Machining working condition state diagnosis system and method based on machining information

Abstract

The invention belongs to the technical field of machining state diagnosis, and discloses a machining condition state diagnosis system and method based on machining information, and the system comprises a data input, preprocessing and analysis module, a machining condition state diagnosis module and a diagnosis result visualization module. The data input, preprocessing and analysis module is used for inputting a processing data set generated in the historical operation process of the processing equipment and preprocessing and analyzing the processing data set; the processing working condition state diagnosis module is used for training and testing a random forest algorithm model by utilizing the preprocessed and analyzed processing data set based on a principal component analysis method and a particle swarm optimization algorithm to obtain a trained processing working condition state diagnosis model; and the diagnosis result visualization module is used for drawing a diagnosis result graph and displaying the diagnosis result graph in a visualization operation interface. According to the method, the machining working condition state of the machining equipment is efficiently and accurately monitored and diagnosed by utilizing the machining data generated in the machining process of the machining equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of machining state diagnosis, and particularly relates to a machining condition state diagnosis system and method based on machining information. Background Art

[0002] Machining condition state diagnosis plays a crucial role in modern manufacturing. By real-time monitoring and analyzing key machining information during the machining process, such as voltage, current, and electric power, potential abnormalities and faults can be detected in a timely manner, thereby avoiding the accumulation and expansion of machining errors, ensuring that the machined workpieces meet the design requirements, and improving product quality.

[0003] At the same time, accurate machining condition state diagnosis can also provide a basis for optimizing process parameters, helping to adjust cutting speed, feed rate, spindle speed, etc., to improve production efficiency and reduce machining costs. In addition, it also has important significance in equipment maintenance, diagnosing the wear trend and fault risk of equipment to guide a reasonable maintenance plan, extend the service life of equipment, and ensure the continuity and stability of production. In the prior art, it is necessary to manually detect the machining equipment regularly for machining condition diagnosis. This way of the prior art has problems such as low diagnosis accuracy and low diagnosis efficiency. Summary of the Invention

[0004] The purpose of the present invention is to effectively utilize the machining data generated by the machining equipment during the machining process, monitor and diagnose the machining condition state of the machining equipment, that is, the operating state of the machining equipment, so as to discover potential faults in advance, reduce the downtime of the machining equipment, improve production efficiency, and reduce maintenance costs.

[0005] In a first aspect, an embodiment of the present invention provides a machining condition state diagnosis system based on machining information. The system includes: a data input, preprocessing and analysis module, a machining condition state diagnosis module, and a diagnosis result visualization module;

[0006] The data input, preprocessing and analysis module is used to input the machining data set generated during the historical operation of the machining equipment, preprocess and analyze the machining data set. Among them, the ways of preprocessing the machining data set include: missing value processing, outlier processing, duplicate value removal processing, normalization processing, and transpose processing;

[0007] The machining condition state diagnosis module is used to train and test a random forest algorithm model based on the principal component analysis method and the particle swarm optimization algorithm, using the preprocessed and analyzed machining data set, to obtain a trained machining condition state diagnosis model;

[0008] The diagnostic result visualization module is used to draw a diagnostic result graph and display the diagnostic result graph in the visualization operation interface; the drawn diagnostic result graph includes: diagnostic errors corresponding to different key hyperparameters of the processing condition state diagnosis model; diagnostic accuracies corresponding to the training process and the testing process of the processing condition state diagnosis model, and classification accuracies corresponding to different classification tasks respectively.

[0009] Optionally, the data entry, preprocessing, and analysis module enters the processing data set generated during the historical operation process of the processing equipment, specifically including:

[0010] Design a visualization operation interface for importing processing data based on the Matlab platform. Among them, the design of the visualization operation interface needs to establish a connection channel between the local file system and the Matlab platform based on the Matlab language;

[0011] Receive the user's selection operation for the processing data file saved in the local file system, respond to the selection operation, read the processing data file, and enter the processing data set in the processing data file.

[0012] Optionally, the data entry, preprocessing, and analysis module preprocesses the processing data set, specifically including:

[0013] Detect whether there are missing values in the processing data set. If there are missing values in the processing data set, calculate the average value of all processing data in the column where the missing value is located in the processing data set, and use the average value to repair the missing value;

[0014] Detect whether there are outliers in the processing data set. If there are outliers in the processing data set, use the average value of the adjacent values of the outliers to replace the outliers;

[0015] Detect whether there are completely duplicate data in the processing data set. If there are completely duplicate data in the processing data set, delete the redundant data in the completely duplicate data;

[0016] Use the normalization method to linearly map the processing data set to the target interval;

[0017] Transpose the processing data set; among them, the transposed processing data set is compatible with the input format of the random forest algorithm model.

[0018] Optionally, the data entry, preprocessing, and analysis module analyzes the first processing data, specifically including:

[0019] Analyze the total number of data in the preprocessed processing dataset, the processing condition status, and the representative numbers corresponding to each type of processing condition status to obtain the first data analysis result, where the processing condition status includes the running state, the standby state, and the warning state;

[0020] Extract the key data features of the preprocessed processing dataset and the quantity of the key data features to obtain the second data analysis result; the key data features include: phase A voltage, phase A current, phase B voltage, phase B current, phase C voltage, phase C current, total power, spindle speed.

[0021] Display the first data analysis result and the second data analysis result on the visualization operation interface.

[0022] Optionally, the processing condition status diagnosis module includes: a dataset division sub-module, a PCA sub-module, a PSO sub-module, and an RF sub-module;

[0023] Dataset division sub-module: Divide the preprocessed and analyzed processing dataset into a training dataset and a test dataset according to the target ratio;

[0024] PCA sub-module, which is used to process the training dataset and the test dataset by using the principal component analysis method. Through coordinate transformation, under the principle of maximizing variance, project the training data and the test data onto a new coordinate to obtain the processed training data and test data;

[0025] PSO sub-module, which is used to continuously iterate the key hyperparameters of the random forest algorithm model by using the particle swarm optimization algorithm until the optimal key hyperparameters of the random forest algorithm model are obtained;

[0026] RF sub-module: Use the optimal key hyperparameters to construct a random forest algorithm model, and train and test the improved random forest algorithm model by using the processed training data and test data.

[0027] Optionally, the processing dataset is the processing data collected by a multi-source data acquisition platform from multiple processing devices of key components of an electro-hydraulic control actuator.

[0028] Optionally, the system further includes: a user permission and security management module;

[0029] The permission and security management module is used to control the access permission of users to the processing condition status diagnosis system, where the access permission includes the login system permission and the password modification permission. Unregistered users need to register, obtain the login system permission after successful registration, and display the human-computer interaction interface of the processing condition status diagnosis system after the user successfully logs in.

[0030] Optionally, the data entry, preprocessing, and analysis module is further configured to enter the current processing data generated during the current operation of the processing equipment, preprocess and analyze the current processing data, and obtain the preprocessed and analyzed current processing data;

[0031] The processing condition state diagnosis module is configured to input the preprocessed and analyzed current processing data into the trained processing condition state diagnosis model to obtain the current processing condition state of the processing equipment;

[0032] The diagnosis result visualization module is configured to draw a diagnosis result graph based on the current processing condition state and display the drawn diagnosis result graph in the visualization operation interface.

