A big data-based tool wear data monitoring system and method

Abstract

The application discloses a kind of cutter wear data monitoring system and method based on big data, belong to big data technical field.The application includes: acquisition and processing module: real-time acquisition cutter wear related data, carry out data cleaning and timeline alignment, extract cutter wear characteristics;Big data storage and analysis module: using HDFS stores cutter wear characteristics, implements data management strategy, introduces data security measures, establishes data index, analyzes the relevance between cutter wear characteristics;Wear rehearsal module: obtain historical data set, train linear regression model to calculate the influence of current cutter wear characteristics on cutter wear state, carry out model evaluation, and predict the features collected in real time;Parameter adjustment module: compare the vector group of predicted cutter wear with time variation with the optimal trend of cutter wear with time variation, and adjust the cutter wear characteristics using gradient optimization algorithm.

Description

Technical Field

[0001] This invention relates to the field of big data technology, specifically to a tool wear data monitoring system and method based on big data. Background Technology

[0002] Big data refers to massive, complex, and high-speed datasets that cannot be captured, managed, processed, and analyzed using traditional data processing tools. Big data typically has three main characteristics: volume, diversity, and speed. Apache Spark is an open-source big data processing framework that provides efficient, versatile, fault-tolerant, and easy-to-use big data processing tools. Spark was originally developed by the AMPLab at UC Berkeley and open-sourced in 2010. Spark's main features include speed, versatility, fault tolerance, ease of use, and a rich ecosystem.

[0003] Tool wear-related data is typically massive, including real-time data collected from various sensors such as vibration frequency, temperature, cutting force, and cutting speed. Traditional data processing tools struggle to effectively handle such large-scale and diverse data, while big data technology offers efficient processing and analysis methods. Tool wear data monitoring requires real-time performance and timely responses to tool conditions. Streaming processing and real-time computing modules in big data technology can meet these real-time processing needs, enabling the system to respond rapidly when tool wear conditions change. Traditional systems, when training models using historical datasets, suffer from relatively low training accuracy due to limitations in algorithms and computing power, making it difficult to accurately predict tool wear conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a tool wear data monitoring system and method based on big data to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] A tool wear data monitoring system based on big data includes:

[0007] Acquisition and Processing Module: Uses sensors to collect tool wear-related data in real time, uses filtering algorithms to clean the raw data, aligns the timelines of data collected from different sensors, and extracts tool wear characteristics;

[0008] Big Data Storage and Analysis Module: Uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes; utilizes the big data processing framework Apache Spark for distributed computing, analyzes the correlation between tool wear characteristics, and outputs the optimal trend of tool wear changes over time;

[0009] Wear Prediction Module: Based on the analysis results obtained from the Big Data Storage and Analysis Module, historical datasets are acquired, and the historical datasets are divided into training sets and datasets. A linear regression model is trained to calculate the impact of the current tool wear characteristics on the tool wear state, the model is evaluated, the real-time collected features are predicted, and a vector group predicting the tool wear change over time is output.

[0010] Parameter adjustment module: The vector set of predicted tool wear over time obtained from the wear prediction module is compared with the optimal trend of tool wear over time obtained from the big data storage and analysis module, and the tool wear characteristics are adjusted using a gradient optimization algorithm.

[0011] The acquisition and processing module extracts tool wear characteristics, which include vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication status, tool shape, and cutting pattern.

[0012] Using a moving average filtering algorithm, the user selects the filter window size, moves the window across the entire data sequence, calculates the average value of the data points within each window, and uses the new moving average value to replace the corresponding position in the original data. For each time point, the filtering algorithm calculates the corresponding filtered value. For vibration frequency, the frequency components of the vibration signal are extracted using Fourier transform. For temperature, cutting force, cutting speed, depth of cut, feed rate, and material hardness, the filtered data is used directly. For cutting time, the filtered time data is accumulated. For lubrication conditions, information is extracted based on lubricating oil pressure and lubricating fluid state. Based on tool shape and cutting pattern, features related to tool shape and cutting pattern are extracted based on the analysis of the vibration signal.

