A numerical control lathe tool wear monitoring method and system

Abstract

The application discloses a numerical control lathe tool wear monitoring method and system, the method comprises the following steps: real-time monitoring of numerical control lathe tool running process data; according to the running process data, the corresponding preset running parameter threshold is matched, when the preset running parameter threshold is monitored, the numerical control lathe tool running process data is transmitted to the edge computing node; using the edge computing node, according to the running process data, the corresponding preset running parameter combination is matched and the running parameter screening is completed; based on the screened running parameter, the key feature selection and extraction are carried out by using the feature selection algorithm, the optimal feature set is obtained and uploaded to the cloud platform; in the cloud platform, according to the running process data, the corresponding tool wear prediction model is matched, the optimal feature set is input into the matched tool wear prediction model, the wear value, the wear degree and the life prediction result are obtained, and the early warning decision and the early warning prompt are carried out. The application realizes efficient, accurate and real-time tool wear monitoring under complex working conditions.

Description

Technical Field

[0001] This application relates to the field of CNC lathe tool wear monitoring technology, and in particular to a method and system for monitoring CNC lathe tool wear. Background Technology

[0002] CNC lathes, as important machine tools, play a crucial role in modern manufacturing. With the continuous improvement of industrial automation, the application range of CNC lathes is becoming increasingly widespread, not only improving production efficiency but also significantly enhancing product quality and stability. However, tool wear on CNC lathes has always been a significant factor restricting their performance. Tool wear can lead to reduced machining accuracy, deteriorated surface quality, and even potential safety accidents; therefore, timely and accurate monitoring of tool wear is of paramount importance.

[0003] In existing technologies, the following methods are commonly used to address the problem of tool wear on CNC lathes: First, the periodic inspection method, which involves manually visually inspecting or periodically measuring changes in tool dimensions to determine the wear condition; second, the vibration / temperature / emission monitoring method, which uses sensors to monitor the vibration signals of the tool during operation, the temperature rise of the tool during cutting, and capture the tiny acoustic signals generated by the tool during cutting, analyzing the changing patterns of these signals to assess the degree of tool wear; and third, the use of intelligent data processing methods, making full use of big data and machine learning technologies to predict and warn of tool loss.

[0004] While the aforementioned methods can effectively monitor tool wear to some extent, they still have some shortcomings. For example, the periodic inspection method relies on the operator's experience and skill level, is easily affected by human factors, and cannot achieve real-time monitoring; although the single vibration / temperature / emission monitoring method can achieve online monitoring, signal interference is significant under complex working conditions, and single signal processing still has deficiencies in measurement accuracy; although intelligent data processing methods can achieve real-time and comprehensive monitoring to a certain extent, most existing prediction methods lack more detailed monitoring of different tool types, different machining materials, and different machining conditions, resulting in the need to improve prediction accuracy. Detailed monitoring often increases the amount of computation, and the lightweight models built into CNC machine tools cannot meet the computational requirements, thus still having shortcomings in achieving timely and accurate monitoring.

[0005] In summary, there is a need to provide a tool wear monitoring method that is efficient, accurate, and real-time, especially one that is adaptable and reliable under complex working conditions. Summary of the Invention

[0006] To achieve efficient, accurate, and real-time tool wear monitoring under complex working conditions and to replace tools in a timely manner, this application provides a method and system for monitoring tool wear on CNC lathes.

[0007] In a first aspect, this application provides a method for monitoring tool wear on a CNC lathe, including:

[0008] Real-time monitoring of CNC lathe tool operation data, including: tool type, machining material, machining conditions, and operating parameters;

[0009] Based on the tool type, machining material, and machining conditions, a corresponding preset operating parameter threshold is matched. When the real-time operating parameter exceeds the preset operating parameter threshold, the CNC lathe tool operation process data at the current moment is transmitted to the edge computing node.

[0010] Using edge computing nodes, corresponding preset operating parameter combinations are matched according to tool type, machining material, and machining conditions, and the operating parameters are filtered based on the matched preset operating parameter combinations. Based on the filtered operating parameters, key features are selected and extracted using a feature selection algorithm to obtain the optimal feature set and upload it to the cloud platform along with the corresponding tool type, machining material, and machining conditions.

[0011] On the cloud platform, a pre-built tool wear prediction model is matched according to the tool type, machining material, and machining condition. The optimal feature set is input into the matched tool wear prediction model to obtain the wear value, wear degree, and life prediction results.

[0012] Based on the obtained wear values, wear levels, and lifespan predictions, early warning decisions are made and early warning prompts are generated.

[0013] By adopting the above scheme, tool operating parameters under complex working conditions are monitored in real time and dynamically matched with preset operating parameter thresholds. Abnormal wear conditions are effectively identified and relevant data is transmitted to edge computing nodes. The edge computing nodes are used to filter and select features of the operating parameters, obtain the optimal feature set, and upload it to the cloud platform. This ensures data quality, reduces data transmission volume, and improves response speed. On the cloud platform, an adaptive matching tool wear prediction model is used to accurately predict the wear degree and life of the tool, improving prediction accuracy. Based on the prediction results, effective early warning decisions are made to assist in timely tool replacement, thereby improving production efficiency and equipment utilization.

[0014] Preferably, the step of matching corresponding preset operating parameter thresholds based on tool type, machining material, and machining conditions includes:

[0015] Collect historical data on the operating parameters and tool wear values ​​of the tool under different tool types, different machining materials, and different combinations of machining conditions. Use deep learning algorithms to obtain the mapping relationship between operating parameters and wear values ​​of the tool under single tool, single machining material, and single or combined machining conditions.

[0016] Based on the constructed mapping relationship, the threshold range of operating parameters under known tool type, known machining material, and known machining conditions at different wear stages is analyzed and obtained; the different wear stages are determined based on the correlation curve generated by calculating the running time and wear value during tool operation; the different wear stages include light wear stage, moderate wear stage, and heavy wear stage;

[0017] The system receives user requests for tool wear monitoring and matches preset threshold ranges for operating parameters at specific wear stages based on these requests. It then determines the corresponding preset operating parameter thresholds based on the matched threshold ranges for the specific wear stages, thus completing the matching of preset operating parameter thresholds according to real-time collected data on tool type, machining material, and machining conditions.

[0018] By adopting the above scheme, historical data is collected and a mapping relationship between operating parameters and tool wear values ​​is established using deep learning algorithms. Then, based on the threshold range of operating parameters at different wear stages and the user's monitoring needs, the preset operating parameter threshold is determined to avoid false alarms or missed alarms in wear monitoring caused by fixed thresholds, and to improve the user experience.

[0019] Preferably, the step of matching the corresponding preset operating parameter combination according to the tool type, machining material, and machining conditions includes:

[0020] Iterate through different tool types, machining materials, and machining condition combinations, and statistically analyze the corresponding historical operating parameters of the tool operation process under a single tool type, a single machining material, and a single or combined machining condition. Set preset operating parameter combinations based on different operating parameters.

[0021] Historical operating parameters and corresponding wear values ​​are filtered out according to each set of preset operating parameter combinations, and training set and test set are divided. The accuracy of the tool wear prediction model generated by training set is obtained according to the test set, and the preset operating parameter combinations are sorted according to the accuracy.

