Metal cutting tool wear monitoring method and system

Abstract

The invention discloses a metal cutting tool wear monitoring method and system, and relates to the technical field of wear monitoring. The method comprises the following steps: monitoring the metal cutting tool in real time to obtain a multi-source sensing sequence; constructing a tool cutting monitoring matrix; inputting the tool cutting monitoring matrix into a tool wear detection channel to obtain a first tool wear coefficient; performing differential detection on the first tool wear coefficient according to a preset cutting scheme, and outputting a wear differential signal; performing image perception correction on the first tool wear coefficient to obtain a second tool wear coefficient; and obtaining a historical tool wear coefficient set of the metal cutting tool, performing wear trend prediction in combination with the second tool wear coefficient, constructing a tool wear trend chart, and performing health management on the metal cutting tool according to the tool wear trend chart. The technical problem of low tool wear monitoring precision in the prior art is solved, and the technical effect of improving the tool wear monitoring precision is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of wear monitoring, and particularly relates to a method and system for monitoring the wear of metal cutting tools. Background Art

[0002] During the metal cutting process, tool wear is one of the key factors affecting machining quality, production efficiency, and manufacturing cost.

[0003] During the metal cutting process, the wear condition of the tool has an important impact on machining efficiency, machining quality, and production cost. Traditional tool wear monitoring methods often rely on the experience of operators and regular manual inspections. This method not only has limited accuracy but also is difficult to detect tool wear in a timely manner. In addition, existing tool wear monitoring systems often only utilize a single sensing information, such as cutting force or vibration signal, which is difficult to comprehensively reflect the tool wear condition. At the same time, due to various interference factors during the cutting process, such as the inhomogeneity of the workpiece material and the change of cutting parameters, a large amount of noise and redundant information are included in the sensing signal, which brings challenges to the accurate monitoring of tool wear. Summary of the Invention

[0004] This application provides a method and system for monitoring the wear of metal cutting tools, which solves the technical problem of low accuracy in tool wear monitoring in the prior art.

[0005] In the first aspect of this application, a method for monitoring the wear of metal cutting tools is provided. The method includes:

[0006] When the metal cutting tool executes a predetermined cutting plan, the metal cutting tool is monitored in real time to obtain a multi-source sensing sequence, where the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence; filtering and denoising are performed according to the multi-source sensing sequence to construct a tool cutting monitoring matrix; the tool cutting monitoring matrix is input into a tool wear detection channel to obtain a first tool wear coefficient; differential detection is performed on the first tool wear coefficient according to the predetermined cutting plan to output a wear differential signal; based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain a second tool wear coefficient; a historical tool wear coefficient set of the metal cutting tool is obtained, and wear trend prediction is performed in combination with the second tool wear coefficient to construct a tool wear trend graph, and health management of the metal cutting tool is performed according to the tool wear trend graph.

[0007] In the second aspect of this application, a system for monitoring the wear of metal cutting tools is provided. The system includes:

[0008] A monitoring module is used to monitor the metal cutting tool in real time when the metal cutting tool executes a predetermined cutting plan, and obtain a multi-source sensing sequence, where the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence; a denoising module is used to perform filtering and denoising according to the multi-source sensing sequence to construct a tool cutting monitoring matrix; a wear detection module is used to input the tool cutting monitoring matrix into a tool wear detection channel to obtain a first tool wear coefficient; a differential detection module is used to perform differential detection on the first tool wear coefficient according to the predetermined cutting plan and output a wear differential signal; a correction module is used to perform image perception correction on the first tool wear coefficient based on the wear differential signal to obtain a second tool wear coefficient; a management module is used to obtain a historical tool wear coefficient set of the metal cutting tool, combine the second tool wear coefficient to perform wear trend prediction, construct a tool wear trend chart, and perform health management on the metal cutting tool according to the tool wear trend chart.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] When the metal cutting tool executes a predetermined cutting plan, the metal cutting tool is monitored in real time to obtain a multi-source sensing sequence, and the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence. Then, filtering and denoising are performed according to the multi-source sensing sequence to construct a tool cutting monitoring matrix. Then, the tool cutting monitoring matrix is input into a tool wear detection channel to obtain a first tool wear coefficient. Next, differential detection is performed on the first tool wear coefficient according to the predetermined cutting plan to output a wear differential signal. Further, based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain a second tool wear coefficient. Finally, a historical tool wear coefficient set of the metal cutting tool is obtained, the second tool wear coefficient is combined to perform wear trend prediction, a tool wear trend chart is constructed, and health management is performed on the metal cutting tool according to the tool wear trend chart. The technical problem of low tool wear monitoring accuracy in the prior art is solved, and the technical effect of improving tool wear monitoring accuracy is achieved. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 It is a schematic flowchart of a method for monitoring the wear of a metal cutting tool provided in an embodiment of this application;

[0013] Figure 2 This is a schematic structural diagram of a metal cutting tool wear monitoring system provided by an embodiment of the present application.

[0014] Explanation of reference numerals: monitoring module 11, denoising module 12, wear detection module 13, differential detection module 14, calibration module 15, management module 16. Specific embodiments

[0015] The present application provides a metal cutting tool wear monitoring method and system, which solves the technical problem of low accuracy of tool wear monitoring in the prior art.

[0016] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0017] It should be noted that the terms "include" and "have" are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0018] Embodiment 1, as Figure 1 shown, the present application provides a metal cutting tool wear monitoring method, wherein the method includes:

[0019] When the metal cutting tool executes a predetermined cutting plan, the metal cutting tool is monitored in real time to obtain a multi-source sensing sequence, and the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence.

