Artificial intelligence-based magnetic lining life prediction method, device, equipment and medium

Abstract

This application relates to a method, apparatus, device, and medium for predicting the lifespan of magnetic liners based on artificial intelligence. It is applied in the field of magnetic liner lifespan prediction technology. The method includes: acquiring initial historical lifespan data and initial customer operating condition data; preprocessing the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data; analyzing the historical lifespan data and the customer operating condition data according to multiple prediction methods to obtain various lifespan prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method, and similar operating condition matching method; verifying the historical errors of various prediction methods based on the historical lifespan data; and fusing the various lifespan prediction information based on the historical errors to obtain target lifespan information. This application has the effect of improving the accuracy of magnetic liner lifespan prediction.

Description

Technical Field

[0001] This application relates to the technical field of magnetic liner life prediction, and in particular to a method, apparatus, equipment and medium for predicting the life of magnetic liner based on artificial intelligence. Background Technology

[0002] Ball mills are key equipment in mining production. With the continuous development of the mining industry, the stable operation of ball mills is crucial for improving production efficiency and reducing production costs. Magnetic liners have been used in ball mills for a long time. They utilize magnetic force to attract broken balls and magnetic materials to form a protective layer, effectively extending the liner's lifespan. This is of great significance for ensuring the stable operation of the ball mill. Reasonably estimating the lifespan of magnetic liners helps companies plan equipment maintenance and replacement in advance, avoiding production stoppages due to liner damage, thereby protecting the company's economic benefits.

[0003] Currently, there are two common practices in the industry for estimating the service life of magnetic liners. One is to rely on empirical formulas, where staff use past experience and some general formulas, combined with some operating parameters of the new customer, to roughly estimate the lifespan of the magnetic liners. The other is a simple analogy method, which refers to the usage of magnetic liners under other similar operating conditions to infer the lifespan of the magnetic liners at the new customer's site. However, due to limitations in work experience and differences in operating parameters across different mines, both methods often struggle to accurately predict the actual service life of magnetic liners. Summary of the Invention

[0004] To improve the accuracy of magnetic liner life prediction, this application provides a method, apparatus, device, and medium for predicting magnetic liner life based on artificial intelligence.

[0005] Firstly, this application provides a method for predicting the lifespan of magnetic liners based on artificial intelligence, employing the following technical solution: An artificial intelligence-based method for predicting the lifespan of magnetic liners includes: Acquire initial historical lifespan data and initial customer operating condition data. The initial historical lifespan data includes initial historical operating condition data, initial historical lifespan, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters. The initial historical lifespan data and the initial customer operating condition data are preprocessed to obtain historical lifespan data and customer operating condition data. The preprocessing includes sample data processing and feature type processing. The historical lifespan data and customer operating condition data are analyzed using multiple prediction methods to obtain various lifespan prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar operating condition matching method. The lifespan prediction information includes empirical lifespan prediction information, model lifespan prediction information, layered wear lifespan prediction information and similar operating condition lifespan prediction information. The historical errors of various prediction methods are verified based on the historical lifetime data. Based on the historical error, the various lifetime prediction information is fused to obtain the target lifetime information.

[0006] By adopting the above technical solutions, comprehensive initial historical lifespan data and initial customer operating condition data can be obtained, providing rich information for subsequent analysis. Preprocessing the initial data can improve data quality and enhance the reliability of subsequent analysis. Analyzing the data using multiple prediction methods yields various lifespan prediction information, allowing for prediction of the magnetic liner's lifespan from different perspectives. Verifying the historical errors of the prediction methods allows for an understanding of their accuracy. Integrating multiple lifespan prediction information based on historical errors to obtain target lifespan information reduces the impact of errors from various prediction methods, improves the accuracy of magnetic liner lifespan prediction, and helps companies plan equipment maintenance and replacement in advance.

[0007] Optionally, the preprocessing of the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data includes: Based on the initial historical lifetime data and the preset feature construction rules, the derived feature data corresponding to the derived features are calculated; The initial historical data is processed to obtain the first historical lifetime data. The initial historical data includes the initial historical lifetime data and the derived feature data. The sample data processing includes missing value imputation, outlier removal, duplicate sample processing, and data normalization. Based on the first historical lifetime data, calculate the first correlation coefficient between every two feature types, the variance inflation factor of each feature type, and the second correlation coefficient between each feature type and lifetime; Target feature types are selected based on the first correlation coefficient, the variance inflation factor, and the second correlation coefficient; The historical lifetime data is determined based on the target feature type and the first historical lifetime data. The initial customer operating condition data is processed as sample data and as feature type data according to the target feature type to obtain the customer operating condition data.

[0008] By adopting the above technical solutions, the calculation of derived feature data can enrich the data information, the sample data processing can improve the integrity, accuracy and consistency of the initial historical data, and the calculation of correlation coefficient and variance inflation factor can help to screen out the target feature types that are more valuable for life prediction. Finally, high-quality historical life data and customer operating condition data are obtained, laying the foundation for accurate prediction of magnetic liner life.