[0022] Based on the preset operating parameter combinations generated under the same conditions of tool type, machining material, and machining conditions, the preset operating parameter combinations corresponding to the tool wear prediction model with the highest accuracy are selected as the matching preset operating parameter combinations. When selecting actual operating parameters, if some operating parameters from the currently matched preset operating parameter combinations are missing from the real-time operating parameters, the next-ranked preset operating parameter combination with an accuracy greater than the preset accuracy is selected as the re-matching preset operating parameter combination. If a preset operating parameter combination with an accuracy greater than the preset accuracy cannot be found by ranking, the preset operating parameter combination with the highest comprehensive score among all preset operating parameter combinations with an accuracy greater than the preset accuracy is selected as the final matching preset operating parameter combination. This completes the matching of corresponding preset operating parameter combinations based on tool type, machining material, and machining conditions, and the corresponding operating parameter selection. The comprehensive score is obtained by weighting the similarity score with the real-time operating parameters and the accuracy score for completing the mapping and supplementation of real-time operating parameters based on missing operating parameters.

[0023] By adopting the above scheme, a comprehensive analysis of historical data is conducted to complete training and testing, determine the optimal preset operating parameter combination to reflect the tool performance under actual working conditions to the greatest extent. Considering that the actual collected data may not be suitable for the optimal preset operating parameter combination, preset operating parameter combinations that meet the accuracy requirements are selected in order. In the case where a preset operating parameter combination that meets the requirements cannot be obtained, the preset operating parameter combination with the highest score is selected by comprehensively judging from the accuracy of supplementing missing data and the similarity score with actual operating parameters. This adaptive screening selects the best operating parameters that can reflect tool wear under the current working conditions, thereby improving the accuracy of subsequent predictions.

[0024] Preferably, the process of selecting and extracting key features based on the filtered operating parameters to obtain the optimal feature set includes:

[0025] When a preset operating parameter threshold is determined based on the threshold range of operating parameters in the light wear stage or the moderate wear stage, the corresponding feature selection algorithm adopts an integrated feature selection algorithm. The integrated feature selection algorithm includes: pre-screening the target features using a variance threshold or mutual information algorithm, and then using a random forest feature importance ranking technique to complete a secondary screening of the features; after selecting and extracting key features using the integrated feature selection algorithm, based on the first preset number of features matching the light wear stage or the moderate wear stage, the same number of features are retained according to the sorting of the screened features.

[0026] When a preset operating parameter threshold is determined based on the threshold range of operating parameters in the severe wear stage, a single feature selection algorithm is adopted for the corresponding feature selection algorithm. The single feature selection algorithm includes: using variance thresholding or mutual information algorithm to filter and sort the importance of target features; after selecting key features using the feature selection algorithm, retaining the same number of features according to the sorted features after filtering, based on the second preset number of features matching the severe wear stage.

[0027] The above scheme selects appropriate feature selection algorithms based on the threshold range of operating parameters at different wear stages. For the light or moderate wear stages, the operating parameters are screened multiple times, retaining only the most important and limited number of features. This improves the selection accuracy of key features and enhances the model's accuracy and robustness. For the severe wear stage, the operating parameters are screened only once to retain a substantial number of features, ensuring the effectiveness of the selected key features and guaranteeing the model's predictive performance under severe wear conditions.

[0028] Preferably, the step of matching the pre-built tool wear prediction model according to the tool type, machining material, and machining conditions includes:

[0029] The tool wear prediction model includes multiple prediction models. Each prediction model includes a multi-branch feature input layer, a multi-branch feature fusion layer, and a multi-branch prediction result output layer. The model is configured to take the output of the branch that outputs the wear value and the output of the branch that outputs the wear degree as the input of the branch that outputs the life prediction result. The model is generated by training a set of historical best feature sets, tool wear values, tool wear degrees, and tool remaining life under a single tool type, a single machining material, and a single or combined machining condition.

[0030] By employing the above scheme, the design of a multi-branch feature input layer, a multi-branch feature fusion layer, and a multi-branch prediction result output layer enables the model to more comprehensively capture and fuse different types of features. Furthermore, by using the wear level output as the feature input for the life prediction result, the model can more effectively transmit feature information and ensure the effectiveness of the output results.

[0031] Preferably, the step of making early warning decisions based on the obtained wear values, wear degree, and lifespan prediction results includes:

[0032] Set different levels of warning rules. Each level of warning rule corresponds to the wear value reaching the preset wear value, the tool wear degree reaching the preset degree, or the tool remaining life being within the preset threshold of the remaining life.

[0033] The urgency of the CNC lathe production plan is obtained, and the warning rules are adjusted accordingly for different urgency levels. The adjusted warning rules include: among the warning rules of the same level, the higher the urgency of the production plan, the higher the tool wear level or the higher the tool wear value or the lower the preset threshold of the remaining tool life.

[0034] Based on the early warning rules, the early warning level corresponding to the obtained wear value, wear degree and life prediction results is determined in order to realize early warning decision-making.

[0035] The above scheme dynamically adjusts the early warning threshold based on the urgency of the CNC lathe production plan, thereby refining the early warning rules and making early warning decisions based on the refined rules.

[0036] Preferred options also include:

[0037] While monitoring the data of the CNC lathe tool's movement process in real time, images of the CNC lathe tool's movement process are simultaneously acquired;

[0038] The wear values, wear degree and life prediction results are statistically obtained, and wear slope-run time curves or life change rate-run time curves corresponding to different wear stages are plotted in real time according to time sequence.

[0039] The system checks in real time whether the wear slope-run time curve does not conform to the trend of the current wear stage or whether the life change rate-run time curve does not conform to the trend of the current wear stage. If they do not conform, the system transmits the CNC lathe tool running process images acquired synchronously to the cloud platform. The system then checks whether a preset event has occurred by querying the tool running process data or analyzing the tool running images. The preset events include tool maintenance events or material idle events. If they exist, no action is taken. If they do not exist, the prediction results of the tool wear prediction model are considered to be pending verification.

[0040] Image recognition technology on a cloud platform is used to analyze images of CNC lathe tool operation process to determine tool wear value, tool wear degree and remaining tool life in order to verify the accuracy of the prediction results of the tool wear prediction model.

[0041] The above scheme performs time series analysis on wear and life changes, and combined with the investigation of preset events, it can promptly detect abnormal trends. Based on the tool operation process images and using image recognition technology, it can confirm the actual wear of the current tool, thereby improving the accuracy and reliability of the prediction results.

[0042] Preferred options also include:

[0043] Real-time monitoring and acquisition of data on the operation of tools in the same batch, wherein the tools in the same batch refer to tools of the same type that process the same material under the same processing conditions;

[0044] Obtain the tool wear prediction output for each tool in the same batch of tools, and obtain the user's satisfaction with the tool wear prediction output for each tool in the same batch of tools. Filter and retain the prediction outputs of several tools whose prediction output satisfaction is greater than the preset satisfaction. Obtain the tool wear prediction output of each retained tool in the same batch of tools and construct the corresponding wear curve. Use the dynamic time warping algorithm to calculate the similarity of the wear curves of the prediction output of tools in the same batch.

[0045] The feature matrix is ​​constructed by extracting features corresponding to tools in the same batch whose wear curves have a similarity greater than the preset similarity of the wear curves. The covariance matrix is ​​calculated, and principal component features with a variance contribution rate greater than the preset contribution rate are selected by eigenvalue decomposition. The weights of each feature value are dynamically allocated according to the contribution rate and fed back to the tool wear prediction model to adjust the feature weight ratio in the tool wear prediction model at the next moment.

[0046] By introducing the above scheme, a batch-volume tool correlation feedback mechanism and a user satisfaction feedback mechanism are introduced to dynamically adjust the weight parameters of each feature in the tool wear prediction model, thereby helping to obtain prediction outputs with high accuracy and meeting user needs.

[0047] Preferably, the process of generating a pre-built tool wear prediction model that matches the tool type, machining material, and machining condition includes:

[0048] For a single tool type, a single machining material, or a combination of single or combined machining conditions, several tool wear prediction models to be trained are designed. The number of tool wear prediction models to be trained is determined to be N+1 based on the number of tools in the same batch N. The feature weight parameters in each tool wear prediction model to be trained are distinguished in advance.

