Heavy turning tool based on acoustic emission sensor and intelligent monitoring system

Abstract

The invention provides a heavy-duty turning tool based on an acoustic emission sensor and an intelligent monitoring system, and relates to the technical field of turning tools, the heavy-duty turning tool comprises a signal processing module, a data analysis module, a calculation module, a model construction module, a simulation module and an execution module; the signal processing module collects the stress sound wave signal and divides the stress sound wave signal into sub-sound wave signals; the data analysis module extracts time-frequency domain characteristic parameters, screens an effective time window and records a running state; the calculation module obtains the impact frequency change rate and the average impact energy to form time sequence data; the model construction module takes time sequence data of previous N windows as input and takes the state of the latest window as a label training state recognition model; the simulation module inputs current and previous N-1 window time sequence data to predict a current state; and the execution module executes corresponding measures according to the prediction state. According to the method, the situation that the accumulated machining time is mistakenly counted in the non-cutting time period is avoided, it is ensured that the lathe tool abrasion analysis time reference is accurate, and the operation state judgment precision and real-time performance are improved through time-frequency domain feature fusion.

Description

Technical Field

[0001] This invention relates to the field of lathe tool technology, specifically to a heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor. Background Technology

[0002] Acoustic emission (AE) is a transient stress wave phenomenon generated by the rapid release of strain energy in solid materials or components due to crack propagation, plastic deformation, or phase transformation. AE is the result of localized energy redistribution within a material or structure, and this redistribution is often an early sign of material or component damage. Acoustic emission technology utilizes this characteristic to conduct online monitoring of early degradation in large energy structure equipment, in order to capture transient and sudden fault signals.

[0003] The existing technology has the following shortcomings: it does not construct a multi-feature differentiation mechanism for "cutting / non-cutting state" and only relies on a single threshold of time domain RMS for judgment. This leads to non-machining periods such as idling and clamping being mistakenly included in the cumulative cutting time, which directly undermines the time benchmark for tool wear analysis. It only stays at the time domain basic feature analysis of acoustic emission signals and lacks in-depth mining of frequency domain features, making it impossible to capture the real-time changes in the tool state during the machining process.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor, comprising:

[0008] The signal processing module is used to track and acquire stress acoustic signals of heavy-duty lathe tool machining with real-time endpoint, retain characteristic frequency bands by bandpass filtering, divide the signal with continuous equal-width time windows, and extract sub-acoustic signals within each time window.

[0009] The data analysis module is used to extract features from the sub-acoustic signals within each time window, obtain time-domain and frequency-domain feature parameters, analyze the time-domain and frequency-domain feature parameters corresponding to each time window, filter out the effective time windows for effective processing, and synchronously record the corresponding running status within each effective processing time window.

[0010] The time-series data construction module is used to obtain the rate of change of the number of impacts and the average impact energy within each effective time window for the time-domain characteristic parameters, thereby constructing the time-series data of the rate of change of the number of impacts and the time-series data of the average impact energy.

[0011] The model building module is used to build a state recognition model based on a long short-term memory network. It takes the time series data of the rate of change of the number of impacts and the average impact energy of the previous N effective time windows as input and the running state of the most recent window among the previous N effective time windows as the label output to train the state recognition model.

[0012] The simulation module is used to input the time-series data of the rate of change of the number of impacts and the average impact energy of the current and the previous N-1 effective time windows into the trained state recognition model to predict the operating state of the heavy-duty turning tool in the current effective time window.

[0013] The execution module is used to perform corresponding measures based on the operating status of the heavy-duty turning tool within the current effective time window.

[0014] Furthermore, the time-domain characteristic parameters include the root mean square value and the number of impacts; the frequency-domain characteristic parameters include the peak frequency and the proportion of high-frequency energy; and the operating status includes normal cutting status, abnormal warning status, and fault shutdown status.