[0013] The big data storage and analysis module implements a data management strategy, performs regular data backups on HDFS, and uses the snapshot function of the Hadoop Distributed File System to create snapshots of tool wear characteristic data; for tool wear characteristic data, a compression algorithm is used to compress the data;

[0014] Regular data backups ensure the system can quickly recover to a previous reliable state in the event of data loss or corruption. Utilizing HDFS's data backup mechanism, tool wear characteristic data can be backed up to other locations by performing regular data snapshots or backup tasks to mitigate the risk of data loss. HDFS's snapshot feature allows capturing the state of data at a specific point in time without affecting real-time data access, facilitating subsequent recovery or comparison. Creating snapshots of tool wear characteristic data and saving them as readable snapshot versions facilitates subsequent retrieval, restoration, or comparison with other versions.

[0015] Data compression helps reduce storage space usage, improve data transmission efficiency, and lower storage costs. Applying appropriate compression algorithms, such as Gzip and Snappy, to tool wear characteristic data can reduce the size of data files. During data transmission and storage, compression algorithms can reduce I / O operations and network bandwidth requirements, thereby improving overall performance.

[0016] Regularly clean up expired or unwanted data to free up storage space. Delete old data based on timestamps or remaining space by establishing a data retention policy. Store tool wear characteristic data in partitions, according to time, tool type, or work task.

[0017] The big data storage and analysis module is configured by the user using a Spark cluster environment, which includes Master nodes and Worker nodes. Spark accesses data in HDFS and allocates resources for distributed computing tasks. Spark SQL, Spark's data loading module, loads tool wear feature data from HDFS into the distributed memory of the Spark cluster.

[0018] Using Spark's machine learning library MLlib, correlation analysis was employed to analyze the relationships between features; Spark's time series library was used to capture time-related patterns and optimal trends, and linear trends from the time series library were applied for time series analysis. The linear trend model was evaluated by calculating the goodness of fit and root mean square error; the results of the linear trend were visualized, and a linear trend line was drawn to reflect the optimal trend of tool wear over time.

[0019] To perform correlation analysis using Spark's machine learning library MLlib, tool wear feature data needs to be loaded from HDFS into a Spark DataFrame. The data needs to be cleaned, missing values ​​removed, and data format consistency ensured. The tool wear features for which correlation analysis is required are then selected. Using MLlib's statistics modules, such as the `Statistics.corr` function, the correlation coefficient matrix between tool wear features is calculated. The correlation coefficient matrix is ​​then interpreted to identify strong or weak correlations between features. The results of the correlation analysis are then saved to HDFS for later use.

[0020] To perform linear trend analysis using Spark's time series library, you need to load tool wear feature time series data from HDFS into a Spark DataFrame. Ensure the timestamps of the time series data are in the correct format, sort them by time, and use Spark's time series library, such as LinearTrend, to calculate the linear trend of tool wear features over time. Calculate metrics such as the R² goodness of fit and root mean square error (RMSE) of the linear trend model. Save the results of the linear trend analysis to HDFS for subsequent analysis and visualization. Use plotting libraries such as Matplotlib or Plotly to draw the linear trend line of tool wear over time, including the actual data points and the linear trend line. Add evaluation metrics such as goodness of fit and RMSE to the visualization results to visually demonstrate the quality of the linear trend model.

[0021] A method for monitoring tool wear data based on big data, comprising the following steps:

[0022] S100: Use sensors to collect tool wear-related data in real time;

[0023] S200: Use filtering algorithms to clean the raw data collected, align the timelines of data collected from different sensors, and extract tool wear characteristics.

[0024] S300 uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes;

[0025] S400 utilizes the big data processing framework Apache Spark for distributed computing to analyze the correlation between tool wear characteristics and output the optimal trend of tool wear over time.

[0026] S500: Based on the analysis results obtained from the big data storage and analysis module, obtain the historical dataset and divide the historical dataset into training set and dataset;

[0027] S600: Train a linear regression model to calculate the impact of the current tool wear characteristics on the tool wear state, and output a vector set predicting the change of tool wear over time;

[0028] S700. Compare the vector set of predicted tool wear over time obtained by the wear prediction module with the optimal trend of tool wear over time obtained by the big data storage and analysis module.

[0029] S800 uses a gradient optimization algorithm to adjust the tool wear characteristics.

[0030] According to step S600, the steps for calculating the influence of the current tool wear characteristics on the tool wear state using a linear regression model are as follows:

[0031] S601. Divide the historical dataset into a training set and a test set. The training set is used to train the linear regression model, and the test set is used to evaluate the performance of the linear regression model.

