical production data to obtain a fine-tuned industrial prediction model, and the industrial prediction model is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0139] In a possible implementation, the fine-tuning module 404 is specifically configured to:

[0140] Determine a second target characteristic parameter and a corresponding result parameter according to the second historical production data, and obtain a second sample data set according to the second target characteristic parameter and the corresponding result parameter;

[0141] Acquire a fine-tuning target preset by the user end, where the fine-tuning target includes at least one of accuracy, recall, and mean absolute percentage error;

[0142] The trained industrial model is loaded, and based on the second sample data set and the fine-tuning target, the trained industrial model is fine-tuned to obtain a fine-tuned industrial model.

[0143] The present embodiment provides a training device for an industrial prediction model, which can execute the training method for an industrial prediction model provided in the above method embodiment. The implementation principle and technical effect thereof are similar, and are not described in detail in the present embodiment.

[0144] This embodiment also provides a parameter prediction device for industrial equipment. Figure 5 A parameter prediction device diagram of an industrial equipment provided by an embodiment of the present invention, such as Figure 5 As shown, the 50 includes:

[0145] The acquisition module 501 is used to acquire the production data of the industrial equipment to be predicted and the corresponding mold, and perform data processing on the production data to obtain characteristic parameters;

[0146] The prediction module 502 is used to input the characteristic parameters into the industrial prediction model to obtain the prediction results output by the industrial prediction model, wherein the industrial prediction model is obtained by the training method of the industrial prediction model described in any one of the above embodiments, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction results are used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

[0147] The present embodiment provides a parameter prediction device for industrial equipment, which can execute the parameter prediction method for industrial equipment provided by the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail in this embodiment.

[0148] Figure 6 Schematic diagram of the hardware of the electronic device provided by the embodiment of the present invention. Figure 6 As shown, the electronic device 60 provided in this embodiment includes: at least one processor 601 and a memory 602. The device 60 also includes a communication component 603. The processor 601, the memory 602 and the communication component 603 are connected via a bus 604.

[0149] In a specific implementation process, at least one processor 601 executes the computer-executable instructions stored in the memory 602, so that at least one processor 601 executes the above method.

[0150] The specific implementation process of the processor 601 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0151] In the above Figure 6 In the illustrated embodiment, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0152] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0153] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0154] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method described above is implemented.

[0155] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0156] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0157] The division of the units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, which may be electrical, mechanical or other forms.

[0158] 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 on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0159] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0160] If the functions 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 this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0161] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0162] So far, the technical solution of the present application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments, and the above embodiments are only used to illustrate the technical solution of the present application rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for training an industrial prediction model, characterized in that: The method comprises: Acquire first historical production data of a plurality of industrial equipment and corresponding molds; Based on the first historical production data, the industrial prediction model is trained to obtain a candidate industrial prediction model that has been trained, wherein the candidate industrial prediction model is a regression model based on the OmniPred framework; Acquire the second historical production data of the industrial equipment to be predicted and the corresponding mold, and based on the second historical production data, fine-tune the model parameters of the candidate industrial prediction model to obtain a fine-tuned industrial prediction model, which is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

2. The method according to claim 1, characterized in that: The step of training the industrial prediction model based on the first historical production data to obtain a trained candidate industrial prediction model includes: Determine a first sample data set according to the first historical production data; Dividing the first sample data set into a training set and a validation set; Based on the training set and the validation set, the industrial prediction model is trained and optimized to obtain a trained candidate industrial prediction model.

3. The method according to claim 2, characterized in that The determining a first sample data set according to the first historical production data includes: Determine initial characteristic parameters and corresponding result parameters according to the first historical production data, wherein the initial characteristic parameters include at least one of equipment parameters, mold parameters and production process parameters of industrial equipment, and the result parameters represent process parameters and / or product parameters corresponding to the initial characteristic parameters; For each characteristic parameter in the initial characteristic parameters, determining a correlation coefficient between the characteristic parameter and the result parameter, wherein the correlation between the characteristic parameter and the result parameter is proportional to the correlation coefficient; The feature parameter whose correlation coefficient is greater than a preset threshold is used as the first target feature parameter, and a first sample data set is obtained according to the target feature parameter and the corresponding result parameter.

4. The method according to claim 1, characterized in that: The fine-tuning of the model parameters of the candidate industrial forecasting model based on the second historical production data to obtain a fine-tuned industrial forecasting model includes: Determine a second target characteristic parameter and a corresponding result parameter according to the second historical production data, and obtain a second sample data set according to the second target characteristic parameter and the corresponding result parameter; Acquire a fine-tuning target preset by the user end, where the fine-tuning target includes at least one of accuracy, recall, and mean absolute percentage error; The trained industrial model is loaded, and based on the second sample data set and the fine-tuning target, the trained industrial model is fine-tuned to obtain a fine-tuned industrial model.

5. The method according to claim 1, characterized in that The obtaining of the second historical production data of the industrial equipment to be predicted and the corresponding mold includes: Obtaining a debugging instruction from a user terminal, wherein the debugging instruction is used to instruct to fine-tune the candidate industrial prediction model for the industrial equipment to be predicted and the corresponding mold; According to the debugging instruction, determining identification information of the industrial equipment to be predicted and the corresponding mold; According to the identification information, second historical production data of the to-be-predicted industrial equipment and the corresponding mold are acquired.

6. A method for predicting parameters of industrial equipment, characterized in that: The method comprises: Acquire production data of the industrial equipment to be predicted and the corresponding mold, and perform data processing on the production data to obtain characteristic parameters; The characteristic parameters are input into an industrial prediction model to obtain a prediction result output by the industrial prediction model, wherein the industrial prediction model is obtained by the training method of the industrial prediction model described in any one of claims 1 to 5, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction result is used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

7. A training device for an industrial prediction model, characterized in that: The device comprises: A first acquisition module, used to acquire first historical production data of a plurality of industrial equipment and corresponding molds; A training module, used for training an industrial prediction model based on the first historical production data to obtain a trained candidate industrial prediction model, wherein the candidate industrial prediction model is a regression model based on an OmniPred framework; A second acquisition module is used to acquire second historical production data of the industrial equipment to be predicted and the corresponding mold; A fine-tuning module is used to fine-tune the model parameters of the candidate industrial prediction model based on the second historical production data to obtain a fine-tuned industrial prediction model, wherein the industrial prediction model is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

8. A parameter prediction device for industrial equipment, characterized in that: The device comprises: An acquisition module is used to acquire production data of the industrial equipment to be predicted and the corresponding mold, and perform data processing on the production data to obtain characteristic parameters; A prediction module, used to input the characteristic parameters into an industrial prediction model to obtain a prediction result output by the industrial prediction model, wherein the industrial prediction model is obtained by the training method of the industrial prediction model described in any one of claims 1 to 5, and is used to predict the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold, and the prediction result is used to indicate the process parameters and / or product parameters of the industrial equipment to be predicted and the corresponding mold.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method according to any one of claims 1 to 5 and / or claim 6.

10. An electronic device, characterized in that: include: at least one processor and memory; wherein, The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the method according to any one of claims 1 to 5 and / or claim 6.

Images