Self-adaptive denoising method suitable for milling cutter state monitoring signal

Abstract

The invention relates to the technical field of milling cutter state monitoring, and provides a self-adaptive denoising method suitable for a milling cutter state monitoring signal, and the method comprises the steps: decomposing an original complex signal into a plurality of IMF components in a milling process through employing a CEEMDAN algorithm, calculating the sample entropy of each IMF component, and carrying out the classification and recombination, obtaining a reconstructed signal and a high-complexity component; a sample entropy is used as a fitness function for measuring signal complexity, and an RIME algorithm is guided to automatically find a parameter combination capable of enabling the VMD decomposition effect to be optimal; performing VMD secondary decomposition on the high-complexity component by using the optimized parameter combination, calculating PCC of the decomposed IMF component and the original signal, and selecting an effective IMF component according to a correlation coefficient weight; and carrying out superposition reconstruction on the screened effective IMF components to generate a de-noised signal with a high signal-to-noise ratio, and taking the de-noised signal as a feature input signal for milling cutter state monitoring. The optimal denoising state is automatically adjusted according to the tool working condition, and the working condition adaptability of the system is improved.

Description

Technical Field

[0001] The embodiments of the present invention generally relate to the field of milling cutter condition monitoring technology, and more particularly to an adaptive noise reduction method suitable for milling cutter condition monitoring signals. Background Technology

[0002] In the field of intelligent manufacturing, the end mill, as a core component of milling, directly determines the quality of the machined workpiece, processing efficiency, and production costs. Monitoring and analyzing signals such as forces and vibrations generated during processing is a key technology for identifying and predicting the wear condition of end mills. However, the signal-to-noise ratio of signals collected on-site is extremely low, often mixed with background noise from multiple sources such as machine tool transmission, cutting fluid impact, and environmental vibration. This noise severely interferes with the extraction of feature information reflecting the true condition of the tool.

[0003] Currently, wavelet denoising, empirical mode decomposition (EMD), variational mode decomposition (VMD), and their improved algorithms are commonly used denoising methods in this field, but all have significant technical bottlenecks: wavelet denoising performance is highly dependent on the selection of wavelet basis functions and threshold rules, lacking adaptability; EMD and its variants are prone to problems such as mode aliasing and endpoint effects, resulting in poor decomposition performance in low signal-to-noise ratio environments, easily leading to distortion of useful signals or noise residue; and the VMD algorithm is limited by parameters. The defects are due to the need to rely on human experience to set them.

[0004] Therefore, developing a denoising method that can adaptively process nonlinear and nonstationary signals from milling and extract useful features with high fidelity in the context of strong noise is of great significance for improving the accuracy of tool condition monitoring. Summary of the Invention

[0005] To address the above issues, this invention employs a frost-ice optimization algorithm that automatically searches for the optimal parameter combination of VMD using sample entropy as the fitness function. It adopts a two-stage denoising strategy combining CEEMDAN preliminary purification with parameter optimization VMD secondary decomposition, and introduces a PCC correlation coefficient screening mechanism to identify effective components in the signal. This effectively supports the accurate judgment of tool wear status, significantly improves the reliability of monitoring results, thereby ensuring the stability of machining quality, improving production efficiency, and reducing the risk of tool wear and production interruption caused by misjudgment.

[0006] According to an embodiment of the present invention, an adaptive denoising method for milling cutter condition monitoring signals is provided.