[0032] S602. Using the training set, with tool wear characteristics as input variables and the vector set predicting tool wear changes over time as output variables, a linear regression model is fitted using a linear regression algorithm, as shown in the following formula:

[0033] ;

[0034] Where Y is a vector set representing the change of tool wear over time; It is the intercept of the linear regression model; These are the coefficients of the linear regression model, representing the degree of influence of each tool wear characteristic on the tool wear state; The input variables are vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication, tool shape, and cutting pattern. This is the error term; n is the number of input variables;

[0035] S603. Use mean squared error, average error, and coefficient of determination to evaluate the model’s fit to the test set data.

[0036] S604. Store the divided training and test set data in HDFS as input to the big data storage and analysis module.

[0037] According to step S603, the method for evaluating the model's fit to the test set data is as follows:

[0038] S603-1. Evaluation is performed using the mean square error, as shown in the following formula:

[0039] ;

[0040] Where m is the number of samples in the test set; It is the i-th vector representing the actual tool wear over time; is the predicted value of tool wear over time for the i-th sample by the linear regression model; MSE is the mean squared error.

[0041] S603-2. Evaluation is performed using the mean absolute error, as shown in the following formula:

[0042] ;

[0043] Where MAE is the mean absolute error;

[0044] S603-3. Evaluation is performed using the coefficient of determination, as shown in the following formula:

[0045] ;

[0046] in, It is the coefficient of determination. It is the mean of the vector of actual tool wear over time;

[0047] S603-4. Calculate the evaluation average G of the mean square error, mean absolute error, and coefficient of determination; user-defined evaluation threshold. ,when If the condition is met, proceed to step S604; otherwise, re-execute step S601. Wherein, G and... All are positive decimals.

[0048] According to step S700, the method for comparing the vector set predicting the change of tool wear over time with the optimal trend of tool wear over time is as follows:

[0049] S701. Obtain the vector set of predicted tool wear over time obtained from the wear prediction module, where each row represents a time point and each column represents a tool wear feature. There are j tool wear features and k time points. Let... Let represent the predicted value of tool wear characteristic k at time point j. Then, the vector set representing the predicted tool wear variation over time is expressed as:

[0050] ;

[0051] Wherein, W is a vector set for predicting the change of tool wear over time, specifically obtained through a linear regression model;

[0052] S702, convert the optimal trend of tool wear over time obtained from the big data storage and analysis module into a vector group form;

[0053] S703. Use root mean square error to compare the vector set predicting tool wear over time with the optimal trend vector set of tool wear over time. The comparison result is used to quantify the difference between the two vector sets.

[0054] S704. Analyze and compare the results to evaluate the consistency between the predicted results and the actual trend.

[0055] According to step S702, the steps to convert it to vector group form are as follows:

[0056] S702-1. Obtain the optimal trend of tool wear over time from the big data storage and analysis module, and select a series of time points, wherein the time points are segmented by hour;

[0057] S702-2. For each selected time point, extract the value of the optimal trend of tool wear over time obtained from the big data storage and analysis module, including various characteristics of tool wear.

[0058] S702-3. Combine the trend values ​​extracted at each time point into a vector to form a vector group, where each vector corresponds to the tool wear state at a time point.

[0059] S702-4. Ensure that the optimal trend vector set of tool wear over time obtained from the big data storage and analysis module is aligned in time with the vector set of predicted tool wear over time obtained from the wear prediction module. If the time granularity is different, interpolation or truncation processing is required.

[0060] According to step S800, the method for adjusting the tool wear characteristics using the gradient optimization algorithm is as follows:

[0061] S801. Based on the comparison results, determine the objective function that needs to be optimized;

[0062] S802. Determine the tool wear characteristics that need to be adjusted, and initialize the parameter values ​​of the tool wear characteristics for the gradient optimization algorithm;

[0063] S803. Using gradient descent, a gradient optimization algorithm, define the update rule for each step. Based on the gradient of the objective function, adjust the parameter values ​​along the negative direction of the gradient to gradually reduce the value of the objective function.

