Mold wear degree prediction method and system, terminal equipment and medium

Abstract

The invention relates to the technical field of wear degree detection, and discloses a mold wear degree prediction method and system, terminal equipment and a medium, and the mold wear degree prediction method comprises the following steps: obtaining historical use data of a mold; processing the historical use data so as to convert the historical use data into target input data related to mold wear degree prediction; constructing a wear prediction model based on the target input data and the corresponding extrusion amount label; and real-time target input data of the mold are obtained, the real-time target input data are input into the wear prediction model, a prediction result output by the wear prediction model is obtained, the prediction result is the real-time extrusion amount of the mold, and the smaller the real-time extrusion amount is, the larger the wear degree of the mold is. According to the method, the wear prediction model is constructed, the wear evolution mode implied in the historical use data is automatically learned, the artificial experience is converted into the quantitative extrusion amount prediction value, and objective, real-time and dynamic evaluation of the wear degree of the mold is achieved.

Description

Technical Field

[0001] This application relates to the field of wear detection technology, and in particular to a method, system, terminal equipment and medium for predicting the wear of molds. Background Technology

[0002] In extrusion processes for various profiles, the die is a critical piece of equipment, and its wear directly affects the dimensional accuracy, surface quality, and production stability of the profiles. Traditionally, judging die wear relies heavily on manual experience, such as observing the degree of wear, which is highly subjective and subject to delays. Currently, industry monitoring of die wear primarily depends on operator experience, such as manually observing the degree of wear and making a rough assessment based on historical replacement cycles. This method is not only highly subjective but also typically only identifies wear after significant damage has already occurred, exhibiting a significant time lag. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, system, terminal device, and medium for predicting mold wear, which can effectively address the subjectivity and lag in mold wear detection.

[0004] In a first aspect, embodiments of this application provide a method for predicting the wear degree of a mold, including: Obtain historical usage data for the mold; The historical usage data is processed to convert it into target input data related to mold wear prediction. A wear prediction model is constructed based on the target input data and the corresponding extrusion amount label; The real-time target input data of the mold is obtained and input into the wear prediction model to obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear of the mold.

[0005] In a first possible embodiment of the first aspect, processing the historical usage data to convert it into target input data related to mold wear prediction includes: The historical usage data is cleaned to obtain cleaned historical usage data. The cleaning process includes deduplication, missing value processing, and noise data processing. Based on the type of the historical usage data after the cleaning process, the historical usage data after the cleaning process is standardized and feature-encoded to obtain the processed historical usage data. The historical usage data is filtered based on the correlation between each processed historical usage data and the predicted wear level of the mold to obtain the target input data. In a second possible embodiment of the first aspect, the step of constructing a wear prediction model based on the target input data and the corresponding extrusion amount label includes: The target input data is divided into multiple groups of time-series input features in chronological order, and each group of time-series input features corresponds to a squeezing amount label. The model architecture for constructing the initial wear prediction model is to divide multiple sets of time-series input features and corresponding extrusion amount labels into training and test sets according to the time sequence. The initial wear prediction model is trained, optimized, and validated based on the training set and the test set to obtain the wear prediction model.

[0006] In a third possible embodiment of the first aspect, the type of the historical usage data after cleaning includes numerical data and non-numerical data, and the standardization and feature encoding of the historical usage data after cleaning according to its type includes: The numerical data is standardized to map it to a target range. The non-numerical data is subjected to one-hot encoding to convert it into a binary vector.

[0007] In a fourth possible embodiment of the first aspect, the filtering of the historical usage data based on the correlation between each of the processed historical usage data and the predicted degree of mold wear includes: The correlation between each processed historical usage data and the predicted wear level of the mold is determined based on information theory and statistical methods. Based on the correlation, historical usage data that is irrelevant to the prediction of mold wear degree is removed, and historical usage data that is relevant to the prediction of mold wear degree is retained to obtain the target input data; The target input data includes the surface quality of the extruded material, die wall thickness, extruded material material, die extrusion breakthrough pressure, die extrusion speed, extruded material temperature, die exit temperature, number of nitriding cycles, and die material. The extrusion quantity label includes theoretical extrusion quantity and actual extrusion quantity.

