Real-time monitoring and early warning method and system for service life of cutter

Abstract

The invention relates to the technical field of machining tools, and provides a tool service life real-time monitoring and early warning method and system.The method comprises the steps that firstly, original acoustic emission signals of a tool working site are obtained, self-adaptive cavitation noise cancellation processing is carried out, and comparison index calculation including entropy change calculation and kurtosis calculation is carried out based on the processed signals; and performing frequency dispersion-attenuation coupling analysis by combining short-time Fourier transform, and extracting a frequency dispersion slope change rate and an energy attenuation ratio. And then, obtaining a current cutter personalized threshold vector identified by the deep learning model from the cloud, comparing the current cutter personalized threshold vector with a comparison index, determining a life remaining ratio, and performing early warning. Through millisecond-level acoustic emission signal analysis, microscopic secondary detection and cloud model self-learning, on the premise of no shutdown and unattended operation, millisecond-level response, extremely low false alarm rate and system seamless integration are achieved, a traditional fixed tool changing period is broken through, and the service life of the tool approaches the real service life limit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machining tool, in particular to a tool life real-time monitoring and early warning method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute the prior art.

[0003] Tool life real-time monitoring is the core link of guaranteeing product quality and equipment efficiency in intelligent manufacturing. With the wide application of high-end processes such as high-speed cutting, five-axis linkage and automatic tool changing in the aviation, automobile and energy industries, the evolution process of tool wear and sudden collapse of blade has been shortened to milliseconds, which puts forward multiple challenges such as rapid response, accurate identification and system integration for the monitoring system. The traditional "fixed tool changing" or "manual inspection" strategy not only cannot adapt to complex machining conditions, but also leads to a large number of tools being scrapped in advance, resulting in serious resource waste. Therefore, developing real-time monitoring technology with high response speed, low false alarm rate and the ability to cooperate with production line system has become the key to improving tool utilization and manufacturing efficiency.

[0004] At present, there are many monitoring methods, but there are three major problems of response delay, high false alarm rate and system isolation. First, the response delay is mainly due to the system relying on offline modeling or low-frequency sampling signals. For example, some schemes use current or torque signals combined with wavelet decomposition and neural networks for wear prediction, which rely on servo system control parameters, have long response path, and need to collect a large number of samples for model updating, making it difficult to achieve millisecond-level response; visual and infrared detection schemes have air lag or image processing delay, and are also difficult to deal with sudden blade collapse in high-speed cutting scenarios. Second, the high false alarm rate is due to the fact that the monitoring signal is too single and has poor anti-interference ability. Taking acoustic emission signals as an example, in high-speed milling, the signal-to-noise ratio is less than 12dB due to the influence of cooling liquid cavitation, electromagnetic interference, etc., and the effective signal loss rate is more than 45%; image recognition schemes are easily disturbed by chip shielding, thermal drift, etc., resulting in a sharp fluctuation in recognition accuracy, poor model generalization ability, and actual effect far lower than laboratory performance. Moreover, the traditional fixed tool changing strategy easily causes the tool to be scrapped before reaching the life limit, resulting in resource waste and increasing production cost. SUMMARY

[0005] To solve the above problems, the present application provides a tool life real-time monitoring and early warning method and system, which uses millisecond-level acoustic emission signal analysis, microscopic secondary detection and cloud model self-learning to achieve millisecond-level response, low false alarm rate and seamless system integration without stopping and unattended, and optimizes the traditional fixed tool changing period, so that the tool life approaches its true service life, providing a solution for tool cutting force control and tool life monitoring.

[0006] To achieve the above object, the present application adopts the following technical solutions: The first aspect of the present application provides a tool life real-time monitoring and early warning method, comprising the following steps: The original acoustic emission signal of the tool working site is acquired for pretreatment, and the pretreated signal is subjected to adaptive cavitation noise cancellation processing; Comparison index calculation, based on the acoustic emission signal after adaptive cavitation noise cancellation processing, band-pass filtering and kurtosis statistical analysis are performed to calculate entropy change And kurtosis ; frequency dispersion-attenuation coupling analysis is performed in combination with short-time Fourier transform to calculate the slope change rate of the modal dispersion curve And energy attenuation ratio ΔE / Δf; The personalized threshold vector of the current tool learned and identified based on the deep learning model is acquired, compared with the calculated comparison indexes, and the remaining proportion of the life is determined to perform early warning; The deep learning model acquires the machining data of the current tool in real time, identifies the individual feature offset of the tool individual relative to the standard tool, adds the set universal reference threshold to the predicted individual feature offset as the personalized threshold vector of the current tool.

[0007] The second aspect of the present application is a tool life real-time monitoring and early warning system, comprising: An acoustic emission sensor and a control terminal, wherein the control terminal is configured to perform the steps of the tool life real-time monitoring and early warning method described above.

[0008] The third aspect of the present application is a tool life real-time monitoring and early warning system, comprising: A noise processing module configured to acquire the original acoustic emission signal of the tool working site for pretreatment, and to perform adaptive cavitation noise cancellation processing on the pretreated signal; A comparison index calculation module configured to calculate the comparison index, based on the acoustic emission signal after adaptive cavitation noise cancellation processing, to perform band-pass filtering and kurtosis statistical analysis, and to calculate entropy change And kurtosis ; frequency dispersion-attenuation coupling analysis is performed in combination with short-time Fourier transform to calculate the slope change rate of the modal dispersion curve And energy attenuation ratio ΔE / Δf; A threshold updating and early warning module configured to acquire the personalized threshold vector of the current tool learned and identified based on the deep learning model, to compare with the calculated comparison indexes, and to determine the remaining proportion of the life to perform early warning; The deep learning model obtains the machining data of the current tool in real time, identifies the individual characteristic offset of the tool individual relative to the standard tool, adds the set general reference threshold value to the predicted individual characteristic offset as the individualized threshold value vector of the current tool.

