A method, apparatus, equipment and storage medium for predicting grinding roller failures.

Abstract

This invention provides a method, apparatus, device, and storage medium for predicting grinding roller faults, comprising: filtering data in the MSE system database according to equipment usage parameters to generate a dataset to be processed; performing dimensionality reduction processing on the dataset to be processed using a piecewise aggregation approximation algorithm to generate dimensionality-reduced data; calling a trained GBDT prediction model to process the dimensionality-reduced data to generate a prediction result; and outputting a warning signal based on the prediction result. This invention aims to address the problems of existing grinding roller fault prediction methods, such as inability to handle certain special data, lack of universality, and the possibility of errors in the output prediction results during actual use, which may lead to the system's inability to quickly issue fault warnings, thereby causing damage to the grinding roller.

Description

Technical Field

[0001] This invention relates to the field of grinding roller fault prediction technology, and specifically to a grinding roller fault prediction method, device, equipment and its storage medium. Background Technology

[0002] Vertical roller mills are ideal large-scale grinding equipment, widely used in industries such as cement, power, metallurgy, chemicals, and non-metallic minerals. They integrate crushing, drying, grinding, grading, and conveying, offering high production efficiency and capable of grinding lumpy, granular, and powdery raw materials into the required powdered materials. The grinding roller, as a key component of the vertical roller mill, directly contacts the material and is primarily used for material crushing and grinding, directly affecting the quality of the tailings powder. The temperature of the grinding roller bearings and the pressure of the grinding roller are crucial parameters. Abnormalities in either can lead to roller damage or, in severe cases, roller stalling, consequently halting the tailings project and causing economic losses for the company. Therefore, in recent years, the prediction of grinding roller failures has received widespread attention from numerous tailings companies and related scholars.

[0003] The processing of feature variables, model selection, and selection of fault warning values ​​in grinding roll fault prediction models directly affect the accuracy of model prediction. The current challenges in grinding roll fault prediction lie in handling large amounts of time-series data, adjusting model parameters, and practical application in MES systems. Current research methods in fault prediction mainly include self-learning, neural networks, and XGBoost. While self-learning grinding roll fault prediction schemes do not rely on expert experience, they cannot predict faults in certain special aspects, such as kurtosis and root mean square. Existing methods combining SOM with small packages and using EWMA control charts to predict rolling linear guide faults are limited by experimental data and lack universality. Existing grinding roll fault prediction schemes based on RNNs and VARs, which are improved recurrent neural network and vector autoregression prediction methods, along with XGBoost prediction models, can predict grinding faults, but in practical use, the output prediction results may contain errors, potentially leading to the system's inability to quickly issue fault warnings and consequently, damage to the grinding rolls.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a grinding roller fault prediction method, device, equipment and storage medium, which can effectively solve the problems of existing grinding roller fault prediction methods being unable to handle some special data, lacking universality, and having errors in the prediction results output during actual use, which may lead to the system being unable to quickly issue fault warnings, thereby causing damage to the grinding roller.

[0006] This invention discloses a method for predicting grinding roller failures, comprising:

[0007] The data in the MSE system database is filtered based on the equipment usage parameters to generate a dataset to be processed.

[0008] The dataset to be processed is dimensionality reduced using a segmented aggregation approximation algorithm to generate dimensionality-reduced data.

[0009] The trained GBDT prediction model is invoked to process the dimensionality-reduced data to generate prediction results;

[0010] Based on the prediction results, an early warning signal is output.

[0011] Preferably, the data in the MSE system database is filtered according to the equipment usage parameters to generate a dataset to be processed, specifically as follows:

[0012] Based on the equipment usage parameter indicators, query the grinding roller component parameter information in the equipment historical data table of the MES system database, and export the grinding roller component parameter information to generate a grinding roller parameter file;

[0013] Data with empty values ​​in the grinding roller parameter file are filtered out, and feature variable data is generated, which includes multiple bearing temperature values ​​and multiple pressure values.