[0064] S804. In each iteration, the parameter values ​​of the tool wear characteristics are updated according to the gradient descent rule, so that the objective function is gradually optimized; this process is repeated until the set stopping condition is reached, wherein the stopping condition is reaching the maximum number of iterations, which is set by the user.

[0065] S805. For each tool wear characteristic, examine whether the optimized parameter values ​​make the predicted values ​​closer to the actual values, and use cross-validation to evaluate the generalization performance.

[0066] S806. Obtain the adjusted tool wear characteristic parameters and adjust the tool wear characteristics.

[0067] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0068] The big data storage and analysis module of this invention uses HDFS to store and manage data, and leverages Apache Spark for distributed computing, enabling more efficient processing of massive amounts of tool wear characteristic data.

[0069] This invention uses Spark's machine learning library MLlib to perform correlation analysis between features, and combines it with a time series library to capture the optimal trend of tool wear over time, thereby gaining a more comprehensive understanding of the dynamic changes in tool wear.

[0070] This invention uses historical datasets to train a linear regression model, evaluates the model, and achieves accurate prediction of real-time acquired features. The parameter adjustment module uses a gradient optimization algorithm to adjust tool wear characteristics in real time, thereby improving the accuracy and adaptability of the model. Attached Figure Description

[0071] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0072] Figure 1 This is a system structure diagram of a tool wear data monitoring system based on big data according to the present invention;

[0073] Figure 2 This is a schematic diagram illustrating the steps of a tool wear data monitoring method based on big data according to the present invention. Detailed Implementation

[0074] 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 some embodiments of the present invention, and not all embodiments. 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.

[0075] Please see Figure 1 and Figure 2 The present invention provides the following technical solution:

[0076] According to one embodiment of the present invention, such as Figure 1The system structure diagram of a tool wear data monitoring system based on big data is shown. The system includes:

[0077] Acquisition and Processing Module: Uses sensors to collect tool wear-related data in real time, uses filtering algorithms to clean the raw data, aligns the timelines of data collected from different sensors, and extracts tool wear characteristics;

[0078] Big Data Storage and Analysis Module: Uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes; utilizes the big data processing framework Apache Spark for distributed computing, analyzes the correlation between tool wear characteristics, and outputs the optimal trend of tool wear changes over time;

[0079] Wear Prediction Module: Based on the analysis results obtained from the Big Data Storage and Analysis Module, historical datasets are acquired, and the historical datasets are divided into training sets and datasets. A linear regression model is trained to calculate the impact of the current tool wear characteristics on the tool wear state, the model is evaluated, the real-time collected features are predicted, and a vector group predicting the tool wear change over time is output.

[0080] Parameter adjustment module: The vector set of predicted tool wear over time obtained from the wear prediction module is compared with the optimal trend of tool wear over time obtained from the big data storage and analysis module, and the tool wear characteristics are adjusted using a gradient optimization algorithm.

[0081] The acquisition and processing module extracts tool wear characteristics, which include vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication status, tool shape, and cutting pattern.

[0082] Using a moving average filtering algorithm, the user selects the filter window size, moves the window across the entire data sequence, calculates the average value of the data points within each window, and uses the new moving average value to replace the corresponding position in the original data. For each time point, the filtering algorithm calculates the corresponding filtered value. For vibration frequency, the frequency components of the vibration signal are extracted using Fourier transform. For temperature, cutting force, cutting speed, depth of cut, feed rate, and material hardness, the filtered data is used directly. For cutting time, the filtered time data is accumulated. For lubrication conditions, information is extracted based on lubricating oil pressure and lubricating fluid state. Based on tool shape and cutting pattern, features related to tool shape and cutting pattern are extracted based on the analysis of the vibration signal.

[0083] The big data storage and analysis module implements a data management strategy, performs regular data backups on HDFS, and uses the snapshot function of the Hadoop Distributed File System to create snapshots of tool wear characteristic data; for tool wear characteristic data, a compression algorithm is used to compress the data;

[0084] Regularly clean up expired or unwanted data to free up storage space. Delete old data based on timestamps or remaining space by establishing a data retention policy. Store tool wear characteristic data in partitions, according to time, tool type, or work task.

[0085] The big data storage and analysis module is configured by the user using a Spark cluster environment, which includes Master nodes and Worker nodes. Spark accesses data in HDFS and allocates resources for distributed computing tasks. Spark SQL, Spark's data loading module, loads tool wear feature data from HDFS into the distributed memory of the Spark cluster.

