A tool wear state monitoring method and system based on multiple types of signals

Abstract

The application provides a tool wear state monitoring method and system based on multiple types of signals, and relates to the technical field of data processing.The method comprises the following steps: acquiring cutting force, acoustic emission signal and vibration signal data, and extracting multiple statistical features from the cutting force and acoustic emission signal data; combining singularity analysis and wavelet transform to extract singularity point features of the vibration signal; constructing a tool wear state monitoring model based on a random forest, using the obtained features to perform preliminary training, and outputting a wear prediction result; if the preliminary prediction result meets a preset condition, acquiring real-time cutting force, acoustic emission signal and vibration signal data, monitoring the wear state of the tool through the model after preliminary training, otherwise, determining key features through feature importance evaluation, and performing fine training on the model using the key features; and monitoring the wear state of the tool through the model after fine training according to the real-time cutting force, acoustic emission signal and vibration signal data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for monitoring tool wear status based on multiple types of signals. Background Technology

[0002] Tool wear is a key factor in machining accuracy and product quality in manufacturing. In precision manufacturing, the wear condition of tools directly affects machining quality, production efficiency, and equipment lifespan. With the continuous development of industrial automation and intelligent manufacturing, tool wear monitoring technology has gradually become an important means to improve production efficiency and reduce costs. Currently, there are various methods for monitoring tool wear, covering multiple types such as acoustic signals, vibration signals, temperature signals, and current signals. With technological advancements, single-signal monitoring is increasingly unable to meet the demands for high precision and efficiency. Therefore, comprehensive monitoring methods based on multiple signal types have gradually become a research hotspot, enabling a more comprehensive and accurate assessment of tool wear conditions.

[0003] With technological advancements, monitoring methods based on the fusion of multiple signal types have gradually emerged in recent years. These methods improve the accuracy and reliability of tool wear monitoring by combining the characteristics of various signals, such as vibration, acoustics, temperature, and current, for comprehensive analysis.

[0004] However, while existing multi-signal fusion methods can compensate for the shortcomings of single signals, they still face problems such as complex signal processing and high computational cost, making it difficult to meet real-time requirements. Furthermore, due to the nonlinearity and uncertainty of the tool wear process, existing methods exhibit low monitoring accuracy in the early stages of wear when signal changes are not significant. Summary of the Invention

[0005] To address the limitations of existing multi-signal fusion methods, which, while able to compensate for the shortcomings of single-signal methods, still suffer from complex signal processing and high computational load, making it difficult to meet real-time requirements. Furthermore, due to the nonlinearity and uncertainty of the tool wear process, existing methods suffer from low monitoring accuracy in the early stages of wear when signal changes are not significant. Therefore, this invention provides a tool wear condition monitoring method and system based on multiple signal types.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First aspect:

[0008] This invention provides a method for monitoring tool wear status based on multiple signal types, comprising:

[0009] S1: Acquire cutting force data, acoustic emission signal data, and vibration signal data of the tool during the machining process;

[0010] S2: Extract multiple statistical features from the cutting force data and the acoustic emission signal data;

[0011] S3: Extract the singularity features of the vibration signal data by combining singularity analysis with wavelet transform;

[0012] S4: Construct a tool wear condition monitoring model based on random forest;

[0013] S5: Using the statistical features and the singular point features as inputs, perform preliminary training on the tool wear state monitoring model and output preliminary tool wear results;

[0014] S6: Determine whether the preliminary tool wear result meets the preset conditions; if yes, proceed to S9; otherwise, proceed to S7.

[0015] S7: Using the pre-trained tool wear state monitoring model, evaluate the feature importance of each statistical feature and the singular point feature to determine the key features affecting the tool wear state.

[0016] S8: Based on the key features, perform fine training on the tool wear condition monitoring model;

[0017] S9: Acquire real-time cutting force data, real-time acoustic emission signal data, and real-time vibration signal data;

[0018] S10: Based on the real-time cutting force data, the real-time acoustic emission signal data, and the real-time vibration signal data, the tool wear state of the tool is monitored using the trained tool wear state monitoring model.

[0019] The second aspect:

[0020] This invention provides a tool wear condition monitoring system based on multiple signal types, comprising:

[0021] processor;

[0022] A memory storing computer-readable instructions, which, when executed by the processor, implement the tool wear condition monitoring method based on multiple signal types as described in the first aspect.