[0008] In a fifth possible embodiment of the first aspect, the step of training, optimizing, and validating the initial wear prediction model based on the training set and the test set to obtain the wear prediction model includes: The initial wear prediction model is iteratively trained based on the training set and preset hyperparameters to obtain the trained initial wear prediction model. Based on the prediction results of the initial wear prediction model after training on the test set, the hyperparameters of the initial wear prediction model after training are adjusted. The prediction deviation of the trained initial wear prediction model for the training set and the test set is determined, and the trained initial wear prediction model is verified based on the prediction deviation to obtain the verified wear prediction model. In a sixth possible embodiment of the first aspect, it further includes: The real-time target input data is collected from the mold data service platform through the model prediction service interface; The collected real-time target input data is input into the wear prediction model through the model prediction service interface.

[0009] Secondly, embodiments of this application provide a mold wear prediction system, comprising: The data processing module is used to acquire historical usage data of the mold, process the historical usage data, and convert the historical usage data into target input data related to the prediction of mold wear degree. The model building module is used to build a wear prediction model based on the target input data and the corresponding extrusion amount label; The wear degree prediction module is used to acquire the real-time target input data of the mold, input the real-time target input data into the wear prediction model, and obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear degree of the mold.

[0010] Secondly, embodiments of this application provide a terminal device, including a memory and a processor. The memory stores a computer program, which executes the aforementioned mold wear prediction method when the processor is running.

[0011] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program that executes the above-described mold wear prediction method when run on a processor.

[0012] The embodiments of this application have the following beneficial effects: This embodiment of a method for predicting mold wear includes: acquiring historical usage data of the mold; processing the historical usage data to convert it into target input data related to mold wear prediction; constructing a wear prediction model based on the target input data and corresponding extrusion amount labels; acquiring real-time target input data of the mold, inputting the real-time target input data into the wear prediction model, and obtaining the prediction result output by the wear prediction model, wherein the prediction result is the real-time extrusion amount of the mold, and the smaller the real-time extrusion amount, the greater the wear of the mold. This application, by constructing a wear prediction model, automatically learns the wear evolution pattern implicit in historical usage data, transforming the originally qualitative manual experience into a quantitative extrusion amount prediction value, thereby achieving an objective, real-time, and dynamic assessment of mold wear and improving the scientific nature of decision-making. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This paper illustrates a first flowchart of the mold wear prediction method according to an embodiment of this application. Figure 2 This paper illustrates a second flowchart of the mold wear prediction method according to an embodiment of this application. Figure 3 A schematic diagram of a third process for predicting mold wear according to an embodiment of this application is shown; Figure 4 A schematic diagram of the mold wear prediction system according to an embodiment of this application is shown.

[0015] Explanation of key component symbols: 200 - Mold wear prediction system; 210 - Data processing module; 220 - Model building module; 230 - Wear prediction module. Detailed Implementation

[0016] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0017] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0018] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0019] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0020] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0021] Figure 1 A flowchart illustrating a mold wear prediction method according to an embodiment of this application is shown. Exemplarily, the mold wear prediction method includes the following steps: S110, obtain historical usage data of the mold.

[0022] In this embodiment, the historical usage data of the mold refers to the multi-dimensional process parameters with time-series characteristics accumulated during the service life of a specific mold or similar molds in the material extrusion production process. This data reflects the usage status and performance evolution of the mold under different working conditions and is a key input for training the mold wear prediction model.

[0023] S120 processes historical usage data to convert it into target input data relevant to mold wear prediction.

[0024] As an example, in the mold wear predictive maintenance scheme, data preprocessing is the first step in model building. It aims to eliminate noise, missing data, inconsistencies and other problems in the original historical usage data, improve data quality and feature representation capabilities, and thus ensure the stability and prediction accuracy of subsequent deep learning model training.

[0025] In one embodiment, such as Figure 2 As shown, the processing of historical usage data specifically includes the following steps: S121, clean the historical usage data to obtain cleaned historical usage data. The cleaning process includes deduplication, missing value processing and noise data processing.