[0014] The feature variable data is mean-processed, and a fault classification label corresponding to the feature variable data is set to generate a dataset to be processed.

[0015] Preferably, the feature variable data is processed to set a fault classification identifier corresponding to the feature variable data, thereby generating a dataset to be processed, specifically as follows:

[0016] The average value of the multiple bearing temperature values ​​is obtained, and the average value is defined as the grinding roller bearing temperature value;

[0017] Based on the equipment usage parameters, the temperature value of the grinding roller bearing, and the pressure value corresponding to the temperature value of the grinding roller bearing, a corresponding fault classification label is set to generate a dataset to be processed.

[0018] Preferably, the dataset to be processed is subjected to dimensionality reduction using a piecewise aggregation approximation algorithm to generate dimensionality-reduced data, specifically as follows:

[0019] According to the formula The dataset to be processed is used to generate dimensionality-reduced data, where m and n are constants. Let B′ be the compression ratio, and B′ = (b′1, b′2, ..., b′). mThe result is the temperature of the roller bearing after dimensionality reduction, B = (b1, b2, ..., b n The result is the roller bearing temperature before dimensionality reduction.

[0020] Preferably, m is less than or equal to n, and m is divisible by n.

[0021] Preferably, the trained GBDT prediction model is used to process the dimensionality-reduced data to generate prediction results, specifically as follows:

[0022] According to the formula T={(x1,y1),(x2,y2),...,(x i ,y i ),...,(x N ,y N The dimensionality-reduced data is processed to generate a training dataset, where x i Refers to b′ i and p′ i ;

[0023] According to the formula The training dataset is initialized, where L(y,f(x)) is the loss function and c is a constant;

[0024] According to the formula Calculate the negative gradient of the loss function in the current GBDT prediction model, where i = 1, 2, ..., N, N is the number of samples, and r mi This represents the negative gradient of the loss function in the current GBDT prediction model.

[0025] For r mi Perform regression tree fitting to generate T m (x);

[0026] According to the formula The training dataset is used to generate prediction results, where m = 1, 2, ..., M, and M is the number of decision trees.

[0027] Preferably, before calling the trained GBDT prediction model to process the dimensionality-reduced data, the method further includes:

[0028] Retrieve historical data from the MSE system database to generate training and test sets;

[0029] The training set data is optimized, and a GBDT model is built. The depth of the regression tree and the number of iterations of the GBDT model are adjusted according to the optimized training set.

[0030] The adjusted GBDT model is tested using the test set, test results are generated, and it is determined whether the test results meet the standards.

[0031] When the test results are deemed to meet the standards, a GBDT prediction model is generated.

[0032] The present invention also discloses a grinding roller failure prediction device, comprising:

[0033] The data processing unit is used to filter the data in the MSE system database according to the equipment usage parameters and generate a dataset to be processed.

[0034] The data dimensionality reduction unit is used to perform dimensionality reduction processing on the dataset to be processed using a segmented aggregation approximation algorithm to generate dimensionality-reduced data.

[0035] The model prediction unit is used to call the trained GBDT prediction model to process the dimensional data to generate prediction results;

[0036] The early warning unit is used to output an early warning signal based on the prediction results.

[0037] The present invention also discloses a grinding roller fault prediction device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a grinding roller fault prediction method as described above.

[0038] The present invention also discloses a readable storage medium, characterized in that it stores a computer program, which can be executed by the processor of the device in which the storage medium is located, to implement a grinding roller fault prediction method as described in any of the above claims.

