Hard and brittle material ultra-precision turning damage monitoring method based on multi-sensor fusion

Abstract

The invention discloses a hard and brittle material ultra-precision turning damage monitoring method based on multi-sensor fusion, and relates to the technical field of precision / ultra-precision machining state monitoring, a multi-sensor signal acquisition device is built on an ultra-precision turning machine tool to capture different types of sensor signals generated in the turning process in real time; performing signal preprocessing operation of minimum entropy deconvolution filtering on the acquired sensor signal; feature extraction is carried out on the signals subjected to minimum entropy deconvolution filtering, and 95 time domain and frequency domain statistical features reflecting turning damage are extracted; by integrating data from different sensors, the precision and reliability of monitoring are remarkably improved, more comprehensive and more accurate information can be obtained by capturing and fusing acoustic characteristics, cutting force and mechanical vibration information in the machining process, so that the abrasion degree of a tool and the damage condition of a workpiece are judged more accurately, and the machining accuracy is improved. And false alarm and missing alarm can be reduced, and the monitoring reliability can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of precision / ultra-precision machining state monitoring, and particularly to a method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion. Background Art

[0002] Hard and brittle materials such as single-crystalline silicon, single-crystalline germanium, glass, silicon carbide, and sapphire, with their excellent physical and chemical properties, have shown great application potential in high-tech fields such as aerospace, electronic engineering, and optics. With the continuous improvement of the requirements for material processing accuracy in modern science and technology, the limitations of traditional grinding processing technology have become increasingly prominent, specifically manifested as problems such as low processing efficiency, high cost, and easy occurrence of thermal damage during the processing. In contrast, ultra-precision turning technology, through the coordinated action of diamond tools, high-precision CNC machine tools, and precision measurement systems, has achieved machining accuracy at the micron and even nanometer levels, while maintaining relatively high processing efficiency. Currently, ultra-precision turning technology has been widely and deeply applied in the field of hard and brittle material processing.

[0003] In the electronics industry, this technology is used to manufacture high-performance alumina ceramic capacitors and insulators, significantly improving the performance and reliability of electronic devices; in the field of infrared optics, ultra-precision turning technology is used to process optical components made of materials such as sapphire, fused quartz, and single-crystalline silicon, such as lenses, mirrors, and windows, which play a crucial role in fields such as infrared detection, imaging, and communication. However, due to the inherent characteristics of hard and brittle materials, such as high hardness, high brittleness, and low thermal conductivity, they are extremely prone to brittle damage such as cracks and spalling during ultra-precision turning processing, thereby affecting the integrity of the processed surface and making it difficult to meet the strict requirements for the performance of hard and brittle materials in high-end applications. Currently, suppressing / eliminating brittle damage generated during the processing of hard and brittle materials has become the core requirement in this field.

[0004] To further improve the processing quality, while increasing production efficiency and reducing manufacturing costs, the application of intelligent monitoring systems in the field of ultra-precision turning is gradually being taken seriously. However, currently, the signal monitoring objects for the ultra-precision turning process mainly focus on tool wear, chatter, and workpiece roughness, etc., and there is still a lack of effective real-time signal monitoring methods to predict damage to the processed surface of materials. In addition, the current signal monitoring process often relies on a single sensor, and the information obtained is relatively one-sided and has large errors. In view of the increasing demand for ultra-precision non-destructive processing of hard and brittle materials, it is of great practical significance to propose a method for in-situ monitoring of processing damage based on multi-sensor signals. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion to solve the problems raised in the above background art.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] A method for monitoring the damage of ultra-precision turning of brittle materials based on multi-sensor fusion, comprising the following steps:

[0008] Step 1, build a multi-sensor signal acquisition device on an ultra-precision lathe, including an acoustic emission sensor, a force sensor and a vibration sensor. The sensors are all arranged at the tool end close to the material processing area to capture different types of sensor signals generated during the turning process in real time;

[0009] Step 2, perform signal preprocessing operations on the collected sensor signals by minimum entropy deconvolution (MED) filtering to enhance the signal recognizability, weaken the adverse effects, and restore the impact components in the filtered signals to the greatest extent;

[0010] Step 3, extract features from the signals filtered by minimum entropy deconvolution, and extract a total of 95 time-domain and frequency-domain statistical features reflecting turning damage, including time-domain statistical indicators such as peak value, mean value, root mean square value, standard deviation, kurtosis, skewness, etc., and frequency-domain statistical indicators such as spectral centroid, mean square frequency, etc.;

[0011] Step 4, use kernel principal component analysis for feature dimension reduction and fusion, and through min-max normalization processing and radial basis function kernel, map the original data to a high-dimensional feature space, calculate the kernel matrix of the feature matrix, perform eigen-decomposition, select the first few principal components whose cumulative contribution rate reaches a certain threshold, and project the original data onto the selected principal components to obtain the data after dimension reduction;

[0012] Step 5, adopt the least squares support vector machine method optimized by genetic algorithm for pattern recognition to identify the material removal pattern of brittle materials during ultra-precision turning and determine the machining damage state.

[0013] A further improvement of the technical solution of the present invention lies in: in the above step 1, the process of capturing the sensor signals during turning is as follows:

[0014] Step 101, install an acoustic emission sensor, a force sensor and a vibration sensor at the tool end of the ultra-precision lathe close to the material processing area to capture the acoustic characteristics, cutting force and mechanical vibration information during the processing respectively, ensure that the sensors are fixed firmly to avoid displacement or vibration during turning, which affects the signal accuracy, and install a signal preprocessing device, including a current amplifier and a filter. Among them, install a current amplifier to enhance the weak signals output by the sensors, and install a filter to reduce noise and interference and improve the signal quality;

[0015] Step 102: Connect the signal line output by the sensor to the data acquisition card of the data acquisition module, ensure the correct connection and compatibility between the data acquisition card and the host computer (computer), configure the sampling rate and resolution of the data acquisition card to ensure that the digital signal can accurately reflect the physical quantity measured by the sensor, and set the input channels of the data acquisition card corresponding to the signal inputs of each sensor.

