A milling cutter abnormality monitoring method based on blade frequency energy visualization

Abstract

A kind of milling cutter abnormal monitoring method based on blade frequency energy visualization, specifically includes the following processes, S1: acquisition milling cutter works each state;S2: the data collected in step S1 is corrected;S3: the data obtained in step S2 is obtained using fast Fourier transform signal 1 to n order conversion harmonic amplitude and phase;S4: the amplitude phase information obtained in step S3 is reconstructed to obtain characteristic signal;S5: the data obtained in step S4 selects a group of data to obtain characteristic signal;S6: the amplitude of each blade is extracted from the image obtained in step S5.This application qualitatively analyzes the fault of milling cutter, and improves the basis for quantitative detection of milling cutter fault;And most of the irrelevant information and noise interference in the signal is removed, the adaptability to complex working conditions is improved, the relative state of each blade of milling cutter can be more clearly and stably characterized, the anti-interference ability and sensitivity of milling cutter abnormal monitoring method are improved, and the favorable technical support for safe and stable work of milling cutter is played.

Description

I: Technical Field

[0001] This invention belongs to the field of tool health monitoring in machining processes, and specifically relates to a method for monitoring milling cutter anomalies based on the visualization of cutting frequency energy. II: Background Technology

[0002] With the rapid development of the world economy and technology, products are evolving towards greater variety, batch production, refinement, and complexity. Milling cutters, as an important machining tool, are widely used in planar, curved surface, and complex cavity machining. In actual operation, due to various reasons such as tool wear and degradation, improper clamping, and process design errors, a series of faults such as wear, chipping, and tool breakage can occur during milling. If ignored, these problems can range from product scrap and wasted production resources to serious threats to the personal safety of production workers.

[0003] To ensure the safe and stable operation of milling cutters, online monitoring technology based on signal processing has become a key means of monitoring milling cutter safety. This mainly includes the following modes: (1) Extracting indicators through time-frequency domain analysis and using a deep convolutional neural network to monitor milling cutter wear; (2) Utilizing a convolutional transform neural network (CNN-transformer neural network). ) Extracting the temporal features of milling cutter data to complete the monitoring of tool wear; (3) After simply extracting a large number of statistical indicators, linear discriminant analysis is used to reduce the dimension of the feature set, and combined with backpropagation (BP) neural network and gray wolf optimization algorithm, the wear monitoring of micro milling cutter is completed. Although the above methods have achieved certain results in the safety of machine tool milling cutters, the above research lacks effective means in characterizing fault features. Since only simple statistical indicators or deep learning algorithms are used for feature extraction, the indicators obtained lack rationality and interpretability. In addition, due to the continuous flushing of cutting fluid and the parallel operation of auxiliary mechanisms during milling cutter operation, a large amount of noise is introduced into the collected signals, which makes the key information of the milling process obscured. Therefore, traditional indicators will inevitably introduce irrelevant variables into the subsequent model, interfering with the accuracy of the final monitoring. In summary, under complex working conditions, exploring a reasonable, interpretable and effective milling cutter anomaly monitoring algorithm that can effectively characterize tool fault features, suppress noise interference, and eliminate irrelevant variables is crucial for the anomaly monitoring of machine tool milling cutters. III: Summary of the Invention

[0004] To overcome the shortcomings of existing milling cutter safety monitoring technologies, as described in the background, this invention provides a milling cutter anomaly monitoring method based on cutting edge frequency energy visualization. This method visualizes the energy of each cutting edge of the milling cutter in an image format, thereby intuitively monitoring the relative changes in energy of each cutting edge and effectively monitoring abnormal conditions of the milling cutter, ensuring its safe and stable operation.