[0049] One of the tool wear prediction models to be trained is used as the global model, and each of the remaining tool wear prediction models to be trained is used as a sub-model and pre-matched with one tool in the same batch of tools.

[0050] The historical best feature set extracted from the historical operation data of each tool in the same batch of tools is used as the corresponding input to the matching sub-model to complete the training of the matching sub-model. After training, the feature weights and gradients of each sub-model are uploaded to the global model. Combined with the federated learning algorithm, the feature weights of each sub-model are aggregated to obtain aggregated feature weights, which are then distributed to each sub-model for iterative updates until the global model converges. The converged global model is then used as the trained tool wear prediction model, generating pre-built tool wear prediction models for single tool type, single machining material, and single or combined machining conditions.

[0051] In order to obtain more accurate tool wear prediction results under single tool type, single machining material, and single or combined machining conditions, the above scheme combines federated learning algorithm to match and design multiple models with different feature weight parameters according to the same batch of tools to obtain the globally optimal tool prediction model, thereby improving prediction accuracy.

[0052] Secondly, this application provides a CNC lathe tool wear monitoring system, comprising:

[0053] The data acquisition module is used to monitor the data of the CNC lathe tool operation process in real time.

[0054] The running data processing module is used to match the corresponding preset running parameter thresholds according to the tool type, machining material, and machining conditions. When the real-time running parameters exceed the preset running parameter thresholds, the CNC lathe tool running process data at the current moment is transmitted to the edge computing node.

[0055] The optimal feature set acquisition module is used to utilize edge computing nodes to match corresponding preset operating parameter combinations based on tool type, machining material, and machining conditions, and to complete the filtering of operating parameters based on the matched preset operating parameter combinations. Based on the filtered operating parameters, the module uses a feature selection algorithm to select and extract key features, obtain the optimal feature set, and upload it along with the corresponding tool type, machining material, and machining conditions to the cloud platform.

[0056] The wear and life prediction module is used on the cloud platform to match the corresponding pre-built tool wear prediction model according to the tool type, machining material and machining conditions. The optimal feature set is input into the matched tool wear prediction model to obtain the wear value, wear degree and life prediction results.

[0057] The early warning decision and prompt generation module is used to combine the production plan with the obtained wear value, wear degree and life prediction results to make early warning decisions and generate early warning prompts.

[0058] By adopting the above scheme and combining edge computing technology and cloud platform technology, we can achieve adaptive screening of operating parameters under different working conditions and acquisition of the optimal feature set, thereby realizing real-time and accurate prediction of the wear state of CNC lathe tools.

[0059] In summary, this application has the following beneficial effects:

[0060] 1. Real-time acquisition of tool operation data and intelligent judgment based on preset operating parameter thresholds ensure rapid response even under complex working conditions; by combining edge computing nodes and cloud platforms, adaptive filtering from operating parameters to operating parameter features is achieved to obtain the optimal feature set and adaptively match the tool wear prediction model to complete real-time, efficient and accurate prediction.

[0061] 2. By using a multi-level early warning mechanism, dynamically adjusting early warning rules based on the urgency of production plans, and generating early warning decisions based on predicted wear values, wear levels, and remaining lifespan, the stability and safety of production are improved.

[0062] 3. Through time series analysis, the changing trend of tool wear was observed. When abnormal trends were identified, images of the CNC lathe tool running process were acquired simultaneously. The accuracy of the tool wear prediction model was verified by analyzing the images, which enhanced the reliability and robustness of the system. Attached Figure Description

[0063] Figure 1 This is a flowchart of the CNC lathe tool wear monitoring method described in a specific embodiment;

[0064] Figure 2 This is a schematic diagram of the CNC lathe tool wear monitoring system described in a specific embodiment. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0066] like Figure 1 As shown in the figure, this application discloses a method for monitoring tool wear on a CNC lathe, the main steps of which include:

[0067] S1. Real-time monitoring of CNC lathe tool operation data.

[0068] Specifically, it monitors CNC lathe tool operation data in real time, including tool type and usage time, machining material, machining conditions, and operating parameters.

[0069] The system can determine the type of tool used in operation by recording the tool model and material, such as drills, turning tools, and milling cutters; and determine the machining material by recording the machining material, such as high-speed steel, cemented carbide, and ceramic materials. It can also collect data on the machining conditions of the lathe tools in real time through an installed sensor network, such as finishing, semi-finishing, and roughing based on machining parameters like tool feed rate and depth of cut. Furthermore, it can collect data on the operating parameters of the lathe tools in real time through the installed sensor network, including parameters reflecting tool wear such as current data, temperature data, vibration signals, sound signals, and cutting force signals.

[0070] In addition, to further ensure the accuracy of the monitoring results, high-precision camera devices are used to simultaneously collect images of the CNC lathe tool's operation process while monitoring the CNC lathe tool's running process data in real time.

[0071] S2. Match preset operating parameter thresholds based on real-time monitoring data of CNC lathe tool operation process.

[0072] Considering that synchronously processing real-time monitoring data of CNC lathe tool operation, wear level and life monitoring requires huge computing power, while the computing resources of CNC lathe are limited, based on actual production experience and actual data, under specific tool materials, specific machining materials and specific machining conditions, we obtain operating parameters that may lead to abnormal situations such as severe or even excessive wear, and set corresponding preset operating parameter thresholds to judge the operating conditions with a risk of severe or even excessive wear.

[0073] For example, when machining aluminum alloy materials, if the cutting force exceeds 100N, it is considered that there may be a risk of severe wear, and a preset threshold for the cutting force is set accordingly.

[0074] S3. Based on the matching preset operating parameter threshold, transmit the operating process data corresponding to the operating parameters that exceed the preset operating parameter threshold to the edge computing node.

[0075] Specifically, when real-time operating parameters are detected to exceed preset operating parameter thresholds, that is, when any type of real-time operating parameter exceeds the corresponding type of preset operating parameter threshold, such as temperature, vibration frequency, or cutting force, the current CNC lathe tool operation process data is transmitted to the edge computing node.

[0076] S4. Use edge computing nodes to filter the acquired CNC lathe tool running process data, obtain the optimal feature set, and upload it to the cloud platform.

[0077] Specifically, to reduce the data transmission burden, edge computing nodes are used to handle the initial processing and filtering of process data, so that only key data points are transmitted, improving data processing efficiency and enabling timely prediction; the specific steps include:

[0078] Considering that different types of cutting tools, under different machining conditions and cutting different materials, have varying degrees of importance in determining wear status, different operating parameters can be used in different tool operation scenarios. For example, cutting force and vibration data are often more important in the roughing stage, while temperature and sound signals are more important in the finishing stage. Therefore, based on actual expert production experience and data, the optimal combination of operating parameters reflecting tool wear under different tool operation scenarios can be obtained. Preset operating parameter combinations suitable for specific tool materials, machining materials, and machining conditions are then stored on various edge computing nodes. Using these edge computing nodes, corresponding preset operating parameter combinations are matched based on tool type, machining material, and machining condition, and the operating parameters are filtered based on the matched preset operating parameter combinations.

[0079] After obtaining the operating parameters through screening, data processing can be performed on the determined operating parameters, including data synchronization, data filtering, data cleaning, and data compensation. For heterogeneous data, the same sampling interpolation or dimensionality reduction processing can be performed.