[0026] In this embodiment, duplicate records may exist in historical usage data. Deduplication is used to remove duplicate data to avoid interference. Missing values ​​may exist in historical usage data. Missing value handling is used to handle missing values ​​in historical usage data through methods such as interpolation, deletion of missing values, or use of default values. Outliers or noise may exist in historical usage data, and appropriate methods need to be used to process this data. Noise data processing is used to handle outliers or noise through methods such as outlier removal and smoothing.

[0027] S122, Based on the type of historical usage data after cleaning, the historical usage data after cleaning is standardized and feature-encoded to obtain the processed historical usage data.

[0028] For example, the types of historical usage data after cleaning include numerical data and non-numerical data. Numerical data are variables with defined dimensions and continuous or discrete numerical ranges, mainly used to describe quantifiable physical quantities. Non-numerical data are categorical or textual variables that cannot be directly used in mathematical operations, used to identify qualitative information such as material and state, for example, color and texture.

[0029] In one implementation, numerical data is standardized to map it to a target interval. To improve the accuracy and efficiency of the algorithm, data standardization is necessary. Data standardization mainly includes the following methods: scaling the data to the [0, 1] interval using min-max standardization; converting the data to a distribution with a mean of 0 and a standard deviation of 1 using Z-score standardization, thus scaling the data to the [0, 1] interval; and performing a logarithmic transformation on the data to scale it to the [0, 1] interval, thereby eliminating skewness in the data distribution.

[0030] In another implementation, non-numerical data is one-hot encoded to convert it into binary vectors. This addresses the challenge of capturing hidden features in deep learning by addressing the lack of numerical continuity among non-numerical data. Therefore, one-hot encoding of non-numerical data is necessary to convert categorical variables into binary vectors and avoid order-related errors. One-hot encoding is a coding method for handling discrete categorical features. Its core principle is to convert each category into a binary vector. The encoding rule is that for data with N categories, a vector of length N is generated, where only the position corresponding to the category is 1, and the rest are 0. For example, in color features, red is encoded as [1, 0, 0], and green as [0, 1, 0].

[0031] S123, based on the correlation between each processed historical usage data and the predicted wear level of the mold, the historical usage data is filtered to obtain the target input data.

[0032] For example, in order to avoid the introduction of irrelevant input features that could affect the overall performance of the model and lead to a waste of resources, it is necessary to determine the correlation between input features and output features, that is, to determine the correlation between historical usage data and the prediction of mold wear.

[0033] In one embodiment, this application can determine the correlation between each processed historical usage data and the mold wear prediction based on information theory and statistical methods. Based on the correlation, historical usage data irrelevant to the mold wear prediction is discarded, while historical usage data relevant to the mold wear prediction is retained to obtain the target input data.

[0034] In this embodiment, the information theory approach evaluates the information contribution of features by measuring information uncertainty, focusing on the degree of information sharing between features and the target. Information theory approaches include, but are not limited to, Mutual Information Maximization (MIM) and information gain. Mutual Information Maximization optimizes model parameters to maximize the amount of shared information between two random variables, thereby enhancing the ability to represent the correlation between variables. A mutual information score of 0 indicates that the two variables are completely independent, while a score greater than 0 indicates a correlation; the larger the value, the greater the dependency. Information gain is a core indicator used for feature selection in decision tree algorithms, measuring the degree to which the uncertainty of the target variable is reduced given a known feature. The mutual information score measures the amount of shared information between two random variables, reflecting the strength of their dependency; a higher mutual information score indicates a stronger correlation between the input and output features. The correlation between historical usage data and mold wear prediction is represented by the mutual information score. Historical usage data with a score greater than zero is retained, while historical usage data with a mutual information score of zero is discarded.

[0035] The statistical methods employed include hypothesis testing or correlation indices to quantify the strength of the association between features and the target variable, and to screen for statistically significant features. These methods include, but are not limited to, Pearson correlation analysis and analysis of variance. Pearson correlation analysis measures the degree of linear correlation between two continuous variables. It requires that the data be approximately normally distributed and that the relationship exhibits a linear trend. The closer the Pearson correlation coefficient is to zero, the weaker the correlation; the closer it is to 1, the stronger the positive correlation; and the closer it is to -1, the stronger the positive correlation. The correlation between historical usage data and mold wear prediction is represented by the Pearson correlation coefficient. Historical usage data with a Pearson correlation coefficient greater than zero are retained, while historical usage data with a Pearson correlation coefficient less than or equal to zero are discarded.