[0007] In a first aspect of the invention, an adaptive denoising method for milling cutter condition monitoring signals is provided. The method includes: Step S01: Use the CEEMDAN algorithm to decompose the original complex signal during the milling process into multiple IMF components, calculate the sample entropy of each IMF component and classify and reorganize them to obtain the reconstructed signal and high-complexity components. Step S02: Using sample entropy as a fitness function to measure signal complexity, guide the RIME algorithm to automatically find the parameter combination that optimizes VMD decomposition. Step S03: Use the optimized parameter combination to perform VMD secondary decomposition on the high-complexity component, calculate the PCC of the decomposed IMF component and the original signal, and select the effective IMF component according to the correlation coefficient weight. Step S04: Superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

[0008] Furthermore, the specific steps of step S01 are as follows: Step S011: Collect raw signals during the milling process using a vibration acceleration sensor and force gauge installed on the machine tool spindle or worktable. ; Step S012: Use the CEEMDAN algorithm to process the original signal. Adaptive decomposition is performed, a process that involves iteratively adding adaptive white noise and extracting modes. The core iterative steps are as follows: , , in, No. The signal to be decomposed after adding noise. Noise figure Indicates the first First-order modal operator, It is Gaussian white noise. This represents the first-order intrinsic mode function of the extracted signal; iterative execution continues until the termination condition is met, ultimately obtaining a series of intrinsic mode function components. ; Step S013: Calculate the sample entropy of each IMF component. The simplified formula for calculating sample entropy is: , in, and Each under similar tolerance Below, the length is and The number of template pairs; using the quartile method, all IMFs are divided into three categories: high complexity components ( ), medium complexity components ( ) and low-complexity components ( ); Step S014: Superimpose the classified low-complexity components and medium-complexity components to form the reconstructed signal after the first purification. Meanwhile, the separated high-complexity components For subsequent processing.

[0009] Furthermore, the step of classifying all IMFs into three categories using the quartile method is as follows: 1) The components described in step S012 Convert to Reconstructed to dimension: ; 2) Matrix The Column and number The columns are denoted as vectors. and The distance between the two is denoted as ,calculate : , In the formula: ; 3) Let For a threshold value, Calculate the threshold condition. The number of them is denoted as And then with Comparison is recorded as : ; 4) Calculation average : ; 5) Increase the dimension to And repeat steps 1)-3), to get: ; 6) Sequence-like for: .

[0010] Furthermore, the specific steps of step S02 are as follows: Step S021: Parameter initialization, setting the population size and maximum number of iterations for the RIME algorithm. VMD parameters and upper and lower bounds of the search and And randomly initialize the positions of the frost particles. ,in ; Step S022: For each particle position The sample entropy value of the corresponding IMF component after VMD decomposition is calculated as the fitness value. Step S023: For each particle, generate a random number. ,like If the fitness value is less than the normalized fitness value of the current particle, then a local perturbation is applied to enhance the algorithm's fine-grained search capability. The formula for local fine-grained search is: ; Step S024: Compare the updated fitness values ​​of each particle and update the global optimal solution. That is, finding the minimum fitness value Parameter combination; if the maximum number of iterations is reached Or it meets the convergence condition Then the optimal parameter combination will be output. Otherwise, continue iterating.

[0011] Furthermore, the specific steps of step S03 are as follows: Step S031: Combine the obtained optimal parameters. Input VMD algorithm for highly complex signals Perform precise decomposition; Step S032: During the initialization phase, set the convergence tolerance to... Maximum number of iterations and initialize modal components. Center frequency and Lagrange multipliers ; Step S033: During the iterative update process, at the... In the next iteration, the three variables are updated sequentially: Modal components are updated according to the formula: , The center frequency is obtained through calculation: , The Lagrange multipliers are then updated accordingly: ; Step S034: After each iteration, check the convergence condition. Check if the conditions are met; if so, stop the iteration. Encapsulate the above process into a VMD module and set the parameter values ​​in step S032. Step S035: In order to effectively identify the feature information related to the tool wear state, the correlation coefficient index of each modal component obtained by decomposition in step S034 is quantitatively analyzed.

[0012] Furthermore, the PCC index calculation for each modal component is used to evaluate the correlation strength between each modal component and the original signal. The calculation steps are as follows: First, calculate the covariance: , Then calculate the standard deviation of the modal components from the original signal: , Then PCC is: , Modal components can be filtered based on PCC: when When the value is greater than 0.5, the component is considered to be strongly correlated with the original signal and contains the main signal characteristics; when the value is less than 0.3... When the correlation coefficient is ≤0.5, it indicates moderate correlation, containing some signal characteristics; when When the correlation coefficient is ≤0.3, it is a weak correlation, and the main component is noise.