[0080] After processing the operating parameters, to reduce the data transmission burden and ensure accurate and timely model prediction, edge computing nodes are used to extract physical features strongly correlated with tool wear, building a high-information dataset, i.e., the optimal feature set, based on the filtered and processed operating data. Specifically, feature selection algorithms can be used for key feature selection and extraction, including: feature extraction from operating parameters, including: extraction of time-domain features (RMS, peak value, impulse factor, etc.) and frequency-domain features (FFT spectrum centroid, wavelet packet energy entropy, etc.) and time-frequency analysis (time-spectrum graph generated by STFT, etc.) from vibration signals / acoustic emission signals; extraction of statistical features (such as mean, variance, interpeak values, etc.) from current signals and cutting force signals; and feature selection using various algorithms such as principal component analysis, variance thresholding, mutual information, or random forest feature importance ranking to retain key features and generate the optimal feature set.

[0081] Finally, the optimal feature set and its corresponding tool type, machining material, machining conditions and other process data are uploaded to the cloud platform.

[0082] S5. On the cloud platform, match the corresponding pre-built tool wear prediction model according to the tool type, machining material, and machining conditions. Input the optimal feature set into the matched tool wear prediction model to obtain the wear value, wear degree and life prediction results.

[0083] Specifically, considering the differences in tool wear states under specific operating scenarios such as combinations of specific tool materials, specific machining materials, and specific machining conditions, in order to more accurately predict adaptive wear results for different operating scenarios, tool wear prediction models matching specific tool materials, specific machining materials, and specific machining conditions can be pre-built and built into a cloud platform. This allows the cloud platform to support the storage and application of multiple different tool wear prediction models. Each pre-built tool wear prediction model can employ various machine learning algorithms, such as support vector machines (SVM) and neural networks, and be trained and generated using the optimal feature set obtained from a large amount of historical data, as well as historical actual wear values, wear degree, and remaining life.

[0084] On the cloud platform, a pre-built tool wear prediction model is matched according to the tool type, machining material, and machining condition. The optimal feature set is input into the matched tool wear prediction model to obtain the wear value, wear degree, and life prediction results.

[0085] S6. Based on the obtained wear value, wear degree and life prediction results, make early warning decisions and generate early warning prompts.

[0086] Specifically, based on the obtained wear level and life prediction results, it can be determined whether the tool needs to be replaced and a corresponding warning can be issued, and a warning message can be generated. In order to realize a multi-protection mechanism, different levels of warning rules are set, such as moderate wear and severe wear, which correspond to different warning levels. Reasonable warning decisions are set according to different warning levels, such as different colored warning indicators, which are transmitted and displayed on the display terminal of the CNC machine tool.

[0087] Using the method described in the above embodiments, data on the operation of CNC lathe tools is collected in real time. Through preset operating parameter threshold matching, edge computing node processing, and cloud platform prediction, efficient, accurate, and real-time tool wear monitoring is achieved.

[0088] In a specific embodiment, to more accurately obtain preset operating parameter thresholds under complex working conditions and ensure accurate wear prediction, a deep learning algorithm is used to establish a mapping relationship between operating parameters and tool wear values, forming an empirical model. This better assists in matching and obtaining preset operating parameter thresholds. The method of matching corresponding preset operating parameter thresholds based on tool type, machining material, and machining conditions specifically includes:

[0089] The system collects historical data on the operating parameters and tool wear values ​​of tools under different tool types, machining materials, and machining conditions. For example, it collects historical operating parameters and tool wear values ​​of milling cutters during the finishing of cemented carbide materials. The collected historical data is clustered to construct a training dataset based on historical data on operating parameters and tool wear values ​​corresponding to specific machining conditions such as single tool, single machining material, and single or combined machining conditions. Deep learning algorithms are then used to obtain the mapping relationship between operating parameters and wear values ​​during tool operation under single tool, single machining material, and single or combined machining conditions. In other words, the system obtains the mapping relationship between operating parameters and wear values ​​as tool operating time increases through training.

[0090] Based on the constructed mapping relationship, the threshold range of operating parameters under known tool type, known machining material, and known machining conditions at different wear stages is analyzed and obtained. Among them, the different wear stages are mainly determined by the correlation curves generated by fitting the tool running time and wear value under historical single tool, single machining material, and single or combined machining conditions. For example, the wear value within the first preset wear value range is identified as the light wear stage, the wear value within the second preset wear value range is identified as the moderate wear stage, and the wear value within the third preset wear value range is identified as the heavy wear stage.

[0091] The preset operating parameter thresholds are determined based on user needs. When the user's requirement for tool wear monitoring is the first requirement, i.e., the user requires high operating performance for worn tools, the operating parameter threshold range for the light wear stage is matched, and the preset operating parameter threshold is determined accordingly. When the user's requirement for tool wear monitoring is the second requirement, i.e., the user requires moderate operating performance for worn tools, the operating parameter threshold range for the medium wear stage is matched, and the preset operating parameter threshold is determined accordingly. When the user's requirement for tool wear monitoring is the third requirement, i.e., the user does not require strict operating performance for worn tools, the operating parameter threshold range for the heavy wear stage is matched, and the preset operating parameter threshold is determined accordingly.

[0092] Furthermore, it is necessary to continuously perform reinforcement learning on the determination of preset operating parameter thresholds to optimize data transmission during operation. Specifically, the operating parameters and tool wear values ​​corresponding to specific machining conditions such as single tool, single machining material, and single or combined machining conditions are periodically supplemented as a supplementary set of training data. This allows for training and optimization of the mapping relationship between operating parameters and wear values ​​during tool operation under the corresponding single tool, single machining material, and single or combined machining conditions. Based on the optimized mapping relationship, the threshold range of operating parameters at different wear stages under known tool type, known machining material, and known machining conditions is re-analyzed and obtained.

[0093] In a specific embodiment, to more accurately obtain preset operating parameter combinations under complex working conditions and ensure accurate wear prediction, a deep learning algorithm is used to obtain the most accurate operating parameter combination corresponding to predicting tool wear under complex working conditions. The method of matching the corresponding preset operating parameter combination based on tool type, machining material, and machining conditions specifically includes:

[0094] Iterate through different tool types, machining materials, and machining condition combinations, and statistically analyze the corresponding historical operating parameters of the tool operation process under a single tool type, a single machining material, and a single or combined machining condition. Set preset operating parameter combinations based on different operating parameters.

[0095] By combining a single tool type, a single machining material, and single or combined machining conditions, historical operating parameters and corresponding wear values ​​are selected according to each set of preset operating parameter combinations, and training sets and test sets are divided. The accuracy of the prediction results of the tool wear prediction model generated by training the training set is obtained based on the test set, and the preset operating parameter combinations under the corresponding combination conditions are sorted according to the order of accuracy.

[0096] Based on the preset operating parameter combinations generated under the same combination conditions according to tool type, machining material, and machining conditions, the preset operating parameter combination corresponding to the tool wear prediction model with the highest accuracy is selected as the matching preset operating parameter combination. For example, the preset operating parameter combinations under the combination conditions of milling cutter-carbide-finishing are sorted as (A,B,D)\(A,B), (B,C,D)\(B,D), and (A,B,D) is selected as the matching preset operating parameter combination.

[0097] When selecting actual operating parameters, some operating parameters may be missing from the currently matched preset operating parameter combinations in the real-time operating parameters. For example, if the actual operating parameters are (B, D), then the next preset operating parameter combination with higher precision is selected as the new preset operating parameter combination to be matched, according to the preset operating parameter combination sorting. For example, if the next preset operating parameter combination (A, B) that meets the precision requirement is still missing, and the preset operating parameter combination (B, D) that matches the actual operating parameters is found, but the precision of this combination is lower than the preset precision requirement, that is, if a preset operating parameter combination with higher precision cannot be found by sorting, then data compensation is used to determine the preset operating parameter combination.