[0039] In summary, the grinding roller fault prediction method, device, equipment, and storage medium provided in this embodiment integrate the PAA algorithm and the GBDT algorithm into a PAA-GBDT algorithm. Starting from the specific causes of grinding roller faults related to bearing temperature and pressure, it performs fault prediction based on real-time data from the MES system. First, it filters characteristic parameters reflecting the working state of the grinding roller from the historical database (tbreport). Second, it processes the filtered bearing temperature and pressure data, then uses PAA to reduce the dimensionality of the bearing temperature and pressure dataset. Next, it sets fault labels based on the actual bearing temperature and pressure warning values. Finally, it uses GBDT to build a prediction model to predict grinding roller faults. The prediction results allow for early inspection of the grinding roller, preventing faults and improving tailings processing efficiency. This fusion algorithm aggregates data, resulting in data that better reflects the actual operating conditions of the grinding roller. Furthermore, PAA-GBDT has strong generalization ability, fast training speed, and prevents overfitting, thus providing better prediction of grinding roller faults. This solves the problems of existing grinding roller fault prediction methods, which cannot handle some special data, lack universality, and produce error results in actual use, which may lead to the system's inability to quickly issue fault warnings and thus damage the grinding roller. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating a grinding roller fault prediction method provided in an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the grinding roller data source provided in an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of the grinding roller provided in an embodiment of the present invention.

[0043] Figure 4 This is a schematic diagram of the data processing flow provided in an embodiment of the present invention.

[0044] Figure 5 This is a schematic diagram comparing the temperature of the grinding roller bearing before and after dimension reduction, provided in an embodiment of the present invention.

[0045] Figure 6 This is a schematic diagram comparing the pressure of the grinding rollers before and after dimension reduction, provided in an embodiment of the present invention.

[0046] Figure 7 This is a schematic diagram of the modeling process of the PAA-GBDT model provided in the embodiment of the present invention.

[0047] Figure 8 is a schematic diagram showing the changes of each sample set with the number of iterations at different tree depths provided in the embodiments of the present invention.

[0048] Figure 9This is a schematic diagram of a grinding roller fault prediction device provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Please see Figure 1 The first embodiment of the present invention provides a method for predicting grinding roller failures, which can be executed by a failure prediction device (hereinafter referred to as the prediction device), specifically by one or more processors within the prediction device, to achieve the following steps:

[0052] S101, Based on the equipment usage parameter indicators, filter the data in the MSE system database to generate a dataset to be processed;

[0053] Specifically, step S101 includes: querying the grinding roller component parameter information in the equipment historical data table of the MES system database according to the equipment usage parameter index, and exporting the grinding roller component parameter information to generate a grinding roller parameter file;

[0054] Data with empty values ​​in the grinding roller parameter file are filtered out, and feature variable data is generated, which includes multiple bearing temperature values ​​and multiple pressure values.

[0055] The feature variable data is mean-processed, and a fault classification label corresponding to the feature variable data is set to generate a dataset to be processed. Specifically, the average value of multiple bearing temperature values ​​is obtained, and the average value is defined as the grinding roller bearing temperature value.

[0056] Based on the equipment usage parameters, the temperature value of the grinding roller bearing, and the pressure value corresponding to the temperature value of the grinding roller bearing, a corresponding fault classification label is set to generate a dataset to be processed.

[0057] It should be noted that the prediction device can be a terminal device (such as a smartphone, smart computer or other smart device) or a server. The prediction device can establish a communication connection with the MES system to realize data interaction.

[0058] Currently available MES systems are crucial for tailings companies' equipment monitoring, fault early warning, and maintenance management. First, the equipment control system uses OPC (Object Linking and Embedding (OLE) for Process Control) data acquisition technology to obtain parameter information from grinding rollers, motors, and hydraulic stations in SPC (Statistical Process Control) and DMCS (Data Mapping Control System). Second, the central control system uses network interfaces to obtain parameters such as grinding roller bearing temperature and grinding roller pressure from the equipment control system. Element UI is a component library based on Vue.js 2.0 designed for developers, designers, and product managers. Finally, the parameter information from the central control system is transmitted to the MES system's MySQL database via Axios technology. Axios is a promise-based HTTP library that encapsulates network requests such as GET and POST. Using the Element UI component library, the Vue framework, and backend-provided interfaces, fault prediction, equipment early warning, and equipment maintenance modules are designed to visualize grinding roller faults, predict future failure rates, and view equipment maintenance records. Grinding roller data sources include... Figure 2 As shown. Among them, the processing of feature variables, model selection, and selection of fault warning values ​​of the grinding roller fault prediction model directly affect the accuracy of the model prediction. Therefore, the difficulty of grinding roller fault prediction lies in how to process a large amount of time series data, adjust model parameters, and apply them to the MES system.