[0009] Optionally, the analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: Obtain the reference life of the magnetic liner; The correction coefficient is determined based on the customer's operating condition data and the preset coefficient adjustment rules; The empirical life prediction information is calculated based on the baseline life and the correction factor. The neural network algorithm is trained based on the historical lifespan data to obtain an AI lifespan prediction model, which incorporates an attention mechanism. The customer's operating condition data is input into the AI ​​lifespan prediction model to obtain the model's lifespan prediction information.

[0010] By adopting the above technical solution, the baseline life of the magnetic liner is obtained and the correction coefficient is determined in combination with the customer's working condition data, thereby calculating the empirical life prediction information. The life of the magnetic liner can be initially estimated based on the actual working conditions. The AI ​​life prediction model with an attention mechanism is trained using historical life data, and the customer's working condition data is input into the AI ​​life prediction model to obtain the model's life prediction information. Artificial intelligence can be used to improve the accuracy of the magnetic liner life prediction.

[0011] Optionally, the analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: Historical maintenance records are determined based on the historical lifespan data, and the historical maintenance records include maintenance time and cumulative wear thickness. A wear curve is fitted based on the maintenance time and the cumulative wear thickness; Based on piecewise linear regression, the first and second boundary points in the wear curve are identified. The first boundary point is the boundary between the running-in period and the steady-state period, and the second boundary point is the boundary between the steady-state period and the acceleration period. The wear thickness and wear rate of each stage are calculated based on the wear curve, wherein the stages include the running-in period, the steady-state period, and the acceleration period; A stage prediction model is constructed based on the stage wear thickness and stage wear rate corresponding to the historical life data of each group. The customer's operating condition data is input into the stage prediction model to obtain the predicted stage wear thickness and predicted stage wear rate for each stage. The duration of each stage is calculated based on the predicted stage wear thickness and the predicted stage wear rate of each stage. The layered wear life prediction information is determined based on the duration of each of the aforementioned stages.

[0012] By adopting the above technical solution, fitting wear curves through historical maintenance records and identifying the boundary points of different stages, the running-in period, steady-state period, and accelerated period of magnetic liner wear can be clearly divided. The stage wear thickness and stage wear rate of each stage are calculated and a stage prediction model is constructed. Based on the customer's working condition data, the predicted stage wear thickness and predicted stage wear rate can be obtained. Furthermore, the duration of each stage can be calculated, and the layered wear life prediction information can be determined, so as to achieve a more accurate layered prediction of the magnetic liner life.

[0013] Optionally, the analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: The feature weights of each target feature type are determined based on the second correlation coefficient between each target feature type and the lifetime; The similarity distance between each of the historical lifetime data and the customer operating condition data is calculated based on the weighted Euclidean distance and the feature weights. The preset number of historical lifespan data with the smallest similarity distance are identified as similar historical data. Calculate the data weight of each of the similar historical data based on the similarity distance; The life prediction information for similar operating conditions is determined based on the data weights and the similar historical data.

[0014] By adopting the above technical solution, the feature weights are determined based on the second correlation coefficient between the target feature type and the lifespan, which can more reasonably measure the impact of each feature on the lifespan. By calculating the similarity distance using the weighted Euclidean distance and feature weights, historical lifespan data that is similar to the customer's working condition data can be accurately identified. After determining the similar historical data, the corresponding data weights are calculated based on the similarity distance, and then the lifespan prediction information for similar working conditions is determined. This can make full use of historical data and improve the accuracy of magnetic liner lifespan prediction.

[0015] Optionally, the step of verifying the historical errors of various prediction methods based on the historical lifetime data includes: The similar historical data are analyzed according to the various prediction methods described above to obtain the historical lifetime prediction values ​​corresponding to the various prediction methods described above. The absolute error of each prediction method is calculated based on the historical lifetime prediction value and the corresponding historical lifetime. The average of the absolute errors corresponding to all the similar historical data is calculated to obtain the historical errors of various prediction methods.

[0016] By adopting the above technical solution, historical lifetime prediction values ​​are obtained by using multiple prediction methods on similar historical data. The absolute error is calculated by combining the historical lifetime, and then the average of the absolute errors corresponding to the similar historical data is calculated as the historical error of each prediction method. This can accurately evaluate the accuracy of various prediction methods, provide a reliable basis for subsequent fusion of lifetime prediction information, and improve the accuracy of magnetic liner lifetime prediction.

[0017] Optionally, the step of fusing various lifetime prediction information based on the historical error to obtain target lifetime information includes: The prediction weights of each prediction method are calculated based on the historical errors of each prediction method. The historical fusion predicted lifetime of each of the similar historical data is calculated based on the predicted weights and the historical lifetime predicted values. Based on the historical fusion predicted lifetime and the historical lifetime, the historical fusion absolute error of each of the similar historical data is calculated. The confidence bias is determined based on the historical fusion absolute error and the preset quantile. The fused predicted life is calculated based on the empirical life prediction information, the model life prediction information, the layered wear life prediction information, the similar working condition life prediction information, and the prediction weights. The confidence interval is determined based on the fusion predicted lifetime and the confidence bias. The target lifetime information is determined based on the fused predicted lifetime and the confidence interval.