[0016] Step 103: Start the data acquisition card and continuously collect the output signals of each sensor during the turning process. After the sensor signals are processed by the current amplifier and filter, they are transmitted to the data acquisition card of the data acquisition module for analog-to-digital (A / D) conversion.

[0017] Step 104: Store the collected digital signals on the hard disk of the host computer (computer), ensure the stability and security during the data transmission process, avoid data loss or damage for subsequent processing and analysis, and install the corresponding signal processing software on the host computer (computer) to receive, store, and preliminarily process the collected digital signals. At the same time, conduct a preliminary check on the data from the aspects of signal integrity, synchronization, and timestamp verification.

[0018] A further improvement of the technical solution of the present invention lies in that: in the said Step 2, the process of signal preprocessing is as follows:

[0019] Step 201: Load the collected force signal, vibration signal, and acoustic emission signal data from the hard disk of the host computer (computer), and perform preprocessing operations on the loaded signal data, such as removing the DC offset, filtering out low-frequency drift, and high-frequency noise.

[0020] Step 202: Set the length L of the inverse filter according to the characteristics and requirements of the signal, initialize the parameters of the inverse filter with random values, and set the objective function as the minimum entropy, and then calculate the impact signal generated during the contact process between the tool and the workpiece. Among them, the longer the length of the inverse filter, the more signal features can be captured, but the computational complexity will also increase, and the minimum entropy means that the entropy of the output signal is the smallest, that is, the information in the signal is the most concentrated.

[0021] Step 203: Use the normalized kurtosis as the objective function to optimize the inverse filter. Through iterative calculation, adjust the parameters of the inverse filter to maximize the kurtosis, so as to achieve the minimum entropy condition, and apply the optimized inverse filter to the original signal, and then determine the optimal inverse filter matrix.

[0022] Step 204: Perform time-domain and frequency-domain analysis on the filtered signal. Observe the changes in the waveform and impulse components of the signal through time-domain analysis, and observe the changes in the signal spectrum and frequency components through frequency-domain analysis. Compare the time-domain waveforms and frequency-domain spectrograms of the signal before and after filtering to verify whether the impulse components are significantly enhanced and whether the signal-to-noise ratio is improved. Compare the time-domain waveforms and frequency-domain spectrograms of the original signal and the filtered signal to visually display the filtering effect, which helps to confirm whether the minimum entropy deconvolution filtering successfully extracts the impulse components in the signal and weakens the adverse effects.

[0023] A further improvement of the technical solution of the present invention lies in that: the calculation formula of the impact signal generated during the contact process between the tool and the workpiece is expressed as:

[0024]

[0025] where y(n) is the impact signal generated during the contact process between the tool and the workpiece, f(n) is the inverse filter, x(n) is the original acquired signal, L is the length of the inverse filter, * is the convolution operation, f in f(l) refers to the convolution / inverse filter set in the minimum entropy deconvolution filtering method, which is equivalent to a convolution kernel, n is the nth sample in the signal time series, and l is the lth coefficient in the convolution kernel;

[0026] Since the larger the kurtosis value (the fourth-order central moment of data normalization) of the signal, the smaller its entropy value, the normalized kurtosis is selected as the objective function in the minimum entropy deconvolution filtering (MED), and its expression is:

[0027]

[0028] When in the formula reaches the maximum value, that is, it satisfies the extreme condition then f(l) can be determined as the optimal inverse filter. Combining with formula (1), the following equation can be obtained:

[0029]

[0030] where is the normalized kurtosis, y(i) represents the ith sample in the output impact signal time series, the x, y, and f functions are the original acquired signal, the output impact signal, and the convolution / inverse filter (convolution kernel) set in the minimum entropy deconvolution filtering (MED) method respectively, n and l represent the nth or lth sample in the signal time series, p represents the pth coefficient in the convolution kernel, and i, n, l, p represent the ith, nth, lth, and pth values in the x, y, and f functions respectively;

[0031] Formula (3) can be written in matrix form:

[0032] f = A -1×b; (4)

[0033] Through iterative calculations using formula (4), the optimal inverse filter matrix is finally determined.

[0034] A further improvement of the technical solution of the present invention lies in: in step 3, the extraction process of the time-domain and frequency-domain statistical features of the turning damage is as follows:

[0035] Step 301: Import the signal data processed by minimum entropy deconvolution filtering into signal analysis software, and extract the turning damage eigenvalue from the filtered signals of different sensors, which are the time-domain and frequency-domain statistical features respectively, providing input variables for subsequent pattern recognition or machine learning models;

[0036] Step 302: Conduct time-domain analysis on the filtered signals of different sensors, and extract 11 time-domain statistical indicators from the time-domain signals, including peak value, mean value, root mean square value, standard deviation, kurtosis, skewness, peak-to-peak value, peak factor, impulse factor, waveform factor, and margin factor;

[0037] Step 303: Perform fast Fourier transform on the time-domain signal to obtain signal spectrum information, and extract 8 frequency-domain statistical indicators from it, including spectrum centroid, mean square frequency, root mean square frequency, frequency variance, spectrum peak value, power spectrum entropy, spectrum kurtosis, and spectrum skewness

[0038] Step 304: Integrate the time-domain and frequency-domain features extracted by each sensor to form a feature vector, and extract a total of 95 statistical features. Among them, the force signal includes forces in the X, Y, and Z directions.

[0039] A further improvement of the technical solution of the present invention lies in: in step 4, the process of obtaining the data after dimensionality reduction is as follows:

[0040] Step 401: Organize the 95 statistical features extracted into a feature matrix, where each row represents a sample and each column represents a feature, and perform Min-Max standardization processing on the feature matrix to scale the value of each feature to the interval [0,1], eliminating the influence of the dimension between different features, making the data at the same order of magnitude, and then performing kernel function calculation;

[0041] Step 402: Select the radial basis function (RBF) as the kernel function. The RBF kernel is also called the Gaussian kernel and is suitable for mapping of non-linear data. Use the selected radial basis function (RBF) kernel function to calculate the kernel function values between each sample point in the original dataset to form a kernel matrix. Among them, each element of the kernel matrix is the kernel function value between two sample points;

[0042] Step 403: Subtract the mean of each row and each column from each element in the kernel matrix to centralize the kernel matrix, ensuring the correct mapping of data in the high-dimensional space, eliminating the offset in the data, and making the subsequent eigenvalue decomposition more accurate.