[0098] Specifically, considering the accuracy of prediction results, data compensation cannot be simply performed randomly. It is necessary to start from the perspective of actual operating parameters and construct mapping relationships between different operating parameters based on the correlation between them. For example, compensation for operating parameter A (acoustic emission signal) can be completed based on the mapping relationship between operating parameter B (cutting force) and A (acoustic emission signal), which in turn can compensate for operating parameter C. The accuracy of mapping and supplementing missing operating parameters in real time is determined according to the strength of the correlation between operating parameters, and a corresponding accuracy score for mapping and supplementing missing operating parameters in real time is obtained. The stronger the correlation, the stronger the accuracy. The similarity between the preset operating parameter combination that completes the compensation for missing operating parameters and the real-time operating parameters is calculated. The similarity score is matched according to the calculation results, and a comprehensive score is obtained by weighted calculation (e.g., the score of (A,B,D) is less than that of (B,C,D)). The preset operating parameter combination with the highest score is determined as the final matched preset operating parameter combination. This completes the matching of the corresponding preset operating parameter combination and the selection of the corresponding operating parameters according to the tool type, machining material, and machining conditions.

[0099] In a specific embodiment, to further improve the selection accuracy of key features, the method, based on the filtering operating parameters, utilizes a feature selection algorithm to select and extract key features to obtain the optimal feature set, including:

[0100] Considering that when the real-time operating parameters exceed the threshold range corresponding to the operating parameters in the light or moderate wear stage, the preset operating parameter threshold is determined, it is highly likely that the current operating tool wear is at a light or moderate wear level, and the possibility of needing to replace the tool is small. This provides a certain prediction fault tolerance, effectively removes redundant and noisy features, and retains only the most important features. Specifically, when the preset operating parameter threshold is determined according to the threshold range corresponding to the operating parameters in the light or moderate wear stage, the corresponding feature selection algorithm adopts an integrated feature selection algorithm. The integrated feature selection algorithm includes: pre-screening the target features using variance thresholding or mutual information algorithm, and then using random forest feature importance ranking technology to complete the secondary screening of features; after selecting and extracting key features using the integrated feature selection algorithm, according to the first preset number of features matching the light wear stage, the same number of features are retained according to the sorting of the screened features.

[0101] Accordingly, considering that when the real-time operating parameters exceed the threshold range corresponding to the operating parameters in the severe wear stage, the preset operating parameter threshold is likely to be in a state of severe wear, requiring tool replacement, resulting in low prediction tolerance and the need to retain an appropriate number of features, specifically, when the preset operating parameter threshold is determined according to the threshold range of the operating parameters in the severe wear stage, a single feature selection algorithm is adopted. The single feature selection algorithm includes: using variance thresholding or mutual information algorithm to filter and sort the importance of target features; after selecting key features using the feature selection algorithm, retaining the same number of features according to the sorted features after filtering, based on the second preset number of features matching the severe wear stage.

[0102] The number of first preset features matching the light wear stage or the moderate wear stage is less than the number of second preset features matching the heavy wear stage.

[0103] In a specific embodiment, to ensure that the designed tool wear prediction model can more comprehensively capture and integrate different types of features, thereby improving the prediction accuracy of tool wear degree and life, the method includes...

[0104] The matching of pre-built tool wear prediction models based on tool type, machining material, and machining conditions includes:

[0105] The tool wear prediction model includes multiple prediction models, each corresponding to a single tool type, a single machining material, and single or combined machining conditions. Each model is designed with a multi-branch feature input layer, a multi-branch feature fusion layer, and a multi-branch prediction result output layer. To further ensure consistency between tool wear values, tool wear degree, and remaining life predictions, and to improve the accuracy of remaining life predictions, a branch outputting wear values ​​and a branch outputting wear degree are provided as inputs to the branch outputting life prediction results. Each prediction model is generated through training using a set of historical best feature sets, tool wear values, tool wear degree, and remaining tool life under a single tool type, a single machining material, and single or combined machining conditions.

[0106] The multi-branch feature fusion layer is designed with multiple sub-branch feature fusion layers. The first sub-branch feature fusion layer performs weighted fusion on each feature input branch in the optimal feature set. The weight ratio is determined according to the importance ranking and quantity in the optimal feature set. The second sub-branch feature fusion layer is designed to fuse each feature input branch in the optimal feature set with the wear value branch and the wear degree branch output respectively, and then perform weighted fusion on the fused content.

[0107] Furthermore, to further optimize the tool wear prediction values, the method also includes: adjusting the multi-branch feature input layer to include the multi-branch feature input layer of the previous moment, the multi-branch feature input layer of the next moment, and the multi-branch feature input layer of the current moment; adjusting the first sub-branch feature fusion layer and the second sub-branch feature fusion layer in the multi-branch fusion layer, both of which complete feature fusion by intermittent cross-fusion, specifically including: during actual feature fusion, fusing the input features of the multi-branch feature input layer of the previous moment with the same type of input features of the multi-branch feature input layer of the current moment to complete feature fusion at a certain moment, and fusing the input features of the multi-branch feature input layer of the next moment with the same type of input features of the multi-branch feature input layer of the current moment to complete feature fusion at the next moment of a certain moment.

[0108] The method further includes: obtaining user satisfaction with the wear level and life prediction results; if the user satisfaction with the wear level and life prediction results is lower than the preset satisfaction level, then it is determined that the pre-built tool wear prediction model matched according to tool type, machining material, and machining condition needs to be optimized and trained. Incremental learning is used to incrementally train the pre-built tool wear prediction model matched according to tool type, machining material, and machining condition, and then the user satisfaction with the wear level and life prediction results is lower than the preset satisfaction level.

[0109] In a specific embodiment, to further ensure the accuracy of the early warning, and in conjunction with existing production plans, to better avoid premature replacement of tools still within their service life or machining accidents caused by failure to detect severe wear in a timely manner, the method of making early warning decisions based on the obtained wear degree and life prediction results further includes:

[0110] For different levels of warning rules, each level corresponds to a tool wear value reaching a preset value, a preset degree of tool wear, or a remaining tool life within a preset threshold range. For example: Level 1 warning corresponds to a tool wear value reaching a first preset value, a tool wear degree reaching a first preset degree (light wear), and a remaining tool life below a first preset threshold (50% of life) but above a second preset threshold (30% of life); Level 2 warning corresponds to a tool wear value reaching a second preset value, a tool wear degree reaching a second preset degree (moderate wear), and a remaining tool life below a second preset threshold (30% of life) but above a third preset threshold (10% of life); Level 3 warning corresponds to a tool wear value reaching a third preset value, a tool wear degree reaching a third preset degree (heavy wear), and a remaining tool life below a third preset threshold (10% of life) but above 0.

[0111] To determine the urgency of the CNC lathe production plan, the warning rules are adjusted accordingly based on different levels of urgency. These adjusted warning rules include: within the same level of warning rules, higher production plan urgency corresponds to higher tool wear values ​​reaching preset tool wear thresholds, higher tool wear levels reaching preset thresholds, and lower remaining tool life thresholds. For example, for general urgency, the original warning rules are maintained; for severe urgency, the corresponding level of warning rules is adjusted. For instance, a first-level warning corresponds to tool wear values ​​reaching the second preset tool wear value, tool wear levels reaching the second preset level (moderate wear), and low remaining tool life. The second level warning corresponds to the tool wear value reaching the third preset value, the tool wear degree reaching the third preset degree (severe wear), and the remaining tool life being lower than the third preset threshold for remaining tool life (10% of life) but higher than the fourth preset threshold for remaining tool life (5% of life); the third level warning corresponds to the tool wear value reaching the fourth preset value, the tool wear degree reaching the fourth preset degree (excessive wear), and the remaining tool life being lower than the fourth preset threshold for remaining tool life (5% of life) but higher than 0.

[0112] Based on the early warning rules, the early warning level corresponding to the obtained wear value, wear degree and life prediction results is determined in order to realize early warning decision-making.