[0018] By adopting the above technical solution, the prediction weights are calculated based on historical errors, and the weights can be allocated in combination with the accuracy of each prediction method, so as to more reasonably integrate lifetime prediction information and obtain more accurate fused lifetime prediction. The accuracy of fused prediction can be evaluated by calculating historical fused lifetime prediction and historical fused absolute error. The confidence bias is determined based on historical fused absolute error and preset quantiles, and the confidence interval can be determined by combining the confidence bias, so as to obtain more accurate and confident magnetic liner lifetime prediction results, thereby improving the accuracy and reliability of magnetic liner lifetime prediction.

[0019] Secondly, this application provides a magnetic liner life prediction device based on artificial intelligence, which adopts the following technical solution: An artificial intelligence-based magnetic liner life prediction device includes: The data acquisition module is used to acquire initial historical life data and initial customer operating condition data. The initial historical life data includes initial historical operating condition data, initial historical life, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters. The preprocessing module is used to preprocess the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data. The preprocessing includes sample data processing and feature type processing. The life prediction module is used to analyze the historical life data and the customer's operating condition data according to multiple prediction methods to obtain multiple life prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar operating condition matching method. The life prediction information includes empirical life prediction information, model life prediction information, layered wear life prediction information and similar operating condition life prediction information. An error verification module is used to verify the historical errors of various prediction methods based on the historical lifetime data. The lifetime determination module is used to fuse various lifetime prediction information based on the historical errors to obtain target lifetime information.

[0020] By adopting the above technical solutions, comprehensive initial historical lifespan data and initial customer operating condition data can be obtained, providing rich information for subsequent analysis. Preprocessing the initial data can improve data quality and enhance the reliability of subsequent analysis. Analyzing the data using multiple prediction methods yields various lifespan prediction information, allowing for prediction of the magnetic liner's lifespan from different perspectives. Verifying the historical errors of the prediction methods allows for an understanding of their accuracy. Integrating multiple lifespan prediction information based on historical errors to obtain target lifespan information reduces the impact of errors from various prediction methods, improves the accuracy of magnetic liner lifespan prediction, and helps companies plan equipment maintenance and replacement in advance.

[0021] Thirdly, this application provides an electronic device that adopts the following technical solution: An electronic device includes a processor coupled to a memory; The memory stores a computer program that can be loaded by a processor and executed by the artificial intelligence-based magnetic liner life prediction method described in any of the first aspects.

[0022] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program capable of being loaded by a processor and executing the artificial intelligence-based magnetic liner lifetime prediction method described in any of the first aspects. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an artificial intelligence-based method for predicting the lifespan of magnetic liners, as provided in an embodiment of this application.

[0024] Figure 2 This is a structural block diagram of an artificial intelligence-based magnetic liner life prediction device provided in an embodiment of this application.

[0025] Figure 3 This is a structural block diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0026] The present application will be further described in detail below with reference to the accompanying drawings.

[0027] This application provides an AI-based method for predicting the lifespan of magnetic liners. This AI-based method can be executed by an electronic device, which can be a server or a terminal device. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet, desktop computer, etc., but is not limited to these.

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0029] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0030] like Figure 1 As shown, an artificial intelligence-based method for predicting the lifespan of magnetic liners is described in the following steps (S101-S105): Step S101: Obtain initial historical lifespan data and initial customer operating condition data.

[0031] Obtain initial historical lifespan data of historical customers and initial customer operating condition data of current customers from staff or database. The initial historical lifespan data includes initial historical operating condition data, initial historical lifespan, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters.

[0032] Initial historical operating data includes initial historical mill data, initial historical ore properties, and initial historical process parameters, such as the ball mill's cylinder diameter and length, the ore's Protodyakonov hardness coefficient, density, feed particle size, and product particle size, as well as the ball mill's throughput, slurry concentration, media ball diameter, filling rate, ball mill speed, and power. Initial historical lifespan is the actual lifespan of the magnetic liners in past use. Initial historical maintenance data records the time when the magnetic liners were first used, the maintenance time for each maintenance, and the cumulative wear thickness during each maintenance.

[0033] The initial mill data, initial ore properties, and initial process parameters all correspond to the aforementioned initial historical mill data, initial historical ore properties, and initial historical process parameters, and will not be repeated here.

[0034] Step S102: Preprocess the initial historical life data and initial customer operating condition data to obtain historical life data and customer operating condition data.

[0035] Preprocess the initial historical lifespan data to obtain historical lifespan data; preprocess the initial customer operating condition data to obtain customer operating condition data.

[0036] Preprocessing includes sample data processing and feature type processing.

[0037] Specifically, the initial historical lifespan data and initial customer operating condition data are preprocessed to obtain historical lifespan data and customer operating condition data, including: calculating derived feature data corresponding to derived features based on the initial historical lifespan data and preset feature construction rules; performing sample data processing on the initial historical data to obtain first historical lifespan data, the initial historical data including initial historical lifespan data and derived feature data, the sample data processing including missing value imputation, outlier removal, duplicate sample processing, and data normalization; calculating the first correlation coefficient between every two feature types, the variance inflation factor of each feature type, and the second correlation coefficient between each feature type and lifespan based on the first historical lifespan data; screening target feature types based on the first correlation coefficient, variance inflation factor, and second correlation coefficient; determining historical lifespan data based on the target feature types and the first historical lifespan data; and performing sample data processing on the initial customer operating condition data and feature type processing according to the target feature types to obtain customer operating condition data.