[0043] Step 404: Perform eigenvalue decomposition on the centralized kernel matrix to obtain eigenvalues and corresponding eigenvectors. Among them, the eigenvalues represent the amount of information contained in each principal component, and the eigenvectors represent the directions of the principal components. According to the magnitudes of the eigenvalues, select multiple (the first few) principal components whose cumulative contribution rate reaches a certain threshold (such as 95% or 99%). These principal components can retain the information of the original data to the greatest extent.

[0044] Step 405: Project the original data onto the selected principal components. By multiplying the kernel matrix with the eigenvectors of the selected principal components, the dimensionality-reduced data is obtained. The dimensionality-reduced data not only retains the main variation information of the original data but also has a greatly reduced dimension, thereby effectively reducing redundant information, improving the computational efficiency of the subsequent model, and outputting the dimensionality-reduced data. The dimensionality-reduced eigenvectors are the final fusion feature set and will be used as the input for the subsequent machine learning model.

[0045] A further improvement of the technical solution of the present invention is that the calculation formula of the radial basis function (RBF) kernel function is:

[0046]

[0047] where K jk is the inner product of sample j and sample k in the high-dimensional feature space, z j and z k are the vector representations of the two samples in the original feature space, and σ is the kernel width parameter.

[0048] A further improvement of the technical solution of the present invention is that in step 5, the determination process of the machining damage state is as follows:

[0049] Step 501: Use the dimensionality-reduced eigenvectors as the input, match the material removal mode and damage state labels corresponding to the hard and brittle materials to obtain the dimensionality-reduced feature set for training and testing the LSSVM model, and divide the feature set into a training set and a testing set, with a ratio of 70% for the training set and 30% for the testing set.

[0050] Step 502: Construct a least squares support vector machine (LSSVM) classification model, select the radial basis function (RBF) kernel function and define the loss function.

[0051] Step 503: Use the genetic algorithm (GA) to optimize the hyperparameters (such as the regularization parameter and kernel width) of the least squares support vector machine (LSSVM) classification model. The genetic algorithm generates multiple candidate solutions through selection, crossover, and mutation operations, and evaluates the fitness of each solution. By iterating the genetic algorithm optimization process until the best hyperparameter combination is found, the classification performance of the least squares support vector machine (LSSVM) classification model is improved;

[0052] Step 504: After determining the best hyperparameters, use the training set data to train the least squares support vector machine (LSSVM) classification model. During the training process, the model learns how to map different input features to the corresponding output categories (i.e., plastic removal or brittle removal). After training, use the test set to evaluate the model, and measure the model performance by calculating metrics such as classification accuracy, recall rate, precision rate, and F1-score. At the same time, the confusion matrix can be used to visualize the classification results of the model;

[0053] Step 505: Input the sample data to be recognized into the trained least squares support vector machine (LSSVM) classification model to obtain the recognition result;

[0054] Step 506: Set the expected value of the eigenvalue under normal turning conditions, combine the eigenvalues of the statistical features in the reduced feature set, calculate the turning damage recognition coefficient, and then analyze the damage state of the hard and brittle material;

[0055] Step 507: According to the recognition result and the damage state analysis result, judge the material removal mode of the hard and brittle material during ultra-precision turning, and according to the recognized material removal mode, combine the machining parameters and process conditions to determine the damage state during the machining process.

[0056] A further improvement of the technical solution of the present invention is that the calculation expression of the turning damage recognition coefficient is:

[0057]

[0058] where D is the turning damage recognition coefficient, representing the damage state, g t is the t-th eigenvalue, which is the statistical feature in the reduced feature set, M is the total number of features, B is the expected value of the eigenvalue under normal turning conditions, determined by experiment or historical data analysis, P is the limit value, representing the maximum allowable deviation between the eigenvalue and the reference value, used to adjust the sensitivity of the exponential function, and the value range of D is between 0 and 1. The closer the value is to 1, the more serious the damage is.

[0059] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is:

[0060] 1. The present invention provides a method for ultra-precision turning damage monitoring of hard and brittle materials based on multi-sensor fusion. By integrating data from different sensors, the accuracy and reliability of monitoring are significantly improved. By capturing acoustic characteristics, cutting forces, and mechanical vibration information during the machining process and fusing them, more comprehensive and accurate information can be obtained, thereby more accurately judging the tool wear degree and workpiece damage condition, helping to reduce false alarms and missed detections, and improving the reliability of monitoring.

[0061] 2. The present invention provides a method for ultra-precision turning damage monitoring of hard and brittle materials based on multi-sensor fusion. By real-time monitoring the tool wear and workpiece damage conditions, potential problems can be detected and processed in a timely manner, avoiding workpiece scrapping and increasing machining costs. At the same time, by optimizing machining parameters, machining efficiency can be improved, machining time and energy consumption can be reduced. In addition, the multi-sensor fusion technology can realize autonomous control of the machining process, reduce manual intervention and operation costs, and can significantly reduce machining costs and improve production efficiency. Brief Description of the Drawings

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0063] Figure 1 It is a schematic diagram of the multi-sensor signal processing flow of the present invention;

[0064] Figure 2 It is a schematic diagram of the multi-sensor signal acquisition device of the ultra-precision turning machine tool of the present invention;

[0065] Figure 3 It is a schematic diagram of the material removal mode during the turning process of hard and brittle materials of the present invention;

[0066] Figure 4 It is a comparison diagram of the original force signal and the force signal after MED filtering of the present invention. Detailed Embodiments