[0015] Furthermore, stress acoustic signals during machining with heavy-duty lathe tools are collected backward from the real-time endpoint, and characteristic frequency band signals are preserved through bandpass filtering. The specific steps are as follows:

[0016] Starting from the real-time point, the stress acoustic signal x(t) is collected backward for a duration of T.

[0017] The frequency domain expression of the stress acoustic signal x(t) is X(f), which is obtained through Fourier transform:

[0018]

[0019] Where X(f) describes the amplitude and phase distribution of each frequency component in x(t);

[0020] The frequency response H(f) of a bandpass filter is defined as:

[0021]

[0022] Where f is the frequency of the frequency domain signal;

[0023] When the spectrum X(f) of the original time-domain signal passes through this bandpass filter, the output spectrum X is... bp (f):

[0024] X bp(f)=X(f)·H(f)

[0025] In the acoustic emission signal of a heavy-duty lathe tool, the contact friction between the tool and the workpiece mainly generates energy in the 20-100kHz range, which is transmitted through the X-ray frequency spectrum. bp (f) Determine whether each frequency component is within the range of 20-100kHz, retain the amplitude and phase information within that frequency band, and suppress other frequency bands.

[0026] Furthermore, the bandpass-filtered stress acoustic signal is segmented using multiple consecutive time windows of equal width, and sub-acoustic signals within each time window are extracted. The specific steps are as follows:

[0027] The filtered spectrum X is obtained by inverse Fourier transform. bp (f) Transform back to the time domain:

[0028]

[0029] Where, x bp (t) is a time-domain signal containing only frequency components from 20 to 100 kHz;

[0030] The time-domain signal x bp (t) is divided into continuous and non-overlapping time windows of fixed width Δt. i , where i is the index of the time window, i∈[1,N], and N is the number of time windows;

[0031] The basis for setting the fixed width Δt:

[0032]

[0033] Among them, f min The lowest cutoff frequency of the bandpass filter is f. min =20kHz;

[0034] For each time window W i Extract its corresponding sub-acoustic signal x i (t):

[0035]

[0036] in, For a rectangular window function, when Within the interval, Right now At this time, the sub-acoustic wave signal x i (t)=x bp (t); other intervals are 0, i.e., x i (t) = 0;

[0037] From the time-domain signal x through a rectangular window bp(t) "cuts out" a segment with a time width of Δt, with the center of the window at t. i Extending to the left and right

[0038] Set the start time t of the i-th time window. i = (i-1)Δt;

[0039] The number of sampling points M contained in each time window:

[0040] M=Δt·f s

[0041] Among them, f s The sampling rate is f, which represents the number of samples collected per second. s One point;

[0042] Set the start and end sampling points of the i-th time window as follows:

[0043] C start,i = (i-1)M

[0044] C end,i =i·M-1

[0045] Among them, C start,i C end,i These are the starting and ending sampling points of the i-th time window, respectively;

[0046] The discrete sub-acoustic signal of the i-th time window is obtained by converting the continuous sub-acoustic signal x. i After mapping (t) to a discrete index interval, the index in [C] is extracted from the original time-domain signal. start,i C end,i The fragment within ], that is,

[0047] x i (n)=x bp (n)n∈[C start,i C end,i ]

[0048] Where, x i (n) represents the discrete sub-acoustic signal in the i-th time window, x bp (n) represents a discrete-time signal, where n is the index of the discrete-time window.

[0049] Furthermore, the time-domain and frequency-domain characteristic parameters corresponding to each time window are analyzed to select the effective time windows for processing. The specific steps are as follows:

[0050] For each time window, it is marked as a valid time window if the following conditions are met simultaneously:

[0051]

[0052] In the formula, RMS i Let RMS be the root mean square value of the i-th time window. idle The root mean square value during idling, RMS fault SC is the root mean square value under fault conditions. i Let SC be the number of impacts in the i-th time window. min SC represents the number of impacts during idling. max f represents the number of impacts during abnormal wear. peak,i fl is the peak frequency of the i-th time window. low f is the peak frequency during idling. high HR is the peak frequency during blade breakage. i Let HR be the proportion of high-frequency energy in the i-th time window. idel This represents the percentage of high-frequency energy during idling.