[0023] Third aspect:

[0024] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the tool wear condition monitoring method based on multiple types of signals as described in the first aspect.

[0025] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0026] (1) In this embodiment of the invention, by extracting statistical features and combining wavelet transform and singularity analysis, the processing flow of the original signal is effectively simplified, and the computational complexity is significantly reduced. The tool wear state monitoring model constructed using random forest can efficiently process large-scale features, thereby avoiding excessive consumption of computing resources and improving the system's real-time performance and processing efficiency.

[0027] (2) In this embodiment of the invention, by initially training the model, effective wear judgment can be provided in the early stage when signal changes are weak, ensuring that preliminary predictions can be made even when signal changes are not significant. Simultaneously, by evaluating the importance of features, key features affecting tool wear are selected, ensuring that the model focuses on those features that can still effectively distinguish wear states when signal changes are small in the early stages. Fine-tuning the model based on these key features further enhances its ability to identify early wear states, thereby significantly improving monitoring accuracy. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating a tool wear condition monitoring method based on multiple signal types provided in an embodiment of the present invention;

[0030] Figure 2 This is a schematic diagram of a tool wear condition monitoring system based on multiple signal types, provided as an embodiment of the present invention. Detailed Implementation

[0031] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0032] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0033] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0034] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0035] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0036] Reference manual attached Figure 1 The diagram shows a flowchart of a tool wear condition monitoring method based on multiple signal types provided by an embodiment of the present invention.

[0037] This invention provides a method for monitoring tool wear condition based on multiple signal types. This method can be implemented by a tool wear condition monitoring device based on multiple signal types, which can be a terminal or a server. The processing flow of the tool wear condition monitoring method based on multiple signal types may include the following steps:

[0038] S1: Acquire cutting force data, acoustic emission signal data, and vibration signal data of the tool during the machining process.

[0039] Specifically, cutting force data is acquired using force sensors, typically triaxial force sensors (or force / torque sensors) to measure cutting forces. These sensors can measure the cutting force on the tool in three directions (X, Y, Z).

[0040] Acoustic emission signal data is acquired in real time using an acoustic emission signal data acquisition system (such as an oscilloscope with appropriate bandwidth or a digital acoustic emission system). Since the acoustic emission signal has a high frequency, the sampling frequency usually needs to be 100kHz or higher to ensure accurate capture of the signal.

[0041] Vibration signal data can be acquired by using accelerometers or vibration sensors (such as piezoelectric sensors) to measure the vibration signals generated by the cutting tool during machining.

[0042] S2: Extract multiple statistical features from cutting force data and acoustic emission signal data.

[0043] In one possible implementation, the statistical characteristics specifically include:

[0044] Maximum, median, mean, standard deviation, peak value, and root mean square of cutting force data.

[0045] The maximum value, median, mean, standard deviation, and root mean square of the acoustic emission signal data.

[0046] In this invention, by extracting multiple statistical features from cutting force data and acoustic emission signal data, complex time-series data can be transformed into concise, structured features, making the model easier to process. Furthermore, different statistical features can describe different aspects of the signal, helping the model to understand the signal more comprehensively.

[0047] S3: By combining singularity analysis with wavelet transform, singular point features of vibration signal data are extracted.

[0048] Singularity analysis is a signal processing technique used to detect abrupt changes, singularities, or irregularities in signals. It is particularly suitable for analyzing signals with non-stationary and nonlinear characteristics, such as mechanical vibration signals, acoustic emission signals, and seismic wave signals.

[0049] Wavelet transform is a powerful signal processing tool that can analyze the characteristics of signals at different scales and frequencies. Unlike the traditional Fourier transform, wavelet transform provides not only frequency domain information but also preserves the time domain information of the signal, making it ideal for processing non-stationary signals such as mechanical vibrations and sound signals.

[0050] In one possible implementation, S3 specifically includes:

[0051] S301: Use wavelet transform to analyze the multi-scale characteristics of vibration signal data.

[0052] Optionally, the wavelet transform is selected as the Mexican Hat wavelet.

[0053] The Mexican Hat wavelet is a commonly used wavelet function with good time-frequency localization properties. It was chosen because it can effectively detect abrupt changes and sharp features in signals, making it particularly suitable for the analysis of vibration signals.