[0005] The technical solution adopted by this invention to solve its technical problem is:

[0006] A method for monitoring milling cutter anomalies based on cutting edge frequency energy visualization is characterized by firstly using the maximization method to achieve whole-cycle sampling to obtain more accurate amplitude and phase information, then reconstructing a feature signal based on the obtained amplitude and phase, aligning the feature signal with a reference signal using the maximum correlation method, and scaling it to polar coordinates to obtain a visualized cutting edge frequency energy monitoring scheme for monitoring milling cutter anomalies. Specifically, the method includes the following steps: S1: Using a sensor installed near the spindle of the milling cutter, single-channel vibration data of the milling cutter under various working states are collected, denoted as... S2: Correct the data collected in step S1, and use the maximization method to achieve integer-cycle sampling of the signal to obtain... and the obtained signal Repeat the extension until the signal reaches three times the sampling frequency length, and denote the resulting signal as x; S3: Use the Fast Fourier Transform to obtain the amplitude and phase of the 1st to nth order frequency-reversed harmonics of the signal from the data obtained in step S2, and denote them as . <amp i alpha i >, where amp i Let alpha be the amplitude of the i-th order rotation frequency. i Let i be the phase of the i-th order frequency, i = 1, 2, ..., n, where n is the number of cutting edges; S4: Reconstruct the characteristic signal from the amplitude and phase information obtained in step S3, denoted as component; S5: Select the characteristic signal component obtained from a set of data obtained in step S4, and label it as comp0, as a reference. Specifically, after acquiring a new set of signals and obtaining the characteristic signal component, calculate the maximum correlation coefficient between the new component and comp0, and use this to perform time-shift correction on the new component to obtain comp0. k and set comp = [comp0, comp1, ..., comp] kThe images are plotted on the same polar coordinate graph, which is called the cutting edge energy distribution graph. The dynamic changes in the cutting edge energy distribution graph can characterize the state of each cutting edge of the milling cutter during the milling process. S6: Extract 2n+2 feature indicators from the image obtained in step S5, including the magnitude of each cutting edge amplitude, relative ratio, minimum angle between cutting edges, and total energy. Use the model to train the model to obtain the machine tool milling cutter anomaly monitoring model for monitoring the milling process.

[0007] Preferably, in step S1, the specific sampling frequency fs = 1000Hz and the sampling duration is approximately 50s.

[0008] Preferably, in step S2, the signal is sampled for an entire cycle using the maximization method. The purpose is to improve the accuracy of subsequent amplitude and phase information calculations, i.e., removing the excess k data points at the end of the signal and setting an observation index. The specific formula is...

[0009] Preferably, in step S3, the formula used to obtain the amplitude and phase of the 1st to nth order frequency harmonics of the signal using the Fast Fourier Transform is: The formula for the amplitude of the i-th order turn frequency involved is amp i =|F(i×fn)|, the formula for the i-th order transition frequency is alpha i = angle(F(i×fn)).

[0010] Preferably, in step S4, the specific feature signal component reconstruction formula is:

[0011] Preferably, in step S6, 60% of the data involved is randomly selected from the samples as the training set, and the remaining samples are used as the test set. The weighted KNN model is used for training. The distance metric of the weighted KNN model is Euclidean distance, the neighbor parameter is set to 10, the square of the Euclidean distance is used as the weight, and the remaining 40% of the data is used as the test set.

[0012] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention mainly utilizes a visual feature signal technology scheme to characterize the variation law of milling cutter faults on the cutting edge, realizes the qualitative analysis of milling cutter faults, and provides a basis for subsequent quantitative detection of milling cutter faults; and utilizes the extension-reconstruction method to extract key order harmonics, remove most of the irrelevant information and noise interference in the signal, greatly improves the adaptability of the method to complex working environments, can more clearly and stably characterize the relative state of each cutting edge of the milling cutter, improves the anti-interference ability and sensitivity of the milling cutter anomaly monitoring method, and provides favorable technical support for the safe and stable operation of milling cutters. IV: Explanation of the attached drawings

[0013] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0014] Figure 1 This is a schematic diagram of the test bench structure.

[0015] Figure 2 These are actual photos of end mills in faulty condition (left: slightly worn end mill; right: chipped end mill).

[0016] Figure 3 The flowchart of this invention.

[0017] Figure 4 Original waveform diagram under normal state of the present invention.

[0018] Figure 5 The time-domain diagram (left) and frequency-domain diagram (right) of the truncated signal of this invention.