[0038] In this embodiment, the database pre-stores preset feature construction rules for derived features (new features obtained by combining or calculating certain features in the initial historical lifetime data). Derived features include, for example, unit power consumption and wear index. The preset feature construction rules are, for example, unit power consumption = power / processing volume. The initial historical lifetime data is analyzed and calculated according to the preset feature construction rules to obtain the derived feature data corresponding to the derived features.

[0039] The initial historical lifespan data (only the initial historical operating condition data) and derived feature data are processed to obtain the first historical lifespan data. Missing value imputation can be performed using methods such as mean imputation and median imputation. Alternatively, the ball mill cylinder diameter and cylinder length can be imputed according to the ball mill model. Outlier removal can be achieved by setting a threshold, and duplicate sample processing can remove duplicate data. Data normalization can unify the data to a specific range.

[0040] Based on the first historical lifespan data corresponding to each feature type, calculate the first correlation coefficient between each pair of feature types, the variance inflation factor of each feature type, and the second correlation coefficient between each feature type and lifespan. The first and second correlation coefficients can be calculated using Pearson correlation coefficient or Spearman correlation coefficient, and the variance inflation factor can be calculated using the variance inflation factor calculation method in multiple linear regression.

[0041] If the first correlation coefficient between two feature types is greater than the preset correlation coefficient (e.g., 0.7), the feature types with the larger variance inflation factor are removed. The remaining feature types are sorted from largest to smallest according to the second correlation coefficient, and the first preset number (e.g., 8) of feature types in the sort are determined as the target feature types. The historical lifespan data includes historical lifespan (i.e., initial historical lifespan), historical maintenance data (i.e., initial historical maintenance data), and initial historical lifespan data after processing the sample data corresponding to the target feature type.

[0042] The initial customer operating condition data of the current customer is processed by sample data processing, including outlier removal (using a set threshold), missing value filling (if there is missing data, staff need to supplement the data), and data normalization (processed according to the normalization parameters of the initial historical life data). At the same time, only the data corresponding to the target feature type in the initial customer operating condition data (after sample data processing) is retained, thus obtaining the customer operating condition data.

[0043] Step S103: Analyze historical lifespan data and customer operating condition data using multiple prediction methods to obtain various lifespan prediction information for the current customer.

[0044] The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar working condition matching method. The life prediction information includes empirical life prediction information, model life prediction information, layered wear life prediction information and similar working condition life prediction information.

[0045] Specifically, historical lifespan data and customer operating condition data are analyzed using various prediction methods to obtain multiple lifespan prediction information for the current customer, including: obtaining the baseline lifespan of the magnetic liner; determining the correction coefficient based on customer operating condition data and preset coefficient adjustment rules; calculating empirical lifespan prediction information based on the baseline lifespan and correction coefficient; training a neural network algorithm based on historical lifespan data to obtain an AI lifespan prediction model, which incorporates an attention mechanism; and inputting customer operating condition data into the AI ​​lifespan prediction model to obtain model lifespan prediction information.

[0046] In this embodiment, the empirical formula method calculates the empirical predicted lifespan using an empirical formula: Empirical predicted lifespan = Baseline lifespan × Product of all correction coefficients. The correction coefficients can include hardness correction coefficients, diameter correction coefficients, throughput correction coefficients, slurry concentration correction coefficients, etc. The database stores the correspondence between customer operating data and various correction coefficients, i.e., preset coefficient adjustment rules. Based on the customer operating data, the values ​​of various correction coefficients are matched from the preset coefficient adjustment rules. Then, the baseline lifespan of the magnetic liner obtained from the staff and the various correction coefficients are substituted into the empirical formula to obtain the empirical predicted lifespan. The empirical lifespan prediction information includes the empirical predicted lifespan, the baseline lifespan, and the values ​​of various correction coefficients.

[0047] AI model prediction involves using a pre-trained AI lifespan prediction model to predict lifespan. Customer operating condition data is input into the AI ​​lifespan prediction model to obtain the predicted lifespan. The predicted lifespan information includes the predicted lifespan and the feature importance distribution during the prediction process. The AI ​​lifespan prediction model is a prediction model trained on a neural network algorithm using historical lifespan data. The AI ​​lifespan prediction model can use algorithms such as Random Forest and XGBoost as regression models, employing a multi-layer perceptron or a specially constructed deep neural network for deep learning, and introducing a multi-head self-attention mechanism in the hidden layers. The multi-head attention mechanism calculates dynamic weights between input features, enabling the network to automatically focus on the feature types (such as hardness, processing load, etc.) that contribute the most to lifespan prediction, while outputting an interpretable feature importance distribution. During model training, historical lifespan data is divided into training, testing, and validation sets according to a preset ratio. Mean squared error (MSE) is used as the loss function. During validation, to prevent overfitting, K-fold cross-validation is used, where K can be 5 or 10.