[0036] Analysis of variance (ANOVA) is used to determine whether there are significant differences in the means of a continuous feature across different class groups. In machine learning feature selection, it is used to evaluate the discriminative power of an input feature on the output target variable. The null hypothesis is that the means of each group are equal (i.e., the feature has no significant impact on the target); the alternative hypothesis is that at least one group has a different mean from the others (the feature has discriminative power). The F-statistic (F-value) is calculated as the ratio of between-group variance to within-group variance. A larger F-value indicates that the between-group difference is much greater than the within-group fluctuation, providing stronger reason to reject the null hypothesis, and signifying a more important feature. A smaller p-value indicates statistical significance. The correlation between historical usage data and mold wear prediction is represented by the F-value. Historical usage data with an F-value greater than a preset threshold can be retained, while historical usage data with an F-value less than or equal to the preset threshold can be removed. This preset threshold can be set according to actual conditions and is not limited here.

[0037] S130, a wear prediction model is constructed based on the target input data and the corresponding extrusion amount label.

[0038] In one embodiment, such as Figure 3 As shown, the specific process of the wear prediction model includes the following steps: S131, the target input data is divided into multiple groups of time-series input features according to the time sequence, and each group of time-series input features corresponds to a squeezing amount label.

[0039] Exemplary target input data includes the surface quality of the extruded material, die wall thickness, extruded material composition, die extrusion breakthrough pressure, die extrusion speed, extruded material temperature, die exit temperature, number of die nitriding cycles, and die material. The extrusion volume label includes both theoretical and actual extrusion volumes. Extruded materials include, but are not limited to, aluminum profiles, copper alloys, and magnesium alloys. The surface quality of the extruded material refers to the state and degree of defects on the outer surface of the extruded product. The die extrusion breakthrough pressure refers to the maximum instantaneous pressure that must be overcome at the beginning of extrusion when the extrusion rod (punch) pushes the billet through the die flow channel. The number of die nitriding cycles refers to the total number of nitriding treatments the die undergoes during its lifespan. The die material determines its core properties such as strength, toughness, heat resistance, and wear resistance, directly affecting its service life. The die material can be steel.

[0040] S132, Construct the model architecture for the initial wear prediction model, and divide multiple sets of time-series input features and corresponding extrusion amount labels into training and test sets according to time order.

[0041] In this embodiment, the model initially adopts an RNN (Recurrent Neural Network) deep learning model. Its core advantage lies in its unique recurrent connection structure, which can store historical time-series information and form a memory for sequential data. In the mold wear prediction scenario, multiple sets of time-series input features and corresponding extrusion amount labels have strong temporal correlation. Through iterative updates of the hidden layer states, the RNN can effectively capture the nonlinear evolution trend of mold wear over time caused by material friction, stress fatigue, and other factors during stamping and other processes.

[0042] In one embodiment, this application constructs a 3-layer RNN stacked architecture based on the high-dimensional input characteristics of wear data. The input layer dimension is consistent with the target input data number, and the output layer is mapped to a single-dimensional real-time extrusion amount prediction through a fully connected layer. The number of hidden layer units adopts a dynamic adjustment strategy, initially set to three times the feature dimension (e.g., 300 units corresponding to 100-dimensional features), and dynamically increased or decreased by 10% during training based on the validation set performance.

[0043] In another implementation, a 7-fold time series cross-validation is used, strictly adhering to the chronological order in dividing the training and test sets to avoid future data leakage. Specifically, the dataset is arranged chronologically, with the first six folds serving as the training set and the seventh fold as the test set.

[0044] S133, the initial wear prediction model is trained, optimized and validated based on the training set and the test set to obtain the wear prediction model.

[0045] In one embodiment, an initial wear prediction model is iteratively trained based on a training set and preset hyperparameters to obtain a trained initial wear prediction model. Based on the prediction results of the trained initial wear prediction model on a test set, the hyperparameters of the trained initial wear prediction model are adjusted. The preset hyperparameters are manually adjusted hyperparameters, including but not limited to batch size and learning rate. This application trains the model based on manually adjusted hyperparameters, and observes the model convergence during training by visualizing the loss curve using TensorBoard. The hyperparameters are continuously adjusted based on the model's test set results to optimize the model's performance. The trained initial wear prediction model is then used to predict on a completely unseen test set to obtain prediction results. If the accuracy of the prediction results is low, the hyperparameters are adjusted until a higher accuracy is achieved.