[0059] The grinding roller is the component that comes into direct contact with the material. It consists of three parts: the grinding roller bearing, the grinding roller shaft, and the grinding roller sleeve. The grinding roller itself is as follows: Figure 3 As shown, 1 represents the grinding roller shaft, 2 represents the grinding roller bearing, and 3 represents the grinding roller sleeve. The grinding roller is mainly used for crushing and grinding. When the bearing temperature and pressure rise abnormally, it will cause damage to the grinding roller bearing, sleeve, and shaft core, which will further cause the grinding roller to stop and affect the tailings processing progress. In actual situations, due to errors by tailings equipment processing personnel, equipment network anomalies, etc., the historical data of the grinding roller obtained from the field may sometimes have missing fields or data redundancy. In this case, data optimization processing is required.

[0060] Please see Figure 4Specifically, in this embodiment, the grinding roller component parameter information in the historical data table (tbreport) of the equipment is first retrieved from the MES system database using a fuzzy query. The retrieved information is then exported and stored in the grinding roller parameter file. Next, the data in the grinding roller parameter file needs to be analyzed and statistically analyzed. Five data features are obtained, but due to network and personnel issues, some null values ​​may occur. Three of the five data features are null values, which do not contribute to grinding roller fault prediction and may even affect the accuracy of the prediction. Therefore, these three useless grinding roller parameters should be deleted so that the data can better reflect the actual operating condition of the grinding roller. The remaining two feature parameters are the grinding roller bearing temperature and pressure. When both values ​​rise abnormally, the lubricating oil will evaporate, causing the grinding roller to lose protection and increasing the risk of damage. Therefore, bearing temperature (B) and pressure (P) are selected as feature variables in the grinding roller fault prediction method.

[0061] In this embodiment, since the grinding roller bearing has three sensing monitoring points, there are three bearing temperatures, namely grinding roller bearing temperature 3 (B3), grinding roller bearing temperature 2 (B2), and grinding roller bearing temperature 1 (B1). To simplify the grinding roller bearing temperature (B), the three grinding roller bearing temperatures need to be summed and averaged, and the result is assigned to B. The calculation formula is B = (B3 + B2 + B1) / 3......(1), and the calculation results are shown in Table 1.

[0062] Table 1:

[0063]

[0064] Taking the normal temperature range of the grinding roller bearing as 80℃ and the normal pressure range of the grinding roller as 10MPa as an example, the parameter range is shown in Table 2. Based on the warning values ​​of each parameter, a fault classification label is set. When B3, B2, B1, and P are all within the normal range, the label value is 0, and in other cases, it is 1. 0 represents normal and 1 represents fault.

[0065] Table 2:

[0066]

[0067] Finally, the processed data is stored in Table 3 to generate the dataset to be processed.

[0068] Table 3:

[0069]

[0070] S102, the dimensionality reduction of the dataset to be processed is performed using a segmented aggregation approximation algorithm to generate dimensionality-reduced data;

[0071] Specifically, step S102 includes: according to the formula The dataset to be processed is used to generate dimensionality-reduced data, where m and n are constants. Let B′ be the compression ratio, and B′ = (b′1, b′2, ..., b′). m The result is the temperature of the roller bearing after dimensionality reduction, B = (b1, b2, ..., b n The result is the roller bearing temperature before dimensionality reduction; m is less than or equal to n, and m is divisible by n.