[0048] Specifically, historical lifespan data and customer operating condition data are analyzed using various prediction methods to obtain multiple lifespan prediction information for the current customer, including: identifying historical maintenance records based on historical lifespan data, including maintenance time and cumulative wear thickness; fitting wear curves based on maintenance time and cumulative wear thickness; identifying the first and second boundary points in the wear curve based on piecewise linear regression, where the first boundary point is the boundary between the running-in period and the steady-state period, and the second boundary point is the boundary between the steady-state period and the acceleration period; calculating the stage wear thickness and stage wear rate for each stage based on the wear curve, where the stages include the running-in period, the steady-state period, and the acceleration period; constructing stage prediction models based on the stage wear thickness and stage wear rate corresponding to each set of historical lifespan data; inputting customer operating condition data into the stage prediction models to obtain the predicted stage wear thickness and predicted stage wear rate for each stage; calculating the stage duration for each stage based on the predicted stage wear thickness and predicted stage wear rate for each stage; and determining layered wear lifespan prediction information based on the stage duration for each stage.

[0049] In this embodiment, the layered wear state method divides the life cycle of the magnetic liner into three stages: running-in period, steady-state period, and acceleration period. The wear rate of each stage is different, and the predicted wear thickness and predicted wear rate of each stage are predicted to obtain the layered predicted life.

[0050] The starting time of each magnetic liner's use, the maintenance time of each overhaul, and the cumulative wear thickness at each overhaul are determined from historical lifespan data. A wear curve for each magnetic liner is plotted, with the two axes representing time and cumulative wear thickness, respectively. The first and second boundary points in the wear curve are identified by methods such as piecewise linear regression or second derivative inflection point detection, thereby obtaining the wear curve corresponding to each stage. The stage wear thickness and stage wear rate are calculated based on the wear curves corresponding to each stage.

[0051] Each set of historical lifespan data corresponds to three stages of stage wear thickness and stage wear rate. The stage wear thickness and stage wear rate corresponding to each set of historical lifespan data are used to train a machine learning model (which can be a random forest model or other machine learning models with the same prediction function) to obtain a stage prediction model. By inputting customer operating condition data into the stage prediction model, the predicted stage wear thickness and predicted stage wear rate for each stage can be obtained. It is worth noting that two stage prediction models can be pre-trained to predict stage wear thickness and predicted stage wear rate, respectively. Stage duration = predicted stage wear thickness / predicted stage wear rate. The stratified predicted lifespan is the sum of the stage durations of the three stages. The stratified wear lifespan prediction information includes the stratified predicted lifespan, the predicted stage wear thickness, the predicted stage wear rate, and the stage duration for each stage.

[0052] Specifically, historical lifespan data and customer operating condition data are analyzed using multiple prediction methods to obtain various lifespan prediction information for the current customer, including: determining the feature weights of each target feature type based on the second correlation coefficient between each target feature type and lifespan; calculating the similarity distance between each historical lifespan data and customer operating condition data based on weighted Euclidean distance and feature weights; identifying a preset number of historical lifespan data with the smallest similarity distance as similar historical data; calculating the data weights of each similar historical data based on the similarity distance; and determining similar operating condition lifespan prediction information based on the data weights and similar historical data.

[0053] The similar operating condition matching method finds similar historical data from historical lifespan data, and predicts the lifespan of the current customer based on similar operating conditions by analyzing the similar historical data.

[0054] In this embodiment, the feature weight of a target feature type is equal to the second correlation coefficient of that target feature type divided by the sum of the second correlation coefficients of all target feature types. The weighted Euclidean distance is calculated as follows: similarity distance = [Σ(w i ×(a i -b i )²)] 1 / 2 , where w i Let a be the feature weight corresponding to the i-th target feature type. i For the historical lifetime data corresponding to the i-th target feature type, b iFor the customer operating condition data corresponding to the i-th target feature type, the similarity distance between each historical lifespan data and the customer operating condition data is calculated using weighted Euclidean distance. The historical lifespan data with the smallest similarity distance (e.g., 5) are determined as similar historical data. The data weight of a similar historical data is the reciprocal of the similarity distance of that similar historical data / the sum of the reciprocals of the similarity distances of all similar historical data. The predicted lifespan of similar operating conditions is the weighted sum of the historical lifespans and data weights in each similar historical data. The lifespan prediction information of similar operating conditions includes the predicted lifespan of similar operating conditions, each similar historical data and its corresponding similarity distance and data weight.

[0055] Step S104: Verify the historical errors of various prediction methods based on historical lifetime data.

[0056] Specifically, the historical errors of various prediction methods are verified based on historical lifetime data, including: analyzing similar historical data according to various prediction methods to obtain the historical lifetime prediction values ​​corresponding to various prediction methods; calculating the absolute errors of various prediction methods based on the historical lifetime prediction values ​​and the corresponding historical lifetimes; and calculating the average of the absolute errors corresponding to all similar historical data to obtain the historical errors of various prediction methods.