[0013] Furthermore, the specific steps of step S04 are as follows: Step S041: Linearly superimpose the selected effective IMF components, i.e., select those with PCC values ​​greater than 0.5. The components, combined with the reconstructed signal obtained from CEEMDAN decomposition, generate a preliminary denoised signal: , in, For the set of indices of the valid components, For the first One effective modal component; Step S042: Calculate the signal-to-noise ratio between the original signal and the reconstructed signal. , In the formula: and These represent the power of the signal and the power of the noise, respectively. Indicates power, For signal length, The first of the original signals One sampling point; Step S043: Calculate the root mean square error: , In the formula: The original signal, The signal after denoising. The signal length; Step S044: Calculate the normalized correlation coefficient: , A comprehensive evaluation of the denoised signal is performed using multiple indicators.

[0014] In a second aspect of the invention, an adaptive noise reduction apparatus for milling cutter condition monitoring signals is provided. The apparatus includes: Signal decomposition module: Used to decompose the original complex signal during milling into multiple IMF components using the CEEMDAN algorithm, calculate the sample entropy of each IMF component and classify and reassemble it to obtain the reconstructed signal and high-complexity components. Parameter optimization module: Used to guide the RIME algorithm to automatically find the parameter combination that optimizes the VMD decomposition effect, using sample entropy as a fitness function to measure signal complexity. Component selection module: Used to perform VMD secondary decomposition on highly complex components using the optimized parameter combination, calculate the PCC of the decomposed IMF components and the original signal, and select effective IMF components based on the correlation coefficient weight. Component reconstruction module: used to superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

[0015] In a third aspect of the invention, an electronic device is provided. The electronic device includes a memory and a processor, the memory storing a computer program, the processor executing the program to implement the method according to a first aspect of the invention.

[0016] In a fourth aspect of the invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method according to a first aspect of the invention.

[0017] This invention utilizes the frost-ice optimization algorithm to automatically search for the optimal parameter combination of VMD using sample entropy as the fitness function. It employs a two-stage denoising strategy combining CEEMDAN preliminary purification with parameter optimization VMD secondary decomposition, and introduces a PCC correlation coefficient screening mechanism to identify effective components in the signal. This effectively supports the accurate judgment of tool wear status, significantly improves the reliability of monitoring results, thereby ensuring the stability of machining quality, improving production efficiency, and reducing the risk of tool wear and production interruption caused by misjudgment.

[0018] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of the present invention, nor is it intended to restrict the scope of the invention. Other features of the invention will become readily apparent from the following description.

[0019] The beneficial effects of this invention are: 1. By using the frost-ice optimization algorithm, the optimal parameter combination of VMD is automatically searched with sample entropy as the fitness function. It can autonomously adjust to the best noise reduction state according to different tools, materials and cutting conditions, which significantly improves the intelligence level and working condition adaptability of the system and overcomes the drawbacks of traditional methods that rely on manual experience and have high debugging costs. 2. A two-stage denoising strategy combining CEEMD initial purification with parameter optimization VMD secondary decomposition is adopted to achieve step-by-step noise removal. On this basis, a PCC correlation coefficient screening mechanism is introduced to ensure the consistency of waveform structure and accurately separate noise from effective modal components, thereby maximizing the preservation of weak state features under strong noise background and achieving the optimal balance between denoising effect and feature fidelity. 3. The signal reconstructed by this invention has a high signal-to-noise ratio and rich tool status information, providing high-quality input for intelligent diagnostic models. It is particularly suitable for complex working conditions in milling, effectively supporting accurate judgment of tool wear status, significantly improving the reliability of monitoring results, thereby ensuring the stability of machining quality, improving production efficiency, and reducing the risk of tool wear and production interruption caused by misjudgment. Attached Figure Description