[0072] Please see Figure 5 Specifically, in this embodiment, taking a data sampling frequency of 30 seconds / time from the historical database as an example, the equipment operates continuously for months unless there is a fault. Therefore, the database contains a large amount of parameter data with relatively small fluctuations, resulting in a lot of redundant data and making data querying very time-consuming. The Piecewise Aggregate Approximation (PAA) algorithm is a feature extraction method based on the average value of segmented sequences. It can effectively return approximate aggregation results that satisfy the confidence interval, reducing data redundancy and simplifying data complexity. Compared with other dimensionality reduction algorithms, it can reduce the dimensionality of massive time-series data in a shorter time, exhibiting better performance advantages and better reflecting the actual operating condition of the grinding roller. Therefore, the grinding roller fault prediction method uses PAA to reduce the dimensionality of the grinding roller feature parameters. Its principle is as follows:

[0073] Let the temperature data of the grinding roller bearing be B = (b1, b2, ..., b n If the length of a given number is n, where n is 2880, and its length is reduced to m dimensions, where m is 200, it can be represented as B′=(b′1,b′2,...,b′). m Its expression is: Where m is less than or equal to n, and m must be divisible by n, compression ratio The time series dimensionality reduction process is shown, where B′ is the result after dimensionality reduction and B is the result before dimensionality reduction.

[0074] Please see Figure 6 Similarly, the processing of grinding roller pressure is done in the same way. Originally, there were 2880 data points for grinding roller bearing temperature and grinding roller pressure. After dimensionality reduction, this was reduced to 200, eliminating a large amount of redundant and invalid data. The time required for data access was also shortened accordingly, demonstrating the excellent performance of the PAA algorithm.

[0075] S103, The trained GBDT prediction model is called to process the dimensionality reduction data to generate prediction results;

[0076] Specifically, step S103 includes: according to the formula T={(x1,y1),(x2,y2),...,(x i ,y i ),...,(x N ,y N The dimensionality-reduced data is processed to generate a training dataset, where x i Refers to b′ i and p′ i ;

[0077] According to the formula The training dataset is initialized, where L(y,f(x)) is the loss function and c is a constant;

[0078] According to the formula Calculate the negative gradient of the loss function in the current GBDT prediction model, where i = 1, 2, ..., N, N is the number of samples, and r mi This represents the negative gradient of the loss function in the current GBDT prediction model.

[0079] For r mi Perform regression tree fitting to generate T m (x);

[0080] According to the formula The training dataset is used to generate prediction results, where m = 1, 2, ..., M, and M is the number of decision trees.

[0081] Specifically, in this embodiment, the dataset is divided into a training set and a test set. First, the parameters are optimized by continuously adjusting the tree depth and the number of iterations to obtain more accurate prediction results. Second, the prediction effect is tested using the test set. If the prediction effect does not meet the requirements, the tree depth and the number of iterations need to be readjusted until the requirements are met. Finally, the GBDT model and prediction results that meet the requirements are saved. GBDT is an iterative decision tree algorithm with strong robustness, powerful fitting ability, and generalization ability. It consists of multiple trees, each commonly referred to as a Classification and Regression Tree (CART). When the model makes predictions, each input feature will traverse all nodes of the decision tree, and each input feature will receive a corresponding predicted value. The final prediction result is the sum of the prediction results of each tree. Its core idea is to improve the accuracy of the prediction results iteratively in the negative gradient direction of the previously built model loss function. The calculation process is as follows:

[0082] First, the training dataset is given by the formula T = {(x1,y1),(x2,y2),...,(x...}. i ,yi ),...,(x N ,y N )}......(3), where x i Refers to b′ i and p′ i .

[0083] Secondly, initialization is performed, using the following formula: Where L(y,f(x))......(4) is the loss function, and c is a constant.

[0084] Finally, repeat the following steps, calculating r for i = 1, 2, ..., N. mi The specific calculation formula is as follows: Where r mi Let N be the negative gradient of the loss function in the current model, and N be the number of samples; for r mi Fit a regression tree to obtain T m (x); then according to the formula The update iterations are performed to obtain a regression tree, where m = 1, 2, ..., M, and M is the number of decision trees.

[0085] S104, based on the prediction result, output a warning signal.

[0086] Specifically, in this embodiment, when a potential equipment failure is predicted, a warning alarm will flash and notify maintenance personnel to go to the work site to inspect and try measures to prevent the failure from occurring.