[0057] In this embodiment, similar historical data are used to predict lifetimes using the various prediction methods described above, resulting in historical lifetime prediction values ​​for each set of similar historical data using various prediction methods. That is, a set of similar historical data corresponds to four historical lifetime prediction values. For a set of similar historical data, the absolute error of each prediction method is the absolute value of the difference between the historical lifetime prediction value and the historical lifetime. The average of all absolute errors corresponding to the same prediction method is calculated to obtain the historical error of each prediction method.

[0058] Step S105: Based on historical errors, fuse various lifetime prediction information to obtain target lifetime information.

[0059] Specifically, various lifetime prediction information is fused based on historical errors to obtain target lifetime information, including: calculating the prediction weights of various prediction methods based on historical errors of various prediction methods; calculating the historical fused predicted lifetime of each similar historical data based on prediction weights and historical lifetime prediction values; calculating the historical fused absolute error of each similar historical data based on historical fused predicted lifetime and historical lifetime; determining the confidence bias based on historical fused absolute error and preset quantiles; calculating the fused predicted lifetime based on empirical lifetime prediction information, model lifetime prediction information, layered wear lifetime prediction information, similar working condition lifetime prediction information, and prediction weights; determining the confidence interval based on the fused predicted lifetime and confidence bias; and determining the target lifetime information based on the fused predicted lifetime and confidence interval.

[0060] In this embodiment, the prediction weight of a prediction method is equal to the reciprocal of the historical error of that prediction method / the sum of the reciprocals of the historical errors of all prediction methods. The historical fusion prediction lifetime of each similar historical data is obtained by weighted summation of the prediction weights of various prediction methods and the historical lifetime prediction values. The historical fusion absolute error of a group of similar historical data is the absolute value of the difference between the historical fusion prediction lifetime and the historical lifetime of the group of similar historical data. All historical fusion absolute errors are sorted from largest to smallest, and the historical fusion absolute error corresponding to the first preset quantile (which can be the 90th quantile) in the sort is taken as the confidence bias.

[0061] The current customer's fusion predicted lifetime is a weighted sum of empirical predicted lifetime, model predicted lifetime, stratified predicted lifetime, and similar operating condition predicted lifetime with their respective prediction weights. The confidence interval is [fusion predicted lifetime - confidence bias, fusion predicted lifetime + confidence bias]. The target lifetime information includes fusion predicted lifetime, confidence interval, and may also include empirical lifetime prediction information, model lifetime prediction information, stratified wear lifetime prediction information, and similar operating condition lifetime prediction information.

[0062] Figure 2 This is a structural block diagram of an artificial intelligence-based magnetic liner life prediction device 200 provided in an embodiment of this application.

[0063] like Figure 2 As shown, the artificial intelligence-based magnetic liner life prediction device 200 mainly includes: The data acquisition module 201 is used to acquire initial historical life data and initial customer operating condition data. The initial historical life data includes initial historical operating condition data, initial historical life, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters. Preprocessing module 202 is used to preprocess the initial historical life data and initial customer operating condition data to obtain historical life data and customer operating condition data. The preprocessing includes sample data processing and feature type processing. The life prediction module 203 is used to analyze historical life data and customer operating condition data according to various prediction methods to obtain various life prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar operating condition matching method. The life prediction information includes empirical life prediction information, model life prediction information, layered wear life prediction information and similar operating condition life prediction information. Error verification module 204 is used to verify the historical errors of various prediction methods based on historical lifetime data; The lifetime determination module 205 is used to fuse various lifetime prediction information based on historical errors to obtain target lifetime information.

[0064] As an optional implementation of this embodiment, the preprocessing module 202 is specifically used to preprocess the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data, including: calculating the derived feature data corresponding to the derived features based on the initial historical lifespan data and preset feature construction rules; performing sample data processing on the initial historical data to obtain the first historical lifespan data, wherein the initial historical data includes the initial historical lifespan data and the derived feature data, and the sample data processing includes missing value imputation, outlier removal, duplicate sample processing, and data normalization; calculating the first correlation coefficient between every two feature types, the variance inflation factor of each feature type, and the second correlation coefficient between each feature type and lifespan based on the first historical lifespan data; filtering the target feature type based on the first correlation coefficient, the variance inflation factor, and the second correlation coefficient; determining the historical lifespan data based on the target feature type and the first historical lifespan data; and performing sample data processing on the initial customer operating condition data and feature type processing according to the target feature type to obtain the customer operating condition data.

[0065] As an optional implementation of this embodiment, the life prediction module 203 is specifically used to analyze historical life data and customer operating condition data according to multiple prediction methods to obtain multiple life prediction information for the current customer, including: obtaining the baseline life of the magnetic liner; determining the correction coefficient based on the customer operating condition data and preset coefficient adjustment rules; calculating empirical life prediction information based on the baseline life and the correction coefficient; training the neural network algorithm based on historical life data to obtain an AI life prediction model, wherein the AI ​​life prediction model introduces an attention mechanism; and inputting the customer operating condition data into the AI ​​life prediction model to obtain model life prediction information.