[0020] The above and other features, advantages, and aspects of the various embodiments of the present invention will become more apparent from the accompanying drawings and the following detailed description. Wherein: Figure 1 A flowchart of an adaptive denoising method for milling cutter condition monitoring signals according to an embodiment of the present invention is shown; Figure 2 (a) A schematic diagram of IMF classification based on sample entropy according to an embodiment of the present invention is shown; Figure 2 (b) A schematic diagram of the fitness value iteration curve according to an embodiment of the present invention is shown; Figure 3 A flowchart of the RIME algorithm optimization according to an embodiment of the present invention is shown; Figure 4 A schematic diagram of VMD decomposition according to an embodiment of the present invention is shown; Figure 5 (a) shows the original signal diagram according to an embodiment of the present invention; Figure 5 (b) shows the original signal frequency domain diagram according to an embodiment of the present invention; Figure 5 (c) A waveform of wavelet-denoised signal according to an embodiment of the present invention is shown; Figure 5 (d) shows a frequency domain diagram of the wavelet-denoised signal according to an embodiment of the present invention; Figure 5(e) A diagram of the CEEMDAN denoised signal according to an embodiment of the present invention is shown; Figure 5 (f) shows a frequency domain diagram of the CEEMDAN denoised signal according to an embodiment of the present invention; Figure 5 (g) A denoised signal diagram of the method according to an embodiment of the present invention is shown; Figure 5 (h) shows a frequency domain diagram of the denoised signal according to an embodiment of the present invention; Figure 6 (a) shows the force evaluation diagram of the denoising algorithm TiAl according to an embodiment of the present invention; Figure 6 (b) shows the vibration evaluation diagram of the denoising algorithm TiAl according to an embodiment of the present invention; Figure 7 A block diagram of an apparatus for adaptive denoising of milling cutter condition monitoring signals according to an embodiment of the present invention is shown; Figure 8 A schematic diagram of an adaptive noise reduction device for milling cutter condition monitoring signals according to an embodiment of the present invention is shown. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] According to an embodiment of the present invention, an adaptive denoising method for milling cutter condition monitoring signals is proposed. The method uses a frost-ice optimization algorithm to automatically search for the optimal parameter combination of VMD (Modular Machine Decomposition) with sample entropy as the fitness function. It employs a two-stage denoising strategy combining CEEMDAN preliminary purification with parameter optimization VMD secondary decomposition. A PCC correlation coefficient screening mechanism is introduced to identify effective components in the signal. This effectively supports accurate judgment of tool wear status, significantly improves the reliability of monitoring results, thereby ensuring machining quality stability, improving production efficiency, and reducing the risk of tool wear and production interruption due to misjudgment.

[0023] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.

[0024] Figure 1 This is a schematic flowchart of an adaptive noise reduction method for milling cutter status monitoring signals according to an embodiment of the present invention. The method includes: Step S01: Use the CEEMDAN algorithm to decompose the original complex signal during the milling process into multiple IMF components, calculate the sample entropy of each IMF component and classify and reorganize them to obtain the reconstructed signal and high-complexity components. Step S02: Using sample entropy as a fitness function to measure signal complexity, guide the RIME algorithm to automatically find the parameter combination that optimizes VMD decomposition. Step S03: Use the optimized parameter combination to perform VMD secondary decomposition on the high-complexity component, calculate the PCC of the decomposed IMF component and the original signal, and select the effective IMF component according to the correlation coefficient weight. Step S04: Superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

[0025] It should be noted that although the operation of the method of the present invention has been described in a specific order in the above embodiments and figures, this does not require or imply that the operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0026] To provide a clearer explanation of the adaptive denoising method applicable to milling cutter status monitoring signals, a specific embodiment will be used for illustration below. However, it is worth noting that this embodiment is only for better illustrating the present invention and does not constitute an undue limitation on the present invention.