[0009] Compared with the prior art, the present application has the following beneficial effects: The monitoring and early warning method of the present application can significantly reduce the response delay of the traditional monitoring method in the high-speed machining scene, shorten the whole process time of signal acquisition, feature calculation and threshold comparison, and meet the millisecond-level real-time requirement. The adaptive cavitation noise cancellation improves the signal-to-noise ratio, reduces the signal loss rate, and effectively improves the system anti-interference capability. The combination of the dispersion-attenuation coupling feature and the double-threshold discrimination algorithm reduces the initial wear false alarm rate. The deep learning model identifies the tool individual offset, dynamically corrects the comparison threshold, has the tool individual adaptation capability, effectively suppresses the false alarm and missed alarm problems caused by manufacturing tolerances and different working conditions, and greatly improves the monitoring accuracy and system adaptability.

[0010] The advantages of the present application and the advantages of the additional aspects will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0011] The drawings accompanying the specification of this application form a part thereof, serve to further provide a further understanding of the present application, and together with the description of the illustrative embodiments thereof, explain the present application, but do not limit the present application.

[0012] Figure 1 is a flow chart of a tool life real-time monitoring and early warning method of embodiment 1 of the present application; Figure 2 is a flow chart of the deep learning model construction and threshold identification of embodiment 1 of the present application; Figure 3 is a flow chart of the acoustic emission signal preprocessing of embodiment 1 of the present application; Figure 4 is a tool life real-time monitoring and early warning system interface diagram of embodiment 3 of the present application. DETAILED DESCRIPTION

[0013] The present application will be further described below in conjunction with the drawings and embodiments.

[0014] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as generally understood by those skilled in the art to which the present application belongs.

[0015] It should be noted that the terminology used herein is for describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features within those embodiments can be combined with each other. The embodiments will now be described in detail with reference to the accompanying drawings.

[0016] Example 1 In one or more of the technical solutions disclosed in the embodiments, such as Figures 1 to 4 As shown, a method for real-time monitoring and early warning of tool life includes the following steps: Step 1: Obtain the original acoustic emission signal from the tool's working environment and preprocess it. Then, perform adaptive cavitation noise cancellation on the preprocessed signal. Step 2: Calculate the comparison index. Based on the acoustic emission signal after adaptive cavitation noise cancellation processing, perform bandpass filtering and kurtosis statistical analysis, and calculate the entropy change. with kurtosis Combining short-time Fourier transform (STFT) with dispersion-attenuation coupling analysis, the rate of change of the slope of the modal dispersion curve is calculated. And the energy decay ratio ΔE / Δf; Step 3: Obtain the personalized threshold vector of the current tool after learning and recognition based on the deep learning model, compare it with the calculated comparison indicators, determine the remaining life ratio and issue an early warning. The deep learning model acquires the machining data of the current tool in real time, identifies the individual feature offset of the tool relative to the standard tool, and adds the set general benchmark threshold to the predicted individual feature offset as the personalized threshold vector of the current tool. Optionally, the deep learning model can be deployed in the cloud or on an edge device; preferably, it can be deployed in the cloud. In this embodiment, an adaptive cavitation noise cancellation unit is integrated into the acoustic emission signal acquisition link to suppress coolant cavitation noise and effectively improve the signal-to-noise ratio. Wear-related entropy changes are extracted based on bandpass filtering and kurtosis analysis. with kurtosis The metrics, combined with Short-Time Fourier Transform (STFT) for dispersion-attenuation coupling analysis, extract the slope of the dispersion curve and the energy attenuation ratio ΔE / Δf of the Lamb wave mode to dynamically characterize the tool wear evolution process. Subsequently, a cloud-based deep learning model continuously receives the machining history data of each tool, identifies the individual feature offset of each tool relative to the standard model, and generates personalized threshold vectors based on a general benchmark threshold, enabling accurate comparison and real-time early warning of the life status of each tool.

[0017] This implementation significantly reduces the response latency of traditional monitoring methods in high-speed machining scenarios, achieving a total time of less than 5ms for signal acquisition, feature calculation, and threshold comparison, meeting millisecond-level real-time requirements. The adaptive cavitation noise cancellation module improves the signal-to-noise ratio by more than 18dB, reducing signal loss rate and effectively enhancing interference resistance. The combination of dispersion-attenuation coupling characteristics and a dual-threshold discrimination algorithm reduces the false alarm rate of initial wear. By identifying individual tool offsets through a deep learning model and dynamically correcting the comparison threshold, the system possesses individual adaptability, effectively suppressing false alarms and missed alarms caused by manufacturing tolerances and different working conditions, greatly improving monitoring accuracy and system adaptability.

[0018] Further technical solutions also include step 4: when the remaining lifespan ratio is lower than the set value, a tool replacement operation is performed, the replaced tool is microscopically identified to determine whether the tool can be reused, and the comparison index calculated by the tool is used as the input value, and the result of microscopic identification is used as the output to construct new training data for rolling training of the deep learning model in the cloud. The current tool processing data includes: comparison indicators calculated from acoustic emission signals collected during tool operation, identification results of whether the replaced tool is reusable obtained by microscopic identification, and the remaining life ratio of reusable tools. Optionally, the training interval can be set, such as one day, two days or longer, to construct a new training set from the tool operation data within the corresponding time period, perform real-time training, and update the personalized threshold vector of the current tool after training.