[0067] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0068] Embodiment 1, as Figures 1 to 4As shown in the figure, the present invention provides a method for monitoring the damage of ultra-precision turning of brittle materials based on multi-sensor fusion, which includes the following steps:

[0069] Step 1: Build a multi-sensor signal acquisition device on an ultra-precision lathe, including an acoustic emission sensor, a force sensor, and a vibration sensor. The sensors are all set near the material processing area at the tool end to capture different types of sensor signals generated during the turning process in real time. Continuously collect the output signals of each sensor during the turning process, transmit the collected signals to a data acquisition card through pre-processing devices such as current amplifiers and filters for A / D conversion, collect the digital signals, and input them to a host computer (computer) for subsequent processing. Install an acoustic emission sensor, a force sensor, and a vibration sensor near the material processing area at the tool end of the ultra-precision lathe to capture the acoustic characteristics, cutting force, and mechanical vibration information during the processing respectively. Ensure that the sensors are firmly fixed to avoid displacement or vibration during the turning process, which may affect the accuracy of the signals. And install signal pre-processing devices, including a current amplifier and a filter. Among them, install a current amplifier to enhance the weak signals output by the sensors, and install a filter to reduce noise and interference and improve the signal quality;

[0070] Furthermore, the cutting force exerted by the tool on the workpiece during the cutting process is monitored in real time through the force sensor. Its change can reflect the interaction between the tool and the workpiece. When brittle damage appears on the workpiece surface, the cutting force will show abnormal fluctuations. The threshold of the abnormal fluctuations depends on the processing material, tool material, and processing parameters. Generally, through multiple experiments on the cutting force under specific materials and cutting conditions, analyze the fluctuation characteristics during normal fluctuations and when brittle damage appears, so as to establish the abnormal fluctuation threshold. According to previous literature, the differences in toughness and brittleness of brittle materials will affect the amplitude of cutting force fluctuations. At the same time, changes in parameters such as cutting speed, feed rate, and cutting depth during the processing process will also affect the amplitude of cutting force fluctuations. In addition, the wear degree of the tool itself will also affect the amplitude of cutting force fluctuations. Usually, the change in the cutting force during normal cutting is within ±5% - ±15%. Exceeding this indicates problems such as excessive tool wear, workpiece damage, or machine tool problems; Monitor the vibration state of the machine tool and the tool during the cutting process through the vibration sensor. The threshold of mechanical vibration is related to the specific processing material, tool material, and working conditions. The acceleration (m / s 2The acceleration value (or g-value), a commonly used measurement index, represents the vibration amplitude during the cutting process. Under normal machining conditions, the acceleration value usually remains within a certain range (such as 0.1 to 0.5 g). If the peak acceleration detected by the vibration sensor exceeds about 1.0 to 1.5 g, it means an increase in the instability of the cutting process and may result in brittle damage. However, the threshold varies depending on materials, tools, machining speed, and process conditions, and needs to be adjusted according to specific applications and experimental results. In addition, the change in the signal frequency components can effectively identify the characteristics of brittle damage. Brittle damage is often accompanied by an increase in high-frequency components (for example, exceeding 5 kHz). Through spectral analysis, if it is found that the amplitude of certain high-frequency components increases significantly, it can be determined that the instability of the cutting process increases and damage appears on the machined surface. By using an acoustic emission sensor to capture the stress wave signals generated inside the workpiece during the cutting process, when brittle damage occurs on the workpiece surface, the lattice structure inside the material changes, and then stress waves are generated. The acoustic emission sensor can detect this signal, and the signal range is determined according to the actual material and working conditions. The degree of brittle damage is judged by the change in the signal. For materials with low brittle damage, the lattice damage inside is relatively small, limited to the surface layer or micro-cracks, and will not significantly reduce the load-bearing capacity and machining quality of the material. Usually, materials with low brittle damage can be used in less demanding applications, such as some structural parts or parts in low-load environments, but they cannot be applied in high-end application fields, such as infrared optical elements or single-crystal silicon reflectors, because such applications require very good surface quality (mirror level) of the material, and even mild surface damage will seriously affect the surface quality. In contrast, medium and high brittle damage indicates that the machining damage has extended to deeper layers, forming large cracks or even material spalling, and the strength, toughness, and wear resistance of the material have decreased significantly, making it unsuitable for applications in precision or high-strength requirement environments, such as high-load bearing components or precision instrument parts. For materials that have reached high brittle damage, they usually lose their original functionality, and their applicability is severely limited, and they may even be unusable;

[0071] Furthermore, connect the signal line output by the sensor to the data acquisition card, ensure the correct connection and compatibility between the data acquisition card and the host computer (computer), configure the sampling rate and resolution of the data acquisition card to ensure that the digital signal can accurately reflect the physical quantity measured by the sensor, and set the input channels of the data acquisition card corresponding to the signal inputs of each sensor. Start the data acquisition card to continuously collect the output signals of each sensor during the turning process. After the sensor signals are processed by the current amplifier and filter, they are transmitted to the data acquisition card of the data acquisition module for analog-to-digital (A / D) conversion. The collected digital signals are stored on the hard disk of the host computer (computer) to ensure the stability and security during the data transmission process, avoid data loss or damage for subsequent processing and analysis. Install the corresponding signal processing software on the host computer (computer) to receive, store, and preliminarily process the collected digital signals. At the same time, conduct a preliminary check on the data from the aspects of signal integrity, synchronization, and timestamp verification;