[0113] In one specific embodiment, to further ensure the accuracy of tool wear prediction, in addition to using operating parameters collected by sensors, image data from the operating process is also used to analyze and verify the accuracy of tool wear prediction. The method further includes:

[0114] While monitoring the data of the CNC lathe tool's movement process in real time, images of the CNC lathe tool's movement process are simultaneously acquired;

[0115] The wear values, wear degree and life prediction results are statistically obtained, and wear slope-run time curves or life change rate-run time curves corresponding to different wear stages are plotted in real time according to time sequence.

[0116] According to the pattern of tool wear, as the running time increases, the wear slope gradually decreases in the light wear stage, exhibits positive and negative changes in the wear slope in the moderate wear stage, and gradually increases in the heavy wear stage. By real-time querying of the wear slope-runtime curve, if any discrepancy is found between the current wear stage and the expected trend, a potential wear prediction error is identified, and the synchronously acquired CNC lathe tool running process images are transmitted to the cloud platform. Similarly, according to the pattern of tool wear, tool life continuously decreases. In the light wear stage, the tool life gradually decreases; in the moderate wear stage, the tool life gradually flattens out; and in the heavy wear stage, the tool life gradually increases. By real-time querying of the life change rate-runtime curve, if any discrepancy is found between the current wear stage and the expected trend, a potential wear prediction error is identified, and the synchronously acquired CNC lathe tool running process images are transmitted to the cloud platform.

[0117] Meanwhile, considering that some events may cause irregular changes, such as: tool maintenance, or the tool not actually cutting due to the absence of material processing due to operational errors in the tool operation production line, or incorrect tool operation parameter assignment, these phenomena are prioritized for investigation. These events are pre-set as preset events. By querying tool operation process data and analyzing tool operation image data, it is determined whether preset events have occurred. If they exist, it indicates that the abnormal trend is caused by preset events and no action is taken. If they do not exist, it is determined that the prediction results of the tool wear prediction model need to be verified.

[0118] Image recognition technology on a cloud platform is used to analyze images of CNC lathe tool operation process to determine the degree of tool wear and life prediction, thereby verifying the accuracy of the prediction results of the tool wear prediction model.

[0119] In one specific embodiment, to further improve the accuracy of tool wear prediction, a user feedback mechanism and a batch feedback mechanism are designed to optimize the weight parameters of each feature in the wear prediction model, thereby improving prediction accuracy. The method further includes:

[0120] Considering the similarity of machining among tools in the same batch, and the differences in the data collected by the sensor network during the machining process, a batch feedback mechanism for tools in the same batch is designed to adjust feature weights and monitor and acquire the data of the machining process of tools in the same batch in real time; wherein, the tools in the same batch refer to tools of the same type that are used to machine the same material under the same machining conditions.

[0121] Accordingly, the optimal feature set of the edge nodes is obtained and input into the tool wear prediction model for prediction. The tool wear prediction output corresponding to each tool in the same batch is obtained. In addition to considering the feedback information of the same batch of tools, the user's satisfaction with the prediction results is also considered. The prediction results of tools with high user satisfaction are selected as much as possible for weight adjustment. Therefore, the user's satisfaction with the tool wear prediction output corresponding to each tool in the same batch is obtained, and the prediction outputs of several tools with prediction output satisfaction greater than the preset satisfaction are selected and retained.

[0122] The tool wear prediction outputs for individual tools from the same batch of tools are obtained and corresponding wear curves are constructed. A dynamic time warping algorithm is used to calculate the similarity of the wear curves predicted from the same batch of tools. Tool wear prediction outputs corresponding to tools with low similarity are removed based on the similarity. Features corresponding to tools from the same batch with wear curve similarity greater than a preset similarity are extracted to construct a feature matrix. The covariance matrix is ​​calculated, and principal component features with a variance contribution rate greater than a preset contribution rate (e.g., features with a contribution rate greater than 85%) are selected through eigenvalue decomposition. The weights of each feature value are dynamically allocated according to the contribution rate and fed back to the tool wear prediction model to adjust the weight ratios of each feature in the tool wear prediction model at the next time step. The dynamic weight allocation formula is as follows:

[0123] In the formula, The weights of the eigenvalues ​​of the i-th principal component are... Let be the eigenvalue of the i-th principal component.

[0124] Furthermore, considering the similarity of machining conditions among tools in the same batch, and the differences in data collected by the sensor network during the machining process, a wear prediction model suitable for all tools in the same batch is trained and constructed using the difference data between different tools under the same machining conditions (maintaining machining condition adaptability) (considering the potential differences in actual data collection). Specifically, the process of generating a pre-built tool wear prediction model that matches the tool type, machining material, and machining condition includes:

[0125] For a single tool type, a single machining material, or a combination of single or combined machining conditions, several tool wear prediction models to be trained are designed. The number of tool wear prediction models to be trained is determined to be N+1 based on the number of tools in the same batch N. The feature weight parameters in each tool wear prediction model to be trained are pre-distinguished. That is, for N tool wear prediction models with the same structure, the feature weight parameters in the model are adjusted to present different features.

[0126] One of the tool wear prediction models to be trained is used as the global model, and each of the remaining tool wear prediction models to be trained is used as a sub-model and pre-matched with one tool in the same batch of tools, that is, each sub-model corresponds to a tool number.

[0127] The historical optimal feature set extracted from the historical operating data of each tool in the same batch is used as input to the matching sub-model to complete the training of the matching sub-model. After training, the feature weights and gradients of each sub-model are uploaded to the global model. Combined with the federated learning algorithm, the feature weights of each sub-model are aggregated to obtain the aggregated feature weights. The specific aggregation formula includes:

[0128] In the formula, For aggregated feature weights, The aggregated feature weights of the previous iteration order, This is the gradient of the sub-model.

[0129] The aggregated feature weights are distributed to each sub-model and iteratively updated until the global model converges. The converged global model is then used as the trained tool wear prediction model, generating pre-built tool wear prediction models for single tool type, single machining material, and single or combined machining conditions.

[0130] like Figure 2 As shown in the figure, this application discloses a CNC lathe tool wear monitoring system, including:

[0131] The data acquisition module 101 is used to monitor the data of the CNC lathe tool operation process in real time.

[0132] The running data processing module 102 is used to match the corresponding preset running parameter thresholds according to the tool type, machining material, and machining conditions. When the real-time running parameters exceed the preset running parameter thresholds, the CNC lathe tool running process data at the current moment is transmitted to the edge computing node.

[0133] The optimal feature set acquisition module 103 is used to utilize edge computing nodes to match corresponding preset operating parameter combinations based on tool type, machining material, and machining conditions, and to complete the filtering of operating parameters based on the matched preset operating parameter combinations; based on the filtered operating parameters, the module uses a feature selection algorithm to select and extract key features, obtain the optimal feature set, and upload it along with the corresponding tool type, machining material, and machining conditions to the cloud platform.

[0134] The wear and life prediction module 104 is used on the cloud platform to match the corresponding pre-built tool wear prediction model according to the tool type, machining material and machining conditions, input the optimal feature set into the matched tool wear prediction model, and obtain the wear value, wear degree and life prediction results.

[0135] The early warning decision and prompt generation module 105 is used to combine the production plan and, based on the obtained wear value, wear degree and life prediction results, make early warning decisions and generate early warning prompts.