[0033] In a second aspect, an embodiment of the present invention provides a processing condition state diagnosis method based on processing information, which is applied to a processing condition state diagnosis system based on processing information. The system includes a data entry, preprocessing, and analysis module, a processing condition state diagnosis module, and a diagnosis result visualization module. The method includes:

[0034] The data entry, preprocessing, and analysis module enters the processing data set generated during the historical operation of the processing equipment, and preprocesses and analyzes the processing data set. Among them, the methods for preprocessing the processing data set include: missing value processing, outlier processing, duplicate value clearing processing, normalization processing, and transposition processing;

[0035] The processing condition state diagnosis module trains and tests the random forest algorithm model by using the preprocessed and analyzed processing data set based on the principal component analysis method and the particle swarm optimization algorithm to obtain the trained processing condition state diagnosis model;

[0036] The diagnosis result visualization module draws a diagnosis result graph and displays the diagnosis result graph in the visualization operation interface; the drawn diagnosis result graph includes: the diagnosis errors corresponding to different key hyperparameters of the processing condition state diagnosis model; the diagnosis accuracies corresponding to the training process and the testing process of the processing condition state diagnosis model, and the classification accuracies corresponding to different classification tasks.

[0037] In a third aspect, an embodiment of the present invention provides an electronic device installed with the processing condition state diagnosis system based on processing information described in the first aspect, including:

[0038] At least one processor;

[0039] A memory for storing instructions executable by the at least one processor;

[0040] Wherein, the at least one processor is configured to execute the instructions to run the processing condition state diagnosis system based on processing information described in the first aspect.

[0041] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to run the processing condition state diagnosis system based on processing information described in the first aspect.

[0042] For the processing condition state diagnosis system provided by the embodiment of the present invention, a data entry, preprocessing and analysis module enters a processing data set generated during the historical operation of a processing device, and preprocesses and analyzes the processing data set; a processing condition state diagnosis module is configured to train and test a random forest algorithm model by using the preprocessed and analyzed processing data set based on the principal component analysis method and the particle swarm optimization algorithm to obtain a trained processing condition state diagnosis model; a diagnosis result visualization module draws a diagnosis result graph and displays the diagnosis result graph in a visualization operation interface. It can be seen that the embodiment of the present invention effectively utilizes the processing data generated during the processing of a processing device, efficiently and accurately monitors and diagnoses the processing condition state of the processing device, that is, the operating state of the processing device, so as to discover potential faults in advance, reduce the downtime of the processing device, improve production efficiency, and reduce maintenance costs. Description of the Drawings

[0043] Figure 1 It is a schematic diagram of the overall structure of a processing condition state diagnosis system based on processing information provided by an embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of the structure of another processing condition state diagnosis system based on processing information provided by an embodiment of the present invention;

[0045] Figure 3 It is a schematic diagram of the working process of a processing condition state diagnosis module provided by an embodiment of the present invention;

[0046] Figure 4 It is a schematic diagram of the structure of the human-computer interaction interface of a processing condition state diagnosis system provided by an embodiment of the present invention;

[0047] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed Embodiments

[0048] The present invention will be described in detail below through embodiments.

[0049] Processing status diagnosis plays a crucial role in modern manufacturing. By real-time monitoring and analyzing key processing information during the processing, such as voltage, current, and electric power, potential abnormalities and faults can be detected in a timely manner, thereby avoiding the accumulation and expansion of processing errors, ensuring that the processed workpieces meet the design requirements, and improving product quality. At the same time, accurate processing status diagnosis can also provide a basis for optimizing process parameters, helping to adjust cutting speed, feed rate, etc., to improve production efficiency and reduce processing costs. In addition, it is also of great significance in equipment maintenance, diagnosing the wear trend and fault risk of equipment to guide a reasonable maintenance plan, extend the service life of the equipment, and ensure the continuity and stability of production.

[0050] The application of machine learning in the field of processing status diagnosis has greatly improved the accuracy and efficiency of diagnosis. Through deep learning from a large amount of historical data and real-time processing information input by real-time sensors, machine learning algorithm models can automatically identify and classify various states during the processing. It can not only reduce the complexity and subjectivity of manual feature extraction but also handle complex non-linear relationships and multi-variable interactions. In addition, machine learning can also perform anomaly detection, timely discover abnormal situations during the processing, and thus take preventive measures to avoid production accidents. This intelligent diagnosis method not only improves processing quality but also optimizes the production process, reduces maintenance costs, and provides strong technical support for the intelligent upgrading of the manufacturing industry.

[0051] To rationally utilize the processing information generated during the processing, accurately monitor the processing status of the equipment, and improve the intelligent level of the machine tool, the embodiment of the present invention proposes a processing condition status diagnosis method and system based on processing information. This system mainly based on the processing information of the processing equipment system collected on the integrated multi-source data acquisition platform, adopts the principal component analysis-particle swarm optimization algorithm-random forest algorithm model as the core diagnosis algorithm model to accurately diagnose the processing status of the processing equipment. By using this system, not only can the operation status of the processing equipment be monitored and diagnosed efficiently and accurately, thereby discovering potential faults in advance, reducing equipment downtime, but also the production efficiency can be improved, the maintenance cost can be reduced, and the comprehensive competitiveness of the enterprise in the fierce market competition can be enhanced.

[0052] For the sake of clear and complete description of the solution, the overall technical solution of the embodiment of the present invention will be elaborated in detail below.

[0053] As Figure 1 shown, a processing condition status diagnosis system based on processing information provided by the embodiment of the present invention may include the following modules:

[0054] The user permission and security management module is used to ensure the security of the system. By establishing login permission settings, it controls the access permissions of users to the system to safeguard system security.

[0055] Among them, the user permission and security management module can manage the login permissions of users, control the access permissions of users to the processing condition status diagnosis system. Unregistered users need to register by entering a username and password. And to avoid errors when users enter passwords, the system requires users to enter the password again for confirmation. After entering the username, password, and password confirmation, the system login permission can be obtained. Then, users can enter the username and password credentials authenticated and authorized by the system through the identity authentication module to successfully log in to the human-machine interaction interface of the processing condition status diagnosis system.

[0056] Among them, the access permissions can include the system login permission and the password modification permission.

[0057] The system login permission is used to control that users can access the processing condition status diagnosis system, manually import the processing data set of the processing equipment, so that the processing condition status diagnosis module can accurately diagnose the processing condition status of the processing equipment, and the diagnosis result visualization module can draw the diagnosis result graph.

[0058] The password modification permission controls that users can modify the password credentials that have been authenticated and authorized by the system to provide a convenient and secure password management service.

[0059] The data entry, preprocessing, and analysis module is used to improve the data quality and visualization degree, support users to import the processing data set of the processing equipment and perform data preprocessing and data analysis on it.

[0060] Among them, the processing information data set of the processing equipment is the processing data collected by the multi-source data acquisition platform from multiple processing equipment of the key components of the electro-hydraulic control actuator.

[0061] The key components of the electro-hydraulic control actuator include pistons, piston rods, flanges, cylinder heads, cylinder blocks, and rod barrels, etc.

[0062] The processing equipment can include CNC lathes, CNC milling machines, drilling machines, and boring machines, etc.