[0019] In step 1, acoustic emission signals can be acquired using an acoustic emission sensor; the acoustic emission sensor is installed within ±15° of the machine tool spindle axis and 60±5mm away from the cutting edge; In step 1, such as Figure 3 The diagram shows the preprocessing steps for acoustic emission signals, including the following: Step 11: Acquire the original acoustic emission signal and perform two-stage processing: charge amplification and voltage amplification; Step 12: Extract the effective cutting signal segment of the tool based on the root mean square value and signal-to-noise ratio of the window signal; Specifically, the acoustic emission signal is intercepted in real time by setting a sliding window with a size of 100 milliseconds, and the root mean square (RMS) and signal-to-noise ratio (SNR) of each window are calculated every 10 milliseconds. The RMS is used to determine whether the signal energy reaches the characteristic threshold of the cutting state, and the spindle speed and feed rate of the machine tool are used for verification. The SNR is used to filter out low-quality signal segments to ensure that the extracted data not only reflects the cutting state but also has sufficient signal quality. Step 13, correspond the acquired acoustic emission signal to the corresponding tool ID to determine the acoustic emission signal of the current tool; Specifically, during initialization or each tool change operation, a unique tool ID (such as T100, T101, etc.) is assigned to each physical tool, and the ID is recorded in the database together with the tool's model, geometric parameters, material information, and other metadata to complete the registration and binding of the tool identity. In the actual machining process, real-time data interaction is performed with the numerical control machine tool control system through the OPC UA interface to dynamically read the activated tool number (ID) on the current spindle; Step 11 is implemented through an amplifier, including a charge amplifier, a voltage amplifier, and an output amplifier. Optionally, the first-stage charge amplifier has a gain of 40 dB, the second-stage voltage amplifier has a gain of 20 dB, and the total gain is 60 dB to avoid oscillation caused by excessively high single-stage gain. The output uses a high-speed video amplifier, which can be AD8397; In the above embodiments, the introduction of two-stage front-end charge amplification and voltage amplification can effectively enhance the quality of the acoustic emission signal, making the weak tool wear signal clearly presented and improving the overall sensitivity and accuracy of the system. The background noise removal and signal segment extraction process can reduce the system misidentification rate and ensure that the extracted data is highly targeted and effective. The main frequency calculation and tool signal mapping mechanism ensures the accurate positioning of the monitoring object, providing support for multi-tool parallel monitoring and improving the monitoring accuracy and reliability of the system as a whole.

[0020] In step 1, the preprocessed acoustic emission signal is subjected to adaptive cavitation noise cancellation processing through the constructed adaptive cavitation noise cancellation unit. The cavitation background noise collected by the reference microphone is used as input, and adaptive filtering methods such as feature basis construction and Hilbert transform are used to generate a cancellation signal with phase inversion characteristics to suppress the cavitation component in the original acoustic emission signal and improve the signal-to-noise ratio; Specifically, the method for adaptive cavitation noise cancellation processing of the preprocessed signal includes the following steps: Step 101, acquire the preprocessed acoustic emission signal and the reference signal r(n) for synchronous acquisition and frequency band filtering to obtain time-domain data with phase alignment and frequency range limited to a set frequency range; Optionally, the frequency range set by the frequency band limitation is 220-320 kHz; Specifically, the reference signal r(n) represents a cavitation noise signal collected by a reference microphone; the reference microphone is placed near a cooling liquid pipeline or outside a machine tool shell, and is used to detect background noise generated by cavitation of the cooling liquid; Optionally, a double-channel synchronous acquisition card can be used to perform hardware-triggered alignment and synchronous sampling on the acoustic emission signal and the reference signal r(n), so as to ensure that the phase error is less than 0.5 µs; for example, the double-channel synchronous acquisition card can be a double-channel synchronous acquisition card with a sampling rate of 1 MHz and a resolution of 16 bits. Step 102, feature base construction: performing spectral analysis and spectral kurtosis discrimination on the reference signal, extracting a set number of main frequency components, and constructing an orthogonal feature base vector group to realize dynamic tracking of the cavitation dominant frequency band; The process of constructing the orthogonal feature base vector group is as follows: Step 1021, performing fast Fourier transform on the reference signal r(n); Specifically, 512-point fast Fourier transform (FFT) is performed; Step 1022, calculating spectral kurtosis to locate the frequency band with the most concentrated cavitation noise energy; Step 1023, extracting a set number of spectral lines with the most energy in the most concentrated frequency band at a set step u1; Specifically, 5 spectral lines with the most energy are extracted at a step u1 of 4 kHz; Step 1024, performing Gram-Schmidt orthogonalization processing on the extracted frequency components corresponding to the spectral lines to generate a feature base vector group; Specifically, when 5 spectral lines are extracted, the 5-dimensional feature base vector group φ=[φ1,φ2,...,φ5] is generated by Gram-Schmidt orthogonalization of the frequency components corresponding to the extracted spectral lines, and the feature base vector group can be updated every 0.512 ms.

[0021] Step 103, performing feature base projection and Hilbert transform on the reference signal to generate an anti-phase time-domain signal opposite in phase to the original cavitation noise, specifically: Step 1031, projecting the reference signal r(n) onto the feature base φ to obtain projection coefficients c(m); Step 1032, performing Hilbert transform on each base vector φ m Performing Hilbert transform to obtain the corresponding complex analytic form; Step 1033, combining the complex analytic forms after Hilbert transform of the corresponding base vectors based on the projection coefficients c(m) to obtain a cavitation noise complex envelope signal , to obtain the instantaneous amplitude A(n) and the instantaneous phase Θ(n) of the cavitation noise; Step 1034, constructing the inverse signal of the cavitation noise complex envelope signal , and generating the inverse cancellation signal through the characteristic base inverse mapping , that is, the cavitation noise inverse time domain signal; Step 104, subtracting the cavitation noise inverse time domain signal from the preprocessed acoustic emission signal collected synchronously to obtain the acoustic emission signal after adaptive cavitation noise cancellation processing; The above steps break through the limitation of traditional filters only for linear superposition, realize the waveform level reconstruction and inverse of non-stationary cavitation noise, and realize the adaptive cancellation of the cooling liquid cavitation noise through the construction of dynamic cavitation frequency tracking and characteristic inverse projection mechanism. Compared with the traditional static filter or fixed reference signal cancellation method, the method has high adaptability to frequency drift and amplitude fluctuation, and the signal-to-noise ratio is improved by more than 18dB in the key noise segment of 200-400kHz, effectively reduces the signal loss rate, improves the system anti-interference ability and wear characteristic extraction precision, and provides a more stable and reliable signal basis for subsequent life judgment.