[0054] Specifically, wavelet transform is applied to vibration signal data to extract signal features from multiple scales (frequency ranges). Through wavelet transform, you can obtain the components of the signal in the low-frequency and high-frequency regions, revealing the multi-scale characteristics of the signal.

[0055] S302: Perform modulus maxima analysis on the wavelet transform coefficients at various scales to determine the number and intensity of maxima points.

[0056] It should be noted that modulus maxima analysis is performed on the coefficients after wavelet transform to extract local maxima. These maxima typically correspond to irregularities or abrupt changes in the signal.

[0057] S303: Calculate the Lipschitz index at each scale based on the number and intensity of maxima.

[0058]

[0059] Where, α q Let N represent the Lipschitz index at the q-th scale. q m represents the number of maxima at the q-th scale. p Let m represent the p-th local maximum. p-1 Let I(m) represent the (p-1)th local maximum. p I(m) represents the intensity of the p-th maximum point. p-1 ) represents the intensity of the (p-1)th maximum point, and log represents the logarithmic function.

[0060] The Lipschitz exponent is a mathematical index used to describe the smoothness or abruptness of a function or signal at a certain point. It is a quantitative description of Lipschitz continuity and is widely used in signal analysis, image processing, and machine learning, especially in dealing with non-stationary signals and abrupt features.

[0061] It should be noted that α q The larger α is, the smoother the signal is at the q-th scale. Conversely, the smaller α is... q The smaller the value, the more pronounced the abrupt changes in the signal at that scale.

[0062] In this invention, the Lipschitz index can quantify the smoothness of a signal. When there are abrupt changes in the signal (such as tool wear, cracks, or equipment failure), the Lipschitz index will decrease significantly, reflecting that the change in the signal is no longer smooth. In this way, the Lipschitz index can help the model automatically identify sudden events or abnormal changes.

[0063] S304: A weighted average of the Lipschitz exponents at various scales is used to obtain the comprehensive Lipschitz exponent of the vibration signal data.

[0064]

[0065] Where HE represents the comprehensive Lipschitz exponent, Q represents the total number of scales after wavelet transform, and N... k This represents the number of maxima at the k-th scale. This represents the sum of the number of maxima at Q scales.

[0066] In this invention, the comprehensive Lipschitz exponent is obtained by weighted averaging of the Lipschitz exponents at various scales. This averaging incorporates abrupt changes in the signal at different scales, providing a more comprehensive signal description. This weighted averaging makes the model more sensitive to changes in different frequency bands of the signal, thereby improving the overall predictive ability.

[0067] S305: Output the composite Lipschitz index as a singularity feature.

[0068] S4: Construct a tool wear condition monitoring model based on random forest.

[0069] Among them, the tool wear condition monitoring model based on Random Forest (RF) is a predictive model that combines machine learning and sensor data, enabling real-time monitoring of tool wear conditions. Random Forest is a powerful ensemble learning method widely used in regression and classification problems. This model is particularly suitable for handling the complex, multi-dimensional data involved in tool wear monitoring.

[0070] In one possible implementation, S4 specifically includes:

[0071] S401: Construct the original dataset based on various statistical features:

[0072] D = {(x1,y1),(x2,y2),…,(x N ,y N )}

[0073] Where D represents the original dataset, x i Let y represent the feature vector of the i-th sample. i Let i represent the target variable for the i-th sample, where i = 1, 2, 3, ..., N, and N represents the number of samples in the dataset.

[0074] S402: Randomly generate multiple bootstrap samples from the original dataset. Each bootstrap sample contains multiple sample data.

[0075] In this invention, the multi-sample generation of self-service samples allows each tree to be trained on an independent sample set, avoiding the problem of overfitting due to certain specific data points.

[0076] S403: Construct multiple decision trees based on each bootstrap sample:

[0077]

[0078] Among them, T b (x,h b ) represents the prediction value of the b-th decision tree on the input data x, where x represents the input data point, and h b Let M represent the tree structure of the b-th decision tree, and let μ represent the total number of leaf nodes in the decision tree. m R represents the average of the target values ​​of all samples in the m-th leaf node. m I() represents the sample set of the m-th leaf node, and I(x∈R) represents the indicator function. m ) = 1 indicates that sample x belongs to leaf node R m ,I(x∈R) m ) = 0 indicates that sample x does not belong to the leaf node R. m .