[0053] Furthermore, for the time-domain characteristic parameters within each effective time window, the rate of change of the number of impacts and the average impact energy within each effective time window are obtained, based on the following formula:

[0054]

[0055] Among them, SCCR j SC represents the rate of change of the number of impacts within the j-th effective time window. j For the number of impacts in the j-th effective time window, SC j-1 Let j be the number of impacts in the (j-1)th effective time window, where j is the index of the effective time window, j∈[2,m], and m is the number of effective time windows;

[0056]

[0057] Among them, E α Let A(t) be the energy of the αth impact, and A(t) be the amplitude of the stress acoustic signal. α1 t α2 Let ASE be the start and end times of the αth impact, respectively. j Let α be the average impact energy within the j-th effective time window, and α be the index of the impact, α∈[1,SC]. j ].

[0058] Furthermore, based on the operating status of the heavy-duty cutting tool within the current effective time window, corresponding measures are implemented. The specific steps are as follows:

[0059] When the operating status is normal cutting state, continue machining;

[0060] An alert is triggered when the running status is in an abnormal warning state;

[0061] Emergency shutdown is initiated when the operating status is a fault shutdown state.

[0062] Compared with the prior art, the beneficial effects of the present invention are:

[0063] This invention retains high-frequency signals through bandpass filtering and segments the signals using continuous equal-width time windows. By combining time-domain and frequency-domain feature parameter analysis to screen effective machining time windows, a multi-feature differentiation mechanism for "cutting / non-cutting states" is constructed. This avoids the miscalculation of non-cutting periods into the cumulative machining time, ensuring the accuracy of the tool wear analysis time reference. Simultaneously, based on the impact frequency change rate and average impact energy of the effective time window to form time-series data, the tool operating state is predicted through a state recognition model, and corresponding measures are implemented. This achieves in-depth mining of time-domain features and fusion analysis of time-domain features, improving the accuracy and real-time performance of tool operating state judgment, and providing reliable technical support for intelligent monitoring and predictive maintenance of cutting tools. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the overall method flow of the present invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0066] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0067] Example 1:

[0068] Please see Figure 1 The present invention provides a technical solution:

[0069] A heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor, comprising:

[0070] The signal processing module is used to collect stress acoustic wave signals during the machining of heavy lathe tools by tracing back from the real time point, retain high-frequency signals through bandpass filtering, and use multiple consecutive equal-width time windows to segment the stress acoustic wave signals after bandpass filtering and extract sub-acoustic wave signals in each time window.

[0071] Based on the above embodiments, stress acoustic signals during machining with heavy-duty lathe tools are collected backward from the real-time endpoint, and characteristic frequency band signals are preserved through bandpass filtering. The specific steps are as follows:

[0072] Starting from the real-time point, the stress acoustic signal x(t) is collected backward for a duration of T, where t is the time variable of the time domain signal;

[0073] The frequency domain expression of the stress acoustic signal x(t) is X(f), which is obtained through Fourier transform:

[0074]

[0075] Where X(f) describes the amplitude and phase distribution of each frequency component in x(t);

[0076] The frequency response H(f) of a bandpass filter is defined as:

[0077]

[0078] Where f is the frequency of the frequency domain signal;

[0079] When the full-band spectrum X(f) of the original time-domain signal passes through this bandpass filter, the output filtered spectrum X is obtained. bp (f):

[0080] X bp (f)=X(f)·H(f)

[0081] In the acoustic emission signal of a heavy-duty lathe tool, the contact friction between the tool and the workpiece mainly generates energy in the 20-100kHz range, which is transmitted through the X-ray frequency spectrum. bp (f) Determine whether each frequency component is within the range of 20-100kHz, retain the amplitude and phase information within that frequency band, and suppress other frequency bands.