[0066] As an optional implementation of this embodiment, the life prediction module 203 is specifically used to analyze historical life data and customer operating condition data according to multiple prediction methods to obtain various life prediction information for the current customer, including: determining historical maintenance records based on historical life data, the historical maintenance records including maintenance time and cumulative wear thickness; fitting a wear curve based on maintenance time and cumulative wear thickness; identifying the first boundary point and the second boundary point in the wear curve based on piecewise linear regression, the first boundary point being the boundary point between the running-in period and the steady-state period, and the second boundary point being the boundary point between the steady-state period and the acceleration period; calculating the stage wear thickness and stage wear rate of each stage based on the wear curve, the stages including the running-in period, the steady-state period, and the acceleration period; constructing a stage prediction model based on the stage wear thickness and stage wear rate corresponding to each set of historical life data; inputting customer operating condition data into the stage prediction model to obtain the predicted stage wear thickness and predicted stage wear rate of each stage; calculating the stage duration of each stage based on the predicted stage wear thickness and predicted stage wear rate of each stage; and determining layered wear life prediction information based on the stage duration of each stage.

[0067] As an optional implementation of this embodiment, the life prediction module 203 is specifically used to analyze historical life data and customer operating condition data according to multiple prediction methods to obtain multiple life prediction information for the current customer, including: determining the feature weight of each target feature type based on the second correlation coefficient between each target feature type and life; calculating the similarity distance between each historical life data and customer operating condition data based on the weighted Euclidean distance and feature weight; determining a preset number of historical life data with the smallest similarity distance as similar historical data; calculating the data weight of each similar historical data based on the similarity distance; and determining similar operating condition life prediction information based on the data weight and similar historical data.

[0068] As an optional implementation of this embodiment, the error verification module 204 is specifically used to verify the historical errors of various prediction methods based on historical lifetime data, including: analyzing similar historical data according to various prediction methods to obtain the historical lifetime prediction values ​​corresponding to various prediction methods; calculating the absolute errors of various prediction methods based on the historical lifetime prediction values ​​and the corresponding historical lifetimes; and calculating the average value of the absolute errors corresponding to all similar historical data to obtain the historical errors of various prediction methods.

[0069] As an optional implementation of this embodiment, the life determination module 205 is specifically used to fuse various life prediction information based on historical errors to obtain target life information, including: calculating the prediction weights of various prediction methods based on the historical errors of various prediction methods; calculating the historical fusion predicted life of each similar historical data based on the prediction weights and historical life prediction values; calculating the historical fusion absolute error of each similar historical data based on the historical fusion predicted life and historical life; determining the confidence bias based on the historical fusion absolute error and preset quantiles; calculating the fusion predicted life based on empirical life prediction information, model life prediction information, layered wear life prediction information, similar working condition life prediction information, and prediction weights; determining the confidence interval based on the fusion predicted life and confidence bias; and determining the target life information based on the fusion predicted life and confidence interval.

[0070] In one example, the module in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0071] For example, when modules in a device can be implemented via a processing element scheduler, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor capable of calling programs. Alternatively, these modules can be integrated together as a system-on-a-chip (SOC).

[0072] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0073] Figure 3 This is a structural block diagram of an electronic device 300 provided in an embodiment of this application.

[0074] like Figure 3 As shown, the electronic device 300 includes a processor 301 and a memory 302, and may further include one or more of an information input / output (I / O) interface 303, a communication component 304, and a communication bus 305.

[0075] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps of the aforementioned artificial intelligence-based magnetic liner life prediction method. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, as well as application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as one or more of 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 storage, flash memory, magnetic disk, or optical disk.

[0076] I / O interface 303 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 304 is used for wired or wireless communication between electronic device 300 and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 304 may include a Wi-Fi component, a Bluetooth component, and an NFC component.

[0077] The electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the artificial intelligence-based magnetic liner lifetime prediction method given in the above embodiments.

[0078] The communication bus 305 may include a path for transmitting information between the aforementioned components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0079] Electronic device 300 may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers, and may also be servers.

[0080] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described artificial intelligence-based magnetic liner life prediction method.

[0081] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0082] The terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0083] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for predicting the lifespan of magnetic liners based on artificial intelligence, characterized in that, include: Acquire initial historical lifespan data and initial customer operating condition data. The initial historical lifespan data includes initial historical operating condition data, initial historical lifespan, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters. The initial historical lifespan data and the initial customer operating condition data are preprocessed to obtain historical lifespan data and customer operating condition data. The preprocessing includes sample data processing and feature type processing. The historical lifespan data and customer operating condition data are analyzed using multiple prediction methods to obtain various lifespan prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar operating condition matching method. The lifespan prediction information includes empirical lifespan prediction information, model lifespan prediction information, layered wear lifespan prediction information and similar operating condition lifespan prediction information. The historical errors of various prediction methods are verified based on the historical lifetime data. Based on the historical error, the various lifetime prediction information is fused to obtain the target lifetime information.