[0079] In one possible implementation, S403 specifically includes:

[0080] S4031: Select multiple splitting variables from the bootstrap sample.

[0081] S4032: Select split points among the splitting variables and divide each splitting variable into left and right subsets based on the split points:

[0082] R1(j,s)={X∣X j ≤s}, R2(j,s)={X|X j >s}

[0083] Where R1 represents the left subset after splitting, R2 represents the right subset after splitting, and X represents a single sample in the dataset. j Let represent the j-th feature of the sample, and s represent the split point.

[0084] In this invention, by randomly selecting splitting variables, it is ensured that the knowledge learned by each tree is different, avoiding the situation where all trees learn the same features, thereby enhancing the generalization ability of the model.

[0085] S4033: Determine the optimal split point with the objective of minimizing the sum of squared residuals for each subset:

[0086]

[0087] Where min represents minimization, μ1 represents the mean of the objective variable in the left subset, and x i Let y represent the feature vector of the i-th sample, R1 represent the left subset after splitting, and y iLet represent the target variable of the i-th sample, μ2 represent the mean of the target variable in the right subset, R2 represent the right subset after splitting, j represent the feature, and s represent the split point.

[0088] In this invention, by selecting the optimal split point (i.e. minimizing the sum of squared residuals of the subsets), the model can split the data into purer subsets, thereby improving the accuracy of predicting tool wear conditions.

[0089] S4034: Based on the optimal split point, split the current node into a left child node and a right child node, and recursively execute steps S4031 to S4034 on the left child node and the right child node.

[0090] S4035: When the number of samples in the bootstrap sample is lower than the preset number or the depth of the tree reaches the maximum depth, stop splitting and complete the construction of each decision tree.

[0091] S404: Combine the decision trees to construct a tool wear condition monitoring model:

[0092]

[0093] in, Let B represent the final prediction value of the random forest for the input data x, where B represents the total number of trees in the random forest, x represents the input data point, and h represents the final prediction value. b Let T represent the tree structure of the b-th decision tree. b (x,h b ) represents the predicted value of the b-th decision tree on the input data x.

[0094] In summary, by recursively splitting nodes, random forests can extract features from data at multiple levels, capture different patterns of tool wear, provide more comprehensive predictions, reduce the bias and variance of individual trees, and make the model's prediction of tool wear more accurate.

[0095] S5: Using various statistical features and singular point features as inputs, the tool wear condition monitoring model is initially trained, and preliminary tool wear results are output.

[0096] S6: Determine whether the preliminary tool wear results meet the preset conditions. If yes, proceed to S9. Otherwise, proceed to S7.

[0097] In one possible implementation, the preset condition is specifically: the prediction accuracy of the preliminary tool wear result is greater than 80%.

[0098] S7: Using the pre-trained tool wear condition monitoring model, evaluate the importance of each statistical feature and singular point feature to determine the key features affecting the tool wear condition.

[0099] Feature importance assessment is a technique in machine learning used to measure the contribution of each feature to the model's predictions. By evaluating feature importance, we can identify which features have a stronger predictive power for the target variable (such as tool wear status), thereby optimizing the model, improving prediction accuracy, and reducing redundant features.

[0100] In one possible implementation, S7 specifically includes:

[0101] S701: Calculate the feature importance evaluation values ​​for each statistical feature and singularity feature using the following formula:

[0102]

[0103] Δ i (s t ,t)=i(t)-p1i(t1)-p2i(t2)

[0104]

[0105] Among them, VarImp(x i ) represents feature x i Importance evaluation value, T represents a decision tree in the random forest, N T v(s) represents the total number of trees in the random forest. t ) represents the feature of the t-th split node, v(s) t )=x i The feature of the t-th split node is obtained by splitting feature xi, p(t) represents the proportion of samples in node t, and Δ i (s t ,t) represents feature x i The reduction in the sum of squared residuals resulting from splitting at the t-th node, where i(t) represents the sum of squared residuals at the t-th node, p1 represents the proportion of samples from the left child node, p2 represents the proportion of samples from the right child node, i(t1) represents the sum of squared residuals from the left child node, i(t2) represents the sum of squared residuals from the right child node, R1 represents the left subset after splitting, R2 represents the right subset after splitting, and x i Let y represent the i-th sample. i Let i represent the target variable for the i-th sample. This represents the mean of the target variable for the samples in the left subset. This represents the mean of the target variable for the samples in the right subset.