[0082] Based on the above embodiments, multiple consecutive time windows of equal width are used to segment the bandpass filtered stress acoustic signal, and sub-acoustic signals within each time window are extracted. The specific steps are as follows:

[0083] The filtered spectrum X is obtained by inverse Fourier transform. bp (f) Transform back to the time domain:

[0084]

[0085] Where, x bp (t) is a time-domain signal containing only frequency components from 20 to 100 kHz;

[0086] The time-domain signal x bp (t) is divided into continuous and non-overlapping time windows of fixed width Δt. i , where i is the index of the time window, i∈[1,N], and N is the number of time windows;

[0087] The basis for setting the fixed width Δt:

[0088]

[0089] Among them, f min The lowest cutoff frequency of the bandpass filter is f. min =20kHz;

[0090] For each time window W i Extract its corresponding sub-acoustic signal x i (t):

[0091]

[0092] in, For a rectangular window function, when Within the interval, Right now At this time, the sub-acoustic wave signal x i (t)=x bp (t); other intervals are 0, i.e., x i (t) = 0;

[0093] From the time-domain signal x through a rectangular window bp (t) "cuts out" a segment with a time width of Δt, with the center of the window at t. i Extending to the left and right

[0094] Set the start time t of the i-th time window. i = (i-1)Δt;

[0095] The number of sampling points M contained in each time window:

[0096] M=Δt·fs s

[0097] Among them, f s The sampling rate is f, which represents the number of samples collected per second. s One point;

[0098] Set the start and end sampling points of the i-th time window as follows:

[0099] C start,i = (i-1)M

[0100] C end,i =i·M-1

[0101] Among them, C start,i C end,i These are the starting and ending sampling points of the i-th time window, respectively;

[0102] The discrete sub-acoustic signal of the i-th time window is obtained by converting the continuous sub-acoustic signal x. i After mapping (t) to a discrete index interval, the index in [C] is extracted from the original time-domain signal. start,i C end,i The fragment within ], that is,

[0103] x i (n)=x bp (n)n∈[C start,i C end,i ]

[0104] Where, x i (n) represents the discrete sub-acoustic signal in the i-th time window, x bp (n) represents a discrete-time signal, where n is the index of the discrete-time window.

[0105] The data analysis module is used to extract features from the sub-acoustic signals within each time window, obtain time-domain and frequency-domain feature parameters, analyze the time-domain and frequency-domain feature parameters corresponding to each time window, filter out the effective time windows for effective processing, and synchronously record the corresponding running status within each effective processing time window.

[0106] Based on the above embodiments, the time-domain characteristic parameters include the root mean square value and the number of impacts; the frequency-domain characteristic parameters include the peak frequency and the proportion of high-frequency energy; and the operating states include normal cutting state, abnormal warning state, and fault shutdown state.

[0107] Based on the above embodiments, feature extraction is performed on the sub-acoustic signals within each time window to obtain the root mean square value, number of impacts, peak frequency, and high-frequency energy ratio. The specific method is as follows:

[0108] The root mean square (RMS) value is the average energy of the subacoustic signal within a time window. Specifically, it is calculated by summing the squares of the signal values ​​at each sampling point within the time window, dividing by the number of sampling points, and then taking the square root. This value reflects the signal's energy intensity; when a tool wears or chipps, the signal energy typically increases significantly.

[0109] The number of impacts is the number of times the signal amplitude exceeds a threshold per unit time, characterizing the stability of the cutting process. First, a dynamic threshold is set (generally 2-3 times the root mean square value of the current time window). The number of times the signal amplitude exceeds this threshold within the time window is counted. Then, this count is divided by the length of the corresponding time window to obtain the number of impacts per unit time. The number of impacts will increase significantly when the cutting process is unstable or the cutting tool breaks.