[0020] When a metal cutting tool executes a predetermined cutting plan, the operating state of the metal cutting tool is monitored in real time by a variety of sensors arranged on the processing equipment to obtain a multi-source sensing sequence characterizing the tool wear state. Specifically, a three-axis dynamic cutting force sensor is installed on the metal cutting tool or the machine tool spindle to collect the cutting force components in the X, Y, and Z directions during the cutting process in real time, and analog-to-digital conversion is performed through a data acquisition module to generate a cutting force time series; an acceleration sensor is arranged near the tool clamping part or the machine tool spindle to obtain the vibration signal generated by the tool due to wear during the cutting process. The vibration signal contains the vibration data in the X, Y, and Z axis directions, and is amplified and filtered through a signal conditioning module to generate a vibration signal time series; an acoustic emission sensor is set near the tool to monitor the high-frequency acoustic signal generated during the interaction between the tool and the workpiece. After the acoustic emission signal is enhanced by a preamplifier, analog-to-digital conversion is performed through a high-speed data acquisition module to generate an acoustic emission signal time series; an infrared temperature sensor or a thermocouple is installed near the tool-workpiece contact area to obtain the temperature change of the tool during the cutting process, and the collected temperature signal is converted into a standard electrical signal and then input into the data acquisition system to generate a temperature signal time series.

[0021] Filter and denoise according to the multi-source sensing sequence to construct a tool cutting monitoring matrix.

[0022] Filter and denoise the multi-source sensing sequence to eliminate environmental noise and system interference, thereby improving the data quality. On this basis, construct a tool cutting monitoring matrix to provide high-quality input data for subsequent tool wear detection.

[0023] Furthermore, filtering and denoising according to the multi-source sensing sequence to construct a tool cutting monitoring matrix includes:

[0024] Perform Kalman filtering on the cutting force sequence to obtain a cutting force matrix; perform band-pass filtering on the vibration signal sequence to obtain a vibration signal matrix; perform band-pass filtering on the acoustic emission signal sequence to obtain an acoustic emission signal matrix; perform Kalman filtering on the temperature signal sequence to obtain a temperature signal matrix; perform time window alignment and splicing according to the cutting force matrix, the vibration signal matrix, the acoustic emission signal matrix, and the temperature signal matrix to generate the tool cutting monitoring matrix.

[0025] Specifically, perform Kalman filtering on the cutting force sequence. By establishing the state space model of the cutting force, the Kalman filtering method gradually estimates the true cutting force value based on the original data measured by the cutting force sensor, removes the errors introduced by noise and inaccurate measurements, and generates a denoised cutting force matrix. For the vibration signal sequence, band-pass filtering is used for processing. The band-pass filter can filter out low-frequency interference and high-frequency noise, and only retain the effective vibration signals related to tool wear during the cutting process, thereby obtaining a vibration signal matrix. For the acoustic emission signal sequence, band-pass filtering is also used to enhance the effective components in the signal and suppress irrelevant low-frequency and high-frequency noise, generating an acoustic emission signal matrix. The temperature signal sequence undergoes Kalman filtering to remove the random fluctuations and environmental impacts in the temperature sensor measurement, ensuring the stability of the temperature signal and generating a temperature signal matrix.

[0026] After completing the above filtering and denoising processing, to ensure the temporal consistency between different sensing data, a time window alignment method is used to perform time synchronization processing on the cutting force matrix, vibration signal matrix, acoustic emission signal matrix, and temperature signal matrix. Specifically, by setting a unified time reference, the data of each matrix is aligned according to the time window to ensure that each sensing signal corresponds accurately in time. Finally, the aligned cutting force matrix, vibration signal matrix, acoustic emission signal matrix, and temperature signal matrix are spliced together according to a predetermined rule to generate a complete tool cutting monitoring matrix, which contains the temporal information of multi-source signals and provides accurate data support for subsequent tool wear analysis and trend prediction.

[0027] Input the tool cutting monitoring matrix into the tool wear detection channel to obtain the first tool wear coefficient.

[0028] The tool wear detection channel includes multiple tool wear detection models, which are constructed based on machine learning and can identify the wear state of the tool according to different types of sensor data (such as signals like cutting force, vibration, acoustic emission, and temperature). Taking the tool cutting monitoring matrix as input data, after being processed by the tool wear detection model, the model will analyze the wear degree of the tool and generate the first tool wear coefficient according to the characteristic information in each signal matrix. The first tool wear coefficient represents the current wear state or wear degree of the tool.

[0029] Furthermore, inputting the tool cutting monitoring matrix into the tool wear detection channel to obtain the first tool wear coefficient includes:

[0030] The tool wear detection channel includes Q tool wear detection models, where Q is a positive integer greater than 1; input the tool cutting monitoring matrix into the Q tool wear detection models to obtain Q tool wear detection coefficients; perform a proportion calculation based on the Q wear detection accuracy coefficients corresponding to the Q tool wear detection models to obtain Q detection output weights; perform a weighted calculation on the Q tool wear detection coefficients according to the Q detection output weights to generate the first tool wear coefficient.