[0063] The processing data can include phase A voltage, phase A current, phase B voltage, phase B current, phase C voltage, phase C current, total power, spindle speed, and specific processing condition status. Among them, the specific processing condition status includes the running state, standby state, and warning state.

[0064] For the sake of clear description of the solution, the main execution steps of the data entry, preprocessing, and analysis module will be introduced in detail in the following embodiments.

[0065] The processing condition state diagnosis module is used to adopt the principal component analysis - particle swarm optimization algorithm - random forest algorithm model as the core diagnosis algorithm model to accurately diagnose the processing condition state of the processing equipment.

[0066] Among them, the processing condition state diagnosis module may include the following sub - modules: the data set division sub - module, the PCA sub - module, the PSO sub - module, and the RF sub - module.

[0067] The data set division sub - module is used to divide the processed processing data set after data entry, pre - processing, and analysis module into a training data set and a test data set according to a ratio of 4:1.

[0068] The PCA sub - module is used to reduce the dimension of the training data set and the test data set by using the principal component analysis method. Through coordinate transformation, following the principle of maximum variance, the training data set and the test data set are projected onto new coordinates, so as to compress the features of multiple dimensions of the training data set and the test data set into lower dimensions. While trying to retain the original features, fewer features are selected to replace the original features, which can effectively reduce the redundancy of data. The ultimate goal is to analyze the training data set and the test data set with fewer main features. Moreover, the features after dimension reduction are not simply some features reduced from the original features, but a fusion of the original features, and finally the purpose of containing most of the original data features with fewer features is achieved.

[0069] The PSO sub - module: uses the particle swarm optimization algorithm (PSO) to optimize the key hyperparameters of the random forest algorithm model, such as the number of decision trees, the depth of decision trees, the minimum number of leaf nodes, etc. PSO can utilize individual experience and group experience to guide the search process. By continuously iteratively updating the positions of particles (i.e., combinations of key hyperparameters), the performance of the random forest algorithm model is gradually improved, effectively avoiding local optima, and improving the generalization ability and diagnostic accuracy of the random forest algorithm model. In addition, the high global search ability of PSO enables it to quickly find the optimal combination of key hyperparameters of the random forest algorithm, reducing the complexity and time cost of manual parameter tuning.

[0070] RF sub-module: After using the particle swarm optimization algorithm to find the optimal combination of the key hyperparameters of the random forest algorithm, the optimal combination of the key hyperparameters is used to construct a random forest algorithm model. Then, the training data set processed by the principal component analysis method is input into the improved random forest model for training. Finally, after the improved random forest algorithm model is trained, the random forest algorithm model is tested using the test data set. After the test data set is input into the random forest algorithm model, the processing condition state corresponding to each test data can be output. Finally, the accuracy rates of the random forest algorithm model on the training data set and the test data set are calculated to evaluate the diagnostic performance of the random forest algorithm model.

[0071] The diagnostic result visualization module is used to improve the convenience of data visualization and display, and draw diagnostic result graphs. Among them, the diagnostic result graphs include: error curve graphs, diagnostic result comparison graphs of the training data set, diagnostic result comparison graphs of the test data set, confusion matrices of the training data set, and confusion matrices of the test data set, and output them to the corresponding coordinate areas, that is, display them in the visualization operation interface.

[0072] Specifically, the diagnostic result visualization module can provide the function of drawing diagnostic result graphs. The diagnostic result graphs include error curve graphs, diagnostic result comparison graphs of the training set, diagnostic result comparison graphs of the test set, confusion matrices of the training set, and confusion matrices of the test set. The system can draw the diagnostic result graphs and output them to the corresponding coordinate areas to improve the convenience of data visualization and display. In the error curve graph, the key hyperparameters of the optimized random forest algorithm model, such as the number of decision trees, can be seen. Within a certain range, as the number of decision trees increases, the trend of the corresponding diagnostic error can be seen.

[0073] The specific diagnostic accuracies of the training data set diagnosis and the test data set diagnosis can be seen in the diagnostic result comparison graph of the training data set and the diagnostic result comparison graph of the test data set respectively; in the confusion matrices of the training data set and the test data set, it can be seen that in the classification tasks of the random forest algorithm model for the three processing condition states in the training data set and the test data set respectively, the corresponding numbers of correct classifications and incorrect classifications and their specific accuracies. Among them, the three processing condition states are the running state, the standby state, and the warning state.

[0074] A processing condition state diagnosis system and method based on processing information provided by an embodiment of the present invention are used to accurately diagnose the processing condition state of a processing device. Its main beneficial effects and advantages can include the following aspects:

[0075] 1. Precise monitoring and fault warning. By means of an integrated multi-source data acquisition platform, multi-source processing data of processing equipment is collected and processed, and a principal component analysis - particle swarm optimization algorithm - random forest algorithm model is adopted to efficiently and accurately monitor and predict the operating status of the processing equipment, and potential faults can be discovered in advance. By real-time analyzing data such as voltage, current, and power during the processing of key components of the electro-hydraulic control actuator, abnormal signals of the processing equipment can be captured in a timely manner, effectively reducing the downtime of the processing equipment and ensuring production continuity.

[0076] 2. Improve production efficiency. Accurate diagnosis of the processing condition status provides a basis for optimizing process parameters. According to the diagnosis results, parameters such as cutting speed and feed rate can be reasonably adjusted to avoid low processing efficiency caused by poor equipment status, thereby improving the overall production efficiency and enabling more qualified products to be produced per unit time.

[0077] 3. Reduce maintenance costs. By knowing the potential faults of the processing equipment in advance, a more reasonable maintenance plan can be formulated, changing passive maintenance to active maintenance. Unnecessary emergency repairs are reduced, maintenance costs and equipment wear are lowered, the service life of the equipment is extended, and additional losses such as workpiece scrapping caused by equipment failures are avoided.

[0078] 4. Improve system usability and security. The user permission and security management module ensures system security. By setting functions such as login permissions and password modification, unauthorized access and malicious tampering are prevented. At the same time, the system provides a visual operation interface, facilitating users to import data, perform diagnostic operations, and view diagnostic results, reducing the operation difficulty and improving the user experience.

[0079] 6. Intuitively display diagnostic results. The diagnostic result visualization module displays diagnostic results in an intuitive graphical way by drawing error curve graphs, comparison graphs of diagnostic results of the training set and test set, confusion matrices, etc. Users can clearly understand the model performance and the classification of processing condition status, facilitating quick decision-making.

[0080] In summary, the processing condition status diagnosis system and method based on processing information provided by the present invention show significant advantages in multiple aspects. Through a precise monitoring and fault warning mechanism, the equipment downtime is effectively reduced, ensuring production continuity; it provides a basis for optimizing process parameters, thereby improving production efficiency; a reasonable maintenance plan is planned to reduce maintenance costs; strong data management capabilities ensure data quality and provide strong support for decision-making; perfect security management and user-friendly design improve system security and user experience; the optimized random forest algorithm model improves diagnostic accuracy; intuitive visual results facilitate users' quick decision-making. This system can fully meet the needs of modern manufacturing for the status diagnosis of processing equipment, enhance the comprehensive competitiveness of enterprises in the fierce market competition, and is of great significance for promoting the intelligent upgrading of the manufacturing industry.