[0110] The peak frequency is the frequency point where signal energy is most concentrated and is related to the inherent vibration of the cutting tool-workpiece system. A Fast Fourier Transform (FFT) is performed on the sub-acoustic signal within a time window to obtain its spectral distribution. The frequency point where energy is most concentrated, i.e., the frequency corresponding to the maximum amplitude spectrum value, is then identified. Tool wear causes changes in the stiffness of the cutting tool-workpiece system, which in turn causes the peak frequency to drift.

[0111] The high-frequency energy percentage is the proportion of high-frequency energy (e.g., 80-100kHz) to the total energy, reflecting the micro-wear of the cutting tool. Specifically, it involves first calculating the energy (sum of squares of the spectral amplitudes) of the high-frequency band and the total characteristic frequency band separately, then dividing the high-frequency energy by the total energy and multiplying by 100%. When micro-wear occurs in the cutting tool, the high-frequency energy decreases due to changes in friction conditions, causing this percentage to drop.

[0112] Based on the above embodiments, the time-domain and frequency-domain characteristic parameters corresponding to each time window are analyzed to select the effective time windows for processing. The specific steps are as follows:

[0113] For each time window, it is marked as a valid time window if the following conditions are met simultaneously:

[0114]

[0115] In the formula, RMS i Let RMS be the root mean square value of the i-th time window. idle The root mean square value during idling, RMS fault SC is the root mean square value under fault conditions. i Let SC be the number of impacts in the i-th time window. min SC represents the number of impacts during idling. max f represents the number of impacts during abnormal wear. peak,i f is the peak frequency of the i-th time window. low f is the peak frequency during idling. high The peak frequency during blade breakage, HR i Let HR be the proportion of high-frequency energy in the i-th time window. idel This represents the percentage of high-frequency energy during idling.

[0116] Based on this, it should be noted that:

[0117] Root mean square (RMS) value during idling idle Root mean square (RMS) value under fault conditions fault Number of impacts during idling (SC) min Impact count during abnormal wear (SC) max Peak frequency f during idling low Peak frequency f during blade breakage high High-frequency energy ratio during idling (HR) idel The method to obtain it is as follows:

[0118] Among them, the root mean square value (RMS) during idling is... idle :

[0119] Ensure the cutting tool is not in contact with the workpiece and the machine tool is idling; collect acoustic emission signals for 10-20 seconds, and calculate the RMS of each window after dividing it into time windows. i After removing outliers, the average of the remaining data is taken as the RMS. idle ;

[0120] Root mean square (RMS) value under fault conditions fault :

[0121] Artificially induced tool failures, such as minor chipping or continued use until severe wear, are investigated; acoustic emission signals under these failure conditions are collected, and the RMS value for each window is calculated. i RMS fault RMS during the fault stabilization period i The average value is 1.5 to 2 times the average RMS value of normal cutting.

[0122] Number of impacts during idling (SC) min :

[0123] Calculate the number of impacts (SC) in each time window during idling. i ; Retrieve all windows SC i The minimum value is taken as SC min .

[0124] Impact count during abnormal wear (SC) max :

[0125] Abnormal wear conditions are determined by measuring tool wear; acoustic emission signals under these conditions are collected, and SC is calculated. i Take the wear stabilization period SC i The average value, or set to 2-3 times the average number of normal cutting impacts.

[0126] Peak frequency f during idling low :

[0127] During idling, perform Fourier transforms on the signals in each time window to calculate the peak frequency f. peak,i ; Retrieve all time windows f peak,i The highest frequency value is used as f low .

[0128] Peak frequency f during blade breakage high :

[0129] The moment of chipping was confirmed by monitoring with a force sensor; signals before and after chipping were collected, and the peak frequency at the time of chipping was extracted; f values ​​from multiple chipping tests were obtained. peak,i The average value, as f high .

[0130] High-frequency energy percentage during idling (HR) idel :

[0131] Calculate the high-frequency energy percentage (HR) for each window during idling. i ; Get HR i The average value as HR idel .