[0086] Using Spark's machine learning library MLlib, correlation analysis was employed to analyze the relationships between features; Spark's time series library was used to capture time-related patterns and optimal trends, and linear trends from the time series library were applied for time series analysis. The linear trend model was evaluated by calculating the goodness of fit and root mean square error; the results of the linear trend were visualized, and a linear trend line was drawn to reflect the optimal trend of tool wear over time.

[0087] To perform correlation analysis using Spark's machine learning library MLlib, tool wear feature data needs to be loaded from HDFS into a Spark DataFrame. The data needs to be cleaned, missing values ​​removed, and data format consistency ensured. The tool wear features for which correlation analysis is required are then selected. Using MLlib's statistics modules, such as the `Statistics.corr` function, the correlation coefficient matrix between tool wear features is calculated. The correlation coefficient matrix is ​​then interpreted to identify strong or weak correlations between features. The results of the correlation analysis are then saved to HDFS for later use.

[0088] According to another embodiment of the invention, such as Figure 2 A schematic diagram illustrating the steps of a tool wear data monitoring method based on big data is shown below. The steps of the tool wear data monitoring method based on big data are as follows:

[0089] S100: Use sensors to collect tool wear-related data in real time;

[0090] S200: Use filtering algorithms to clean the raw data collected, align the timelines of data collected from different sensors, and extract tool wear characteristics.

[0091] S300 uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes;

[0092] S400 utilizes the big data processing framework Apache Spark for distributed computing to analyze the correlation between tool wear characteristics and output the optimal trend of tool wear over time.

[0093] S500: Based on the analysis results obtained from the big data storage and analysis module, obtain the historical dataset and divide the historical dataset into training set and dataset;

[0094] S600: Train a linear regression model to calculate the impact of the current tool wear characteristics on the tool wear state, and output a vector set predicting the change of tool wear over time;

[0095] S700. Compare the vector set of predicted tool wear over time obtained by the wear prediction module with the optimal trend of tool wear over time obtained by the big data storage and analysis module.

[0096] S800 uses a gradient optimization algorithm to adjust the tool wear characteristics.

[0097] Based on step S200, the following partial data was obtained:

[0098] Timestamp: 12:00 PM, Vibration frequency: 200 Hz, Temperature: 80°C, Cutting force: 50 N, Cutting speed: 300 m / min, Depth of cut: 5 mm, Feed rate: 0.2 mm / rev, Cutting time: 10 min, Material hardness: HRC40, Lubrication: Normal, Tool shape: Tapered, Cutting mode: Continuous cutting;

[0099] Timestamp: 12:05 PM, Vibration frequency: 210 Hz, Temperature: 82°C, Cutting force: 55 N, Cutting speed: 310 m / min, Depth of cut: 6 mm, Feed rate: 0.25 mm / rev, Cutting time: 12 min, Material hardness: HRC42, Lubrication: Normal, Tool shape: Tapered, Cutting mode: Continuous cutting;

[0100] Timestamp: 12:10 PM, Vibration frequency: 220 Hz, Temperature: 85°C, Cutting force: 60 N, Cutting speed: 320 m / min, Depth of cut: 7 mm, Feed rate: 0.3 mm / rev, Cutting time: 15 min, Material hardness: HRC45, Lubrication: Normal, Tool shape: Circular, Cutting mode: Interrupted cutting;

[0101] According to step S500, 90% of the historical dataset is used for training, and 10% is used for testing. The purpose of this is to better utilize the large-scale data to train the model, while still having enough test data to evaluate the model's generalization ability.

[0102] According to step S600, the steps for calculating the influence of the current tool wear characteristics on the tool wear state using a linear regression model are as follows:

[0103] S601. Divide the historical dataset into a training set and a test set. The training set is used to train the linear regression model, and the test set is used to evaluate the performance of the linear regression model.

[0104] S602. Using the training set, with tool wear characteristics as input variables and the vector set predicting tool wear changes over time as output variables, a linear regression model is fitted using a linear regression algorithm, as shown in the following formula:

[0105] ;

[0106] Where Y is a vector set representing the change of tool wear over time; It is the intercept of the linear regression model; These are the coefficients of the linear regression model, representing the degree of influence of each tool wear characteristic on the tool wear state; The input variables are vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication, tool shape, and cutting pattern. This is the error term; n is the number of input variables;

[0107] S603. Use mean squared error, average error, and coefficient of determination to evaluate the model’s fit to the test set data.