[0046] In another embodiment, this application determines the prediction bias of the initial wear prediction model after training on the training set and the test set. Based on the prediction bias, the initial wear prediction model after training is validated to obtain a validated wear prediction model. The validation set includes data from all previous iterations to ensure that the model learns the evolution of the time series. In this embodiment, an early stopping mechanism is triggered when the validation set loss does not decrease significantly for 10 consecutive iterations to avoid overfitting. Actual optimization results show that after 300 iterations, the wear prediction error of the control model on the test set is reduced. Evaluation metrics such as MAE (Mean Absolute Error) are introduced to quantify the prediction bias of the mold wear model. If the prediction bias is small, the wear prediction model passes validation.

[0047] S140: Obtain the real-time target input data of the mold, input the real-time target input data into the wear prediction model, and obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear of the mold.

[0048] For example, when a batch of molds wears out, based on practical experience, factors such as the deviation between theoretical and actual extrusion amounts, wall thickness, and surface quality of the extruded profile have a strong influence on mold wear. The wear prediction model can infer the current wear level of the mold based on these factors and give the real-time extrusion amount of the mold on the material, such as 20 tons. At this time, the on-site staff can judge whether maintenance is required based on the tolerable extrusion amount (such as 10 tons). For example, if it is less than 10 tons, the technicians will be prompted to inspect and repair the mold wear.

[0049] In one embodiment, real-time target input data is collected from a mold data service platform via a model prediction service interface; the collected real-time target input data is then input into a wear prediction model via the model prediction service interface. Exemplarily, the model prediction service interface is an interface designed based on the input-output characteristics of the mold wear model. The transmission protocol of this model prediction service interface is HTTP (Hypertext Transfer Protocol), and the request method is a POST request. After the wear prediction model is validated, it is deployed to a material production system, such as an aluminum profile production system, and integrated with the mold data service platform to obtain the target input data. The mold data service platform includes, but is not limited to, DaaS (Data as a Service), MES (Manufacturing Execution System), and asset cloud systems, which transmit the data to the mold wear model via the model prediction service interface for prediction, predicting and outputting the current compressible allowance of the mold.

[0050] In this embodiment, deep learning technology is used to capture the wear patterns of the die during continuous processing, based on historical and current data from the material production process. This accurately predicts the extrusion amount before the die wear reaches its limit or exceeds the theoretical weight requirement. By learning the wear increment sequence of the die in different processing batches, the model can identify the characteristic differences between early die wear and later rapid wear stages, while also uncovering the potential correlation between the target input data and die wear. This dynamic modeling capability upgrades traditional experience-based threshold-based maintenance to data-driven predictive maintenance, helping workers make scientific maintenance decisions before the die performance deteriorates to a critical point, thus reducing the risk of sudden failures.

[0051] Figure 4 A schematic diagram of a mold wear prediction system 200 according to an embodiment of this application is shown. Exemplarily, the mold wear prediction system 200 includes: The data processing module 210 is used to acquire historical usage data of the mold and process the historical usage data to convert it into target input data related to the prediction of mold wear.

[0052] The model building module 220 is used to build a wear prediction model based on the target input data and the corresponding extrusion amount label.

[0053] The wear prediction module 230 is used to acquire real-time target input data of the mold, input the real-time target input data into the wear prediction model, and obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear of the mold.

[0054] It is understood that the system in this embodiment corresponds to the mold wear prediction method in the above embodiment, and the options in the above embodiment are also applicable to this embodiment, so they will not be described again here.

[0055] This application also provides a terminal device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor, by running the computer program, causes the terminal device to perform the functions of the various modules in the above-described mold wear prediction method or the above-described mold wear prediction system.

[0056] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0057] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving execution instructions.