[0072] Step 2, perform signal preprocessing operations of minimum entropy deconvolution (MED) filtering on the collected sensor signals (force signals, vibration signals, and acoustic emission signals) to enhance the signal recognizability, weaken the adverse effects, restore the impact components in the filtered signal to the greatest extent, achieve a significant improvement in the signal-to-noise ratio. Load the collected force signal, vibration signal, and acoustic emission signal data from the hard disk of the host computer (computer), and perform preprocessing operations of removing the DC offset, filtering out low-frequency drift and high-frequency noise on the loaded signal data to ensure the accuracy and reliability of the data. Among them, for removing the DC offset, calculate the mean value of each signal and subtract this mean value from the signal to remove the DC offset. For filtering out low-frequency drift and high-frequency noise, use a band-pass filter to filter the signal to filter out low-frequency drift and high-frequency noise. Set the length L of the inverse filter according to the characteristics and requirements of the signal, initialize the parameters of the inverse filter with random values, and set the objective function as the minimum entropy. Then calculate the impact signal generated during the contact process between the tool and the workpiece. Among them, the longer the length of the inverse filter, the more signal features can be captured, but the computational complexity will also increase. And the minimum entropy means that the entropy of the output signal is the smallest, that is, the information in the signal is the most concentrated. Use the normalized kurtosis as the objective function to optimize the inverse filter. Through iterative calculation, adjust the parameters of the inverse filter to maximize the kurtosis, so as to achieve the minimum entropy condition, and apply the optimized inverse filter to the original signal. Then determine the optimal inverse filter matrix, perform time-domain and frequency-domain analysis on the filtered signal. Through time-domain analysis, observe the waveform of the signal and the change of the impact components. Through frequency-domain analysis, observe the spectrum of the signal and the change of the frequency components. And compare the time-domain waveform and frequency-domain spectrogram of the signal before and after filtering to verify whether the impact components are significantly enhanced and whether the signal-to-noise ratio is improved. Compare the time-domain waveform and frequency-domain spectrogram of the original signal and the filtered signal to visually display the filtering effect, which helps to confirm whether the minimum entropy deconvolution filtering successfully extracts the impact components in the signal and weakens the adverse effects;

[0073] Further, the calculation formula for the impact signal generated during the contact process between the tool and the workpiece is expressed as:

[0074]

[0075] where y(n) is the impact signal generated during the contact process between the tool and the workpiece, f(n) is the inverse filter, x(n) is the original collected signal, L is the length of the inverse filter, * is the convolution operation, f in f(l) refers to the convolution / inverse filter set in the minimum entropy deconvolution filtering method, which is equivalent to a convolution kernel, n is the nth sample in the signal time series, and l is the lth coefficient in the convolution kernel;

[0076] Since the larger the kurtosis value of the signal (the fourth-order central moment of data normalization), the smaller its entropy value, the normalized kurtosis is selected as the objective function in the minimum entropy deconvolution (MED) filter, and its expression is as follows:

[0077]

[0078] When in the formula reaches the maximum value, that is, it satisfies the extreme condition then it can be determined that f(l) is the optimal inverse filter. By combining with formula (1), the following equation can be obtained:

[0079]

[0080] where is the normalized kurtosis, y(i) represents the i-th sample in the output impulse signal time series, the x, y, and f functions are the original acquired signal, the output impulse signal, and the convolution / inverse filter (convolution kernel) set in the minimum entropy deconvolution (MED) method respectively, n and l represent the n-th or l-th sample in the signal time series, p represents the p-th coefficient in the convolution kernel, and i, n, l, p represent the i-th, n-th, l-th, and p-th values in the x, y, and f functions respectively;

[0081] Formula (3) can be written in matrix form:

[0082] f = A -1 × b; (4)

[0083] Through iterative calculations using formula (4), the optimal inverse filter matrix is finally determined;

[0084] Step 3: Extract features from the signal after minimum entropy deconvolution filtering. A total of 95 time-domain and frequency-domain statistical features reflecting turning damage are extracted, including time-domain statistical indicators such as peak value, mean value, root mean square value, standard deviation, kurtosis, skewness, etc., and frequency-domain statistical indicators such as spectral centroid, mean square frequency, etc. Import the signal data processed by minimum entropy deconvolution filtering into the signal analysis software to extract turning damage feature values for the filtered signals of different sensors, which are time-domain and frequency-domain statistical features respectively, providing input variables for subsequent pattern recognition or machine learning models. Conduct time-domain analysis on the filtered signals of different sensors, and extract 11 time-domain statistical indicators from the time-domain signals, including peak value, mean value, root mean square value, standard deviation, kurtosis, skewness, peak-to-peak value, peak factor, impulse factor, waveform factor, and margin factor. Among them, the peak value is the maximum value in the signal, reflecting the impact intensity of the signal; the mean value is the average value of the signal, used to evaluate the overall level of the signal; the root mean square value is the square root of the average value of the square of the signal, used to measure the energy level of the signal; the standard deviation is the square root of the average value of the square of the difference between the signal value and its mean value, reflecting the dispersion degree of the signal; the kurtosis is the ratio of the fourth-order central moment of the signal to the square of the variance, used to describe the sharpness of the signal; the skewness is the ratio of the third-order central moment of the signal to the cube root of the variance, used to describe the symmetry of the signal; the peak-to-peak value is calculated as the difference between the maximum value and the minimum value in the signal; the peak factor is the ratio of the peak value to the root mean square value, used to evaluate the impact characteristics of the signal; the impulse factor is the ratio of the peak value to the average value, reflecting the impulse characteristics of the signal; the waveform factor is the ratio of the root mean square value to the average value, describing the waveform characteristics of the signal; the margin factor is the ratio of the peak value to the absolute average value of the signal, used to evaluate the dynamic range of the signal. Conduct fast Fourier transform on the time-domain signal to obtain signal spectrum information, and extract 8 frequency-domain statistical indicators from it, including spectral centroid, mean square frequency, root mean square frequency, frequency variance, spectral peak, power spectral entropy, spectral kurtosis, spectral skewness. Among them, the spectral centroid is the weighted average frequency of the signal spectrum, reflecting the central position of the signal in the frequency domain; the mean square frequency is the weighted average frequency of the square of the signal spectrum, used to measure the frequency distribution characteristics of the signal; the root mean square frequency is the root mean square value of the spectrum; the frequency variance is used to measure the dispersion degree of the spectrum distribution; the spectral peak is used to determine the maximum value in the spectrum and its corresponding frequency; the power spectral entropy is used to measure the complexity of the spectrum; the spectral kurtosis is a statistic used to describe the sharpness of the spectral peak; the spectral skewness is used to measure the asymmetry of the spectrum distribution. Integrate the time-domain and frequency-domain features extracted by each sensor to form a feature vector, and extract a total of 95 statistical features. Among them, the force signal includes forces in three directions: X, Y, and Z;