2. The method according to claim 1, characterized in that, The preprocessing of the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data includes: Based on the initial historical lifetime data and the preset feature construction rules, the derived feature data corresponding to the derived features are calculated; The initial historical data is processed to obtain the first historical lifetime data. The initial historical data includes the initial historical lifetime data and the derived feature data. The sample data processing includes missing value imputation, outlier removal, duplicate sample processing, and data normalization. Based on the first historical lifetime data, calculate the first correlation coefficient between every two feature types, the variance inflation factor of each feature type, and the second correlation coefficient between each feature type and lifetime; Target feature types are selected based on the first correlation coefficient, the variance inflation factor, and the second correlation coefficient; The historical lifetime data is determined based on the target feature type and the first historical lifetime data. The initial customer operating condition data is processed as sample data and as feature type data according to the target feature type to obtain the customer operating condition data.

3. The method according to claim 1, characterized in that, The analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: Obtain the reference life of the magnetic liner; The correction coefficient is determined based on the customer's operating condition data and the preset coefficient adjustment rules; The empirical life prediction information is calculated based on the baseline life and the correction factor. The neural network algorithm is trained based on the historical lifespan data to obtain an AI lifespan prediction model, which incorporates an attention mechanism. The customer's operating condition data is input into the AI ​​lifespan prediction model to obtain the model's lifespan prediction information.

4. The method according to claim 1, characterized in that, The analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: Historical maintenance records are determined based on the historical lifespan data, and the historical maintenance records include maintenance time and cumulative wear thickness. A wear curve is fitted based on the maintenance time and the cumulative wear thickness; Based on piecewise linear regression, the first and second boundary points in the wear curve are identified. The first boundary point is the boundary between the running-in period and the steady-state period, and the second boundary point is the boundary between the steady-state period and the acceleration period. The wear thickness and wear rate of each stage are calculated based on the wear curve, wherein the stages include the running-in period, the steady-state period, and the acceleration period; A stage prediction model is constructed based on the stage wear thickness and stage wear rate corresponding to the historical life data of each group. The customer's operating condition data is input into the stage prediction model to obtain the predicted stage wear thickness and predicted stage wear rate for each stage. The duration of each stage is calculated based on the predicted stage wear thickness and the predicted stage wear rate of each stage. The layered wear life prediction information is determined based on the duration of each of the aforementioned stages.

5. The method according to claim 2, characterized in that, The analysis of the historical lifespan data and the customer operating condition data using multiple prediction methods yields various lifespan prediction information for the current customer, including: The feature weights of each target feature type are determined based on the second correlation coefficient between each target feature type and the lifetime; The similarity distance between each of the historical lifetime data and the customer operating condition data is calculated based on the weighted Euclidean distance and the feature weights. The preset number of historical lifespan data with the smallest similarity distance are identified as similar historical data. Calculate the data weight of each of the similar historical data based on the similarity distance; The life prediction information for similar operating conditions is determined based on the data weights and the similar historical data.

6. The method according to claim 5, characterized in that, The verification of the historical errors of various prediction methods based on the historical lifetime data includes: The similar historical data are analyzed according to the various prediction methods described above to obtain the historical lifetime prediction values ​​corresponding to the various prediction methods described above. The absolute error of each prediction method is calculated based on the historical lifetime prediction value and the corresponding historical lifetime. The average of the absolute errors corresponding to all the similar historical data is calculated to obtain the historical errors of various prediction methods.

7. The method according to claim 6, characterized in that, The process of fusing various lifetime prediction information based on the historical error to obtain target lifetime information includes: The prediction weights of each prediction method are calculated based on the historical errors of each prediction method. The historical fusion predicted lifetime of each of the similar historical data is calculated based on the predicted weights and the historical lifetime predicted values. Based on the historical fusion predicted lifetime and the historical lifetime, the historical fusion absolute error of each of the similar historical data is calculated. The confidence bias is determined based on the historical fusion absolute error and the preset quantile. The fused predicted life is calculated based on the empirical life prediction information, the model life prediction information, the layered wear life prediction information, the similar working condition life prediction information, and the prediction weights. The confidence interval is determined based on the fusion predicted lifetime and the confidence bias. The target lifetime information is determined based on the fused predicted lifetime and the confidence interval.

8. A magnetic liner life prediction device based on artificial intelligence, characterized in that, include: The data acquisition module is used to acquire initial historical life data and initial customer operating condition data. The initial historical life data includes initial historical operating condition data, initial historical life, and initial historical maintenance data. The initial customer operating condition data includes initial mill data, initial ore properties, and initial process parameters. The preprocessing module is used to preprocess the initial historical lifespan data and the initial customer operating condition data to obtain historical lifespan data and customer operating condition data. The preprocessing includes sample data processing and feature type processing. The life prediction module is used to analyze the historical life data and the customer's operating condition data according to multiple prediction methods to obtain multiple life prediction information for the current customer. The prediction methods include empirical formula method, AI model prediction method, layered wear state method and similar operating condition matching method. The life prediction information includes empirical life prediction information, model life prediction information, layered wear life prediction information and similar operating condition life prediction information. An error verification module is used to verify the historical errors of various prediction methods based on the historical lifetime data. The lifetime determination module is used to fuse various lifetime prediction information based on the historical errors to obtain target lifetime information.

9. An electronic device, characterized in that, Includes a processor, which is coupled to a memory; The processor is configured to execute a computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It includes a computer program or instructions that, when run on a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

Images