[0031] The tool wear detection channel includes Q tool wear detection models, where Q is a positive integer greater than 1; input the tool cutting monitoring matrix into the Q tool wear detection models to obtain Q tool wear detection coefficients. According to the performance of each tool wear detection model, calculate the wear detection accuracy coefficient of each model. Specifically, use a set of test data sets with known wear states to evaluate the Q tool wear detection models. The test data set contains tool samples with various wear degrees, and the wear states of each sample are labeled; by inputting the test data into each model, obtain its prediction results and compare the prediction results with the actual wear states; then, quantitatively evaluate the performance of each model according to standard evaluation indicators (such as accuracy, precision, recall, F1-score, etc.), specifically calculate the performance of each model on the test set; based on these evaluation results, calculate the wear detection accuracy coefficient of each model, usually by comprehensively considering the performance of each evaluation indicator with weights and assigning different weights according to the overall performance of the model. For example, if a certain model performs excellently on the test set with high accuracy and F1-score, then the accuracy coefficient of this model is relatively high, reflecting its high prediction reliability. Subsequently, to ensure the comparability of the accuracy coefficients of different models, they are usually normalized, adjusting the accuracy coefficients of all models to between 0 and 1 to ensure that the sum of all accuracy coefficients is 1. According to the Q wear detection accuracy coefficients, perform a proportion calculation to obtain Q detection output weights, and the detection output weights reflect the relative contributions of each tool wear detection model in the final wear assessment. According to the Q detection output weights, perform a weighted calculation on the Q tool wear detection coefficients to generate the final first tool wear coefficient. Through this weighted calculation method, comprehensively consider the prediction results of different models to ensure that the obtained first tool wear coefficient can accurately reflect the wear state of the tool, providing a reliable basis for subsequent wear differential detection, wear trend prediction, and tool health management.

[0032] Among them, each tool wear detection model uses different algorithms and feature processing methods. Based on the input tool cutting monitoring matrix, it extracts feature information from signals such as cutting force, vibration, acoustic emission, and temperature, and evaluates the tool wear degree based on this feature information, and then outputs the corresponding wear detection coefficient. Specifically, first, multi-source sensor signals such as cutting force, vibration, acoustic emission, and temperature are collected, and these signals are preprocessed, including filtering and denoising, outlier removal, and data normalization, etc.; after preprocessing, the signal data is used as the input matrix and enters the tool wear detection model; signal processing technologies (such as Fourier transform, wavelet transform, time-frequency analysis, etc.) are used to extract key feature information from each signal, such as the oscillation frequency of the cutting force signal, the frequency characteristics of the vibration signal, the intensity and frequency distribution of the acoustic emission signal, the thermal deformation of the temperature signal, etc.; historical tool wear data is used to train multiple tool wear detection models. During the training process, appropriate machine learning algorithms (such as support vector machines, decision trees, neural networks, etc.) are selected, and the model is optimized through calibration data to ensure that it can accurately predict the tool wear state according to the signal characteristics; after training, the model is evaluated through cross-validation and evaluation metrics (such as accuracy, recall rate, F1-score, etc.) to ensure that the model can effectively identify the tool states with different wear degrees. The trained and evaluated model can be officially applied to the cutting process. Through the input tool cutting monitoring matrix, the model can output the corresponding wear detection coefficient in real time, providing an accurate basis for subsequent wear analysis and health management.

[0033] Perform differential detection on the first tool wear coefficient according to the predetermined cutting plan, and output a wear differential signal.

[0034] Performing differential detection on the first tool wear coefficient based on a predetermined cutting plan, that is, making a theoretical prediction of the wear condition of the metal cutting tool, comparing the tool wear prediction coefficient with the first tool wear coefficient to obtain a wear detection differential coefficient to characterize the deviation between the current tool wear condition and the theoretical prediction value; if the wear detection differential coefficient exceeds the wear detection differential threshold, it indicates that the actual tool wear condition has deviated from the normal prediction range, and there may be abnormal wear or working condition fluctuations. At this time, a wear differential signal is generated to provide a warning message.

[0035] Furthermore, performing differential detection on the first tool wear coefficient according to the predetermined cutting plan and outputting a wear differential signal includes:

[0036] Perform wear prediction on the metal cutting tool according to the predetermined cutting scheme to obtain a tool wear prediction coefficient; calculate a wear detection differential coefficient based on the tool wear prediction coefficient and the first tool wear coefficient; determine whether the wear detection differential coefficient is greater than or equal to a wear detection differential threshold; if the wear detection differential coefficient is greater than or equal to the wear detection differential threshold, generate the wear differential signal.

[0037] First, based on the predetermined cutting scheme, the wear trend of the metal cutting tool is theoretically predicted, and a tool wear prediction model is established. The tool wear prediction coefficient is calculated by combining historical cutting data, tool material characteristics, cutting process parameters and the wear law of tools under similar working conditions. Subsequently, the first tool wear coefficient obtained by real-time monitoring is compared with the tool wear prediction coefficient, and the wear detection differential coefficient is calculated (that is, the absolute value of the difference between the first tool wear coefficient and the tool wear prediction coefficient is calculated) to measure the deviation between the actual wear of the tool and the theoretical prediction value. Next, the wear detection differential threshold is set, and it is determined whether the wear detection differential coefficient is greater than or equal to the threshold. If the wear detection differential coefficient exceeds the set threshold, it indicates that the actual wear of the tool may be abnormal, such as increased wear, changes in cutting conditions or process deviations. At this time, a wear differential signal is generated to provide an early warning prompt, so that the system can adjust the cutting parameters, optimize the processing process or take maintenance measures, thereby improving the service life and processing quality of the tool and ensuring the stability and safety of the cutting process.