[0132] The time-series data construction module is used to obtain the rate of change of the number of impacts and the average impact energy within each effective time window for the time-domain characteristic parameters, thereby constructing the time-series data of the rate of change of the number of impacts and the time-series data of the average impact energy.

[0133] Based on the above embodiments, for the time-domain characteristic parameters within each effective time window, the rate of change of the number of impacts and the average impact energy within each effective time window are obtained, according to the following formula:

[0134]

[0135] Among them, SCCR j SC represents the rate of change of the number of impacts within the j-th effective time window. j For the number of impacts in the j-th effective time window, SC k-1 Let j be the number of impacts in the (j-1)th effective time window, where j is the index of the effective time window, j∈[2,m], and m is the number of effective time windows;

[0136]

[0137] Among them, E α Let A(t) be the energy of the αth impact, and A(t) be the amplitude of the stress acoustic signal. α1 t α2 Let ASE be the start and end times of the αth impact, respectively. j Let α be the average impact energy within the j-th effective time window, and α be the index of the impact, α∈[1,SC].j ].

[0138] The model building module is used to build a state recognition model based on a long short-term memory network. It takes the time series data of the rate of change of the number of impacts of the heavy lathe tool and the time series data of the average impact energy of the previous N effective time windows as input, and the running state of the most recent effective time window among the previous N effective time windows as the label output to train the state recognition model.

[0139] Based on the above embodiments, the state recognition model based on LSTM (Long Short-Term Memory) takes time-series data as input and captures the dynamic evolution characteristics of the state of heavy-duty lathe tools through memory units. The core components are as follows:

[0140] Input layer:

[0141] The time series data of the rate of change of the number of impacts and the time series data of the average impact energy for the first N effective time windows are used to form a two-dimensional time series sequence;

[0142] The two-dimensional time series is converted into a three-dimensional tensor, where: the first dimension is the number of samples (e.g., 1000 training samples); the second dimension is the time step (e.g., N=20, indicating that the features of the first 20 windows are used); and the third dimension is the feature dimension (2 features: rate of change of impact number and average impact energy).

[0143] LSTM layer:

[0144] A two-layer LSTM network is used: the first layer returns the sequence, and the second layer extracts global features.

[0145] The first LSTM layer has 64 neurons and retains the hidden state at all time steps. It is used to capture short-term temporal correlations and the impact energy fluctuations between adjacent windows.

[0146] The second LSTM layer has 32 neurons and extracts long-term dependent features, showing the continuous upward trend of impact energy during tool wear.

[0147] Fully connected hidden layer:

[0148] Two fully connected layers, both using the ReLU activation function, map the temporal features of the LSTM output to a high-dimensional abstract space, enhancing the ability to represent key states;

[0149] Output layer:

[0150] Classification task (normal / warning / fault): 3 neurons, softmax activation function, output probability distribution of each state;

[0151] The process of training the state recognition model is as follows:

[0152] The rate of change of impact number and average impact energy are standardized to eliminate the influence of dimensions; the feature sequences of the first N windows are extracted in chronological order to form a three-dimensional input tensor.

[0153] Training set: 70% of the data, used to learn model parameters, including LSTM weights and fully connected layer biases;

[0154] Validation set: 10% of the data, with hyperparameters dynamically adjusted during training, including learning rate and number of LSTM neurons, to prevent overfitting;

[0155] Test set: 20% of the data, used to independently evaluate the model's generalization ability, including the accuracy of recognizing new tool states;

[0156] Cross-entropy loss is used to measure the difference between the distribution of predicted probabilities and true labels; the Adam optimizer is used in conjunction with a learning rate decay strategy, such as reducing the learning rate to 50% of its original value when the validation set loss remains unchanged for 5 consecutive rounds.

[0157] Iterative process: Batch input of 3D time series data, LSTM layer processes the features of each window sequentially according to time step; calculate the predicted value through forward propagation, and update the network weights through backpropagation; calculate the loss and evaluation metric on the validation set after each round of training;

[0158] Training stops when the mean squared error of the validation set falls within the interval [0, 0.01].