[0081] After elaborating on the overall technical solution of the embodiments of the present invention in detail, the processing condition state diagnosis system based on processing information provided by the embodiments of the present invention will be elaborated in detail below.

[0082] As Figure 2 As shown in the figure, a processing condition state diagnosis system 200 based on processing information provided by an embodiment of the present invention may include the following modules, namely: a data entry, preprocessing, and analysis module 201, a processing condition state diagnosis module 202, and a diagnosis result visualization module 203;

[0083] The data entry, preprocessing, and analysis module 201 is used to enter the processing data set generated during the historical operation of the processing equipment and preprocess and analyze the processing data set.

[0084] Among them, the processing data set is the processing data collected by the multi-source data acquisition platform from multiple processing equipment of key components of the electro-hydraulic control actuator. The ways of preprocessing the processing data set include: missing value processing, outlier processing, duplicate value clearing processing, normalization processing, and transpose processing.

[0085] Among them, the data entry function in the data entry, preprocessing, and analysis module supports users to import the processing data set of the processing equipment by means of manual input.

[0086] As an implementation manner of an embodiment of the present invention, the data entry, preprocessing, and analysis module enters the processing data set generated during the historical operation of the processing equipment, which may specifically include the following two steps, namely step a1 and step a2:

[0087] Step a1, design a visualization operation interface for importing processing data based on the Matlab platform. Among them, the design of the visualization operation interface needs to establish a connection channel between the local file system and the Matlab platform based on the Matlab language.

[0088] Step a2, receive the selection operation of the user for the processing data file saved in the local file system, respond to the selection operation, read the processing data file, and enter the processing data set in the processing data file.

[0089] Specifically, the data entry, preprocessing, and analysis module designs a visualization operation interface for importing data based on the Matlab platform. Users can manually import the processing data set of the processing equipment through this visualization operation interface. Among them, the design of this visualization operation interface needs to establish a connection channel between the local file system and the Matlab platform based on the Matlab language, allowing users to browse the local file system, select the Excel file saved with the processing data set of the processing equipment, and read the content of the selected Excel file as the original processing data set.

[0090] For the data entry function, the following operations can be performed using the Matlab language:

[0091] [file,path] = uigetfile({'*.xlsx;*.xls','Excel file (*.xlsx,*.xls)';

[0092] '*.*','All files (*.*)'},'Select the Excel file to be imported');

[0093] fullPath = Build_complete_file_path(path,file);

[0094] raw_data = readtable(fullPath);

[0095] Among them, Build_complete_file_path is a custom function, whose main function is to build a complete file path, and dataTable = readtable(fullPath); is to use the readtable function to read the data of the Excel file and store the data in the table data type raw_data as the original processed data set.

[0096] As an implementation of the embodiment of the present invention, the data entry, preprocessing, and analysis module preprocesses the processed data set, which may specifically include the following steps, namely steps b1 to b5:

[0097] Step b1, detect whether there are missing values in the processed data set. If there are missing values in the processed data set, calculate the average value of all the processed data in the column where the missing values are located, and use the average value to repair the missing values.

[0098] Step b2, detect whether there are outliers in the processed data set. If there are outliers in the processed data set, use the average value of the adjacent values of the outliers to replace the outliers.

[0099] Step b3, detect whether there are completely duplicate data in the processed data set. If there are completely duplicate data in the processed data set, delete the redundant data in the completely duplicate data.

[0100] Step b4, use the normalization method to linearly map the processed data set to the target interval.

[0101] Step b5, transpose the processed data set. Among them, the transposed processed data set is compatible with the input format of the random forest algorithm model.

[0102] Specifically, the data preprocessing function in the data entry, preprocessing, and analysis module performs operations such as missing value processing, outlier processing, duplicate value removal, normalization, and transposition on the original processed data set. By correcting data defects and optimizing data structures, it can not only significantly improve data quality and analysis reliability, but also prevent large errors in the random forest algorithm model caused by incomplete information, thereby providing a standardized and highly reliable data basis for subsequent modeling, mining, and diagnosis.

[0103] For the missing value processing in the data preprocessing function, the following Matlab language can be used to perform the operation:

[0104]

[0105] Among them, "missingValues = ismissing(raw_data);" is to use the ismissing function to judge whether each element in the original data set raw_data is a missing value and store the logical judgment result in missingValues. "raw_data = fillmissing(raw_data,'movmean', 'EndValues', 'nearest');" is to use the fillmissing function to replace the missing values with the mean value of the column with missing values in the original data set.

[0106] For the outlier processing in the data preprocessing function, the following Matlab language can be used to perform the operation:

[0107]

[0108] Among them, dataNumeric stores all column data except the last column extracted from the original data raw_data, mean_value stores the mean value of each column in dataNumeri, standard_deviation is a custom function whose main function is to calculate the standard deviation sigma of each column in dataNumeric, zScores stores the absolute value of the Z-score of each data point, "outliers = zScores > threshold;" is the logic for determining outliers, and "column(outliersInColumn) = meanNeighbor(outliersInColumn);" is to replace the outliers with the mean value of adjacent values.

[0109] For the duplicate value removal processing in the data preprocessing function, the following Matlab language can be used to perform the operation:

[0110] raw_data = unique(raw_data, 'rows','stable');

[0111] Among them, "rows" is for comparison by row, that is, only when all elements of a certain row are exactly the same as those of other rows is it regarded as a duplicate. "stable" is to ensure that the order of the final result is the same as the order in which these rows first appear in the original data, rather than the default sorting result.

[0112] For the normalization process in the data preprocessing function, the following operations can be performed through Matlab language:

[0113] [normalized_input_train, input_norm_setting] = mapminmax(input_train, 0, 1);

[0114] normalized_input_test = mapminmax('apply', input_test, input_norm_setting);

[0115] normalized_output_train = output_train;

[0116] Among them, normalized_input_train is the training data set after maximum-minimum normalization, and normalized_input_test is the test data set after maximum-minimum normalization.

[0117] For the transpose process in the data preprocessing function, the following operations can be performed through Matlab language:

[0118] normalized_input_train = normalized_input_train';

[0119] normalized_input_test = normalized_input_test';

[0120] normalized_output_train = normalized_output_train';

[0121] Among them, the symbol "'" is the transpose symbol.

[0122] As an implementation manner of the embodiment of the present invention, the data entry, preprocessing, and analysis module analyzes the first processed data, and specifically may include the following steps, namely step c1 to step c3:

[0123] Step c1: Analyze the total number of data, the processing condition status, and the representative numbers corresponding to each type of processing condition status in the preprocessed processing dataset to obtain the first data analysis result. Here, the processing condition status includes the running status, standby status, and warning status.

[0124] Step c2: Extract the key data features and the number of key data features of the preprocessed processing dataset to obtain the second data analysis result. The key data features include: phase A voltage, phase A current, phase B voltage, phase B current, phase C voltage, phase C current, total power, and spindle speed.

[0125] Step c3: Display the first data analysis result and the second data analysis result on the visualization operation interface.