[0038] Furthermore, performing wear prediction on the metal cutting tool according to the predetermined cutting scheme to obtain a tool wear prediction coefficient includes:

[0039] Read the tool pre-wear coefficient of the metal cutting tool, wherein the tool pre-wear coefficient is the tool wear coefficient of the metal cutting tool before executing the predetermined cutting scheme; perform networking association of tools of the same model according to the metal cutting tool to determine a set of tools of the same family; perform a tool wear increase sample search on the set of tools of the same family according to the predetermined cutting scheme to obtain a tool wear increase search set; perform centralized value calculation on the tool wear increase search set to obtain a predicted tool wear increase; perform an incentive adjustment calculation on the tool pre-wear coefficient according to the predicted tool wear increase to obtain the tool wear prediction coefficient.

[0040] First, the tool pre-wear coefficient of the metal cutting tool is read. The tool pre-wear coefficient is the tool wear coefficient of the tool before executing the predetermined cutting plan to ensure the accuracy of the prediction benchmark. Then, based on the specifications, materials, manufacturing batches and use environment of the metal cutting tool, the same model of tools are networked and associated to build a set of tools of the same family to make full use of the historical wear data of similar tools and improve the prediction accuracy. On this basis, according to the predetermined cutting plan, the tool wear increase samples that meet the current cutting conditions are retrieved from the set of tools of the same family to form a tool wear increase retrieval set. For example, after executing the predetermined cutting plan, the historical tool wear coefficient increases by 20%. Subsequently, the tool wear increase retrieval set is analyzed, and the concentrated value, such as the mean, median or weighted average, is calculated using statistical methods to obtain the predicted tool wear increase. Finally, according to the predicted tool wear increase, the tool pre-wear coefficient is calculated by incentive adjustment to obtain the tool wear prediction coefficient, where the tool wear prediction coefficient = tool pre-wear coefficient + tool pre-wear coefficient × predicted tool wear increase.

[0041] Based on the wear differential signal, the first tool wear coefficient is subjected to image perception correction to obtain the second tool wear coefficient.

[0042] Based on the wear differential signal, the real-time image data of the metal cutting tool is called. The real-time image is obtained by a vision detection system installed on the processing equipment. The vision detection system may include a high-resolution industrial camera, a light source system, and an image acquisition module, and continuously captures the tool at a preset image acquisition frequency to ensure the real-time and integrity of the wear state. Subsequently, preprocessing is performed on the obtained real-time tool image, including but not limited to image enhancement, noise removal, contrast adjustment, and edge detection. Among them, image enhancement can use the Retinex algorithm to enhance the surface details of the tool, noise removal can use Gaussian filtering for smoothing, and edge detection can detect the boundary features of the tool edge wear area through Canny or Roberts operators to improve the recognizability of the tool wear area. Based on the preprocessed tool image, a wear feature perception model is constructed. This model can be trained using deep learning methods (such as convolutional neural network CNN or ResNet). During the training process, a large dataset of labeled tool wear images is used, and the model parameters are optimized based on the supervised learning method, enabling it to automatically extract the features of the tool wear area and output the image perception tool wear coefficient, which characterizes the wear degree of the tool at the visual level. Then, the first tool wear coefficient is fused with the image perception tool wear coefficient, and the first tool wear correction coefficient and the second tool wear correction coefficient are obtained through proportion calculation to measure the weights of different detection methods and ensure the reliability of the final wear assessment. Next, based on the first tool wear correction coefficient and the second tool wear correction coefficient, weighted calculation is performed on the first tool wear coefficient and the image perception tool wear coefficient. The weights can be dynamically adjusted based on historical detection data, model confidence, and tool usage environment to obtain the corrected second tool wear coefficient, thereby improving the accuracy of wear assessment, enabling the tool wear monitoring system to more accurately predict the tool life, and providing a more reliable decision-making basis for tool health management.

[0043] Furthermore, based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain the second tool wear coefficient, including:

[0044] Based on the wear differential signal, the real-time tool image of the metal cutting tool is retrieved; the wear characteristics of the metal cutting tool are perceived according to the real-time tool image to obtain the image perception tool wear coefficient; adaptive correction parameter calculation is performed based on the first tool wear coefficient and the image perception tool wear coefficient to obtain the first tool wear correction coefficient and the second tool wear correction coefficient; based on the first tool wear correction coefficient and the second tool wear correction coefficient, weighted calculation is performed on the first tool wear coefficient and the image perception tool wear coefficient to generate the second tool wear coefficient.

[0045] First, based on the wear differential signal, the real-time image of the metal cutting tool is called. This image is collected by a high-precision industrial camera installed on the machine tool and combined with an adjustable light source system to ensure image clarity and stability. Then, preprocess the real-time image of the tool, including image denoising, contrast enhancement, and edge detection. Mean filtering or Gaussian filtering methods can be used for denoising, histogram equalization technology can be used for contrast enhancement, and the Canny operator or Sobel operator can be used for edge detection to accurately extract the tool wear area. Subsequently, analyze the preprocessed image based on a deep learning model (such as the CNN or Transformer architecture), extract the key feature information of the wear area, and match it with the existing tool wear database to generate an image-aware tool wear coefficient. Then, calculate according to the relative contribution degrees of the first tool wear coefficient and the image-aware tool wear coefficient, and use an adaptive weighting strategy to determine the first tool wear correction coefficient and the second tool wear correction coefficient. Specifically, α 2 = 1 - α 1 , where, W 1 is the first tool wear coefficient, W 2 is the image-aware tool wear coefficient, α 1 is the first tool wear correction coefficient, α 2 is the second tool wear correction coefficient, λ is a weight adjustment parameter used to control the influence degree of the wear coefficient difference on the weight, and δ is a balance threshold used to avoid drastic changes in the weight caused by small errors. Finally, according to the first tool wear correction coefficient and the second tool wear correction coefficient, perform weighted calculation on the first tool wear coefficient and the image-aware tool wear coefficient to generate the second tool wear coefficient; specifically, W corrected = α 1 W 1 + α 2 W 2 , where, W corrected is the second tool wear coefficient, W 1 is the first tool wear coefficient, W 2 is the image-aware tool wear coefficient, α 1 is the first tool wear correction coefficient, α 2 is the second tool wear correction coefficient.