[0085] Step 4: Use kernel principal component analysis (KPCA) for feature dimensionality reduction and fusion. Through min-max normalization and the radial basis function (RBF) kernel, map the original data to a high-dimensional feature space, calculate the kernel matrix of the feature matrix, perform eigen-decomposition, select the first few principal components whose cumulative contribution rate reaches a certain threshold, project the original data onto the selected principal components, and then obtain the data after dimensionality reduction. Organize the 95 extracted statistical features into a feature matrix, where each row represents a sample and each column represents a feature. Then perform Min-Max normalization on the feature matrix to scale the value of each feature to the interval [0, 1], eliminate the influence of the dimension between different features, and make the data at the same order of magnitude. Then perform kernel function calculation. Select the radial basis function (RBF) as the kernel function. The RBF kernel is also called the Gaussian kernel and is suitable for mapping non-linear data. Use the selected radial basis function (RBF) kernel function to calculate the kernel function values between each sample point in the original dataset to form a kernel matrix. Each element in the kernel matrix is the kernel function value between two sample points. Subtract the mean of the corresponding row and column from each element in the kernel matrix to achieve the centering process of the kernel matrix, ensure the correct mapping of the data in the high-dimensional space, eliminate the offset in the data, and make the subsequent eigen-decomposition more accurate. Perform eigen-value decomposition on the centered kernel matrix to obtain the eigen-values and the corresponding eigen-vectors. Among them, the eigen-value represents the amount of information contained in each principal component, and the eigen-vector represents the direction of the principal component. According to the size of the eigen-values, select multiple (the first few) principal components whose cumulative contribution rate reaches a certain threshold (such as 95% or 99%). These principal components can retain the information of the original data to the greatest extent. Project the original data onto the selected principal components. By multiplying the kernel matrix by the eigen-vectors of the selected principal components, obtain the data after dimensionality reduction. The data after dimensionality reduction not only retains the main variation information of the original data but also greatly reduces the dimension, thus effectively reducing redundant information and improving the calculation efficiency of the subsequent model. Then output the data after dimensionality reduction. The eigen-vector after dimensionality reduction is the final fused feature set and will be used as the input for the subsequent machine learning model;

[0086] Further, the calculation formula of the radial basis function (RBF) kernel function is:

[0087]

[0088] Where K jk is the inner product of sample j and sample k in the high-dimensional feature space, z j and z k are the vector representations of the two samples in the original feature space, and σ is the kernel width parameter;

[0089] Step 5: Use the least squares support vector machine (LSSVM) method optimized by genetic algorithm for pattern recognition to identify the material removal pattern of brittle materials during ultra-precision turning and determine the machining damage state.

[0090] Example 2: As Figures 1 to 4 shown, based on Example 1, the present invention provides a technical solution: Preferably, in Step 5, the determination process of the machining damage state is as follows:

[0091] Take the dimension-reduced feature vector as the input, match the material removal pattern and damage state label corresponding to the brittle material to obtain the dimension-reduced feature set, which is used to train and test the LSSVM model, and divide the feature set into a training set and a test set, with a ratio of 70% for the training set and 30% for the test set;

[0092] Furthermore, the removal patterns of brittle materials in turning are divided into two categories: brittle removal and plastic removal. Plastic removal specifically refers to the phenomenon that during ultra-precision turning operations, the material undergoes plastic deformation under the cutting action of the tool and then forms continuous chips instead of fractures. This process involves the rearrangement of the internal atomic or molecular structure of the material and usually occurs below the yield strength threshold of the material. In contrast, the brittle removal pattern describes the way in which brittle materials are removed by the initiation, propagation, and final fragmentation of cracks under the action of the tool. When the cutting depth of the tool is large or the applied cutting force exceeds a certain limit, brittle materials tend to undergo brittle fracture, resulting in the generation of a large number of fragments and the formation of cracks. In this removal pattern, the machined surface often exhibits rough features, accompanied by significant cracks and fragmentation marks, thus seriously affecting the surface quality. In view of this, by using advanced signal monitoring technology to accurately determine or predict the material removal pattern of brittle materials during ultra-precision machining, it is possible to effectively evaluate whether there is damage on the machined surface of the material;

[0093] In addition, the damage states are divided into: no damage / near no damage, low-degree damage, medium-degree damage, and high-degree damage. No damage / near no damage: The machined surface of the material is completely removed by plastic means, without surface or subsurface cracks; Low-degree damage: There are few microscopic damages on the material surface, the removal pattern is mainly plastic removal, accompanied by a small number of microcracks; Medium-degree damage: There are obvious cracks and small pits on the surface, the material removal pattern is a mixed removal of brittle and plastic, with local brittle fractures; High-degree damage: There are obvious deep cracks, spalling, or severe microscopic structure damage on the material surface, and the removal pattern is mainly brittle fracture;