[0087] In summary, the aforementioned grinding roll fault prediction method suffers from small fluctuations and a large volume of grinding roll bearing temperature and pressure data. Therefore, the PAA algorithm is introduced to remove redundant data. Furthermore, because the acquired data is continuous, linear, and time-series, and some grinding roll bearing temperature and pressure data are missing, and the GBDT algorithm is suitable for processing data with these characteristics, PAA and GBDT are integrated to better predict grinding roll faults. The specific implementation process is as follows:

[0088] First, the feature parameters are optimized according to the above data processing flow. Second, the processed grinding roller bearing temperature B and grinding roller pressure P are input into PAA. The grinding roller bearing temperature B and P are segmented and aggregated for approximate dimensionality reduction using formula (2) to obtain B′ and P′. Finally, B′, P′ and fault label are used as inputs to GBDT. Fault prediction of the grinding roller is achieved by formula (3) to formula (7) and by continuously adjusting the number of iterations and tree depth of the model.

[0089] Please see Figure 7In one possible embodiment of the present invention, before calling the trained GBDT prediction model to process the dimensionality-reduced data, the method further includes:

[0090] Retrieve historical data from the MSE system database to generate training and test sets;

[0091] The training set data is optimized, and a GBDT model is built. The depth of the regression tree and the number of iterations of the GBDT model are adjusted according to the optimized training set.

[0092] The adjusted GBDT model is tested using the test set, test results are generated, and it is determined whether the test results meet the standards.

[0093] When the test results are deemed to meet the standards, a GBDT prediction model is generated.

[0094] Specifically, in this embodiment, taking the modeling using historical data from a certain company as an example, the predictive performance of the PAA-GBDT model is trained and verified. The input features are grinding roller bearing temperature and grinding roller pressure, and the output features are grinding roller bearing temperature, grinding roller pressure, and grinding roller fault identifier. The experimental simulation tool is Python. The sample set used in the experiment consists of real data from a tailings treatment system from October to December 2021 and April 2022. The data for the three months of 2021 are grinding roller operating data, and the data for April 2022 are grinding roller non-operating data. The sample information is shown in Table 4.

[0095] The sample set was divided into a training set and a test set in a 7:3 ratio.

[0096] Table 4:

[0097] The PAA-GBDT algorithm involves numerous parameters in its modeling, and these parameters affect the model's prediction performance. Parameter tuning is essential to improving the model's prediction accuracy. The main parameters involved in the model are shown in Table 5. After adjustment, while keeping other parameters constant, the tree depth (max_depth) and the number of iterations (n_estimators) were changed. The prediction accuracy was compared with the number of iterations at different tree depths.

[0098] The tree depth was set to 1 to 5, and the maximum number of iterations was 30. The prediction results of each model were output, as shown in Figure 8. The prediction results show that the trends of the four sample sets are roughly the same. Sample set 1, with a tree depth of 5, can predict the fault situation well after 21 iterations. As the tree depth decreases, the number of iterations required increases accordingly. When the depth is 1, 30 iterations are still insufficient to achieve the desired prediction effect. Sample sets 2 and 3 are roughly the same as sample set 1. When the depth is 5, the number of iterations for sample sets 2, 3, and 4 are 22, 21, and 23, respectively. Similarly, when the tree depth is 1, 30 iterations for each sample set are also insufficient to achieve a good prediction effect. Therefore, the model depth should be set to 5, and the number of iterations should be 23.

[0099] Table 5:

[0100]

[0101] In this embodiment, all models use the same input features, training set, and test set, and all parameters are optimal, exhibiting good generalization ability. Figure 8 compares the prediction results of each model across sample sets 1 to 4, demonstrating their prediction accuracy. The comparison shows that the training set results of each model generally have an accuracy of over 90%, indicating that each model can be used for grinding roll fault prediction. However, PAA-GBDT has a higher accuracy compared to other models: 99.3% for sample 1, 99.4% for sample 2, 98.6% for sample 3, and 99.4% for sample 4. The PAA-GBDT model performs best across the four sample sets, demonstrating its ability to effectively predict grinding roll faults, thus contributing to fault prevention and control and promoting tailings production.