[0046] Furthermore, performing wear feature perception on the metal cutting tool according to the real-time image of the tool to obtain an image-aware tool wear coefficient includes:

[0047] Perform adaptive wear feature convolution on the real-time tool image to obtain a tool wear convolution image; align the historical tool wear convolution image set and the historical image-aware tool wear coefficient set to obtain an image-aware tool wear evaluation construction set; perform supervised training on the ResNet network according to the image-aware tool wear evaluation construction set, and obtain a tool wear evaluation loss coefficient every time a predetermined number of training times is reached; if the tool wear evaluation loss coefficient is less than the tool wear evaluation loss threshold, generate an image-aware tool wear evaluation model; input the tool wear convolution image into the image-aware tool wear evaluation model, and output the image-aware tool wear coefficient.

[0048] First, based on the real-time tool image, use a convolutional neural network (CNN) to construct an adaptive wear feature convolution module to extract multi-scale features from the real-time tool image to obtain a tool wear convolution image. Among them, the adaptive wear feature convolution module adopts a multi-layer convolution kernel combination structure to enhance the ability to capture wear features of different scales; then, based on historical tool wear data, align the historical tool wear convolution image set and the historical image-aware tool wear coefficient set to form an image-aware tool wear evaluation construction set. Among them, the historical image-aware tool wear coefficient set is the wear evaluation result obtained based on actual measurement data or calibration data; then, use a deep residual neural network (ResNet) to perform supervised training on the image-aware tool wear evaluation construction set. Every time a predetermined number of training times is reached, calculate the tool wear evaluation loss coefficient. The tool wear evaluation loss coefficient is calculated based on the mean square error (MSE) loss function or the cross-entropy loss function to measure the deviation between the model prediction result and the true wear coefficient; if the tool wear evaluation loss coefficient is less than the tool wear evaluation loss threshold, it is determined that the training has converged, and an image-aware tool wear evaluation model is generated. Among them, the tool wear evaluation loss threshold is set according to experimental calibration or adaptive adjustment to ensure that the evaluation accuracy of the model meets the predetermined requirements; finally, input the tool wear convolution image into the image-aware tool wear evaluation model, and generate and output the image-aware tool wear coefficient based on the feature mapping and fully connected layer output of the model to characterize the wear degree of the metal cutting tool.

[0049] Furthermore, performing adaptive wear feature convolution on the real-time tool image to obtain a tool wear convolution image includes:

[0050] Perform Retinex enhancement processing on the real-time image of the tool to obtain the first tool image; perform background separation on the first tool image according to the Fisher criterion function to obtain the second tool image; construct an adaptive wear feature convolution kernel according to the second tool image, and perform wear feature convolution on the second tool image according to the adaptive wear feature convolution kernel to obtain the initial tool wear convolution image; introduce the Roberts operator to perform wear area edge feature detection on the second tool image to obtain the wear edge detection operator; perform edge enhancement on the initial tool wear convolution image according to the wear edge detection operator to generate the tool wear convolution image.

[0051] First, perform Retinex enhancement processing on the real-time image of the tool to obtain the first tool image. Among them, the Retinex enhancement processing improves the local contrast and detail visibility of the image through multi-scale decomposition and reflection component estimation, thereby enhancing the recognizability of the tool wear area. Then, perform background separation on the first tool image based on the Fisher criterion function to obtain the second tool image. Among them, the Fisher criterion function is used to measure the between-class variance between the tool area and the background area, and the image is binarized by maximizing the between-class variance to accurately extract the tool wear area. Next, construct an adaptive wear feature convolution kernel according to the second tool image. Among them, the adaptive wear feature convolution kernel dynamically adjusts the size and weight of the convolution kernel by analyzing the texture features, gray gradients, and spatial distribution patterns of the tool wear area to enhance the ability to extract wear features, and perform wear feature convolution on the second tool image according to the adaptive wear feature convolution kernel to obtain the initial tool wear convolution image. Further, introduce the Roberts operator to perform wear area edge feature detection on the second tool image to obtain the wear edge detection operator. Among them, the Roberts operator is a gradient-based edge detection method that accurately captures the boundary features of the tool wear area by calculating the gradient change rate of the image pixels. Finally, perform edge enhancement on the initial tool wear convolution image according to the wear edge detection operator. Among them, edge enhancement technology is used to highlight the boundary information of the wear area to improve the contour clarity of the wear area and generate the tool wear convolution image for subsequent calculation and analysis of the tool wear evaluation model.

[0052] Furthermore, obtaining the second tool wear coefficient includes:

[0053] Judge whether the second tool wear coefficient is greater than or equal to the tool wear threshold; if the second tool wear coefficient is greater than or equal to the tool wear threshold, generate a tool wear warning signal.

[0054] After obtaining the second tool wear coefficient, a threshold judgment is made on the second tool wear coefficient. Specifically, a tool wear threshold is obtained, which is a safety wear threshold set based on historical tool wear data and tool failure mechanism modeling, and it is judged whether the second tool wear coefficient is greater than or equal to the tool wear threshold; if the second tool wear coefficient is greater than or equal to the tool wear threshold, a tool wear warning signal is generated. Among them, the tool wear warning signal is used to indicate that the current tool has reached or exceeded the allowable wear limit, and trigger corresponding tool replacement, machining parameter adjustment or equipment shutdown protection measures to prevent the machining accuracy from decreasing or the equipment from being damaged due to severe tool wear.