[0106] In this invention, by analyzing the importance of features, we can intuitively understand how the model makes decisions. For tool wear monitoring models, we can clearly identify which features (such as the mean of cutting force, the standard deviation of vibration, etc.) are most critical for predicting wear conditions, thus providing valuable information for maintenance personnel. Meanwhile, irrelevant features or features with little impact on the target variable will have their importance underestimated, effectively removing redundant features, thereby simplifying subsequent modeling and improving computational efficiency.

[0107] It should be noted that Δ i (s t The larger the value of ,t), the more helpful the feature is in predicting the target variable.

[0108] S702: Sort the importance evaluation values ​​of each feature in descending order, and select the feature corresponding to the highest importance evaluation value as the key feature.

[0109] In this invention, by using only features with strong predictive power for the target variable (tool wear state), the model can reduce noise interference and focus on key features, thereby improving the overall predictive performance of the model. Simultaneously, identifying which features (such as certain statistical characteristics in cutting force, vibration, or acoustic emission signals) have the greatest impact on wear prediction can provide valuable guidance for further tool management and maintenance. Using these features first improves the efficiency of immediate decision-making.

[0110] S8: Based on key features, perform fine training on the tool wear condition monitoring model.

[0111] In one possible implementation, the method for finely training the tool wear condition monitoring model is as follows:

[0112] With the goal of minimizing the mean square error loss function, an improved particle swarm optimization algorithm is used to finely train the tool wear condition monitoring model.

[0113] The mean squared error loss function is as follows:

[0114]

[0115] Where MSE represents the mean squared error loss function, y i Let represent the true target value of the i-th sample, and let represent . Let N represent the predicted value of the i-th sample, and N represent the total number of samples in the dataset.

[0116] Optionally, the improved particle swarm optimization algorithm specifically includes:

[0117] S801: Initialize the particle swarm, which contains Togo particles, each representing a feasible solution.

[0118] S802: Improve the fitness function of the particle swarm optimization algorithm with the inverse of the mean squared error loss function.

[0119] S803: Calculate the fitness function value for each particle in the swarm.

[0120] S804: Update the position of the particle with the highest fitness value to the global optimum.

[0121] S805: Determine the neighborhood of each particle, and based on the particle information within the neighborhood, determine the local optimal position of the particle with the highest fitness value within the neighborhood.

[0122] In this invention, by guiding the particle to its local optimal position and sharing neighborhood information, the particle's search capability within a local area can be improved. The particle adjusts its position based on the particle with the highest fitness value in its neighborhood, thereby improving the accuracy of the local search.

[0123] S806: Update the particle's velocity and position using the following formula:

[0124] V u =ωV u +c1r1(Pbset u -X u )+c2r2(Gbset-X u )

[0125] Among them, V u Let ω represent the particle velocity, ω represent the inertial weight, c1 and c2 represent the acceleration constants, and r1 and r2 represent random numbers. Pbest u Gbest represents the historical best position of the u-th particle, and Gbest represents the global best position.

[0126] S807: Generate a random number and determine if the generated random number is less than the preset mutation probability. If yes, proceed to S808. Otherwise, proceed to S809.

[0127] S808: Using the mutation operation, update the particle's position using the following formula, and then proceed to S812:

[0128] T uv =X uv +A(Lbest-X uv )

[0129] Among them, T uv X represents the position of the u-th particle after mutation in the v-th dimension. uv This represents the current position information of the u-th particle in the v-th dimension, and A represents the magnification factor.

[0130] S809: Apply a random permutation function to the local optimal locations to increase the diversity of the solution space:

[0131] Lbest := permuting(Lbest)

[0132] Here, Lbest represents the optimal solution position found by the particle in its neighborhood, and permuting() represents the random permutation function.

[0133] In this invention, by introducing mutation operations and random permutation of the local solution space, the diversity of the solution space can be maintained during the optimization process, the global search capability can be enhanced, and it is helpful to find a better solution.