[0108] S604. Store the divided training and test set data in HDFS as input to the big data storage and analysis module.

[0109] According to step S603, the method for evaluating the model's fit to the test set data is as follows:

[0110] S603-1. Evaluation is performed using the mean square error, as shown in the following formula:

[0111] ;

[0112] S603-2. Evaluation is performed using the mean absolute error, as shown in the following formula:

[0113] ;

[0114] S603-3. Evaluation is performed using the coefficient of determination, as shown in the following formula:

[0115] ;

[0116] in, It is the coefficient of determination. It is the mean of the vector of actual tool wear over time. ;

[0117] S603-4, Calculate the evaluation average of mean square error, mean absolute error, and coefficient of determination. G=5, user-defined evaluation threshold. =5, At that time, proceed to step S604.

[0118] According to step S700, the method for comparing the vector set predicting the change of tool wear over time with the optimal trend of tool wear over time is as follows:

[0119] S701. Obtain the vector set of predicted tool wear over time obtained from the wear prediction module, where each row represents a time point and each column represents a tool wear feature. There are j tool wear features and k time points. Let... Let represent the predicted value of tool wear characteristic k at time point j. Then, the vector set representing the predicted tool wear variation over time is expressed as:

[0120] ;

[0121] Wherein, W is a vector set for predicting the change of tool wear over time, specifically obtained through a linear regression model;

[0122] S702, convert the optimal trend of tool wear over time obtained from the big data storage and analysis module into a vector group form;

[0123] S703. Use root mean square error to compare the vector set predicting tool wear over time with the optimal trend vector set of tool wear over time. The comparison result is used to quantify the difference between the two vector sets.

[0124] S704. Analyze and compare the results to evaluate the consistency between the predicted results and the actual trend.

[0125] According to step S702, the steps to convert it to vector group form are as follows:

[0126] S702-1. Obtain the optimal trend of tool wear over time from the big data storage and analysis module, and select a series of time points, wherein the time points are segmented by hour;

[0127] S702-2. For each selected time point, extract the value of the optimal trend of tool wear over time obtained from the big data storage and analysis module, including various characteristics of tool wear.

[0128] S702-3. Combine the trend values ​​extracted at each time point into a vector to form a vector group, where each vector corresponds to the tool wear state at a time point.

[0129] S702-4. Ensure that the optimal trend vector set of tool wear over time obtained from the big data storage and analysis module is aligned in time with the vector set of predicted tool wear over time obtained from the wear prediction module. If the time granularity is different, interpolation or truncation processing is required.

[0130] According to step S800, the method for adjusting the tool wear characteristics using the gradient optimization algorithm is as follows:

[0131] S801. Based on the comparison results, determine the objective function that needs to be optimized;

[0132] S802. Determine the tool wear characteristics that need to be adjusted, and initialize the parameter values ​​of the tool wear characteristics for the gradient optimization algorithm;

[0133] S803. Using gradient descent, a gradient optimization algorithm, define the update rule for each step. Based on the gradient of the objective function, adjust the parameter values ​​along the negative direction of the gradient to gradually reduce the value of the objective function.

[0134] S804. In each iteration, update the parameter values ​​of the tool wear characteristics according to the gradient descent rule, so that the objective function is gradually optimized; repeat this process until the set stopping condition is reached, the stopping condition is reaching the maximum number of iterations, which is set by the user to 5000.

[0135] S805. For each tool wear characteristic, examine whether the optimized parameter values ​​make the predicted values ​​closer to the actual values, and use cross-validation to evaluate the generalization performance.

[0136] S806. Obtain the adjusted tool wear characteristic parameters and adjust the tool wear characteristics.