[0022] In step 2, based on the acoustic emission signal after adaptive cavitation noise cancellation processing, band-pass filtering and characteristic statistical analysis are performed to calculate the entropy change and the kurtosis The process includes the following steps: Step 21, applying band-pass filtering to the acoustic emission signal after adaptive cavitation noise cancellation processing; a band-pass filter of 50-400kHz can be used; Step 22, setting a sliding window, and in each sliding window, calculating the kurtosis to measure the sharpness of the signal; and calculating the entropy change to measure the complexity of the signal, and the higher the entropy, the more dispersed the energy is; The kurtosis The calculation formula is: ; Wherein, is the number of sampling points in the sliding window, is the signal amplitude of the i-th sampling point in the window, is the mean value of the signal in the window, is the standard deviation of the signal in the window; The kurtosis value significantly increases (for example, >5.0) indicates that there are a large number of sudden impacts in the signal, which usually corresponds to the collapse of the tool, micro-cracks and other sharp damage. The kurtosis The significant increase in value can be achieved by setting a threshold value, such as setting the threshold value to be no less than 5.0, and when greater than the set threshold value, it is judged to be a significant increase; The calculation formula of the entropy change is: ; ; Wherein, the wavelet packet energy probability distribution represents the proportion of the energy of the kth wavelet packet node (i.e. frequency band) in the jth layer decomposition in the total energy of the layer; the wavelet packet entropy represents the Shannon entropy calculated on the jth layer decomposition; the reference entropy represents the jth layer wavelet packet entropy calculated in a reference state or normal state; The continuous rise indicates that the signal energy becomes dispersed from concentrated, reflecting the progressive wear of the tool. When exceeds the wavelet packet entropy threshold value, it indicates that the tool enters the abnormal wear stage, and the wavelet packet entropy threshold value can be set to be greater than 40%; in this embodiment, 40% can be used as a general reference threshold value of the wavelet packet entropy threshold value; Optionally, the sliding window is set to 100ms, and a group of is obtained every 100ms, which is used to identify the cutting state change or abnormal vibration event.

[0023] In step 2, based on the acoustic emission signal after adaptive cavitation noise cancellation processing, the frequency dispersion-attenuation coupling analysis is carried out combined with short-time Fourier transform (STFT), and the slope change rate of the modal dispersion curve and the wavelet packet energy attenuation rate ratio are calculated. The process includes the following steps: Step 201, performing short-time Fourier transform (STFT) on the acoustic emission signal to obtain a frequency-time distributed spectrum curve; Specifically, the key analysis parameters of short-time Fourier transform (STFT) are set as: Hanning window (length 256 points), and overlap rate 75%. This configuration can provide a frequency resolution of about 1.95kHz under a sampling rate of 2MHz, which is sufficient to capture the subtle spectral structure changes caused by tool wear.

[0024] Step 202, extracting the S0 / A0 modal dispersion curve of the Lamb wave in the spectrum curve, and calculating the slope change rate of the modal dispersion curve: ; Wherein, represents the slope of the Lamb wave dispersion curve; ​represents the interval time interval; S0 is a symmetric mode, A0 is an anti-symmetric mode; Slope change rate The steepness of the time evolution of the dispersion characteristics of the stress wave propagation, when k disp >0.18ms / kHz, indicating that the tool has entered an abnormal wear stage.

[0025] Step 203, calculate the wavelet packet energy attenuation rate ratio in the set frequency band ; The set frequency band can be 300-400 kHz; ; Wherein, represents the energy difference of the set frequency band; represents the frequency difference of the set frequency band; Slope change rate of modal dispersion curve , used to represent the change of modal propagation characteristics, wavelet packet energy attenuation rate ratio Then it is used to quantify the linear attenuation rate of signal energy with the increase of frequency.

[0026] In step 3, the generation method of the personalized threshold vector of the current tool includes the following steps: Step 31, according to the information of the model, coating and material hardness of the current tool, the reference threshold vector of the current tool is obtained from the cloud ; ; Wherein, represents the kurtosis reference threshold, represents the entropy change reference threshold, represents the slope change rate reference threshold of the modal dispersion curve, represents the wavelet packet energy attenuation rate ratio reference threshold; Optionally, the reference threshold vector is set for each type of tool, a reference table is constructed, and a MySQL reference table is stored in the cloud; Through offline establishment, for each type of tool, the acoustic emission signals of the tool in the standard working condition from the new tool to the complete wear process are collected, the moment when the flank wear of each tool reaches the critical value (VB=0.15 mm) is determined by offline microscope measurement, the acoustic emission characteristic values corresponding to the critical point are extracted, and the characteristic values of all samples of the same type are statistically analyzed (such as taking the average value), to obtain the reference threshold vector of the tool of the same type.