[0094] Construct a least squares support vector machine (LSSVM) classification model, select the radial basis function (RBF) kernel function and define the loss function. The LSSVM classification model learns the classification boundary of the data by solving the least squares optimization problem with a regularization term. By minimizing the squared loss function, the problem is transformed into solving a system of linear equations, significantly improving the computational efficiency. Use the genetic algorithm (GA) to optimize the hyperparameters (such as the regularization parameter and kernel width) of the least squares support vector machine (LSSVM) classification model. The genetic algorithm generates multiple candidate solutions through selection, crossover, and mutation operations, and evaluates the fitness of each solution (based on the classification accuracy obtained from cross-validation). The global search mechanism of the genetic algorithm effectively avoids the deficiencies of manual adjustment and reduces the risk of local optimal solutions. The self-adaptability of the genetic algorithm enables it to flexibly adjust parameters on different datasets, thereby enhancing the adaptability and generalization ability of the model. Through iterative genetic algorithm optimization processes until the best hyperparameter combination is found, the classification performance of the least squares support vector machine (LSSVM) classification model is improved. After determining the best hyperparameters, use the training set data to train the least squares support vector machine (LSSVM) classification model. During the training process, the model learns how to map different input features to the corresponding output categories (i.e., plastic removal or brittle removal). After training, use the test set to evaluate the model. Measure the model performance by calculating metrics such as classification accuracy, recall, precision, and F1-score. At the same time, the confusion matrix can be used to visualize the classification results of the model. Input the sample data to be recognized into the trained least squares support vector machine (LSSVM) classification model to obtain the recognition result. Set the expected value of the eigenvalue under normal turning conditions, combine the eigenvalues of the statistical features in the reduced feature set, calculate the turning damage recognition coefficient, and then analyze the damage state of the hard and brittle material. According to the recognition result and the analysis result of the damage state, judge the material removal mode of the hard and brittle material during ultra-precision turning, and based on the recognized material removal mode, combine the machining parameters and process conditions to determine the damage state during the machining process;

[0095] Furthermore, the calculation expression of the turning damage recognition coefficient is:

[0096]

[0097] where D is the turning damage recognition coefficient, representing the damage state, g t is the t-th eigenvalue, which is the statistical feature in the reduced feature set, M is the total number of features, B is the expected value of the eigenvalue under normal turning conditions, determined by experiment or historical data analysis, P is the limit value, representing the maximum allowable deviation between the eigenvalue and the reference value, used to adjust the sensitivity of the exponential function. The value range of D is between 0 and 1, and the value closer to 1 indicates more serious damage. Represents the sum of the square roots of all eigenvalues, which is used to capture the overall trend of the eigenvalues. Is a reciprocal function, combined with the natural logarithm, which is used to adjust the influence of the summation result and make the formula more robust to outliers. Is a Sigmoid function, which is used to map the deviation between the eigenvalue and the reference value to between 0 and 1. The closer the value is to 1, the greater the deviation from the reference value. When Is close to B, D is close to 0, indicating no damage or very low damage level. When Is significantly greater than B, D is close to 1, indicating a very high damage level.

[0098] As described above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims.

Claims

1. A damage monitoring method for ultra-precision turning of hard and brittle materials based on multi-sensor fusion, characterized in that: The following steps are involved: Step 1: Build a multi-sensor signal acquisition device on the ultra-precision turning machine tool to capture different types of sensor signals generated during the turning process in real time; Step 2, performing a signal preprocessing operation of minimum entropy deconvolution filtering on the collected sensor signal; Step 3, feature extraction is performed on the signal after minimum entropy deconvolution filtering, and a total of 95 time domain and frequency domain statistical features reflecting turning damage are extracted; Step 4: Use kernel principal component analysis to perform feature dimensionality reduction and fusion, map the original data to a high-dimensional feature space, calculate the kernel matrix of the feature matrix, perform feature decomposition, and then obtain the reduced-dimensional data; In step 5, the least squares support vector machine method optimized by genetic algorithm is used for pattern recognition to identify the material removal pattern of hard and brittle materials during ultra-precision turning and determine the machining damage state.

2. The method for monitoring damage of ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 1 is characterized in that: In step 1, the process of capturing the sensor signal during turning is as follows: Step 101, installing an acoustic emission sensor, a force sensor, and a vibration sensor at the tool end of the ultra-precision turning machine tool near the material processing area to capture the acoustic characteristics, cutting force, and mechanical vibration information during the processing, and installing a signal preprocessing device, including a current amplifier and a filter; Step 102, connect the signal line output by the sensor to the data acquisition card, make the data acquisition card correctly connected and compatible with the host computer, configure the sampling rate and resolution of the data acquisition card, and set the input channel of the data acquisition card to correspond to the signal input of each sensor; Step 103, start the data acquisition card to continuously collect the output signals of each sensor during the turning process, and after the sensor signals are processed by the current amplifier and the filter, they are transmitted to the data acquisition card of the data acquisition module for analog-to-digital conversion; Step 104, the collected digital signal is stored on the host computer hard disk, and the corresponding signal processing software is installed on the host computer to receive, store and preliminarily process the collected digital signal. At the same time, the data is preliminarily checked from the perspective of signal integrity, synchronization and timestamp verification.

3. The method for monitoring damage of ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 2 is characterized in that: In step 2, the signal preprocessing process is: Step 201, loading the collected force signal, vibration signal and acoustic emission signal data from the host computer hard disk, and performing preprocessing operations on the loaded signal data to remove DC offset, filter low-frequency drift and high-frequency noise; Step 202, setting the length L of the inverse filter according to the characteristics and requirements of the signal, initializing the parameters of the inverse filter using random values, and setting the objective function to minimum entropy, thereby calculating the impact signal generated by the contact process between the tool and the workpiece; Step 203, using normalized kurtosis as the objective function to optimize the inverse filter, adjusting the parameters of the inverse filter through iterative calculation to maximize the kurtosis, thereby achieving the minimum entropy condition, and applying the optimized inverse filter to the original signal to determine the optimal inverse filter matrix; Step 204, perform time domain and frequency domain analysis on the filtered signal, observe the changes in the waveform and impulse components of the signal through time domain analysis, observe the changes in the spectrum and frequency components of the signal through frequency domain analysis, and compare the time domain waveform and frequency domain spectrogram of the signal before and after filtering to verify whether the impulse components are significantly enhanced and whether the signal-to-noise ratio is improved.