[0136] In one specific embodiment, the system further includes:

[0137] The operation data acquisition module 101 is also used to simultaneously acquire images of the CNC lathe tool operation process while monitoring the CNC lathe tool operation process data in real time;

[0138] The wear and life prediction verification module 106 is used to statistically analyze the acquired wear values, wear degree, and life prediction results, and to plot the wear slope-run time curve or life change rate-run time curve corresponding to different wear stages in real time according to the time sequence. It also queries in real time whether the wear slope-run time curve or the life change rate-run time curve does not conform to the change trend of the current wear stage. If not, it transmits the synchronously acquired CNC lathe tool running process image to the cloud platform, and determines whether a preset event has occurred by querying the tool running process data or analyzing the tool running image. The preset events include tool maintenance events or material idle events. If they exist, no processing is performed; if they do not exist, the prediction result of the tool wear prediction model is deemed to need verification. The cloud platform's image recognition technology is used to analyze the CNC lathe tool running process image to determine the tool wear value, tool wear degree, and remaining tool life to verify the accuracy of the tool wear prediction model's prediction results.

[0139] In one specific embodiment, the running data acquisition module 101 in the system is also used to monitor and acquire the running process data of the same batch of tools in real time. The same batch of tools refers to tools that use the same tool type to process the same material under the same processing conditions.

[0140] The wear and life prediction module 104 is further configured to obtain the tool wear prediction output corresponding to each tool in the same batch of tools, and obtain the user's satisfaction with the tool wear prediction output corresponding to each tool in the same batch of tools, filter and retain the prediction outputs of several tools whose prediction output satisfaction is greater than the preset satisfaction; obtain the tool wear prediction output of each retained tool in the same batch of tools and construct wear curves accordingly, and use a dynamic time warping algorithm to calculate the similarity of the wear curves of the prediction outputs of the same batch of tools; obtain the features extracted from the same batch of tools whose wear curve similarity is greater than the preset similarity of the wear curves to construct a feature matrix, calculate the covariance matrix, and filter the principal component features whose variance contribution rate is greater than the preset contribution rate through eigenvalue decomposition, and dynamically allocate the weight of each feature value according to the contribution rate and feed it back to the tool wear prediction model to adjust the feature weight ratio in the tool wear prediction model at the next moment.

[0141] This application also discloses a computer-readable storage medium.

[0142] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the CNC lathe tool wear monitoring method described above. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0143] This application also discloses a computer device.

[0144] Specifically, the computer device includes a memory and a processor, and the memory stores a computer program that can be loaded by the processor and executed as described above for monitoring the wear of CNC lathe tools.

[0145] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for monitoring tool wear on a CNC lathe, characterized in that, include: Real-time monitoring of CNC lathe tool operation data, including: tool type, machining material, machining conditions, and operating parameters; Based on the tool type, machining material, and machining conditions, a corresponding preset operating parameter threshold is matched. When the real-time operating parameters exceed the preset operating parameter threshold, the CNC lathe tool operation process data at the current moment is transmitted to the edge computing node. Using edge computing nodes, corresponding preset operating parameter combinations are matched according to tool type, machining material, and machining conditions, and the operating parameters are filtered based on the matched preset operating parameter combinations. Based on the filtered operating parameters, key features are selected and extracted using feature selection algorithms to obtain the optimal feature set and upload it to the cloud platform along with the corresponding tool type, machining material, and machining conditions. On the cloud platform, a pre-built tool wear prediction model is matched according to the tool type, machining material, and machining condition. The optimal feature set is input into the matched tool wear prediction model to obtain the wear value, wear degree, and life prediction results. Based on the obtained wear values, wear levels, and life prediction results, early warning decisions are made and early warning prompts are generated; The preset operating parameter thresholds include: Collect historical data on the operating parameters and tool wear values ​​of the tool under different tool types, different machining materials, and different combinations of machining conditions. Use deep learning algorithms to obtain the mapping relationship between operating parameters and wear values ​​of the tool under single tool, single machining material, and single or combined machining conditions. Based on the constructed mapping relationship, the threshold range of operating parameters under known tool type, known machining material, and known machining conditions at different wear stages is analyzed and obtained; the different wear stages are determined based on the correlation curve generated by calculating the running time and wear value during tool operation; the different wear stages include light wear stage, moderate wear stage, and heavy wear stage; The system receives user requests for tool wear monitoring and matches a preset threshold range for operating parameters at a specific wear stage based on those requests. It then determines the corresponding preset operating parameter thresholds based on the matched threshold range for the specific wear stage, thus completing the matching of the preset operating parameter thresholds according to the real-time collected data on tool type, machining material, and machining conditions. The preset operating parameter combination includes: Iterates through different tool types, machining materials, and machining condition combinations, and statistically analyzes the corresponding historical operating parameters of the tool operation process under a single tool type, a single machining material, and a single or combined machining condition. It then sets preset operating parameter combinations based on different operating parameters. Historical operating parameters and corresponding wear values ​​are filtered out according to each set of preset operating parameter combinations, and training set and test set are divided. The accuracy of the tool wear prediction model generated by training set is obtained according to the test set, and the preset operating parameter combinations are sorted according to the accuracy. Based on the preset operating parameter combinations generated under the same conditions of tool type, machining material, and machining conditions, the preset operating parameter combinations corresponding to the tool wear prediction model with the highest accuracy are selected as the matching preset operating parameter combinations. When selecting actual operating parameters, if some operating parameters from the currently matched preset operating parameter combinations are missing from the real-time operating parameters, the next-ranked preset operating parameter combination with an accuracy greater than the preset accuracy is selected as the new matching preset operating parameter combination. If a preset operating parameter combination with an accuracy greater than the preset accuracy cannot be found by ranking, the preset operating parameter combination with the highest comprehensive score among all preset operating parameter combinations with an accuracy greater than the preset accuracy is selected as the final matching preset operating parameter combination. This completes the matching of corresponding preset operating parameter combinations based on tool type, machining material, and machining conditions, and the selection of corresponding operating parameters. The comprehensive score is obtained by weighted calculation based on the similarity score with the real-time operating parameters and the accuracy score for completing the mapping and supplementation of real-time operating parameters based on missing operating parameters.

2. The method for monitoring tool wear on a CNC lathe according to claim 1, characterized in that, Based on the selected operating parameters, a feature selection algorithm is used to select and extract key features to obtain the optimal feature set, including: When a preset operating parameter threshold is determined based on the threshold range of operating parameters in the light wear stage or the moderate wear stage, the corresponding feature selection algorithm adopts an integrated feature selection algorithm. The integrated feature selection algorithm includes: pre-screening the target features using a variance threshold or mutual information algorithm, and then using a random forest feature importance ranking technique to complete a secondary screening of the features; after selecting and extracting key features using the integrated feature selection algorithm, based on the first preset number of features matching the light wear stage or the moderate wear stage, the same number of features are retained according to the sorting of the screened features. When a preset operating parameter threshold is determined based on the threshold range of operating parameters in the severe wear stage, a single feature selection algorithm is adopted for the corresponding feature selection algorithm. The single feature selection algorithm includes: using variance thresholding or mutual information algorithm to filter and sort the importance of target features; after selecting key features using the feature selection algorithm, retaining the same number of features according to the sorted features after filtering, based on the second preset number of features matching the severe wear stage.

3. The method for monitoring tool wear on a CNC lathe according to claim 1, characterized in that, Based on tool type, machining material, and machining conditions, corresponding pre-built tool wear prediction models are matched, including: The tool wear prediction model includes multiple prediction models. Each prediction model includes a multi-branch feature input layer, a multi-branch feature fusion layer, and a multi-branch prediction result output layer. The model is configured to take the output of the branch that outputs the wear value and the output of the branch that outputs the wear degree as the input of the branch that outputs the life prediction result. The model is generated by training a set of historical best feature sets, tool wear values, tool wear degrees, and tool remaining life under a single tool type, a single machining material, and a single or combined machining condition.