[0102] In summary, in practical applications, the equipment management module of the MES system obtains the grinding roller parameter operation information from a certain MES system backend database through the Swagger interface. Using Element UI and AXIOS technology, the PAA-GBDT prediction model is integrated into the MES system as a sub-module for equipment management, encompassing fault prediction, equipment early warning, and equipment maintenance. The fault prediction sub-module predicts future grinding roller failures. If a fault is predicted, an alarm flashes, and maintenance personnel are notified to go to the work site to inspect and try preventative measures. If the maintenance personnel cannot handle the issue, the information is uploaded to a higher level for expert inspection and maintenance. The maintenance information is recorded in the equipment maintenance sub-module, which improves equipment maintenance speed, equipment utilization, and tailings processing efficiency.

[0103] In short, the proposed grinding roll failure prediction method starts from the key parameters affecting the grinding roll and combines the temporal characteristics of the grinding roll parameters to propose a grinding roll failure prediction model integrating PAA and GBDT. By comparing the prediction results of the GBDT model and the XGBoost model, it can be seen that the model integrating PAA and GBDT has better prediction performance and can predict grinding roll failures. When the algorithm is applied to the MES system, it can predict future grinding roll failures, enabling early inspection of the grinding rolls, avoiding failures, improving equipment utilization, promoting the efficiency of tailings treatment, and driving the economic development of tailings companies.

[0104] Please see Figure 9 A second embodiment of the present invention provides a grinding roller fault prediction device, comprising:

[0105] Data processing unit 201 is used to filter data in the MSE system database according to equipment usage parameter indicators and generate a dataset to be processed;

[0106] The data dimensionality reduction unit 202 is used to perform dimensionality reduction processing on the dataset to be processed using a segmented aggregation approximation algorithm to generate dimensionality-reduced data.

[0107] The model prediction unit 203 is used to call the trained GBDT prediction model to process the dimensional data to generate prediction results;

[0108] The early warning unit 204 is used to output an early warning signal based on the prediction result.

[0109] A third embodiment of the present invention provides a grinding roller fault prediction device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a grinding roller fault prediction method as described in any of the above embodiments.

[0110] A fourth embodiment of the present invention provides a readable storage medium, characterized in that it stores a computer program that can be executed by a processor of the device in which the storage medium is located, to implement a grinding roller fault prediction method as described in any of the above claims.

[0111] Exemplary examples show that the computer program described in the third and fourth embodiments of the present invention can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in implementing a grinding roller fault prediction device. For example, the apparatus described in the second embodiment of the present invention.

[0112] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the aforementioned grinding roller fault prediction method, connecting various parts of the method via various interfaces and lines.

[0113] The memory can be used to store the computer program and / or modules. The processor implements various functions of a grinding roller fault prediction method by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, text conversion function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0114] If the implemented module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0115] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0116] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions that fall within the scope of the present invention are within the scope of protection of the present invention.

Claims

1. A method for predicting grinding roller failures, characterized in that, include: The data in the MSE system database is filtered based on the equipment usage parameters to generate a dataset to be processed. The dataset to be processed is dimensionality reduced using a segmented aggregation approximation algorithm to generate dimensionality-reduced data. The trained GBDT prediction model is invoked to process the dimensionality-reduced data to generate prediction results; Based on the prediction results, an early warning signal is output; The data in the MSE system database is filtered based on equipment usage parameters to generate a dataset to be processed, specifically: Based on the equipment usage parameter indicators, query the grinding roller component parameter information in the equipment historical data table of the MES system database, and export the grinding roller component parameter information to generate a grinding roller parameter file; Data with empty values ​​in the grinding roller parameter file are filtered out, and feature variable data is generated, which includes multiple bearing temperature values ​​and multiple pressure values. The feature variable data is mean-processed, and a fault classification label corresponding to the feature variable data is set to generate a dataset to be processed; The dataset to be processed is subjected to dimensionality reduction using a piecewise aggregation approximation algorithm to generate dimensionality-reduced data, specifically as follows: According to the formula The dataset to be processed is used to generate dimensionality-reduced data, where m and n are constants. The compression ratio is... The results show the temperature of the roller bearing after dimensionality reduction. The results show the roller bearing temperature before dimensionality reduction.