4. The method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 3 is characterized in that: The calculation formula of the impact signal generated by the contact process between the tool and the workpiece is expressed as: Among them, y(n) is the impact signal generated by the contact process between the tool and the workpiece, f(n) is the inverse filter, x(n) is the original acquisition signal, L is the length of the inverse filter, * is the convolution operation, the f in f(l) refers to the convolution / inverse filter set in the minimum entropy deconvolution filtering method, which is equivalent to a convolution kernel, n is the nth sample in the signal time series, and l is the lth coefficient in the convolution kernel; Since the larger the kurtosis value of the signal, the smaller its entropy value, the normalized kurtosis is selected as the objective function in the minimum entropy deconvolution filter, and its expression is: When the formula When the maximum value is reached, the extreme value condition is satisfied. Then it can be determined that f(l) is the best inverse filter, and the following equation can be obtained by combining formula (1): in, is the normalized kurtosis, y(i) represents the i-th sample in the output impulse signal time series, x, y, and f functions are the original acquisition signal, the output impulse signal, and the convolution / inverse filter set in the minimum entropy deconvolution filtering method, n and l represent the n-th or l-th sample in the signal time series, p represents the p-th coefficient in the convolution kernel, and i, n, l, and p represent the i-th, n-th, l-th, and p-th values ​​in the x, y, and f functions, respectively; Formula (3) can be written in matrix form: f=A -1 ×b; (4) Formula (4) is used to perform iterative calculations and finally determine the optimal inverse filter matrix.

5. The method for monitoring damage of ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 4 is characterized in that: In step 3, the extraction process of the statistical characteristics of turning damage in time domain and frequency domain is as follows: Step 301, importing the signal data processed by minimum entropy deconvolution filtering into the signal analysis software, extracting the turning damage characteristic values ​​of the filtered signals of different sensors, which are respectively the time domain and frequency domain statistical characteristics; Step 302, performing time domain analysis on the filtered signals of different sensors, and extracting 11 time domain statistical indicators including peak value, mean value, root mean square value, standard deviation, kurtosis, skewness, peak-to-peak value, peak factor, pulse factor, waveform factor, and margin factor from the time domain signals; Step 303, performing fast Fourier transform on the time domain signal to obtain signal spectrum information, from which eight frequency domain statistical indicators including spectrum center of gravity, mean square frequency, root mean square frequency, frequency variance, spectrum peak, power spectrum entropy, spectrum kurtosis, and spectrum skewness are extracted; Step 304 , integrating the time domain and frequency domain features extracted by each sensor to form a feature vector, extracting a total of 95 statistical features, wherein the force signal includes forces in three directions: X, Y, and Z.

6. The method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 5 is characterized in that: In step 4, the process of obtaining the data after dimensionality reduction is as follows: Step 401, organize the extracted 95 statistical features into a feature matrix, where each row represents a sample and each column represents a feature, and perform Min-Max normalization on the feature matrix so that the value of each feature is scaled to the interval [0,1], eliminating the impact of the dimensions between different features, so that the data is at the same order of magnitude, and then perform kernel function calculation; Step 402, selecting a radial basis function as a kernel function, using the selected radial basis function, calculating the kernel function value between each sample point in the original data set to form a kernel matrix, wherein each element of the kernel matrix is ​​a kernel function value between two sample points; Step 403, subtract the mean of the corresponding row and the corresponding column from each element in the kernel matrix to achieve centralization processing of the kernel matrix; Step 404, performing eigenvalue decomposition on the centralized kernel matrix to obtain eigenvalues ​​and corresponding eigenvectors, and selecting a plurality of principal components whose cumulative contribution rates reach corresponding thresholds according to the magnitude of the eigenvalues; Step 405, projecting the original data onto the selected principal component, obtaining the reduced-dimensional data by multiplying the kernel matrix with the eigenvector of the selected principal component, and outputting the reduced-dimensional data. The reduced-dimensional eigenvector is the final fusion feature set.

7. The method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 6 is characterized in that: The calculation formula of the radial basis function kernel function is: Among them, K jk is the inner product of sample j and sample k in the high-order feature space, z j and z k is the vector representation of the two samples in the original feature space, and σ is the kernel width parameter.

8. The method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 7 is characterized in that: In step 5, the process of determining the machining damage state is as follows: Step 501, using the reduced-dimensional feature vector as input, matching the material removal mode and damage state label corresponding to the hard and brittle material, obtaining the reduced-dimensional feature set, and dividing the feature set into a training set and a test set, wherein the removal mode of the hard and brittle material in turning processing is divided into two categories: brittle removal and plastic removal, and the damage state is divided into: no damage / near no damage, low degree damage, medium degree damage and high degree damage; Step 502, constructing a least squares support vector machine classification model, selecting a radial basis function and defining a loss function; Step 503, using a genetic algorithm to optimize the hyperparameters of the least squares support vector machine classification model, the genetic algorithm generates multiple candidate solutions through selection, crossover and mutation operations, and evaluates the fitness of each solution, and iterates the genetic algorithm optimization process until the best hyperparameter combination is found; Step 504, after determining the optimal hyperparameters, the least squares support vector machine classification model is trained using the training set data, and after the training is completed, the model is evaluated using the test set; Step 505, inputting the sample data to be identified into the trained least squares support vector machine classification model to obtain the identification result; Step 506, setting the expected value of the characteristic value under normal turning conditions, combining the characteristic value of the statistical feature in the feature set after dimensionality reduction, calculating the turning damage identification coefficient, and then analyzing the damage state of the hard and brittle material; Step 507, judging the material removal pattern of the hard and brittle material during the ultra-precision turning process based on the identification results and the damage state analysis results, and judging the damage state during the machining process based on the identified material removal pattern in combination with the machining parameters and process conditions.

9. The method for monitoring damage in ultra-precision turning of hard and brittle materials based on multi-sensor fusion according to claim 8 is characterized in that: The calculation expression of the turning damage identification coefficient is: Where D is the turning damage identification coefficient, which indicates the damage state, g t is the tth eigenvalue, which is the statistical feature in the feature set after dimensionality reduction. M is the total number of features. B is the expected value of the eigenvalue under normal turning conditions. P is the limit value, which represents the maximum allowable deviation between the eigenvalue and the reference value. The value range of D is between 0 and 1. The closer the value is to 1, the more serious the damage.

Images