[0126] Specifically, the data analysis function in the data entry, preprocessing, and analysis module can accurately analyze the preprocessed processing dataset, analyze the total number of data, the types of processing condition status, and their representative numbers in the preprocessed processing dataset, and can extract key data features (including representative symbols, Chinese meanings, and their units), and count the number of key data features. Then, display the data analysis result in the text box of the message bar on the visualization operation page to facilitate users to more intuitively understand the detailed information of the preprocessed processing dataset.

[0127] For the data analysis function, the following Matlab language can be used to perform the operation:

[0128] num_row = size(raw_data,1);

[0129] num_column = size(raw_data,2);

[0130] analysis_data = zeros(num_row,num_column);

[0131] analysis_data = analysis_data(1:end,:);

[0132] statusList = {...

[0133] '1: Running'

[0134] '2: Standby'

[0135] '3: Alarm'

[0136] };

[0137] num_data = size(analysis_data, 1);

[0138] data_total_number = num2str(num_data);

[0139] analysis_data = analysis_data(randperm(num_data), :);

[0140] num_feature = size(analysis_data, 2) - 1;

[0141] features = num2str(num_feature);

[0142] featuresList = {...

[0143] 'UA: Voltage of Phase A, Unit: V'

[0144] 'UB: Voltage of Phase B, Unit: V'

[0145] 'UC: Voltage of Phase C, Unit: V'

[0146] 'IA: Current of Phase A, Unit: A'

[0147] 'IB: Current of Phase B, Unit: A'

[0148] 'IC: Current of Phase C, Unit: A'

[0149] 'P: Total Active Power, Unit: KW'

[0150] 'n: Spindle Speed, Unit: r / min'

[0151] };

[0152] Among them, num_row is the number of rows of the original data set, num_column is the number of columns of the original data set, statusList is the three processing condition states and their representative symbols, num_feature is the number of input features of the original data set, and featuresList is the specific meanings, representative symbols and units of 8 key data features.

[0153] The processing condition state diagnosis module 202 is used to train and test the random forest algorithm model using the preprocessed and analyzed processing data set based on the principal component analysis method and the particle swarm optimization algorithm, and obtain the trained processing condition state diagnosis model.

[0154] As an implementation manner of an embodiment of the present invention, the processing condition state diagnosis module includes: a data set division sub-module, a PCA sub-module, a PSO sub-module, and an RF sub-module;

[0155] Data set division sub-module: Divide the processed and analyzed processing data set into a training data set and a test data set according to a target ratio;

[0156] PCA sub-module, which is used to process the training data set and the test data set by using the principal component analysis method. Through coordinate transformation, under the principle of maximizing variance, project the training data and the test data onto a new coordinate to obtain the processed training data and test data;

[0157] PSO sub-module, which is used to continuously iterate the key hyperparameters of the random forest algorithm model by using the particle swarm optimization algorithm until the optimal key hyperparameters of the random forest algorithm model are obtained;

[0158] RF sub-module: Construct a random forest algorithm model by using the optimal key hyperparameters, and train and test the random forest algorithm model by using the processed training data and test data.

[0159] Specifically, as Figure 3 shown, the data set division sub-module is used to divide the processed processing data set into a training data set and a test data set according to a ratio of 4:1.

[0160] For the data set division sub-module, the following Matlab language can be used to perform operations:

[0161] num_trainproportion = 0.8;

[0162] num_input_train_row = num_row * num_trainproportion;

[0163] input_train = row(1:num_input_train_row, 1:num_feature);

[0164] input_train = table2array(input_train);

[0165] input_train = input_train';

[0166] num_output_train_row = num_row * num_trainproportion;

[0167] output_train_col = num_feature + 1;

[0168] output_train = row(1:num_output_train_row, output_train_col);

[0169] output_train = table2array(output_train);

[0170] output_train = output_train';

[0171] num_trainsize = size(input_train, 2);

[0172] input_test_row = num_trainsize + 1;

[0173] input_test = row(input_test_row:end, 1:num_feature);

[0174] input_test = table2array(input_test);

[0175] input_test = input_test';

[0176] output_test_row = num_trainsize + 1;

[0177] output_test_col = num_feature + 1;

[0178] output_test = row(output_test_row:end, output_test_col);

[0179] output_test = table2array(output_test);

[0180] output_test = output_test';

[0181] num_testsize = size(input_test, 2);

[0182] Among them, num_trainproportion is the training dataset proportion coefficient, num_input_train_row is the number of rows of the training dataset, input_train stores the input features of the training dataset, output_train stores the output features of the training dataset, num_trainsize is the number of samples in the training dataset, input_test stores the input features of the test dataset, output_test stores the output features of the test dataset, and num_testsize is the number of samples in the test dataset.

[0183] The PCA sub-module is used to reduce the dimensionality of the training dataset and the test dataset by using the principal component analysis method. Through coordinate transformation, following the principle of maximum variance, the training dataset and the test dataset are projected onto new coordinates, thereby compressing the information of multiple dimensions of the training dataset and the test dataset into lower dimensions. While trying to retain the original features, fewer features are selected to replace the original features, which can effectively reduce data redundancy. The ultimate goal is to analyze the training dataset and the test dataset with fewer main features. Moreover, the features after dimensionality reduction are not simply some features reduced from the original features, but integrate the information of the original features, ultimately achieving the goal of being able to contain most of the information of the original data with fewer features.

[0184] For the PCA module, the following operations can be performed through Matlab language:

[0185] explainedVarianceRatioThreshold = 0.95;

[0186] [coeff,score,~,~,explained] = pca(input_train');

[0187] cumulativeVarianceRatio = cumsum(explained) / sum(explained);

[0188] numComponents = find(cumulativeVarianceRatio >=

[0189] explainedVarianceRatioThreshold,1,'first');

[0190] input_train_pca = score(:,1:numComponents)';

[0191] input_test_pca = (input_test' * coeff(:, 1:numComponents))';

[0192] Among them, "explainedVarianceRatioThreshold = 0.95;" means that the variance retention threshold is set to 95%, and "[coeff, score, ~, ~, explained] = pca(input_train');" means performing PCA dimensionality reduction on the transposed training data to obtain the principal component coefficient matrix coeff, the score matrix score, and the variance contribution rate explained of each principal component. cumulativeVarianceRatio is the cumulative variance ratio, numComponents is the number of principal components that satisfy the cumulative variance ratio ≥ variance retention threshold, input_train_pca is the training data set after PCA dimensionality reduction, and input_test_pca is the test data set after PCA dimensionality reduction.

[0193] The PSO sub-module is used to optimize the key hyperparameters of the random forest algorithm model using the particle swarm optimization algorithm, such as the number of decision trees, the depth of decision trees, the minimum number of leaf nodes, etc. PSO can utilize individual experience and group experience to guide the search process. By continuously iteratively updating the positions of particles (i.e., parameter combinations), the performance of the random forest algorithm model can be gradually improved, effectively avoiding local optima and improving the generalization ability and diagnostic accuracy of the random forest algorithm model. In addition, the high global search ability of PSO enables it to quickly find the best combination of key hyperparameters of the random forest algorithm, reducing the complexity and time cost of manual parameter tuning.

[0194] The RF sub-module is used to construct a random forest algorithm model using the best combination of key hyperparameters of the random forest algorithm found by the particle swarm optimization algorithm. Then, the training set processed by the principal component analysis method is input into the optimized random forest algorithm model for training. Finally, after the model training is completed, the test data set is diagnosed, and the accuracy rates of the model on the training data set and the test data set are calculated to evaluate its diagnostic performance.