[0055] Obtain the historical tool wear coefficient set of the metal cutting tool, combine the second tool wear coefficient to predict the wear trend, construct a tool wear trend chart, and perform health management on the metal cutting tool according to the tool wear trend chart.

[0056] First, collect the historical tool wear coefficient set of the metal cutting tool during previous cutting processes. This historical wear coefficient set contains data records of tool wear conditions in multiple different cutting cycles; then, combine the second tool wear coefficient, that is, the current wear evaluation result, and through data analysis and modeling methods, predict the wear trend and predict the wear development trend of the tool in a certain period of time in the future. This process can use methods such as time series analysis, regression analysis, neural network, etc. to predict the wear trend. Exemplarily, a long short-term memory neural network (LSTM) is used to predict tool wear; collect historical tool wear data and organize it in chronological order as time series data; in order to improve the prediction accuracy, other relevant operating parameters (such as cutting force, temperature, vibration, etc.) can be combined as input features for processing; then, divide the data set into a training set, a validation set and a test set. By training the training set, the LSTM model can learn the long-term and short-term dependence relationships in the time series data and predict the wear development trend of the tool in a certain period of time in the future; during the training process, the LSTM network optimizes the model parameters by minimizing the prediction error (such as mean square error); after the training is completed, the model can output the predicted value of the future wear coefficient based on the historical wear data and relevant operating parameters.

[0057] Next, according to the predicted wear trend, a tool wear trend graph is constructed to show the wear change process of the tool in different cutting cycles and the predicted future wear changes, which helps analyze the remaining time of the tool life and the development speed of wear. Based on the tool wear trend graph, health management of the metal cutting tool is carried out. By monitoring the changes in the tool wear trend graph, the wear state of the tool can be evaluated in real time, and the time of tool failure can be predicted in advance, so as to reasonably arrange the tool replacement cycle, optimize the cutting parameters, ensure the stable performance of the tool during the machining process, and avoid the decline of product quality and the reduction of production efficiency caused by excessive tool wear.

[0058] In summary, the embodiments of the present application have at least the following technical effects:

[0059] When the metal cutting tool executes a predetermined cutting plan, the metal cutting tool is monitored in real time to obtain a multi-source sensing sequence, and the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence. Then, filtering and denoising are performed according to the multi-source sensing sequence to construct a tool cutting monitoring matrix. Next, the tool cutting monitoring matrix is input into the tool wear detection channel to obtain the first tool wear coefficient. Then, according to the predetermined cutting plan, differential detection is performed on the first tool wear coefficient to output a wear differential signal. Further, based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain the second tool wear coefficient. Finally, a historical tool wear coefficient set of the metal cutting tool is obtained, and wear trend prediction is performed in combination with the second tool wear coefficient to construct a tool wear trend graph, and health management of the metal cutting tool is carried out according to the tool wear trend graph. The technical problem of low tool wear monitoring accuracy in the prior art is solved, and the technical effect of improving the tool wear monitoring accuracy is achieved.

[0060] Embodiment 2, based on the same inventive concept as a metal cutting tool wear monitoring method in the foregoing embodiment, as Figure 2 shown, the present application provides a metal cutting tool wear monitoring system, wherein the system includes:

[0061] The monitoring module 11 is used to monitor the metal cutting tool in real time when the metal cutting tool executes a predetermined cutting plan, and obtain a multi-source sensing sequence, where the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence; the denoising module 12 is used to perform filtering and denoising according to the multi-source sensing sequence to construct a tool cutting monitoring matrix; the wear detection module 13 is used to input the tool cutting monitoring matrix into a tool wear detection channel to obtain a first tool wear coefficient; the differential detection module 14 is used to perform differential detection on the first tool wear coefficient according to the predetermined cutting plan and output a wear differential signal; the correction module 15 is used to perform image perception correction on the first tool wear coefficient based on the wear differential signal to obtain a second tool wear coefficient; the management module 16 is used to obtain the historical tool wear coefficient set of the metal cutting tool, combine the second tool wear coefficient to predict the wear trend, construct a tool wear trend graph, and perform health management on the metal cutting tool according to the tool wear trend graph.

[0062] Further, the wear detection module 13 is used to execute the following method:

[0063] The tool wear detection channel includes Q tool wear detection models, where Q is a positive integer greater than 1; input the tool cutting monitoring matrix into the Q tool wear detection models to obtain Q tool wear detection coefficients; perform a proportion calculation according to the Q wear detection accuracy coefficients corresponding to the Q tool wear detection models to obtain Q detection output weights; perform a weighted calculation on the Q tool wear detection coefficients according to the Q detection output weights to generate the first tool wear coefficient.

[0064] Further, the differential detection module 14 is used to execute the following method:

[0065] Perform wear prediction on the metal cutting tool according to the predetermined cutting plan to obtain a tool wear prediction coefficient; calculate a wear detection differential coefficient according to the tool wear prediction coefficient and the first tool wear coefficient; determine whether the wear detection differential coefficient is greater than or equal to a wear detection differential threshold; if the wear detection differential coefficient is greater than or equal to the wear detection differential threshold, generate the wear differential signal.