[0027] A specific setting example, ; Step 32, calculate each comparison index of all acoustic emission signals of the current tool before the current time, calculate the statistics of each comparison index, input the pre-trained deep learning model for recognition, and obtain the individual feature offset. Specifically, the cloud deep learning model can adopt a lightweight CNN model; the individual feature offset, i.e., the individual feature offset of the tool individual relative to the standard tool, is represented as: wherein, represents the kurtosis offset; represents the entropy change offset; represents the dispersion slope offset; represents the energy attenuation rate offset; the offset vector Each component in the offset vector can be positive or negative; a positive offset indicates that the tool individual has higher resistance in the corresponding feature dimension relative to the standard tool, and the alarm threshold needs to be increased to fully exploit its service life. A negative offset indicates that the tool individual has poor resistance, and the alarm threshold needs to be reduced to ensure machining safety.

[0028] Step 33, add the reference threshold to the predicted individual feature offset to generate a personalized threshold vector suitable for the current tool : The obtained personalized threshold vector is sent to the tool working site, compared with the calculated comparison indicators, and a warning is given based on the comparison result; The above embodiment can fully exploit the feature offset information generated by the tool individual in actual application, break through the problem that the traditional fixed threshold method cannot be compatible with manufacturing tolerances, coating differences, and material fluctuations, and realize tool-level differentiated judgment logic. By dynamically generating a personalized threshold vector, the system can accurately identify the life stage according to the actual wear evolution characteristics, significantly reduce false positives and false negatives caused by inaccurate thresholds, improve the intelligent and adaptive level of the overall monitoring system, and ensure maximum utilization of tool resources and high-reliability processing.

[0029] To realize intelligent prediction of the feature offset trend of the tool individual, a supervised learning-based individual feature offset prediction model is constructed and trained in the cloud. The model can predict the multi-dimensional threshold offset vector of the new tool relative to the standard tool according to the acoustic emission features of the new tool in the initial processing stage, to support subsequent adaptive threshold setting and personalized life modeling.

[0030] Further, step 32 further includes a process of training the cloud deep learning model, including the following steps: Step 321, acquire acoustic emission signals in the historical working operation of the tool to construct a training data set; Specifically, a large amount of historical tool state data under multiple production lines and multiple working conditions is acquired, covering full life cycle records of different models and batches, and each data corresponds to an independent tool;​​ Step 322, Input Feature Construction: Calculate the corresponding comparison indicators, including entropy change, from the acquired acoustic emission signals of the tool during operation. , cliff Rate of change of slope of modal dispersion curve And the wavelet packet energy attenuation ratio ΔE / Δf; the calculation process for this step is the same as in step 2; Step 323: Calculate the statistics for each comparison indicator feature, and concatenate all the statistics into a high-dimensional vector as input data for the deep learning model; Optional statistics include mean, variance, maximum, minimum, skewness, and trend slope; Step 324, Label Construction: For each tool, obtain the time when it reaches the wear critical value using offline microscopic measurement methods. The stable value of the feature within the sliding window is then compared with the standard reference threshold vector of the same type of tool. Subtracting them creates the offset label: The wear threshold can be set as the tool wear amount. ; Specifically, the method for calculating the eigenvalue stability is to extract from... The characteristic stability value is obtained by averaging the comparison index values ​​of each acoustic emission signal within a time window around the given time. ; The formula for calculating the offset label is: ; Step 324: Input the input data into the deep learning model and train the deep learning model with the offset labels as the output. Step 325: Calculate the training loss based on the training results, iterate the training until the training cutoff condition is met, and then perform cross-validation to obtain the trained deep learning model; optionally, 5-fold cross-validation can be used. Specifically, such as Figure 2 The diagram shows the online monitoring closed loop, demonstrating the real-time monitoring and adaptation process executed by the industrial control computer at the edge: After a new tool begins machining, the edge node collects its initial signals and extracts features. On one hand, it calls the baseline model distributed from the cloud; on the other hand, it inputs the initial features into a deep learning model to predict the individual feature offset Δ of the tool. Finally, the general benchmark thresholds distributed from the cloud will be... Compared with the predicted Δ Add them together to generate a personalized threshold that applies only to the current tool. This is used for subsequent real-time threshold determination and lifetime prediction. The arrows indicate the data interaction between the cloud database and edge nodes, including the distribution of model parameters and thresholds, realizing the working principle of cloud-based collaboration.

[0031] Further technical solutions, in step 3, the personalized threshold vector of the current tool is compared with the calculated comparison indicators, and the method for determining the remaining proportion of life includes the following steps: Step 301, when the comparison index exceeds the corresponding threshold value in the personalized threshold vector, it is determined that an abnormality occurs, and the normalized over-standard value of the current comparison index value relative to the personalized threshold value is calculated; Specifically, when And ; or, And , it is determined that an abnormality occurs, and the system immediately captures the over-standard characteristic value of the current window.

[0032] The normalized over-standard value of the current characteristic value relative to its personalized threshold value is calculated, and the calculation formula is as follows: ; ; ; ; Step 302, based on the set weight, the obtained normalized over-standard value is weighted and fused to calculate the comprehensive score ; ; ; Among them, and are weight coefficients; Step 303, the comprehensive score is mapped to obtain the percentage of tool remaining life; Specifically, a mapping function can be constructed to map the comprehensive score to obtain the tool life remaining proportion, and the mapping function is specifically: ; In addition, in order to improve the robustness of the algorithm, a limiting condition can also be introduced: ; The maximum safety score threshold set by the limiting condition is used to forcibly truncate the abnormal score value.