[0195] For the RF sub-module, the following Matlab language operations can be performed:

[0196] numTrees = round(global_best_pos(1));

[0197] minLeafSize = round(global_best_pos(2));

[0198] maxDepth = round(global_best_pos(3));

[0199] RF_final_model = TreeBagger(numTrees,...

[0200] normalized_input_train, normalized_output_train,...

[0201] 'Method', 'classification', 'MinLeafSize', minLeafSize,...

[0202] 'MaxNumSplits', maxDepth,...

[0203] 'OOBPred', 'On', 'OOBVarImp', 'On');

[0204] predict_train = predict(RF_final_model, normalized_input_train);

[0205] predict_test = predict(RF_final_model, normalized_input_test);

[0206] predict_train = str2double(predict_train);

[0207] predict_test = str2double(predict_test);

[0208] output_train = cell2mat(output_train);

[0209] output_test = cell2mat(output_test);

[0210] accuracy_train = sum((predict_train' == output_train)) / num_trainsize * 100;

[0211] accuracy_test = sum((predict_test' == output_test)) / num_testsize * 100;

[0212] [output_train, sortedIndex_train] = sort(output_train);

[0213] [output_test, sortedIndex_test] = sort(output_test);

[0214] predict_train = predict_train(sortedIndex_train);

[0215] predict_test = predict_test(sortedIndex_test);

[0216] Among them, numTrees, minLeafSize, and maxDepth are the optimal parameters obtained through the particle swarm optimization algorithm. RF_final_model is a random forest classification model. The training set is used for model training, and then the training set and the test set are diagnosed, and the diagnostic accuracies of the training set and the test set are calculated.

[0217] The diagnostic result visualization module 203 is used to draw diagnostic result graphs and display the diagnostic result graphs in the visualization operation interface. The drawn diagnostic result graphs include: diagnostic errors corresponding to different key hyperparameters of the processing condition state diagnosis model; diagnostic accuracies corresponding to the training process and the test process of the processing condition state diagnosis model, and classification accuracies corresponding to different classification tasks.

[0218] Specifically, the diagnostic result visualization module can provide the function of drawing diagnostic result graphs. The diagnostic result graphs include error curve graphs, comparison graphs of diagnostic results of the training data set, comparison graphs of diagnostic results of the test data set, confusion matrices of the training data set, and confusion matrices of the test data set. The system can draw diagnostic result graphs and output them to the corresponding coordinate areas to improve the convenience of data visualization and display. In the error curve graph, the key hyperparameters of the optimized random forest algorithm model, such as the number of decision trees, can be seen. Within a certain range, as the number of decision trees increases, the trend of the corresponding error can be seen; the specific accuracies of the training set diagnosis and the test set diagnosis can be seen in the comparison graph of diagnostic results of the training data set and the comparison graph of diagnostic results of the test data set respectively; in the confusion matrices of the training data set and the test data set, the number of correct classifications and incorrect classifications and their specific accuracies in the classification tasks of the principal component analysis - particle swarm optimization algorithm - random forest algorithm model for the three processing condition states in the training set and the test set respectively can be seen.

[0219] For the drawing of the error curve in the diagnostic result graph, the following operations can be performed using Matlab language:

[0220] xlabel(app.errorcurve,'Number of decision trees');

[0221] ylabel(app.errorcurve,'Error');

[0222] title(app.errorcurve,'Error Curve');

[0223] plot(app.errorcurve,1:numTrees,oobError(RF_final_model),'b-','LineWidth',1);

[0224] xlim(app.errorcurve,[1,numTrees]);

[0225] app.errorcurve is the name of the coordinate area component for displaying the error curve. Its X-axis label is set to the number of decision trees, ranging from 1 to 50. The Y-axis label is set to error, and the chart title is Error Curve. "plot(app.errorcurve,1:numTrees,oobError(RF_final_model),'b-','LineWidth',1);" is for drawing the error curve and setting the line style to solid line, color to blue, and line width to 1 unit.

[0226] For the drawing of the comparison graph of the training set diagnostic results in the diagnostic result graph, the following operations can be performed using Matlab language:

[0227] plot(app.predicttraining,1:num_trainsize,output_train,

[0228] 'r-*','LineWidth',3,'DisplayName','True Value');

[0229] hold(app.predicttraining,'on');

[0230] plot(app.predicttraining,1:num_trainsize,predict_train,

[0231] 'b--o','LineWidth',1,'DisplayName','Diagnostic Value');

[0232] hold(app.predicttraining, 'off');

[0233] legend(app.predicttraining,'show');

[0234] xlabel(app.predicttraining, 'Diagnostic samples');

[0235] ylabel(app.predicttraining, 'Diagnostic results');

[0236] string1 = {'Comparison of diagnostic results in the training set'; ['Accuracy = 'num2str(accuracy_train)'%']};

[0237] title(app.predicttraining, string1);

[0238] Among them, app.predicttraining is the name of the coordinate area component that shows the comparison graph of the diagnostic results in the training set. Its X-axis label is set to Diagnostic samples, the Y-axis label is set to Diagnostic results, and the chart title is Comparison of diagnostic results in the training set.

[0239] "plot(app.predicttraining, 1:num_trainsize, output_train, 'r-*', 'LineWidth', 3, 'DisplayName', 'True value');" is to plot the true value curve of the training set and set the line style to solid line, color to red, data point marker symbol to asterisk, and the width of the line to 3 units. "plot(app.predicttraining, 1:num_trainsize, predict_train, 'b--o', 'DisplayName', 'Diagnostic value');" is to plot the diagnostic value curve of the training set and set the line style to dashed line, color to blue, and data point marker symbol to circle.

[0240] For the drawing of the comparison graph of the diagnostic results of the test data set in the diagnostic result graph, the following Matlab language operations can be performed:

[0241] plot(app.predicttest, 1:num_testsize, output_test,

[0242] 'r-*','LineWidth',3,'DisplayName','True Value');

[0243] hold(app.predicttest,'on');

[0244] plot(app.predicttest,1:num_testsize,predict_test,

[0245] 'b--o','LineWidth',1,'DisplayName','Diagnostic Value');

[0246] hold(app.predicttest,'off');

[0247] legend(app.predicttest,'show');

[0248] xlabel(app.predicttest,'Diagnostic Samples');

[0249] ylabel(app.predicttest,'Diagnostic Results');

[0250] string2 = {'Comparison of Diagnostic Results for Test Set'; ['Accuracy = 'num2str(accuracy_test)'%']};

[0251] title(app.predicttest,string2);

[0252] app.predicttest is the name of the axes component for displaying the comparison graph of diagnostic results for the test set. Its X-axis label is set to Diagnostic Samples, its Y-axis label is set to Diagnostic Results, and the chart title is Comparison of Diagnostic Results for Test Set.

[0253] "plot(app.predicttest, 1:num_testsize, output_test, 'r-*', 'LineWidth', 3, 'DisplayName', 'True Value');" plots the curve of the true values in the test set and sets the line style to solid, color to red, data point marker symbol to asterisk, and line width to 3 units. "plot(app.predicttest, 1:num_testsize, predict_test, 'b--o', 'LineWidth', 1, 'DisplayName', 'Diagnostic Value');" plots the curve of the diagnostic values in the training set and sets the line style to dashed, color to blue, data point marker symbol to circle, and line width to 1 unit.