[0137] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0138] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A tool wear data monitoring system based on big data, characterized in that: include: Acquisition and Processing Module: Uses sensors to collect tool wear-related data in real time, uses filtering algorithms to clean the raw data, aligns the timelines of data collected from different sensors, and extracts tool wear characteristics; The acquisition and processing module extracts tool wear characteristics, which include vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication status, tool shape, and cutting pattern. Using a moving average filtering algorithm, the user selects the filter window size, moves the window across the entire data sequence, calculates the average value of the data points within each window, and uses the new moving average value to replace the corresponding position in the original data. For each time point, the filtering algorithm calculates the corresponding filtered value. For vibration frequency, the frequency components of the vibration signal are extracted using Fourier transform. For temperature, cutting force, cutting speed, depth of cut, feed rate, and material hardness, the filtered data is used directly. For cutting time, the filtered time data is accumulated. For lubrication conditions, information is extracted based on lubricating oil pressure and lubricating fluid state. Based on tool shape and cutting pattern, features related to tool shape and cutting pattern are extracted based on the analysis of the vibration signal. Big Data Storage and Analysis Module: Uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes; utilizes the big data processing framework Apache Spark for distributed computing, analyzes the correlation between tool wear characteristics, and outputs the optimal trend of tool wear changes over time; The big data storage and analysis module implements a data management strategy, performs regular data backups on HDFS, and uses the snapshot function of the Hadoop Distributed File System to create snapshots of tool wear characteristic data; for tool wear characteristic data, a compression algorithm is used to compress the data; Regularly clean up expired or unwanted data to free up storage space. By formulating data retention policies, delete old data based on data timestamps or remaining space. Store tool wear characteristic data in partitions according to time, tool type, or work task. The big data storage and analysis module is configured by the user using a Spark cluster environment, which includes Master nodes and Worker nodes. Spark accesses data in HDFS and allocates resources for distributed computing tasks. Spark SQL, Spark's data loading module, loads tool wear feature data from HDFS into the distributed memory of the Spark cluster. Using Spark's machine learning library MLlib, correlation analysis is employed to analyze the relationships between features; Spark's time series library is used to capture time-related patterns and optimal trends, and linear trends from the time series library are applied for time series analysis. The linear trend model is evaluated by calculating the goodness of fit and root mean square error. Visualize the results of the linear trend and draw a linear trend line to reflect the optimal trend of tool wear over time; Wear Prediction Module: Based on the analysis results obtained from the Big Data Storage and Analysis Module, historical datasets are acquired, and the historical datasets are divided into training sets and datasets. A linear regression model is trained to calculate the impact of the current tool wear characteristics on the tool wear state, the model is evaluated, the real-time collected features are predicted, and a vector group predicting the tool wear change over time is output. Parameter adjustment module: The vector set of predicted tool wear over time obtained from the wear prediction module is compared with the optimal trend of tool wear over time obtained from the big data storage and analysis module, and the tool wear characteristics are adjusted using a gradient optimization algorithm.

2. A tool wear data monitoring method based on big data, applied to the tool wear data monitoring system based on big data as described in any one of claims 1, characterized in that: The steps are as follows: S100: Use sensors to collect tool wear-related data in real time; S200: Use filtering algorithms to clean the raw data collected, align the timelines of data collected from different sensors, and extract tool wear characteristics. S300 uses HDFS to store tool wear characteristics obtained from the acquisition and processing module, implements data management strategies, introduces data security measures, and establishes data indexes; S400 utilizes the big data processing framework Apache Spark for distributed computing to analyze the correlation between tool wear characteristics and output the optimal trend of tool wear over time. S500: Based on the analysis results obtained from the big data storage and analysis module, obtain the historical dataset and divide the historical dataset into training set and dataset; S600: Train a linear regression model to calculate the impact of the current tool wear characteristics on the tool wear state, and output a vector set predicting the change of tool wear over time; According to step S600, the steps for calculating the influence of the current tool wear characteristics on the tool wear state using a linear regression model are as follows: S601. Divide the historical dataset into a training set and a test set. The training set is used to train the linear regression model, and the test set is used to evaluate the performance of the linear regression model. S602. Using the training set, with tool wear characteristics as input variables and the vector set predicting tool wear changes over time as output variables, a linear regression model is fitted using a linear regression algorithm, as shown in the following formula: ； Where Y is a vector set representing the change of tool wear over time; It is the intercept of the linear regression model; These are the coefficients of the linear regression model, representing the degree of influence of each tool wear characteristic on the tool wear state; The input variables are vibration frequency, temperature, cutting force, cutting speed, depth of cut, feed rate, cutting time, material hardness, lubrication, tool shape, and cutting pattern. This is the error term; n is the number of input variables; S603. Use mean squared error, average error, and coefficient of determination to evaluate the model’s fit to the test set data. S604. Store the divided training and test set data in HDFS as input to the big data storage and analysis module. S700. Compare the vector set of predicted tool wear over time obtained by the wear prediction module with the optimal trend of tool wear over time obtained by the big data storage and analysis module. According to step S700, the method for comparing the vector set predicting the change of tool wear over time with the optimal trend of tool wear over time is as follows: S701. Obtain the vector set of predicted tool wear over time obtained from the wear prediction module, where each row represents a time point and each column represents a tool wear feature. There are j tool wear features and k time points. Let... Let represent the predicted value of tool wear characteristic k at time point j. Then, the vector set representing the predicted tool wear variation over time is expressed as: ； Wherein, W is a vector set for predicting the change of tool wear over time, specifically obtained through a linear regression model; S702, convert the optimal trend of tool wear over time obtained from the big data storage and analysis module into a vector group form; S703. Use root mean square error to compare the vector set predicting tool wear over time with the optimal trend vector set of tool wear over time. The comparison result is used to quantify the difference between the two vector sets. S704. Analyze and compare the results to evaluate the consistency between the predicted results and the actual trend; S800 uses a gradient optimization algorithm to adjust the tool wear characteristics.