[0058] This application also provides a computer-readable storage medium for storing the computer program used in the aforementioned terminal device. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0059] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0060] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0061] If the aforementioned functions are implemented as software functional modules 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 this application, in essence, or the part that contributes to the prior art, or a part of the 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 to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for predicting the wear degree of a mold, characterized in that, include: Obtain historical usage data for the mold; The historical usage data is processed to convert it into target input data related to mold wear prediction. A wear prediction model is constructed based on the target input data and the corresponding extrusion amount label; The real-time target input data of the mold is obtained and input into the wear prediction model to obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear of the mold.

2. The method for predicting mold wear according to claim 1, characterized in that, The step of processing the historical usage data to convert it into target input data related to mold wear prediction includes: The historical usage data is cleaned to obtain cleaned historical usage data. The cleaning process includes deduplication, missing value processing, and noise data processing. Based on the type of the historical usage data after the cleaning process, the historical usage data after the cleaning process is standardized and feature-encoded to obtain the processed historical usage data. The historical usage data is filtered based on the correlation between each processed historical usage data and the predicted wear level of the mold to obtain the target input data.

3. The method for predicting mold wear according to claim 1, characterized in that, The step of constructing a wear prediction model based on the target input data and the corresponding extrusion amount label includes: The target input data is divided into multiple groups of time-series input features in chronological order, and each group of time-series input features corresponds to a squeezing amount label. The model architecture for constructing the initial wear prediction model is to divide multiple sets of time-series input features and corresponding extrusion amount labels into training and test sets according to the time sequence. The initial wear prediction model is trained, optimized, and validated based on the training set and the test set to obtain the wear prediction model.

4. The method for predicting mold wear according to claim 2, characterized in that, The types of historical usage data after cleaning include numerical data and non-numerical data. The standardization and feature encoding of the historical usage data after cleaning, based on its type, includes: The numerical data is standardized to map it to a target range. The non-numerical data is subjected to one-hot encoding to convert it into a binary vector.

5. The method for predicting mold wear according to claim 2, characterized in that, The filtering of the historical usage data based on the correlation between each processed historical usage data and the predicted wear level of the mold includes: The correlation between each processed historical usage data and the predicted wear level of the mold is determined based on information theory and statistical methods. Based on the correlation, historical usage data that is irrelevant to the prediction of mold wear degree is removed, and historical usage data that is relevant to the prediction of mold wear degree is retained to obtain the target input data; The target input data includes the surface quality of the extruded material, die wall thickness, extruded material material, die extrusion breakthrough pressure, die extrusion speed, extruded material temperature, die exit temperature, number of nitriding cycles, and die material. The extrusion quantity label includes theoretical extrusion quantity and actual extrusion quantity.

6. The method for predicting mold wear according to claim 3, characterized in that, The process of training, optimizing, and validating the initial wear prediction model based on the training set and the test set to obtain the wear prediction model includes: The initial wear prediction model is iteratively trained based on the training set and preset hyperparameters to obtain the trained initial wear prediction model. Based on the prediction results of the initial wear prediction model after training on the test set, the hyperparameters of the initial wear prediction model after training are adjusted. The prediction deviation of the trained initial wear prediction model for the training set and the test set is determined, and the trained initial wear prediction model is verified based on the prediction deviation to obtain the verified wear prediction model.

7. The method for predicting mold wear according to claim 1, characterized in that, Also includes: The real-time target input data is collected from the mold data service platform through the model prediction service interface; The collected real-time target input data is input into the wear prediction model through the model prediction service interface.

8. A mold wear prediction system, characterized in that, include: The data processing module is used to acquire historical usage data of the mold, process the historical usage data, and convert the historical usage data into target input data related to the prediction of mold wear degree. The model building module is used to build a wear prediction model based on the target input data and the corresponding extrusion amount label; The wear degree prediction module is used to acquire the real-time target input data of the mold, input the real-time target input data into the wear prediction model, and obtain the prediction result output by the wear prediction model. The prediction result is the real-time extrusion amount of the mold. The smaller the real-time extrusion amount, the greater the wear degree of the mold.

9. A terminal device, characterized in that, It includes a memory and a processor, the memory storing a computer program that executes the mold wear prediction method according to any one of claims 1 to 7 when the processor is running.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when run on a processor, executes the mold wear prediction method according to any one of claims 1 to 7.

Images