[0134] S810: The experimental location for generating particles is determined using the following formula:

[0135] Mutant = X + F(Lbest - X)

[0136] Where Mutant represents the position of the mutated particle, X represents the current position of the particle, Lbest represents the local best position, and F represents the amplification factor.

[0137] S811: Use the crossover operator to perform a crossover operation on the current particle to generate the solution for the next generation, then proceed to S812.

[0138] S812: Calculate the fitness of each particle and update the global optimal position based on the particle's fitness.

[0139] S813: Determine if the maximum number of iterations is met. If yes, output the optimal result. Otherwise, return to S805.

[0140] S9: Acquire real-time cutting force data, real-time acoustic emission signal data, and real-time vibration signal data.

[0141] S10: Based on real-time cutting force data, real-time acoustic emission signal data, and real-time vibration signal data, the tool wear condition is monitored using the trained tool wear condition monitoring model.

[0142] Optionally, the tool wear condition can be categorized as light wear, moderate wear, and heavy wear.

[0143] Specifically, real-time cutting force data, real-time acoustic emission signal data, and real-time vibration signal data are acquired. Based on the feature importance analysis results, key features affecting tool wear status are identified. The key features are used as the main focus, while the other features are used as supplementary features. These are then input into the trained tool wear status monitoring model, and the tool wear status is output.

[0144] In this invention, by selecting key features based on feature importance analysis and inputting them into the trained tool wear condition monitoring model, the model's prediction accuracy and computational efficiency can be significantly improved. This not only optimizes the model's performance and reduces redundant features and computational burden, but also enhances the model's stability and robustness, avoiding overfitting. Simultaneously, by focusing on the most relevant features, more accurate and flexible decision support can be provided for real-time monitoring of tool wear conditions, reducing maintenance costs and improving the system's real-time response capability and adaptability.

[0145] In one possible implementation, the process after S10 includes:

[0146] Record the tool wear condition and generate a tool wear condition report.

[0147] In this invention, real-time recording of tool wear status provides valuable historical data for subsequent analysis. By tracking the wear changes of each tool throughout its entire lifespan, it is possible to clearly trace back to any specific point in time, helping engineers understand wear trends and historical events. This is of great value for later maintenance decisions and operational optimization.

[0148] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0149] (1) In this embodiment of the invention, by extracting statistical features and combining wavelet transform and singularity analysis, the processing flow of the original signal is effectively simplified, and the computational complexity is significantly reduced. The tool wear state monitoring model constructed using random forest can efficiently process large-scale features, thereby avoiding excessive consumption of computing resources and improving the system's real-time performance and processing efficiency.

[0150] (2) In this embodiment of the invention, by initially training the model, effective wear judgment can be provided in the early stage when signal changes are weak, ensuring that preliminary predictions can be made even when signal changes are not significant. Simultaneously, by evaluating the importance of features, key features affecting tool wear are selected, ensuring that the model focuses on those features that can still effectively distinguish wear states when signal changes are small in the early stages. Fine-tuning the model based on these key features further enhances its ability to identify early wear states, thereby significantly improving monitoring accuracy.

[0151] Reference manual attached Figure 2 The diagram shows a schematic of the structure of a tool wear condition monitoring system based on multiple signal types provided by the present invention.

[0152] The present invention also provides a tool wear condition monitoring system 20 based on multiple signal types, applied to the above-mentioned tool wear condition monitoring method based on multiple signal types, comprising:

[0153] Processor 201.

[0154] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, they implement the tool wear condition monitoring method based on multiple types of signals as described in the method embodiment.

[0155] The tool wear condition monitoring system 20 based on multiple signal types provided by the present invention can perform the tool wear condition monitoring method based on multiple signal types described above and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0156] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0157] (1) In this embodiment of the invention, by extracting statistical features and combining wavelet transform and singularity analysis, the processing flow of the original signal is effectively simplified, and the computational complexity is significantly reduced. The tool wear state monitoring model constructed using random forest can efficiently process large-scale features, thereby avoiding excessive consumption of computing resources and improving the system's real-time performance and processing efficiency.