[0254] For the plotting of the confusion matrix of the training dataset in the diagnostic result graph, the following operations can be performed using Matlab language:

[0255] fig = figure('Visible', 'off');

[0256] cm = confusionchart(output_train, predict_train);

[0257] cm.Title = 'Confusion Matrix for Train Data';

[0258] cm.ColumnSummary = 'column-normalized';

[0259] cm.RowSummary = 'row-normalized';

[0260] f = getframe(fig);

[0261] img = f.cdata;

[0262] close(fig);

[0263] lines4 = findall(app.Confusiontraining, 'Type', 'Line');

[0264] delete(lines4);

[0265] imshow(img, 'Parent', app.Confusiontraining);

[0266] Among them, "cm = confusionchart(output_train, predict_train);" is to generate a confusion matrix chart cm using the confusionchart function, with its title set to Confusion Matrix for TrainData, the column summary set to column-normalized, and the row summary set to row-normalized. "imshow(img, 'Parent', app.Confusiontraining);" is to display the image img capturing the confusion matrix of the training set in the coordinate area component named app.Confusiontraining using the imshow function.

[0267] For the drawing of the confusion matrix of the test set in the diagnostic result graph, the following operations can be performed through the Matlab language:

[0268] fig1 = figure('Visible', 'off');

[0269] cm = confusionchart(output_test, predict_test);

[0270] cm.Title = 'Confusion Matrix for Test Data';

[0271] cm.ColumnSummary = 'column-normalized';

[0272] cm.RowSummary = 'row-normalized';

[0273] f1 = getframe(fig1);

[0274] img1 = f1.cdata;

[0275] close(fig1);

[0276] lines5 = findall(app.Confusiontest, 'Type', 'Line');

[0277] delete(lines5);

[0278] imshow(img1, 'Parent', app.Confusiontest);

[0279] Among them, "cm = confusionchart(output_test, predict_test);" is to generate a confusion matrix chart cm using the confusionchart function, with its title set to Confusion Matrix for TestData, column summary set to column-normalized, and row summary set to row-normalized. "imshow(img1, 'Parent', app.Confusiontest);" is to display the image img1 capturing the confusion matrix of the test set in the coordinate area component named app.Confusiontest using the imshow function.

[0280] Based on the above embodiments, in one implementation, the system may further include a user permission and security management module;

[0281] The permission and security management module is used to control the access permissions of users to the processing condition state diagnosis system. Among them, the access permissions include the system login permission and the password modification permission. Unregistered users need to register. After successful registration, they can obtain the system login permission. After the user successfully logs in, the human-computer interaction interface of the processing condition state diagnosis system will be displayed.

[0282] Specifically, the user permission and security management module manages the user login permission and controls the access permissions of users to the processing condition state diagnosis system. Unregistered users need to register by entering a username and password. To avoid errors when the user enters the password, the system requires the user to enter the password again for confirmation. After entering the username, password, and password confirmation, the system login permission can be obtained. Then, the registered user can enter the username and password credentials authenticated and authorized by the system through the authentication module and successfully log in to the human-computer interaction interface of the processing condition state diagnosis system, that is, the visualization operation interface, as Figure 4 shown.

[0283] Meanwhile, to further safeguard account security, a password modification function is set up to facilitate users to flexibly adjust according to their own situations. For example, if they forget the original password or find it difficult to remember, they can modify it to something easier to remember for convenient login. When users modify their passwords, the system will first require them to enter the currently used username and password for identity verification. This is to ensure that only the real account holder can perform the password modification operation and prevent others from maliciously tampering. If the current password is entered incorrectly, the system will prompt the user to re-enter it to increase security. After successful identity verification, the user enters the password modification page. The new password must meet the strength requirements of being at least 8 characters long and containing uppercase letters, lowercase letters, numbers, and special characters to help users set a secure password. To avoid errors when users enter the new password, the system requires the user to re-enter the new password for confirmation. The two entered passwords must be exactly the same, otherwise the system will prompt the user to re-enter. When the user completes the input and confirmation of the new password and clicks the modify password button, the system will encrypt the new password and store it in the database to complete the password modification operation.

[0284] Based on the above embodiments, in one implementation, the data entry, preprocessing, and analysis module is further configured to enter the current processing data generated during the current operation of the processing equipment, preprocess and analyze the current processing data to obtain the preprocessed and analyzed current processing data.

[0285] The processing condition state diagnosis module is configured to input the preprocessed and analyzed current processing data into the trained processing condition state diagnosis model to obtain the current processing condition state of the processing equipment.

[0286] The diagnosis result visualization module is configured to draw a diagnosis result graph based on the current processing condition state and display the drawn diagnosis result graph in the visualization operation interface.

[0287] Specifically, by inputting the current processing data generated during the current operation of the processing equipment into the processing condition state diagnosis system, the data entry, preprocessing, and analysis module can preprocess and analyze the current processing data, input the preprocessed and analyzed current processing data into the trained processing condition state diagnosis model to obtain the current processing condition state of the processing equipment, that is, it can monitor the processing condition state of the processing equipment in real time, draw the current processing condition state obtained by real-time monitoring into a diagnosis result graph, and display the diagnosis result graph in the visualization operation interface. In this way, users can timely and intuitively see the processing condition state of the processing equipment.

[0288] The processing condition status diagnosis system provided by the embodiment of the present invention includes a data input, preprocessing, and analysis module that inputs a processing data set generated during the historical operation of a processing device, and preprocesses and analyzes the processing data set; a processing condition status diagnosis module that, based on the principal component analysis method and the particle swarm optimization algorithm, uses the preprocessed and analyzed processing data set to train and test a random forest algorithm model to obtain a trained processing condition status diagnosis model; and a diagnosis result visualization module that draws a diagnosis result graph and displays the diagnosis result graph in a visualization operation interface. It can be seen that the embodiment of the present invention effectively utilizes the processing data generated by the processing device during the processing process, efficiently and accurately monitors and diagnoses the processing condition status of the processing device, that is, the operating status of the processing device, thereby discovering potential faults in advance, reducing the downtime of the processing device, improving production efficiency, and reducing maintenance costs.

[0289] The embodiment of the present invention also provides a processing condition status diagnosis method based on processing information, which is applied to a processing condition status diagnosis system based on processing information. The system includes a data input, preprocessing, and analysis module, a processing condition status diagnosis module, and a diagnosis result visualization module. The method includes:

[0290] The data input, preprocessing, and analysis module inputs a processing data set generated during the historical operation of a processing device, and preprocesses and analyzes the processing data set. Among them, the methods for preprocessing the processing data set include: missing value processing, outlier processing, duplicate value clearing processing, normalization processing, and transposition processing;

[0291] The processing condition status diagnosis module, based on the principal component analysis method and the particle swarm optimization algorithm, uses the preprocessed and analyzed processing data set to train and test a random forest algorithm model to obtain a trained processing condition status diagnosis model;

[0292] The diagnosis result visualization module draws a diagnosis result graph and displays the diagnosis result graph in a visualization operation interface; the drawn diagnosis result graph includes: diagnosis errors corresponding to different key hyperparameters of the processing condition status diagnosis model; diagnosis accuracies corresponding to the training process and the testing process of the processing condition status diagnosis model, respectively, and classification accuracies corresponding to different classification tasks, respectively.