3. The tool wear data monitoring method based on big data according to claim 2, characterized in that: According to step S603, the method for evaluating the model's fit to the test set data is as follows: S603-1. Evaluation is performed using the mean square error, as shown in the following formula: ； Where m is the number of samples in the test set; It is the i-th vector representing the actual tool wear over time; is the predicted value of tool wear over time for the i-th sample by the linear regression model; MSE is the mean squared error. S603-2. Evaluation is performed using the mean absolute error, as shown in the following formula: ； Where MAE is the mean absolute error; S603-3. Evaluation is performed using the coefficient of determination, as shown in the following formula: ； in, It is the coefficient of determination. It is the mean of the vector of actual tool wear over time; S603-4. Calculate the evaluation average G of the mean square error, mean absolute error, and coefficient of determination; user-defined evaluation threshold. ,when If the condition is met, proceed to step S604; otherwise, re-execute step S601. Wherein, G and... All are positive decimals.

4. The tool wear data monitoring method based on big data according to claim 2, characterized in that: According to step S702, the steps to convert it to vector group form are as follows: S702-1. Obtain the optimal trend of tool wear over time from the big data storage and analysis module, and select a series of time points, wherein the time points are segmented by hour; S702-2. For each selected time point, extract the value of the optimal trend of tool wear over time obtained from the big data storage and analysis module, including various characteristics of tool wear. S702-3. Combine the trend values ​​extracted at each time point into a vector to form a vector group, where each vector corresponds to the tool wear state at a time point. S702-4. Ensure that the optimal trend vector set of tool wear over time obtained from the big data storage and analysis module is aligned in time with the vector set of predicted tool wear over time obtained from the wear prediction module. If the time granularity is different, interpolation or truncation processing is required.

5. The tool wear data monitoring method based on big data according to claim 2, characterized in that: According to step S800, the method for adjusting the tool wear characteristics using the gradient optimization algorithm is as follows: S801. Based on the comparison results, determine the objective function that needs to be optimized; S802. Determine the tool wear characteristics that need to be adjusted, and initialize the parameter values ​​of the tool wear characteristics for the gradient optimization algorithm; S803. Using gradient descent, a gradient optimization algorithm, define the update rule for each step. Based on the gradient of the objective function, adjust the parameter values ​​along the negative direction of the gradient to gradually reduce the value of the objective function. S804. In each iteration, the parameter values ​​of the tool wear characteristics are updated according to the gradient descent rule, so that the objective function is gradually optimized. Repeat this process until a set stopping condition is reached, which is the maximum number of iterations, set by the user. S805. For each tool wear characteristic, examine whether the optimized parameter values ​​make the predicted values ​​closer to the actual values, and use cross-validation to evaluate the generalization performance. S806. Obtain the adjusted tool wear characteristic parameters and adjust the tool wear characteristics.

Images