[0027] The following specific example will further illustrate the adaptive noise reduction method applicable to milling cutter condition monitoring signals: Step S01: Use the CEEMDAN algorithm to decompose the original complex signal during milling into multiple IMF components, calculate the sample entropy of each IMF component, and classify and reassemble them to obtain the reconstructed signal. and high complexity components .

[0028] Step S011: Collect raw signals during the milling process using a vibration acceleration sensor and force gauge installed on the machine tool spindle or worktable. .

[0029] In this embodiment, the experiment used a CAMUMC850 ultrasonic-assisted three-axis machining center; the tool used was an EH1301L four-groove carbide end mill with a cutting diameter of 10mm, a radius of 0.5mm, and a cutting length of 25mm, for vertical milling; the workpiece used was a γ-TiAl alloy with a diameter of 74.8mm and a length of 199mm; it was mounted in a worktable fixture during the experiment, and wet milling was performed using cutting fluid; during the experiment, vibration signals were monitored by an IGTech A26F100T01C system, and cutting forces were measured using a Kistler 5167A three-axis force gauge. All signals were sampled at a frequency of 20kHz, and the original signal... The data is acquired and stored in real time online via a data acquisition card. The acquired data covers six channels of signals, including three-axis cutting force and three-axis vibration. The specific milling parameters are shown in Table 1. Table 1

[0030] Step S012: Use the CEEMDAN algorithm to process the original signal. Adaptive decomposition is performed, a process that involves iteratively adding adaptive white noise and extracting modes. The core iterative steps are as follows: , , in, No. The signal to be decomposed after adding noise. Noise figure Indicates the first First-order modal operator, It is Gaussian white noise. This represents the first-order intrinsic mode function of the extracted signal; iterative execution continues until the termination condition is met, ultimately obtaining a series of intrinsic mode function components. ; Step S013: Calculate the sample entropy of each IMF component. The simplified formula for calculating sample entropy is: , in, and Each under similar tolerance Below, the length is and The number of template pairs; using the quartile method, all IMFs are divided into three categories: high complexity components ( ), medium complexity components ( ) and low-complexity components ( ).

[0031] The specific steps for classifying all IMFs into three categories are as follows: 1. The components described in step S012 Convert to Reconstructed to dimension: ; 2. Matrix The Column and number The columns are denoted as vectors. and The distance between the two is denoted as ,calculate : , In the formula: ; 3. Let For a threshold value, Calculate the threshold condition. The number of them is denoted as And then with Comparison is recorded as : ; 4. Calculation average : ; 5. Increase the dimension to And repeat steps 1-3 to get: ; 6. Sequence-like for: , Sample Entropy The larger the value of , the higher the complexity of the time series; conversely, the smaller the value of sample entropy. A smaller value indicates a simpler time series structure and stronger regularity. The average sample entropy under different parameter combinations is calculated, and its quartiles (25th, 50th, and 75th) are used to classify IMFs. , .

[0032] For example, attached Figure 2(a) shows that a vibration signal, after being decomposed by CEEMDAN adaptive decomposition, was broken down into 7 independent IMFs. Analysis based on sample entropy results shows that each component exhibits significant differences under different scale parameters (m, r). These components were classified into three categories using the quartile method: Co-IMF1, Co-IMF2, and Co-IMF3, with Co-IMF1 having a quartile value of 1.197. Referring to the criteria in Table 1, this component was identified as a high-complexity signal component.

[0033] Step S014: Superimpose the classified low-complexity components and medium-complexity components to form the reconstructed signal after the first purification. Meanwhile, the separated high-complexity components For subsequent processing.

[0034] In this embodiment, step S014 will attach Figure 2 (a) The low-complexity Co-IMF3 and medium-complexity Co-IMF2 components after classification are superimposed to form the reconstructed signal after the first purification. Meanwhile, the highly complex Co-IMF1 component (mainly containing high-frequency noise) was separated. , for subsequent processing.