[0293] The embodiment of the present invention also provides an electronic device 500, which is installed with the processing condition status diagnosis system based on processing information described in the embodiment of the present invention, as Figure 5 shown, including:

[0294] At least one processor 501;

[0295] A memory 502 for storing instructions executable by the at least one processor;

[0296] Wherein, the at least one processor is configured to execute the instructions to run the processing condition state diagnosis system based on processing information according to the embodiments of the present invention.

[0297] The embodiments of the present invention also provide a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can run the processing condition state diagnosis system based on processing information according to the embodiments of the present invention.

[0298] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Without departing from the principle and spirit of the present invention, those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A processing condition state diagnosis system based on processing information, characterized in that The system includes: data entry, preprocessing and analysis module, processing condition status diagnosis module, and diagnosis result visualization module; The data entry, preprocessing and analysis module is used to enter the processing data set generated during the historical operation of the processing equipment, and preprocess and analyze the processing data set. Among them, the methods for preprocessing the processing data set include: missing value processing, outlier processing, duplicate value removal processing, normalization processing, and transposition processing; The processing condition status diagnosis module is used to train and test the random forest algorithm model based on the principal component analysis method and the particle swarm optimization algorithm, using the preprocessed and analyzed processing data set, to obtain a trained processing condition status diagnosis model; The diagnosis result visualization module is used to draw a diagnosis result graph and display the diagnosis result graph in the visualization operation interface; the drawn diagnosis result graph includes: the diagnosis error corresponding to different key hyperparameters of the processing condition status diagnosis model; the diagnosis accuracy corresponding to the training process and the testing process of the processing condition status diagnosis model, and the classification accuracy corresponding to different classification tasks.

2. The system according to claim 1, characterized in that The data entry, preprocessing and analysis module enters the processing data set generated during the historical operation of the processing equipment, specifically including: Design a visualization operation interface for importing processing data based on the Matlab platform. Among them, the design of the visualization operation interface needs to establish a connection channel between the local file system and the Matlab platform based on the Matlab language; Receive the user's selection operation for the processing data file saved in the local file system, respond to the selection operation, read the processing data file, and enter the processing data set in the processing data file.

3. The system according to claim 1, wherein The data entry, preprocessing and analysis module preprocesses the processing data set, specifically including: Detect whether there are missing values in the processing data set. If there are missing values in the processing data set, calculate the average value of all processing data in the column where the missing values are located in the processing data set, and use the average value to repair the missing values; Detect whether there are outliers in the processing data set. If there are outliers in the processing data set, use the average value of the adjacent values of the outliers to replace the outliers; Detect whether there are completely duplicate data in the processing data set. If there are completely duplicate data in the processing data set, delete the redundant data in the completely duplicate data; Use the normalization method to linearly map the processing data set to the target interval; Perform matrix transposition on the processing data set; among them, the transposed processing data set is compatible with the input format of the random forest algorithm model.

4. The system according to claim 2, wherein The data entry, preprocessing and analysis module analyzes the first processing data, specifically including: Analyze the total number of data, processing condition status, and the representative numbers corresponding to each type of processing condition status in the preprocessed processing data set, to obtain the first data analysis result. Among them, the processing condition status includes operating status, standby status, and warning status; Extract the key data features of the processed machining dataset and the quantity of the key data features, obtaining a second data analysis result; the key data features include: phase A voltage, phase A current, phase B voltage, phase B current, phase C voltage, phase C current, total power, spindle speed. Display the first data analysis result and the second data analysis result on a visualization operation interface.

5. The system according to any one of claims 1 to 4, wherein The machining condition state diagnosis module includes: a dataset division sub-module, a PCA sub-module, a PSO sub-module, and an RF sub-module. Dataset division sub-module: Divide the preprocessed and analyzed machining dataset into a training dataset and a test dataset according to a target ratio. PCA sub-module, which is used to process the training dataset and the test dataset by using the principal component analysis method. Through coordinate transformation, following the principle of maximum variance, project the training data and the test data onto a new coordinate to obtain the processed training data and test data. PSO sub-module, which is used to continuously iterate the key hyperparameters of the random forest algorithm model by using the particle swarm optimization algorithm until the optimal key hyperparameters of the random forest algorithm model are obtained. RF sub-module: Use the optimal key hyperparameters to construct a random forest algorithm model, and train and test the random forest algorithm model with the processed training data and test data.

6. The system according to any one of claims 1 to 4, characterized in that, The machining dataset is machining data collected by a multi-source data acquisition platform from multiple machining devices of key components of an electro-hydraulic control actuator.

7. The system according to any one of claims 1 to 4, characterized in that, The system further includes: a user permission and security management module. The permission and security management module is used to control the access permissions of users to the machining condition state diagnosis system. Among them, the access permissions include the permission to log in to the system and the permission to modify the password. Unregistered users need to register. After successful registration, they can obtain the permission to log in to the system, and after the user successfully logs in, the man-machine interaction interface of the machining condition state diagnosis system is displayed.

8. The system according to any one of claims 1 to 4, characterized in that The data entry, preprocessing, and analysis module is further used to enter the current machining data generated during the current operation of the machining device, preprocess and analyze the current machining data, and obtain the preprocessed and analyzed current machining data. The machining condition state diagnosis module is used to input the preprocessed and analyzed current machining data into the trained machining condition state diagnosis model to obtain the current machining condition state of the machining device. The diagnosis result visualization module is used to draw a diagnosis result graph based on the current machining condition state and display the drawn diagnosis result graph on the visualization operation interface.

9. A machining condition state diagnosis method based on machining information, characterized in that, Applied to a machining condition state diagnosis system based on machining information, the system includes a data entry, preprocessing, and analysis module, a machining condition state diagnosis module, and a diagnosis result visualization module. The method includes: The data entry, preprocessing and analysis module enters the processing data set generated during the historical operation of the processing equipment, and preprocesses and analyzes the processing data set. Among them, the methods for preprocessing the processing data set include: missing value processing, outlier processing, duplicate value clearing processing, normalization processing and transposition processing; The processing condition state diagnosis module trains and tests the random forest algorithm model based on the principal component analysis method and the particle swarm optimization algorithm, using the preprocessed and analyzed processing data set, to obtain the trained processing condition state diagnosis model; The diagnosis result visualization module draws a diagnosis result graph and displays the diagnosis result graph in the visualization operation interface; the drawn diagnosis result graph includes: the diagnosis errors corresponding to different key hyperparameters of the processing condition state diagnosis model; the diagnosis accuracies corresponding to the training process and the testing process of the processing condition state diagnosis model, and the classification accuracies corresponding to different classification tasks respectively.

10. An electronic device, characterized in that, Installed with the processing condition state diagnosis system based on processing information according to any one of claims 1 to 8, including: At least one processor; A memory for storing instructions executable by the at least one processor; Wherein, the at least one processor is configured to execute the instructions to run the processing condition state diagnosis system based on processing information according to any one of claims 1 to 8.

Images