[0066] Further, the differential detection module 14 is used to execute the following method:

[0067] Read the tool pre-wear coefficient of the metal cutting tool, wherein the tool pre-wear coefficient is the tool wear coefficient of the metal cutting tool before executing the predetermined cutting scheme; perform networking association of tools of the same model according to the metal cutting tool to determine a set of tools of the same family; perform a tool wear increase sample search on the set of tools of the same family according to the predetermined cutting scheme to obtain a tool wear increase search set; perform centralized value calculation on the tool wear increase search set to obtain a predicted tool wear increase; perform an incentive adjustment calculation on the tool pre-wear coefficient according to the predicted tool wear increase to obtain the tool wear prediction coefficient.

[0068] Furthermore, the correction module 15 is used to perform the following method:

[0069] Based on the wear differential signal, a real-time tool image of the metal cutting tool is retrieved; wear characteristics of the metal cutting tool are sensed according to the real-time tool image to obtain an image-perceived tool wear coefficient; adaptive correction parameter calculation is performed based on the first tool wear coefficient and the image-perceived tool wear coefficient to obtain a first tool wear correction coefficient and a second tool wear correction coefficient; based on the first tool wear correction coefficient and the second tool wear correction coefficient, a weighted calculation is performed on the first tool wear coefficient and the image-perceived tool wear coefficient to generate the second tool wear coefficient.

[0070] Furthermore, the correction module 15 is used to perform the following method:

[0071] Adaptive wear feature convolution is performed according to the real-time image of the tool to obtain a tool wear convolution image; the historical tool wear convolution image set and the historical image-perceived tool wear coefficient set are aligned to obtain an image-perceived tool wear assessment construction set; supervised training is performed on the ResNet network according to the image-perceived tool wear assessment construction set, and a tool wear assessment loss coefficient is obtained after each predetermined number of trainings; if the tool wear assessment loss coefficient is less than a tool wear assessment loss threshold, an image-perceived tool wear assessment model is generated; the tool wear convolution image is input into the image-perceived tool wear assessment model, and the image-perceived tool wear coefficient is output.

[0072] Furthermore, the correction module 15 is used to perform the following method:

[0073] Perform Retinex enhancement processing on the real-time image of the tool to obtain the first tool image; perform background separation on the first tool image according to the Fisher criterion function to obtain the second tool image; construct an adaptive wear feature convolution kernel according to the second tool image, and perform wear feature convolution on the second tool image according to the adaptive wear feature convolution kernel to obtain the initial tool wear convolution image; introduce the Roberts operator to perform wear area edge feature detection on the second tool image to obtain the wear edge detection operator; perform edge enhancement on the initial tool wear convolution image according to the wear edge detection operator to generate the tool wear convolution image.

[0074] Further, the denoising module 12 is used to execute the following method:

[0075] Perform Kalman filtering on the cutting force sequence to obtain the cutting force matrix; perform band-pass filtering on the vibration signal sequence to obtain the vibration signal matrix; perform band-pass filtering on the acoustic emission signal sequence to obtain the acoustic emission signal matrix; perform Kalman filtering on the temperature signal sequence to obtain the temperature signal matrix; perform time window alignment and splicing according to the cutting force matrix, the vibration signal matrix, the acoustic emission signal matrix, and the temperature signal matrix to generate the tool cutting monitoring matrix.

[0076] Further, the management module 16 is used to execute the following method:

[0077] Judge whether the second tool wear coefficient is greater than or equal to the tool wear threshold; if the second tool wear coefficient is greater than or equal to the tool wear threshold, generate a tool wear warning signal.

[0078] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0079] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0080] This specification and the accompanying drawings are merely illustrative of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A metal cutting tool wear monitoring method, characterized in that: The method comprises: When the metal cutting tool executes a predetermined cutting plan, the metal cutting tool is monitored in real time to obtain a multi-source sensing sequence, wherein the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence; Perform filtering and denoising according to the multi-source sensing sequence to construct a tool cutting monitoring matrix; Inputting the tool cutting monitoring matrix into a tool wear detection channel to obtain a first tool wear coefficient; Performing differential detection on the first coefficient of tool wear according to the predetermined cutting scheme, and outputting a wear differential signal; Based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain a second tool wear coefficient; A historical tool wear coefficient set of the metal cutting tool is obtained, wear trend prediction is performed in combination with the second tool wear coefficient, a tool wear trend graph is constructed, and health management of the metal cutting tool is performed according to the tool wear trend graph.

2. A metal cutting tool wear monitoring method as claimed in claim 1, characterized in that: Inputting the tool cutting monitoring matrix into the tool wear detection channel to obtain the first tool wear coefficient includes: The tool wear detection channel includes Q tool wear detection models, where Q is a positive integer greater than 1; Inputting the tool cutting monitoring matrix into the Q tool wear detection models to obtain Q tool wear detection coefficients; Calculate the proportions of Q wear detection accuracy coefficients corresponding to the Q tool wear detection models to obtain Q detection output weights; The Q tool wear detection coefficients are weightedly calculated according to the Q detection output weights to generate the first tool wear coefficient.

3. A metal cutting tool wear monitoring method as claimed in claim 1, characterized in that: The method comprises: performing differential detection on the first coefficient of tool wear according to the predetermined cutting scheme and outputting a wear differential signal, comprising: Predicting the wear of the metal cutting tool according to the predetermined cutting scheme to obtain a tool wear prediction coefficient; Calculating a wear detection differential coefficient according to the tool wear prediction coefficient and the tool wear first coefficient; Determining whether the wear detection differential coefficient is greater than or equal to a wear detection differential threshold; If the wear detection differential coefficient is greater than or equal to the wear detection differential threshold, the wear differential signal is generated.