[0035] Step S02: Using sample entropy as a fitness function to measure signal complexity, guide the RIME algorithm to automatically find the parameter combination that optimizes VMD decomposition. .

[0036] Step S021: Parameter initialization, setting the population size of the RIME algorithm to 5 and the maximum number of iterations. VMD parameters and upper and lower bounds of the search and And randomly initialize the positions of the frost particles. ,in ; Step S022: For each particle position (representing a set) (Combined), calculate the sample entropy value of the corresponding IMF component after VMD decomposition as the fitness value. Figure 3 The diagram illustrates the particle position update process, which is shown below: 1. For any combination of parameters Perform VMD decomposition to obtain IMF components The VMD objective function is: , In the formula: After decomposition One modal component; The number of modal components ( ); This refers to the set of center frequencies corresponding to the modal functions. For the Dijkstra function; This is a convolution operation; ; 2. For each IMF component, set the embedding dimension. Time delay Calculate its sample entropy : , Then calculate the number of matches. , that is, subsequence and The Euclidean distance is less than the threshold. ( Pick The number of samples (0.15 times the standard deviation). And calculate the normalized cross-match probability. Finally, calculate the sample entropy. ; 3. Construct a fitness function based on sample entropy. The fitness function is used to quantify parameter combinations. The corresponding VMD decomposition effect is defined as the sum of the entropies of all IMF samples after decomposition (the lower the overall complexity, the better the decomposition effect). The specific calculation is as follows: , 4. Based on the current iteration number Total number of iterations The ratio is used to calculate the adhesion coefficient. When random number At that time, the particle positions are adjusted according to the frost particle position update formula: , In the formula, The updated position of the particle. For Frost Ice Population The first optimal crystal in the middle One particle, The degree of adhesion is a random number in the range (0,1). A random number in the range (-1, 1); and These are represented as the upper and lower bounds of the particle, respectively. Indicates the direction of particle movement; As environmental factors, Used to control the number of segments in the step function; Step S023: For each particle, generate a random number. ,like Less than the current particle's normalized fitness value, i.e. Then, a local perturbation is applied to enhance the algorithm's fine-grained search capability. The local fine-grained search formula is: ; Step S024: Compare the updated fitness values ​​of each particle and update the global optimal solution. That is, finding the minimum fitness value Parameter combination; if the maximum number of iterations is reached Or it meets the convergence condition Then the optimal parameter combination will be output. Otherwise, continue iterating.

[0037] The attachment is obtained through step S024. Figure 2 (b) Results: The RIME algorithm converged after 70 iterations, and the final fitness value stabilized at 2.164, at which point the results were obtained. , The optimal combination of parameters.

[0038] Step S03: Utilize the optimized parameter combination For high complexity components Perform VMD secondary decomposition, calculate the PCC (Pearson correlation coefficient) between the decomposed IMF components and the original signal, and select the effective IMF components based on the correlation coefficient weights. The specific steps are as follows.

[0039] Step S031: Combine the obtained optimal parameters , Input VMD algorithm for highly complex signals Perform precise decomposition; Step S032: During the initialization phase, set the convergence tolerance to... Maximum number of iterations and initialize modal components. Center frequency and Lagrange multipliers ; Step S033: During the iterative update process, at the... In the next iteration, the three variables are updated sequentially: Modal components are updated according to the formula: , The center frequency is obtained through calculation: , The Lagrange multipliers are then updated accordingly: ; Step S034: After each iteration, check the convergence condition. Check if the conditions are met; if so, stop the iteration. Encapsulate the above process into a VMD module and set the parameter values ​​from step S032. Final algorithm output. Modal components arranged in descending order of center frequency Each modal component has a finite bandwidth and a specific center frequency, thus enabling precise characterization of the signal's time-frequency distribution. For example... Figure 4 As shown, three variational eigenmode components were successfully decomposed. .