[0019] Figure 6 The present invention uses a maximization method to achieve an integer period sampling map.

[0020] Figure 7 The time domain (left) and frequency domain (right) plots of the positive signal of this invention.

[0021] Figure 8 Characteristic signal diagrams of the normal state and slight wear of the present invention.

[0022] Figure 9 Energy distribution diagram of the cutting edge during machining under various conditions.

[0023] Figure 10 Training results of the invention milling cutter anomaly monitoring model.

[0024] Figure 11 The test results of the end mill anomaly monitoring in this invention are shown in the figure. 5. Detailed Implementation Methods

[0025] Figure 3 As shown, a method for monitoring milling cutter anomalies based on cutting edge frequency energy visualization first utilizes the maximization method to achieve whole-cycle sampling to obtain more accurate amplitude and phase information. Then, a feature signal is reconstructed based on the obtained amplitude and phase. The feature signal is then aligned with a reference signal using the maximum correlation method and scaled to polar coordinates, resulting in a visualized cutting edge frequency energy monitoring scheme for monitoring milling cutter anomalies on machine tools. This invention was tested on a 3-DOF FANUC small machining center. The machine tool's electric spindle had a maximum speed of 24,000 rpm and a maximum load of 300 kg. KISLER accelerometers were used, magnetically mounted on the sides of the spindle. A simplified diagram of the test bench structure is shown below. Figure 1As shown, the experimental milling cutter was a 4-flute cutter. A total of five states were tested and verified: normal state, slightly worn state, severely worn state, chipped edge, and severely chipped edge. The operating speed was 3000 rpm, and the cutting depth was 0.2 mm. A photograph of the actual tool under fault conditions is shown below. Figure 2 As shown. The specific workflow is as follows.

[0026] Figure 3 As shown in the diagram, in step one, an accelerometer installed near the milling cutter spindle is used to collect single-channel vibration data of the milling cutter under various states. The sampling frequency is fs = 1000Hz, and the sampling time is approximately 50s. Figure 4 As shown. To increase the sample size, the collected data was divided into segments of 1 second each, resulting in several samples, denoted as samples 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 ... The truncated signal is as follows Figure 5 As shown; from Figure 5 The analysis revealed significant energy dissipation in the signal spectrum, with key harmonics, particularly frequency shift and higher-order harmonics, being quite ambiguous. This interferes with subsequent amplitude and phase extraction, easily leading to inaccurate feature extraction and ultimately misjudgment of the tool status. Therefore, further processing is required in the next step. Step two, extract... any sample For example, for To improve the accuracy of subsequent amplitude and phase information calculations, a maximization method is used to achieve full-cycle sampling of the signal, i.e., removing the extra k data points at the end of the signal. The observation index is set as shown in formula (1):

[0027]

[0028] Where, ω i For the frequency of interest, considering the need to calculate the amplitude and phase from the transition frequency to the blade frequency, ω is taken in this embodiment. i =50, 100, 150, 200Hz, i.e., 1st to 4th order rotation frequencies, where N is... The data length, and the index varies with k, as follows: Figure 6 As shown. From Figure 6 Find the k value corresponding to the maximum value in the example data. In the middle, when 192 points are deleted from the end, full-cycle sampling can be achieved, and the signal after deletion is denoted as x. * To further refine the spectrum, the signal x obtained through deletion correction will be... * Repeat the extension until the signal reaches at least three times the sampling frequency length. The example data was repeated four times, and the resulting signal is denoted as x. Figure 7 As shown. From Figure 7 It can be found that, compared to Figure 5The corrected signal has clearer amplitudes at key frequencies and its energy dissipation is effectively reduced.