4. The method for monitoring tool wear on a CNC lathe according to claim 1, characterized in that, Based on the obtained wear values, wear levels, and lifespan prediction results, early warning decisions include: Set different levels of warning rules. Each level of warning rule corresponds to the wear value reaching the preset wear value, the tool wear degree reaching the preset degree, or the remaining tool life being within the preset threshold of the remaining life. The urgency of the CNC lathe production plan is obtained, and the warning rules are adjusted accordingly for different urgency levels. The adjusted warning rules include: among the warning rules of the same level, the higher the urgency of the production plan, the higher the tool wear level or the higher the tool wear value or the lower the preset threshold of the remaining tool life. Based on the early warning rules, the early warning level corresponding to the obtained wear value, wear degree and life prediction results is determined in order to realize early warning decision-making.

5. The method for monitoring tool wear on a CNC lathe according to claim 1, characterized in that, Also includes: While monitoring the data of the CNC lathe tool's movement process in real time, images of the CNC lathe tool's movement process are simultaneously acquired; The wear values, wear degree and life prediction results are statistically obtained, and wear slope-run time curves or life change rate-run time curves corresponding to different wear stages are plotted in real time according to time sequence; The system checks in real time whether the wear slope-run time curve does not conform to the trend of the current wear stage or whether the life change rate-run time curve does not conform to the trend of the current wear stage. If they do not conform, the system transmits the CNC lathe tool running process images acquired synchronously to the cloud platform. The system then checks whether a preset event has occurred by querying the tool running process data or analyzing the tool running images. The preset events include tool maintenance events or material idle events. If they exist, no action is taken. If they do not exist, the prediction results of the tool wear prediction model are considered to be pending verification. Image recognition technology on a cloud platform is used to analyze images of CNC lathe tool operation process to determine tool wear value, tool wear degree and remaining tool life in order to verify the accuracy of the prediction results of the tool wear prediction model.

6. The method for monitoring tool wear on a CNC lathe according to claim 1, characterized in that, Also includes: Real-time monitoring and acquisition of data on the operation of tools in the same batch, wherein the tools in the same batch refer to tools of the same type that process the same material under the same processing conditions; Obtain the tool wear prediction output for each tool in the same batch of tools, and obtain the user's satisfaction with the tool wear prediction output for each tool in the same batch of tools. Filter and retain the prediction outputs of several tools whose prediction output satisfaction is greater than the preset satisfaction. Obtain the tool wear prediction output of each retained tool in the same batch of tools and construct the corresponding wear curve. Use the dynamic time warping algorithm to calculate the similarity of the wear curves of the prediction output of tools in the same batch. The feature matrix is ​​constructed by extracting features corresponding to tools in the same batch whose wear curves have a similarity greater than the preset similarity of the wear curves. The covariance matrix is ​​calculated, and principal component features with a variance contribution rate greater than the preset contribution rate are selected by eigenvalue decomposition. The weights of each feature value are dynamically allocated according to the contribution rate and fed back to the tool wear prediction model to adjust the feature weight ratio in the tool wear prediction model at the next moment.

7. The method for monitoring tool wear on a CNC lathe according to claim 6, characterized in that, The process of generating a pre-built tool wear prediction model based on tool type, machining material, and machining conditions includes: For a single tool type, a single machining material, or a combination of single or combined machining conditions, several tool wear prediction models to be trained are designed. The number of tool wear prediction models to be trained is determined to be N+1 based on the number of tools in the same batch N. The feature weight parameters in each tool wear prediction model to be trained are distinguished in advance. One of the tool wear prediction models to be trained is used as the global model, and each of the remaining tool wear prediction models to be trained is used as a sub-model and pre-matched with one tool in the same batch of tools. The historical best feature set extracted from the historical operation data of each tool in the same batch of tools is used as the corresponding input to the matching sub-model to complete the training of the matching sub-model. After training, the feature weights and gradients of each sub-model are uploaded to the global model. Combined with the federated learning algorithm, the feature weights of each sub-model are aggregated to obtain aggregated feature weights, which are then distributed to each sub-model for iterative updates until the global model converges. The converged global model is then used as the trained tool wear prediction model, generating pre-built tool wear prediction models for single tool type, single machining material, and single or combined machining conditions.

8. A CNC lathe tool wear monitoring system, characterized in that, include: The data acquisition module is used to monitor the data of the CNC lathe tool operation process in real time. The running data processing module is used to match the corresponding preset running parameter thresholds according to the tool type, machining material, and machining conditions. When the real-time running parameters exceed the preset running parameter thresholds, the CNC lathe tool running process data at the current moment is transmitted to the edge computing node. The preset operating parameter thresholds include: collecting historical operating parameters and tool wear values ​​corresponding to tool operation under different tool types, different machining materials, and different machining condition combinations; using deep learning algorithms to obtain the mapping relationship between operating parameters and wear values ​​under single tool, single machining material, and single or combined machining condition conditions; based on the constructed mapping relationship, analyzing and obtaining the threshold range of operating parameters at different wear stages under known tool types, known machining materials, and known machining condition conditions; the different wear stages are determined based on the correlation curve generated by calculating the running time and wear value during tool operation; the different wear stages include light wear stage, moderate wear stage, and heavy wear stage; receiving user requests for tool wear monitoring, matching the preset threshold range of operating parameters at specific wear stages according to the user's requests for tool wear monitoring; determining the preset operating parameter thresholds corresponding to the matched threshold ranges of operating parameters at specific wear stages, thereby completing the matching of corresponding preset operating parameter thresholds based on real-time collected tool type, machining material, and machining condition data; The optimal feature set acquisition module utilizes edge computing nodes to match corresponding preset operating parameter combinations based on tool type, machining material, and machining condition, and completes the filtering of operating parameters based on the matched preset operating parameter combinations. Based on the filtered operating parameters, a feature selection algorithm is used to select and extract key features, obtain the optimal feature set, and upload it along with the corresponding tool type, machining material, and machining condition to the cloud platform. The preset operating parameter combinations include: traversing different tool types, machining materials, and machining condition combinations, statistically analyzing the operating parameters corresponding to the tool operation process under different historical tool types, machining materials, and single or combined machining conditions, and setting preset operating parameter combinations based on different operating parameters; filtering historical operating parameters and corresponding wear values ​​according to each set of preset operating parameter combinations, dividing them into training and test sets, obtaining the accuracy of the tool wear prediction model trained using the training set based on the test set, and arranging the preset operating parameter combinations in order of accuracy; and sorting the preset operating parameter combinations according to tool type, machining material, machining condition, and machining condition. The preset operating parameter combinations generated under the same working conditions are sorted. The preset operating parameter combination corresponding to the tool wear prediction model with the highest accuracy is selected as the matching preset operating parameter combination. When selecting actual operating parameters, if some operating parameters from the currently matched preset operating parameter combination are missing from the real-time operating parameters, the next-ranked preset operating parameter combination with an accuracy greater than the preset accuracy is selected as the re-matching preset operating parameter combination. If a preset operating parameter combination with an accuracy greater than the preset accuracy cannot be found by ranking, the preset operating parameter combination with the highest comprehensive score among all preset operating parameter combinations with an accuracy greater than the preset accuracy is selected as the final matching preset operating parameter combination. This completes the matching of corresponding preset operating parameter combinations based on tool type, machining material, and machining conditions, and the selection of corresponding operating parameters. The comprehensive score is obtained by weighted calculation based on the similarity score with real-time operating parameters and the accuracy score for completing the mapping and supplementation of real-time operating parameters based on missing operating parameters. The wear and life prediction module is used on the cloud platform to match the corresponding pre-built tool wear prediction model according to the tool type, machining material and machining conditions. The optimal feature set is input into the matched tool wear prediction model to obtain the wear value, wear degree and life prediction results. The early warning decision and prompt generation module is used to combine the production plan with the obtained wear value, wear degree and life prediction results to make early warning decisions and generate early warning prompts.

Images