[0040] Step S035: To effectively identify feature information related to tool wear state, the correlation coefficients of each modal component obtained from step S034 are quantitatively analyzed. The PCC index is used to evaluate the correlation strength between each modal component and the original signal. The calculation steps are as follows: First, calculate the covariance: , Then calculate the standard deviation of the modal components from the original signal: , The correlation coefficient is then: , Modal components can be screened based on correlation coefficients: when When the value is greater than 0.5, the component is considered to be strongly correlated with the original signal and contains the main signal characteristics; when the value is less than 0.3... When the correlation coefficient is ≤0.5, it indicates moderate correlation, containing some signal characteristics; when When the correlation coefficient is ≤0.3, it indicates a weak correlation, with noise being the main component. Specific calculation results show: and The PCC values ​​of the components and the original signal were 0.76 and 0.62, respectively, showing a significant correlation. The PCC value of the component is only 0.15. Considering the generally low correlation characteristic of high-frequency noise components in signal processing, it is determined that... The noise is the dominant component, and it is removed.

[0041] Step S04: Superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

[0042] Step S041: Linearly superimpose the selected effective IMF components, i.e., select those with significant correlation. , Components, i.e., PCC values ​​greater than 0.5. The components, combined with the reconstructed signal obtained from CEEMDAN decomposition, generate a preliminary denoised signal: , in, For the set of indices of the valid components, For the first One effective modal component.

[0043] Figure 5 The vibration signal denoising effect was demonstrated: In signal processing and analysis during milling, sensor noise exhibits significant high-frequency characteristics, while the effective signal components and their energy are mainly distributed in the low-frequency region composed of the Tooth Passing Frequency (TPF) and its harmonics. The fundamental frequency and its integer harmonics components concentrate on characterizing the dynamic characteristics of the cutting process. Based on the process parameters given in Table 1 of step S01: tool speed n = 3000 r / min, number of tool teeth N = 4, the theoretical value of the Tooth Passing Frequency can be calculated using the following formula: , Using the 200Hz over-tooth frequency as a key reference, through in-depth analysis of the original vibration time-domain signal and its frequency-domain energy distribution, such as... Figure 5 (a) Figure 5 As shown in (b), a significant energy concentration characteristic can be observed in the high-frequency range of 6200Hz to 7400Hz in the original signal, specifically manifested as 31st, 33rd, and 37th times the over-tooth frequency components. This high-frequency noise may originate from sensor electronic interference or resonance phenomena in the mechanical system. Notably, the energy intensity of this high-frequency noise is significantly higher than that of the low-frequency effective signals reflecting tool wear characteristics (such as the 200Hz fundamental frequency and its 6th and 8th harmonics). The presence of high-frequency noise severely interferes with the extraction of the periodic contact characteristics between the tool and the workpiece during the cutting process, thereby significantly reducing the accuracy of model feature recognition.

[0044] The three-level decomposition method based on the traditional db4 wavelet basis is used to denoise the processed signal, which can effectively separate noise from the effective components, such as... Figure 5 (c) Figure 5 As shown in (d), the algorithm first decomposes the noisy signal into wavelet coefficients of different frequency bands, and uses a hard threshold to zero out the high-frequency coefficients, thereby suppressing high-frequency interference; at the same time, it retains the core frequency bands in the low-frequency band that reflect the characteristics of tool wear (such as the 200 Hz fundamental frequency and 6 Hz). This method uses wavelet coefficients (such as TPF and other low-order harmonics) to achieve noise reduction. While simple and fast, it effectively eliminates high-frequency random noise by truncating the wavelet coefficients using a global threshold. However, when there is partial overlap between the high-frequency band of the signal and the effective characteristic frequency band, it is prone to misjudging weak characteristic signals as noise and removing them.

[0045] Compared to traditional wavelet denoising, CEEMDAN is more adaptable and stable in nonlinear and non-stationary signal processing. Through adaptive decomposition, it breaks down the signal into independent IMF components, effectively filtering out high-frequency noise and avoiding signal distortion caused by over-filtering in traditional methods. Figure 5 (e) Figure 5 As shown in (f), after processing, the waveform of the signal is closer to the original signal and is smoother overall, while retaining effective information well. Time-frequency analysis shows that the reconstructed signal effectively retains most of the effective components. However, the noise energy in the high-frequency part is still relatively high.