[0029] Figure 3 As shown in the figure, in step three, the data from step two are used to obtain the amplitude and phase of the signal's first to nth order frequency harmonics using fast Fourier transform, as shown in formula (2):

[0030]

[0031] Where F(n) is the frequency domain signal, x(t) represents the t-th sampling point of signal x, and j is the imaginary unit; denoted as […]. <amp i alpha i >, then the amplitude of the i-th order frequency is as shown in formula (3):

[0032] amp i =|F(i×fn)| (3)

[0033] Where fn = 50Hz is the rotation frequency, i = 1, 2, ... n, and n is the number of cutting edges of the tool, n = 4;

[0034] The phase of the i-th order frequency is shown in formula (4):

[0035] alpha i =angle(F(i×fn)) (4)

[0036] Here, angle(·) represents calculating the complex phase angle, that is, calculating the angle between the complex vector and the positive direction of the x-coordinate axis. The amplitude and phase information obtained in this step are used in step four to accurately reconstruct the feature signal in order to obtain the fault information hidden in the signal.

[0037] Figure 3 As shown in the figure, in step four, the characteristic signal is reconstructed based on the amplitude and phase information obtained in step three. First, a sequence θ is constructed. hat To simplify subsequent calculations as much as possible while ensuring signal clarity, the sequence length is controlled to be one period. The sequence length is set to 500, as shown in formula (5):

[0038] θ hat = [0 2π / 499 … 2π] T (5)

[0039] in, The sequence length m = 500; considering that the magnitude of the frequency conversion amplitude has little effect on characterizing the difference in the blade state and will mask more detailed information in the feature signal, the feature signal component reconstruction starts from the second harmonic of the frequency conversion. As shown in formula (6):

[0040]

[0041] The feature signal component is reconstructed as a superposition of simple harmonic signals. To facilitate the next step of analysis, the minimum value of component is set to 0. Therefore, the feature signal extracted from the example data is as follows: Figure 8 As shown. The characteristic signal obtained in this step removes most of the environmental noise interference and the influence of other frequency components, and contains rich fault information, providing stable and reliable support for subsequent anomaly monitoring.

[0042] Figure 3 As shown in step five, the feature signal component obtained from the first set of data of the aforementioned five states is selected and labeled as comp0, serving as the baseline. The subsequent new components are then cross-correlated with comp0, as shown in formula (7):

[0043]

[0044] Find the k corresponding to the maximum value of R(k), and use it to perform time-shift correction on the new component to obtain the comp. k ; set comp = [comp0, comp1, ..., comp k Plotted on the same polar coordinate graph, such as Figure 9 As shown. From Figure 9 It can be easily observed that as the fault appears and worsens, the four "edges" that were originally clearly distinguishable in the edge energy distribution diagram gradually become blurred, and the overall amplitude increases. This phenomenon indicates that as the milling cutter fault occurs, the energy distribution of each cutting edge changes dramatically. By observing the dynamic changes in the edge energy distribution diagram, the condition of the milling cutter can be monitored.

[0045] Figure 3 As shown in step six, to assist in monitoring the milling cutter status, ten feature indicators are extracted from the image, including the magnitude of each cutting edge width, relative ratio, minimum inter-cutting angle, and total energy. 60% of the data is randomly selected from the samples as the training set, and the remaining samples are used as the test set. A weighted KNN model is then used for training. The weighted KNN model uses Euclidean distance as the distance metric, with a neighbor parameter set to 10, and uses the square of the Euclidean distance as the weight. The confusion matrix of the training result is shown below. Figure 10 As shown. Using the remaining 40% of the data for testing, the test results are as follows. Figure 11 As shown in the figure, the test accuracy is as high as 99.07%. To eliminate randomness, the above operation was repeated 10 times, and the average test accuracy was 98.45%. This proves that the method described in this invention can effectively identify abnormal conditions of milling cutters.