[0159] The simulation module is used to input the time series data of the rate of change of impact number and the time series data of the average impact energy of the heavy-duty turning tool in the current and the previous N-1 effective time windows into the trained state recognition model to predict the operating state of the heavy-duty turning tool in the current effective time window.

[0160] The execution module is used to perform corresponding measures based on the operating status of the heavy-duty cutting tool within the current effective time window.

[0161] Based on the above embodiments, corresponding measures are implemented according to the operating status of the heavy-duty turning tool within the current effective time window. The specific steps are as follows:

[0162] When the operating status is normal cutting state, continue machining;

[0163] An alert is triggered when the running status is in an abnormal warning state;

[0164] Emergency shutdown is initiated when the operating status is a fault shutdown state.

[0165] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0166] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by software, electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0167] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0168] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor, characterized in that: include: The signal processing module is used to track and acquire stress acoustic signals of heavy-duty lathe tool machining with real-time endpoint, retain characteristic frequency bands by bandpass filtering, divide the signal with continuous equal-width time windows, and extract sub-acoustic signals within each time window. The data analysis module is used to extract features from the sub-acoustic signals within each time window, obtain time-domain and frequency-domain feature parameters, analyze the time-domain and frequency-domain feature parameters corresponding to each time window, filter out the effective time windows for effective processing, and synchronously record the corresponding running status within each effective processing time window. The time-series data construction module is used to obtain the rate of change of the number of impacts and the average impact energy within each effective time window for the time-domain characteristic parameters, thereby constructing the time-series data of the rate of change of the number of impacts and the time-series data of the average impact energy. The model building module is used to build a state recognition model based on a long short-term memory network. It takes the time series data of the rate of change of the number of impacts and the average impact energy of the previous N effective time windows as input and the running state of the most recent window among the previous N effective time windows as the label output to train the state recognition model. The simulation module is used to input the time-series data of the rate of change of the number of impacts and the average impact energy of the current and the previous N-1 effective time windows into the trained state recognition model to predict the operating state of the heavy-duty turning tool in the current effective time window. The execution module is used to perform corresponding measures based on the operating status of the heavy-duty turning tool within the current effective time window.

2. The heavy-duty lathe tool and intelligent monitoring system based on acoustic emission sensor according to claim 1, characterized in that: The time-domain characteristic parameters include the root mean square value and the number of impacts; Frequency domain characteristic parameters include peak frequency and high-frequency energy ratio; operating status includes normal cutting status, abnormal warning status, and fault shutdown status.

3. The heavy-duty lathe tool and intelligent monitoring system based on acoustic emission sensor according to claim 2, characterized in that: The stress acoustic signal during machining with a heavy-duty lathe tool is collected by tracing back from the real-time endpoint, and the characteristic frequency band signal is preserved by bandpass filtering. The specific steps are as follows: Starting from the real-time point, the stress acoustic signal x(t) is collected backward for a duration of T. The frequency domain expression of the stress acoustic signal x(t) is X(f), which is obtained through Fourier transform: Where X(f) describes the amplitude and phase distribution of each frequency component in x(t); The frequency response H(f) of a bandpass filter is defined as: Where f is the frequency of the frequency domain signal; When the spectrum X(f) of the original time-domain signal passes through this bandpass filter, the output spectrum X is... bp (f): X bp (f)=X(f)·H(f) In the acoustic emission signal of a heavy-duty lathe tool, the contact friction between the tool and the workpiece generates energy in the 20-100kHz range, which is transmitted through the X-ray frequency spectrum. bp (f) Determine whether each frequency component is within the range of 20-100kHz, retain the amplitude and phase information within that frequency band, and suppress other frequency bands.