4. A metal cutting tool wear monitoring method as claimed in claim 3, characterized in that: Predicting the wear of the metal cutting tool according to the predetermined cutting scheme to obtain a tool wear prediction coefficient includes: Reading a tool pre-wear coefficient of the metal cutting tool, wherein the tool pre-wear coefficient is a tool wear coefficient of the metal cutting tool before executing the predetermined cutting plan; Networking and associating tools of the same model according to the metal cutting tools to determine a set of tools of the same family; Performing a tool wear increase sample search on the set of tools of the same family according to the predetermined cutting scheme to obtain a tool wear increase search set; Performing concentrated value calculation based on the tool wear increase retrieval set to obtain a predicted tool wear increase; The tool wear prediction coefficient is obtained by performing an incentive adjustment calculation on the tool pre-wear coefficient according to the predicted tool wear increase.

5. A metal cutting tool wear monitoring method as claimed in claim 1, characterized in that: Based on the wear differential signal, image perception correction is performed on the first tool wear coefficient to obtain a second tool wear coefficient, including: Based on the wear differential signal, retrieving a real-time tool image of the metal cutting tool; Perceiving the wear characteristics of the metal cutting tool according to the real-time image of the tool, and obtaining the image-perceived tool wear coefficient; Perform adaptive correction parameter calculation according to the first tool wear coefficient and the image-perceived tool wear coefficient to obtain a first tool wear correction coefficient and a second tool wear correction coefficient; According to the first tool wear correction coefficient and the second tool wear correction coefficient, a weighted calculation is performed on the first tool wear coefficient and the image-perceived tool wear coefficient to generate the second tool wear coefficient.

6. A metal cutting tool wear monitoring method as claimed in claim 5, characterized in that: The wear characteristics of the metal cutting tool are sensed according to the real-time image of the tool to obtain the image-perceived tool wear coefficient, including: Performing adaptive wear feature convolution according to the real-time image of the tool to obtain a tool wear convolution image; Aligning the historical tool wear convolution image set and the historical image-aware tool wear coefficient set to obtain the image-aware tool wear assessment construction set; Performing supervised training on the ResNet network according to the image-perceived tool wear assessment construction set, and obtaining a tool wear assessment loss coefficient after each predetermined number of trainings; If the tool wear assessment loss coefficient is less than the tool wear assessment loss threshold, generating an image-perceived tool wear assessment model; The tool wear convolution image is input into the image-perceived tool wear assessment model, and the image-perceived tool wear coefficient is output.

7. A metal cutting tool wear monitoring method as claimed in claim 6, characterized in that: Performing adaptive wear feature convolution according to the real-time image of the tool to obtain a tool wear convolution image, including: Performing Retinex enhancement processing on the real-time image of the tool to obtain a first tool image; Performing background separation on the first tool image according to the Fisher criterion function to obtain a second tool image; According to the second tool image, an adaptive wear feature convolution kernel is constructed, and wear feature convolution is performed on the second tool image according to the adaptive wear feature convolution kernel to obtain an initial convolution image of tool wear; Introducing the Roberts operator to perform edge feature detection of the wear area of ​​the second tool image to obtain a wear edge detection operator; The tool wear initial convolution image is edge enhanced according to the wear edge detection operator to generate the tool wear convolution image.

8. A metal cutting tool wear monitoring method as claimed in claim 1, characterized in that: Filtering and denoising are performed according to the multi-source sensing sequence to construct a tool cutting monitoring matrix, including: Performing Kalman filtering on the cutting force sequence to obtain a cutting force matrix; Performing bandpass filtering on the vibration signal sequence to obtain a vibration signal matrix; Performing bandpass filtering on the acoustic emission signal sequence to obtain an acoustic emission signal matrix; Performing Kalman filtering on the temperature signal sequence to obtain a temperature signal matrix; The tool cutting monitoring matrix is ​​generated by performing time window alignment and splicing according to the cutting force matrix, the vibration signal matrix, the acoustic emission signal matrix and the temperature signal matrix.

9. A metal cutting tool wear monitoring method as claimed in claim 1, characterized in that: Obtain the second tool wear coefficient, including: Determining whether the second tool wear coefficient is greater than or equal to a tool wear threshold; If the second tool wear coefficient is greater than or equal to the tool wear threshold, a tool wear warning signal is generated.

10. A metal cutting tool wear monitoring system, characterized in that: A metal cutting tool wear monitoring method for implementing any one of claims 1 to 9, the system comprising: A monitoring module, used for real-time monitoring of the metal cutting tool when the metal cutting tool executes a predetermined cutting plan to obtain a multi-source sensing sequence, wherein the multi-source sensing sequence includes a cutting force sequence, a vibration signal sequence, an acoustic emission signal sequence, and a temperature signal sequence; A denoising module, used for filtering and denoising according to the multi-source sensing sequence to construct a tool cutting monitoring matrix; A wear detection module, used for inputting the tool cutting monitoring matrix into a tool wear detection channel to obtain a first tool wear coefficient; A differential detection module, used for performing differential detection on the first coefficient of tool wear according to the predetermined cutting scheme, and outputting a wear differential signal; A correction module, configured to perform image perception correction on the first tool wear coefficient based on the wear differential signal to obtain a second tool wear coefficient; A management module is used to obtain a historical tool wear coefficient set of the metal cutting tool, perform wear trend prediction in combination with the second tool wear coefficient, construct a tool wear trend graph, and perform health management of the metal cutting tool according to the tool wear trend graph.

Images