[0046] Figure 3As shown above, this invention utilizes visualized feature signals to characterize the variation law of milling cutter faults on the cutting edge, realizing qualitative analysis of milling cutter faults and providing a basis for subsequent quantitative detection of milling cutter faults. By using the extension-reconstruction method, key order harmonics are extracted, removing most of the irrelevant information and noise interference in the signal, greatly improving the adaptability of the method to complex working conditions, enabling a clearer and more stable characterization of the relative state of each cutting edge of the milling cutter, and improving the anti-interference ability and sensitivity of the milling cutter anomaly monitoring method.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for monitoring milling cutter anomalies based on cutting edge frequency energy visualization, characterized in that, First, the maximization method is used to achieve whole-cycle sampling to obtain more accurate amplitude and phase information. Then, the characteristic signal is reconstructed based on the obtained amplitude and phase. The maximum correlation method is used to align the characteristic signal with the reference signal and scale it to polar coordinates, resulting in a visualized cutting edge frequency energy monitoring scheme for machine tool milling cutter anomaly monitoring. The specific process includes the following steps: S1: Single-channel vibration data of the milling cutter under various working states is collected using a sensor installed near the spindle of the machine tool milling cutter, denoted as... S2: Correct the data collected in step S1, and use the maximization method to achieve integer-cycle sampling of the signal to obtain... and the obtained signal Repeat the extension process until the signal reaches three times the sampling frequency length; the resulting signal is denoted as... S3: Use the Fast Fourier Transform (FFT) to obtain signal 1 from the data obtained in step S2. n The amplitude and phase of the first-order turning harmonic are denoted as follows: ,in For the first i The amplitude of the first-order turn frequency, For the first i Phase of the first-order frequency transition, ,in n S4: Reconstruct the characteristic signal from the amplitude and phase information obtained in step S3, denoted as [missing information]. component ; S5: Step S4: Obtain the characteristic signal obtained by selecting a set of data. component Marked as As a benchmark, specifically, when a new set of signals is acquired, characteristic signals are obtained. component Then, calculate the new component and The maximum correlation coefficient, and based on this, the new component Perform time-shift correction to obtain and will When plotted on the same polar coordinate graph, it is called the cutting edge energy distribution graph. By observing the dynamic changes in the cutting edge energy distribution graph, the state of each cutting edge of the milling cutter during the milling process can be characterized. S6: Extract the magnitude of each blade amplitude, relative ratio, minimum inter-blade angle, and total energy from the image obtained in step S5. n Using two feature indicators and a weighted KNN model for training, a machine tool milling cutter anomaly monitoring model is obtained to monitor the milling process.

2. The milling cutter anomaly monitoring method based on cutting edge frequency energy visualization according to claim 1, characterized in that, In step S1, the specific sampling frequency fs =1000 Hz The sampling time is 50 seconds.

3. The milling cutter anomaly monitoring method based on cutting edge frequency energy visualization according to claim 1, characterized in that, In step S2, the maximization method is used to achieve integer-cycle sampling of the signal. The purpose is to improve the accuracy of subsequent amplitude and phase information calculations, that is, to remove redundant data at the end of the signal. k Set observation indicators for each data point. index The specific formula is , N The number of sampling points for the original signal. n This is the number of sampling point serial numbers. For the first i The angular frequency of the first-order harmonic. ,in Main spindle frequency The sampling frequency.

4. The milling cutter anomaly monitoring method based on cutting edge frequency energy visualization according to claim 1, characterized in that, In step S3, the signal from 1 to 2 is obtained using the Fast Fourier Transform. n The formulas used to apply to the amplitude and phase of a first-order turning frequency harmonic are: The first one involved i The formula for the amplitude of the first-order rotation frequency is: , No. i The formula for the first-order frequency is: , N For frequency conversion harmonic order, For the modulo operation of complex numbers, The frequency domain index corresponding to the spindle rotation frequency. , Let i be the phase of the i-th order reciprocating harmonic. Phase angle operation for complex numbers.

5. The milling cutter anomaly monitoring method based on cutting edge frequency energy visualization according to claim 1, characterized in that, In step S4, specific feature signals component The reconstruction formula is , i It is the harmonic order. n The number of cutting edges of the tool. The normalized phase angle variable represents the phase angle within one principal axis rotation cycle.

6. The milling cutter anomaly monitoring method based on cutting edge frequency energy visualization according to claim 1, characterized in that, In step S6, 60% of the data is randomly selected from the samples as the training set, and the remaining samples are used as the test set. The weighted KNN model is used for training. The distance metric used in the weighted KNN model is Euclidean distance, the neighbor parameter is set to 10, the square of the Euclidean distance is used as the weight, and the remaining 40% of the data is used as the test set.

Images