2. The method for predicting grinding roller failure according to claim 1, characterized in that, The feature variable data is processed by setting fault classification labels corresponding to the feature variable data, and a dataset to be processed is generated, specifically as follows: The average value of the multiple bearing temperature values ​​is obtained, and the average value is defined as the grinding roller bearing temperature value; Based on the equipment usage parameters, the temperature value of the grinding roller bearing, and the pressure value corresponding to the temperature value of the grinding roller bearing, a corresponding fault classification label is set to generate a dataset to be processed.

3. The method for predicting grinding roller failure according to claim 1, characterized in that, m is less than or equal to n, and m is divisible by n.

4. The method for predicting grinding roller failure according to claim 1, characterized in that, The trained GBDT prediction model is invoked to process the dimensionality-reduced data to generate prediction results, specifically as follows: According to the formula The dimensionality-reduced data is processed to generate a training dataset, wherein, Reference and ; According to the formula The training dataset is initialized, wherein, For loss function, It is a constant; According to the formula Calculate the negative gradient of the loss function in the current GBDT prediction model, where, N is the number of samples. The negative gradient of the loss function is the value of the current GBDT prediction model; right Perform fitting regression tree processing to generate ; According to the formula The training dataset is used to perform calculations and generate prediction results, wherein, And M is the number of decision trees.

5. The method for predicting grinding roller failure according to claim 1, characterized in that, Before calling the trained GBDT prediction model to process the dimensionality-reduced data, the following steps are also included: Retrieve historical data from the MSE system database to generate training and test sets; The training set data is optimized, and a GBDT model is built. The depth of the regression tree and the number of iterations of the GBDT model are adjusted according to the optimized training set. The adjusted GBDT model is tested using the test set, test results are generated, and it is determined whether the test results meet the standards. When the test results are deemed to meet the standards, a GBDT prediction model is generated.

6. A grinding roller fault prediction device, characterized in that, include: The data processing unit is used to filter the data in the MSE system database according to the equipment usage parameters and generate a dataset to be processed. The data dimensionality reduction unit is used to perform dimensionality reduction processing on the dataset to be processed using a segmented aggregation approximation algorithm to generate dimensionality-reduced data. The model prediction unit is used to call the trained GBDT prediction model to process the dimensional data to generate prediction results; An early warning unit is used to output an early warning signal based on the prediction result; The data in the MSE system database is filtered based on equipment usage parameters to generate a dataset to be processed, specifically: Based on the equipment usage parameter indicators, query the grinding roller component parameter information in the equipment historical data table of the MES system database, and export the grinding roller component parameter information to generate a grinding roller parameter file; Data with empty values ​​in the grinding roller parameter file are filtered out, and feature variable data is generated, which includes multiple bearing temperature values ​​and multiple pressure values. The feature variable data is mean-processed, and a fault classification label corresponding to the feature variable data is set to generate a dataset to be processed; The dataset to be processed is subjected to dimensionality reduction using a piecewise aggregation approximation algorithm to generate dimensionality-reduced data, specifically as follows: According to the formula The dataset to be processed is used to generate dimensionality-reduced data, where m and n are constants. The compression ratio is... The results show the temperature of the roller bearing after dimensionality reduction. The results show the roller bearing temperature before dimensionality reduction.

7. A grinding roller fault prediction device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements a grinding roller failure prediction method as described in any one of claims 1 to 5.

8. A readable storage medium, characterized in that, The storage medium contains a computer program that can be executed by a processor of the device in which the storage medium is located, to implement a grinding roller failure prediction method as described in any one of claims 1 to 5.

Images