[0033] In the above embodiment, in order to realize the accurate evaluation of the remaining life of the tool in the machining process, the system constructs a comprehensive scoring mechanism based on the difference between the current window acoustic emission characteristic value and the personalized threshold vector, and maps the score to the remaining life percentage R through a segmented linear function, in order to drive the early warning system and assist the operator in decision-making. In this embodiment, based on the obtained remaining life proportion, a warning is performed, and a three-level response mechanism is adopted: Level III warning: the remaining life proportion is not greater than 30%, yellow LED flickering can be performed, the flickering frequency is 2Hz, and the duty cycle is 50%; Level II warning: the remaining life proportion is not greater than 10%, level II warning is executed; the remaining life and the tool coordinates are pushed to the mobile terminal; Level I warning: when the life is zero, level I warning is executed; an emergency stop instruction is sent to the PLC to lock the machine tool to prevent workpiece damage; at the same time, a CSV event record is automatically generated, including timestamp, tool ID, personalized threshold, entropy , kurtosis , stop reason and other information, the event record log is encrypted and backed up in the cloud every day, and quality traceability is supported.

[0034] In step 4, when the remaining life proportion is lower than the set value, the tool changing operation is performed, the tool changed is micro-identified to determine whether the tool can be reused, and the comparison index calculated for the tool is taken as an input value, the micro-identified result is taken as an output, and new training data is constructed to perform rolling training on the deep learning model in the cloud; When the remaining life proportion is lower than the set value, the tool changing operation is performed, and the method for micro-identifying the tool changed to determine whether the tool can be reused comprises the following steps: Step 41, for the tool changed, a tool changing tool picture photographed by a microscope camera is obtained; Specifically, after the level I warning triggers the tool changing, the old tool is transferred to an offline station by an engineer, a 4K high-definition microscope camera (48MP resolution, 60FPS low delay) is used to photograph the wear area of the rear tool face, and the wear amount (accuracy ±0.02mm) is automatically measured. Wear amount measurement algorithm: Step 42, the tool changing tool picture is subjected to grayscale processing; Step 43, the image subjected to grayscale processing is subjected to Gaussian filtering; Step 44, for the image subjected to filtering, Canny edge detection and contour extraction are performed, a minimum circumscribed rectangle frame is marked, and a target area image of the tool to be identified is obtained; Step 45, based on the obtained target area image, the wear belt length and the depth are calculated; The wear belt length calculation method, specifically: a mapping relationship between an image pixel coordinate system and a physical space coordinate system is established through a preset calibration plate, a conversion coefficient of pixels to millimeters is determined ; in the image rectangular coordinate system, the original cutting edge contour of the tool is taken as a geometric reference, and the starting boundary point and the terminal boundary point of the wear area are automatically identified ; and calculate the Euclidean distance as the length of the wear band : ; depth calculation ) method, specifically: Based on digital image processing and curve fitting technology, the precise quantification of depth parameters is performed. Based on the discrete point set of the un-worn area of the tool, the least squares method is used to fit the standard profile curve of the original cutting edge ; feature point extraction is performed on the lower boundary profile of the wear area, and the actual profile curve of the wear area is obtained through curve fitting ; along the length direction of the wear band set the sampling point sequence }, for each sampling point, calculate the wear depth in the vertical direction: ; Take the maximum value to calculate the depth as the depth calculation value: ; wherein, is the conversion factor from pixels to millimeters.

[0035] Specifically, 5 consecutive images can be collected to calculate and , and bilinear interpolation is used to improve the calculation accuracy; Step 46, based on the length of the wear band and the depth, the remaining life is judged, and the actual remaining life percentage is calculated; ; wherein, is the maximum allowed wear depth, which can be set to 0.15mm; is the total length of the tool cutting edge; After the above image processing, if > 10%, it is marked as "reusable tool" and re-entered into the warehouse, if , the tool is determined to be scrapped, and is marked as "scrap"; the remaining life data, model, used time, measured wear value, etc. of the reusable tool are written into the cloud MySQL database through the API interface. These microscopic measured data and macro acoustic emission characteristic data together constitute the training samples of the deep learning model, which are used to optimize the prediction accuracy of individual feature offset.

[0036] Optionally, the newly added data can be further utilized in the early morning every day through a lightweight fine-tuning mechanism, such as using the PyTorch framework, and the learning rate is set to 1x10 -3, batch size is set to 32, the cloud-side deep learning model is optimized in a rolling manner to continuously reduce the prediction error and dynamically update the personalized threshold vector.

[0037] The embodiment constructs a sensor, analysis, judgment, control, and learning integrated closed-loop tool life real-time monitoring and early warning logic architecture, which is different from the loose structure of the existing technology, in which each functional module is isolated, the response is lagging, and manual intervention is required. The edge industrial computer continuously collects acoustic emission signals, and the feature extraction, threshold judgment, and three-level early warning instruction triggering are completed instantly through local calculation; after early warning, the 4K microscopic vision module is further used to collect old tool wear data and return it to the cloud database, and the fine-tuning of the transfer learning model is automatically executed daily to realize the dynamic evolution of the benchmark threshold. The architecture forms a continuous closed loop, a millisecond-level closed loop, and a self-adaptive closed loop among physical signal collection, real-time decision logic, device control response, and cloud model updating, and realizes the leap of tool life monitoring from static judgment to dynamic feedback learning.

[0038] To illustrate the effect of the above early warning method, a simulation test is performed, as follows: The traditional double-threshold value is set as: K>5.0, ΔHj>40%. The embodiment adds a frequency dispersion-attenuation coupling threshold, which is set as: the slope change rate of the frequency dispersion curve >0.18ms / kHz, ΔE / Δf>4.5dB / 100kHz.

[0039] Experimental data: when the cutting length reaches 5500m, the tool face wear V B is equal to about 0.062mm. At this time, K=4.1, ΔHj=33%, the traditional algorithm does not trigger early warning; however, the embodiment adds a comparison index, the slope change rate of the frequency dispersion curve reaches 0.18ms / kHz, ΔE / Δf=5.1dB / 100kHz, successfully triggering the II-level early warning (life ≤10%), which proves the higher sensitivity of the new model to initial wear.