[0158] (2) In this embodiment of the invention, by initially training the model, effective wear judgment can be provided in the early stage when signal changes are weak, ensuring that preliminary predictions can be made even when signal changes are not significant. Simultaneously, by evaluating the importance of features, key features affecting tool wear are selected, ensuring that the model focuses on those features that can still effectively distinguish wear states when signal changes are small in the early stages. Fine-tuning the model based on these key features further enhances its ability to identify early wear states, thereby significantly improving monitoring accuracy.

[0159] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0160] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0161] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0162] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0163] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0164] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0165] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0166] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0167] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0168] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0169] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0170] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes 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.

[0171] This invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the tool wear condition monitoring method based on multiple types of signals as described in the method embodiment.

[0172] The present invention provides a computer-readable storage medium that can implement the steps and effects of the tool wear condition monitoring method based on multiple types of signals in the above-described method embodiments. To avoid repetition, the present invention will not elaborate further.

[0173] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0174] (1) In this embodiment of the invention, by extracting statistical features and combining wavelet transform and singularity analysis, the processing flow of the original signal is effectively simplified, and the computational complexity is significantly reduced. The tool wear state monitoring model constructed using random forest can efficiently process large-scale features, thereby avoiding excessive consumption of computing resources and improving the system's real-time performance and processing efficiency.

[0175] (2) In this embodiment of the invention, by initially training the model, effective wear judgment can be provided in the early stage when signal changes are weak, ensuring that preliminary predictions can be made even when signal changes are not significant. Simultaneously, by evaluating the importance of features, key features affecting tool wear are selected, ensuring that the model focuses on those features that can still effectively distinguish wear states when signal changes are small in the early stages. Fine-tuning the model based on these key features further enhances its ability to identify early wear states, thereby significantly improving monitoring accuracy.

[0176] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0177] The following points need to be explained:

[0178] (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0179] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the present invention; that is, these drawings are not drawn to actual scale. It is understood that when an element such as a layer, film, region, or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element, or there may be intermediate elements.

[0180] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0181] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring tool wear condition based on multiple signal types, characterized in that, include: S1: Acquire cutting force data, acoustic emission signal data, and vibration signal data of the tool during the machining process; S2: Extract multiple statistical features from the cutting force data and the acoustic emission signal data; S3: Extract the singularity features of the vibration signal data by combining singularity analysis with wavelet transform; S4: Construct a tool wear condition monitoring model based on random forest; S5: Using the statistical features and the singular point features as inputs, perform preliminary training on the tool wear state monitoring model and output preliminary tool wear results; S6: Determine whether the preliminary tool wear result meets the preset conditions; If so, proceed to S9; Otherwise, proceed to S7; S7: Using the pre-trained tool wear state monitoring model, evaluate the feature importance of each statistical feature and the singular point feature to determine the key features affecting the tool wear state. S8: Based on the key features, perform fine training on the tool wear condition monitoring model; S9: Acquire real-time cutting force data, real-time acoustic emission signal data, and real-time vibration signal data; S10: Based on the real-time cutting force data, the real-time acoustic emission signal data, and the real-time vibration signal data, the tool wear state of the tool is monitored using the trained tool wear state monitoring model.

2. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, The statistical characteristics specifically include: The maximum value, median, mean, standard deviation, peak value, and root mean square of the cutting force data; The maximum value, median, mean, standard deviation, and root mean square of the acoustic emission signal data.

3. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, S3 specifically includes: S301: Use wavelet transform to analyze the multi-scale characteristics of the vibration signal data; S302: Perform modulus maxima analysis on the wavelet transform coefficients at various scales to determine the number and intensity of maxima points; S303: Calculate the Lipschitz index at each scale based on the number and intensity of the maxima: Where, α q Let N represent the Lipschitz index at the q-th scale. q m represents the number of maxima at the q-th scale. p Let m represent the p-th local maximum. p-1 Let I(m) represent the (p-1)th local maximum. p I(m) represents the intensity of the p-th maximum point. p-1 ) represents the intensity of the (p-1)th maximum point, and log represents the logarithmic function; S304: Calculate the weighted average of the Lipschitz indices at each scale to obtain the comprehensive Lipschitz index of the vibration signal data. Where HE represents the comprehensive Lipschitz exponent, Q represents the total number of scales after wavelet transform, and N... k This represents the number of maxima at the k-th scale. This represents the sum of the number of maxima on Q scales; S305: Output the comprehensive Lipschitz index as the singularity feature.

4. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, S4 specifically includes: S401: Construct the original dataset based on each of the aforementioned statistical features; S402: Randomly generate multiple bootstrap samples from the original dataset; wherein each bootstrap sample contains multiple sample data; S403: Based on each of the aforementioned bootstrap samples, construct multiple decision trees: Among them, T b (x,h b ) represents the prediction value of the b-th decision tree on the input data x, where x represents the input data point, and h b Let M represent the tree structure of the b-th decision tree, and let μ represent the total number of leaf nodes in the decision tree. m R represents the average of the target values ​​of all samples in the m-th leaf node. m I() represents the sample set of the m-th leaf node, and I(x∈R) represents the indicator function. m ) = 1 indicates that sample x belongs to leaf node R m ,I(x∈R) m ) = 0 indicates that sample x does not belong to the leaf node R. m ; S404: Combine the decision trees to construct the tool wear condition monitoring model: in, Let B represent the final prediction value of the random forest for the input data x, where B represents the total number of trees in the random forest, x represents the input data point, and h represents the final prediction value. b Let T represent the tree structure of the b-th decision tree. b (x,h b ) represents the predicted value of the b-th decision tree on the input data x.

5. The tool wear condition monitoring method based on multiple signal types according to claim 4, characterized in that, Specifically, S403 includes: S4031: Select multiple splitting variables from the bootstrap sample; S4032: Select a split point among the split variables and divide each split variable into two subsets, left and right, based on the split point; S4033: Determine the optimal split point with the objective of minimizing the sum of squared residuals for each subset: Where min represents minimization, μ1 represents the mean of the objective variable in the left subset, and x i Let y represent the feature vector of the i-th sample, R1 represent the left subset after splitting, and y i Let represent the target variable of the i-th sample, μ2 represent the mean of the target variable in the right subset, R2 represent the right subset after splitting, j represent the feature, and s represent the split point; S4034: Based on the optimal split point, divide the current node into a left child node and a right child node, and recursively execute steps S4031 to S4034 on the left child node and the right child node. S4035: When the number of samples in the self-service sample is lower than the preset number or the depth of the tree reaches the maximum depth, stop splitting and complete the construction of each decision tree.

6. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, The preset condition is specifically: the prediction accuracy of the preliminary tool wear result is greater than 80%.

7. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, Specifically, S7 includes: S701: Calculate the feature importance evaluation value of each of the statistical features and the singular point features using the following formula: Δ i (s t ,t)=i(t)-p1i(t1)-p2i(t2) Among them, VarImp(x i ) represents feature x i Importance evaluation value, T represents a decision tree in the random forest, N T v(s) represents the total number of trees in the random forest. t ) represents the feature of the t-th split node, v(s) t ) = x i The feature of the t-th split node is obtained by splitting feature xi, p(t) represents the proportion of samples in node t, and Δ i (s t ,t) represents feature x i The reduction in the sum of squared residuals resulting from splitting at the t-th node, where i(t) represents the sum of squared residuals at the t-th node, p1 represents the proportion of samples from the left child node, p2 represents the proportion of samples from the right child node, i(t1) represents the sum of squared residuals from the left child node, i(t2) represents the sum of squared residuals from the right child node, R1 represents the left subset after splitting, R2 represents the right subset after splitting, and x i Let y represent the i-th sample. i Let i represent the target variable for the i-th sample. This represents the mean of the target variable for the samples in the left subset. This represents the mean of the target variable for the samples in the right subset; S702: Sort the importance evaluation values ​​of each feature in descending order, and select the feature corresponding to the feature with the highest importance evaluation value as the key feature.

8. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, The specific method for fine-tuning the tool wear condition monitoring model is as follows: With the goal of minimizing the mean square error loss function, the tool wear condition monitoring model is finely trained using an improved particle swarm optimization algorithm.

9. The tool wear condition monitoring method based on multiple signal types according to claim 1, characterized in that, Following S10, the following is also included: The wear condition of the tool is recorded, and a tool wear condition report is generated.

10. A tool wear condition monitoring system based on multiple signal types, characterized in that, include: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the tool wear condition monitoring method based on multiple types of signals as described in any one of claims 1 to 9.

Images