4. The heavy-duty lathe tool and intelligent monitoring system based on acoustic emission sensor according to claim 3, characterized in that: The bandpass-filtered stress acoustic signal is segmented using multiple consecutive time windows of equal width, and the sub-acoustic signal within each time window is extracted. The specific steps are as follows: The filtered spectrum X is obtained by inverse Fourier transform. bp (f) Transform back to the time domain: Where, x bp (t) is a time-domain signal containing only frequency components from 20 to 100 kHz; The time-domain signal x bp (t) is divided into continuous and non-overlapping time windows of fixed width Δt. i , where i is the index of the time window, i∈[1,N], and N is the number of time windows; The basis for setting the fixed width Δt: Among them, f min The lowest cutoff frequency of the bandpass filter is f. min =20kHz; For each time window W i Extract its corresponding sub-acoustic signal x i (t): in, For a rectangular window function, when Within the interval, Right now At this time, the sub-acoustic wave signal x i (t)=x bp (t); other intervals are 0, i.e., x i (t) = 0; From the time-domain signal x through a rectangular window bp (t) "cuts out" a segment with a time width of Δt, with the center of the window at t. i Extending to the left and right Set the start time t of the i-th time window. i = (i-1)Δt; The number of sampling points M contained in each time window: M=Δt·f s Among them, f s The sampling rate is f, which represents the number of samples collected per second. s One point; Set the start and end sampling points of the i-th time window as follows: C start,i =(i-1)M C end,i =i·M-1 Among them, C start,i C end,i These are the starting and ending sampling points of the i-th time window, respectively; The discrete sub-acoustic signal of the i-th time window is obtained by converting the continuous sub-acoustic signal x. i After mapping (t) to a discrete index interval, the index in [C] is extracted from the original time-domain signal. start,i C end,i The fragment within ], that is, x i (n)=x bp (n)n∈[C start,i ,C end,i ] Where, x i (n) represents the discrete sub-acoustic signal in the i-th time window, x bp (n) represents a discrete-time signal, where n is the index of the discrete-time window.

5. The heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor according to claim 4, characterized in that: For each time window, the time-domain and frequency-domain characteristic parameters are analyzed to select the effective time windows for processing. The specific steps are as follows: For each time window, it is marked as a valid time window if the following conditions are met simultaneously: In the formula, RMS i Let RMS be the root mean square value of the i-th time window. idle The root mean square value during idling, RMS fault SC is the root mean square value under fault conditions. i Let SC be the number of impacts in the i-th time window. min SC represents the number of impacts during idling. max f represents the number of impacts during abnormal wear. peak,i fl is the peak frequency of the i-th time window. low f is the peak frequency during idling. high HR is the peak frequency during blade breakage. i Let HR be the proportion of high-frequency energy in the i-th time window. idel This represents the percentage of high-frequency energy during idling.

6. The heavy-duty lathe tool and intelligent monitoring system based on acoustic emission sensor according to claim 2, characterized in that: For the time-domain characteristic parameters within each effective time window, the rate of change of the number of impacts and the average impact energy within each effective time window are obtained, based on the following formulas: Among them, SCCR j SC represents the rate of change of the number of impacts within the j-th effective time window. j For the number of impacts in the j-th effective time window, SC j-1 Let j be the number of impacts in the (j-1)th effective time window, where j is the index of the effective time window, j∈[2,m], and m is the number of effective time windows; Among them, E α Let A(t) be the energy of the αth impact, and A(t) be the amplitude of the stress acoustic signal. α1 t α2 Let ASE be the start and end times of the αth impact, respectively. j Let α be the average impact energy within the j-th effective time window, and α be the index of the impact, α∈[1,SC]. j ].

7. The heavy-duty lathe tool and intelligent monitoring system based on an acoustic emission sensor according to claim 2, characterized in that: Based on the operating status of the heavy-duty cutting tool within the current effective time window, implement corresponding measures. The specific steps are as follows: When the operating status is normal cutting state, continue machining; An alert is triggered when the running status is in an abnormal warning state; Emergency shutdown is initiated when the operating status is a fault shutdown state.

Images