[0040] Then, three new tools of the same type (T100, T101, T102) are selected for testing: First, the benchmark is established: under the standard working condition, the general benchmark threshold is measured as: k dispth =0.18ms / kHz, (ΔE / Δf) th =5.0dB / 100kHz; Initial collection is performed: after running for 10 minutes, the initial feature averages of the three tools are: the standard tool T100: [0.02ms / kHz, 0.8dB / 100kHz]; the naturally high T101: [0.08ms / kHz, 2.1dB / 100kHz]; The inherently low T102 level: [-0.01ms / kHz, -0.5dB / 100kHz]; The initial features are input into the deep learning model to predict the offset. : T101: ; The personalized threshold of T101 = [0.18 + 0.06, 5.0 + 1.3] = [0.24ms / kHz, 6.3dB / 100kHz]; T102: ; The personalized threshold for T102 is [0.18-0.03, 5.0-0.7] = [0.15ms / kHz, 4.3dB / 100kHz]. T101 cutting tool at wear amount V B When the wear is 0.10 mm, the eigenvalue is [0.22, 5.9]. Using a general threshold (0.18, 5.0) will result in serious false alarms; using its specific threshold (0.24, 6.3) will be considered normal until the wear worsens to V. B The warning is accurately triggered only when the feature value is 0.14mm and the threshold value is [0.25, 6.5]. Similarly, the T102 tool avoids the risk of missed detections due to excessively high threshold values. It is evident that the deep learning correction mechanism in this embodiment can adapt to the characteristics of different individuals, significantly improving monitoring accuracy without drastically changing the system architecture.

[0041] Example 2 Based on Embodiment 1, this embodiment of the present invention provides a real-time tool life monitoring and early warning system, including: an acoustic emission sensor, a signal processing module, a database, an early warning execution module, a microscopic vision device, and a control terminal, wherein the control terminal is configured to execute the steps of the real-time tool life monitoring and early warning method described in Embodiment 1. Acoustic emission sensor: used to collect processing signals; Signal processing module: Includes a two-stage design of charge amplifier and voltage amplifier, 1V / pC sensitivity, 1.6kHz-1MHz bandwidth, and single 12-24V power supply; Database: Pre-stores baseline thresholds and personalized thresholds for different tool models to provide a basis for early warning; Early warning execution module: Triggers corresponding alarms and control actions based on early warning instructions.

[0042] Micro-vision device: used for non-contact measurement of the worn area of the rear face of the old tool after tool replacement; the module can include a 48MP industrial camera, a telecentric lens, and a ring light source, has a 60FPS low-delay image acquisition capability, a measurement accuracy of ±0.02mm, and supports automatic identification of the length of the wear band and the depth of the wear mark; the image and the measurement result are uploaded to the industrial computer through the USB3.0 interface, if the wear amount is less than 0.15mm, it is marked as "reusable", and the data is written to the cloud MySQL database for model fine-tuning and tool life limit mining.

[0043] Embodiment 3 Based on embodiment 1, the embodiment provides a tool life real-time monitoring and early warning system, comprising: A noise processing module configured to acquire and preprocess the original acoustic emission signals of the tool working site, and to perform adaptive cavitation noise cancellation processing on the preprocessed signals; A comparison index calculation module configured to calculate comparison indexes, perform band-pass filtering and kurtosis statistical analysis on the acoustic emission signals after adaptive cavitation noise cancellation processing, and calculate entropy and kurtosis ; combined with short-time Fourier transform, frequency dispersion-attenuation coupling analysis is performed to calculate the energy attenuation ratio ΔE / Δf; A threshold updating and early warning module configured to acquire the individualized threshold vector of the current tool learned and identified based on the deep learning model, compare the individualized threshold vector with the calculated comparison indexes, determine the remaining proportion of the life, and perform early warning; The deep learning model acquires the machining data of the current tool in real time, identifies the individual feature offset of the tool individual relative to the standard tool individual, adds the set universal reference threshold to the predicted individual feature offset vector, and generates an individualized threshold vector applicable to the current tool.

[0044] It should be noted that each module in the embodiment corresponds to each step in embodiment 1, and the specific implementation process is the same, which will not be repeated here.

[0045] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0046] The above describes the specific embodiments of the present application in conjunction with the drawings, but is not a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications or changes made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A tool life real-time monitoring and early warning method, characterized in that, The method comprises the following steps: acquiring original acoustic emission signals of a tool working site for preprocessing, and performing adaptive cavitation noise cancellation processing on the preprocessed signals; Comparison metrics were calculated, and bandpass filtering and kurtosis statistical analysis were performed on the acoustic emission signal after adaptive cavitation noise cancellation processing to calculate entropy change. with kurtosis Combining short-time Fourier transform with dispersion-attenuation coupling analysis, the rate of change of the slope of the modal dispersion curve is calculated. And the energy decay ratio ΔE / Δf; acquiring an individualized threshold vector of the current tool learned and identified based on a deep learning model, comparing the individualized threshold vector with each comparison index calculated, and determining a remaining life proportion to perform early warning; the deep learning model acquires machining data of the current tool in real time, identifies individual feature offset of the tool individual relative to a standard tool, adds a set universal reference threshold to the predicted individual feature offset to serve as the individualized threshold vector of the current tool.

2. The tool life real-time monitoring and early warning method according to claim 1, characterized in that: when the remaining life proportion is lower than a set value, a tool replacement operation is performed, microscopic identification is performed on the replaced tool, it is determined whether the tool can be reused, and the comparison index calculated for the tool is taken as an input value, the microscopic identification result is taken as an output, and new training data is constructed to perform rolling training on the deep learning model in the cloud; the method for performing microscopic identification on the replaced tool to determine whether the tool can be reused comprises the following steps: for the replaced tool, a tool replacement picture taken by a microscope camera is acquired; the tool replacement picture is subjected to grayscale processing; the grayscale image is subjected to Gaussian filtering; for the filtered image, Canny edge detection and contour extraction are performed, a minimum circumscribed rectangle is marked, and a target region image of the tool to be identified is obtained; based on the obtained target region image, the length and depth of the wear band are calculated; based on the length and depth of the wear band, the remaining life is determined, and the actual remaining life percentage is calculated.

3. The tool life real-time monitoring and early warning method according to claim 1, characterized in that: the method for performing adaptive cavitation noise cancellation processing on the preprocessed signals comprises the following steps: the preprocessed acoustic emission signals are synchronously collected with a reference signal r(n) and subjected to frequency band filtering to obtain time domain data with phase alignment and frequency band limited in a set frequency range; spectrum analysis and spectral kurtosis discrimination are performed on the reference signal, a set number of main frequency components are extracted, and an orthogonal characteristic basis vector group is constructed; the reference signal is subjected to characteristic basis projection and Hilbert transform to generate an anti-phase time domain signal with opposite phase to the original cavitation noise; the synchronously collected preprocessed acoustic emission signals are subtracted by the cavitation noise anti-phase time domain signal to obtain acoustic emission signals subjected to adaptive cavitation noise cancellation processing.

4. The tool life real-time monitoring and early warning method according to claim 3, characterized in that: the process of constructing the orthogonal characteristic basis vector group is specifically as follows: fast Fourier transform is performed on the reference signal r(n); the frequency band with the most concentrated cavitation noise energy is calculated; a set number of spectral lines with the largest energy are extracted in the most concentrated frequency band with a set step size.

5. The tool life real-time monitoring and early warning method according to claim 3, characterized in that: the process of generating the anti-phase time domain signal with opposite phase to the original cavitation noise by performing characteristic basis projection and Hilbert transform on the reference signal is specifically as follows: the reference signal r(n) is projected onto the characteristic basis φ to obtain projection coefficients c(m). for each basis vector φ of the feature basis φ m performing a Hilbert transform to obtain a corresponding complex analytic form; The complex analytic form after the Hilbert transform of the corresponding basis vector is weighted and combined based on the projection coefficient c(m) to obtain the cavitation noise complex envelope signal, and the instantaneous amplitude A(n) and the instantaneous phase θ(n) of the cavitation noise are obtained. constructing an inverse signal of the cavitation noise complex envelope signal and generating an anti-phase cancellation signal by inverse mapping of the characteristic basis i.e. the cavitation noise anti-phase time domain signal.

6. The tool life real-time monitoring and early warning method of claim 1, wherein: Based on the acoustic emission signal after adaptive cavitation noise cancellation processing, band-pass filtering and feature statistical analysis are performed, and entropy change is calculated and kurtosis The process includes the following steps: A band-pass filter is applied to the acoustic emission signal after adaptive cavitation noise cancellation processing; A sliding window is set, and within each sliding window, the kurtosis is calculated and wavelet packet entropy variation .

7. The tool life real-time monitoring and early warning method of claim 1, wherein: Based on the acoustic emission signal after adaptive cavitation noise cancellation processing, combined with short-time Fourier transform for dispersion-attenuation coupling analysis, the slope change rate of modal dispersion curve is calculated And the process of wavelet packet energy attenuation rate ratio , including the following steps: A short-time Fourier transform is performed on the acoustic emission signal to obtain a frequency-time distribution spectrum curve; Extract the modal dispersion curve of the Lamb wave in the spectrum curve, calculate the slope change rate of the modal dispersion curve, and calculate the wavelet packet energy attenuation rate ratio in the set frequency band .

8. The tool life real-time monitoring and early warning method of claim 1, wherein: The generation method of the personalized threshold vector of the current tool includes the following steps: According to the information of the model, coating, and material hardness of the current tool, a reference threshold vector of the current tool is obtained ; Obtain each comparison index calculated from all acoustic emission signals before the current time of the current tool, calculate the statistics of each comparison index, input the pre-trained deep learning model for identification, and obtain the individual feature offset; The reference threshold is added to the predicted individual characteristic offset to generate a personalized threshold vector applicable to the current tool T new at the current time t.

9. A tool life real-time monitoring and early warning system, characterized in that, It includes: An acoustic emission sensor and a control terminal, wherein the control terminal is configured to perform the steps of the tool life real-time monitoring and early warning method of any one of claims 1-8.

10. A tool life real-time monitoring and early warning system, characterized in that, It includes: A noise processing module configured to obtain and preprocess the original acoustic emission signal of the tool working site, and perform adaptive cavitation noise cancellation processing on the preprocessed signal; The comparison index calculation module is configured to calculate a comparison index based on an acoustic emission signal after adaptive cavitation noise cancellation processing, perform band-pass filtering and kurtosis statistical analysis, calculate entropy change and kurtosis ; combined with short-time Fourier transform, perform dispersion-attenuation coupling analysis, calculate the slope change rate of the modal dispersion curve and energy attenuation ratio ΔE / Δf; A threshold updating and early warning module configured to obtain the personalized threshold vector of the current tool identified based on the deep learning model, compare it with each comparison index calculated, determine the remaining proportion of the life, and perform early warning; The deep learning model obtains the machining data of the current tool in real time, identifies the individual feature offset of the tool individual relative to the standard tool, adds the set universal reference threshold to the predicted individual feature offset as the personalized threshold vector of the current tool.

Images