[0046] This method combines the high integrity and low reconstruction error of CEEMDAN decomposition, the global search capability of the RIME algorithm, and the parameter optimization mechanism of VMD. It adaptively determines the optimal parameter combination of VMD by using sample entropy as the fitness function. The noise reduction effect is as follows: Figure 5 (g) Figure 5 As shown in (h), the algorithm significantly improves the denoising effect while preserving the effective signal.

[0047] The decomposition process of force signals is the same as that of vibration signals, and will not be repeated here.

[0048] Step S042: Calculate the signal-to-noise ratio (SNR) between the original signal and the reconstructed signal: , In the formula: and These represent the power of the signal and the power of the noise, respectively. Indicates power, For signal length, The first of the original signals One sampling point.

[0049] Step S043: Calculate the root mean square error (RMSE): , In the formula: The original signal, The signal after denoising. This is the signal length.

[0050] Step S044: Calculate the normalized correlation coefficient (NCC): , A comprehensive evaluation of the denoised signal is performed using multiple indicators.

[0051] Figure 6 The performance evaluation of three denoising algorithms was presented. Based on the SNR value, the algorithm of this invention performed best, significantly higher than the other two algorithms, indicating its strongest noise suppression capability, especially its outstanding denoising effect in the high-frequency region. Figure 5 (g) Figure 5 (h) As shown in the green box diagram. Regarding RMSE values, both CEEMDAN and the algorithm of this invention are small and similar, indicating that both have high denoising accuracy. In NCC value analysis, CEEMDAN retains 14% of the vibration signal during processing. The critical harmonic TFP has a high NCC value; while in force signal processing, the NCC values ​​of both CEEMDAN and the algorithm of this invention are close to 1, indicating that both algorithms can effectively preserve the signal's information and achieve high signal fidelity. Overall, the algorithm of this invention demonstrates significant advantages in noise reduction, signal smoothness, and preservation of critical information, making it an efficient and reliable noise reduction method.

[0052] Based on the same inventive concept, this invention also proposes an adaptive noise reduction device suitable for milling cutter condition monitoring signals. The implementation of this device can be found in the implementation of the method described above, and repeated details will not be elaborated further. Figure 7 As shown, the device 100 includes: Signal decomposition module 101: Used to decompose the original complex signal during milling into multiple IMF components using the CEEMDAN algorithm, calculate the sample entropy of each IMF component and classify and reorganize it to obtain the reconstructed signal and high-complexity components. Parameter optimization module 102: Used to guide the RIME algorithm to automatically find the parameter combination that makes the VMD decomposition effect optimal, using sample entropy as a fitness function to measure signal complexity. Component selection module 103: used to perform VMD secondary decomposition on highly complex components using the optimized parameter combination, calculate the PCC of the decomposed IMF components and the original signal, and select effective IMF components based on the correlation coefficient weight. Component reconstruction module 104: used to superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

[0053] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0054] like Figure 8As shown, the device includes a central processing unit (CPU), which can perform various appropriate actions and processes based on computer program instructions stored in read-only memory (ROM) or loaded from storage units into random access memory (RAM). The RAM can also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0055] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0056] The processing unit executes the various methods and processes described above, such as method steps S01 to S04. For example, in some embodiments, method steps S01 to S04 may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of method steps S01 to S04 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute method steps S01 to S04 by any other suitable means (e.g., by means of firmware).

[0057] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload programmable logic devices (CPLDs), and so on.

[0058] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0059] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0060] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the invention. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0061] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. An adaptive noise reduction method for milling cutter condition monitoring signals, characterized in that, The method includes: Step S01: Use the CEEMDAN algorithm to decompose the original complex signal during the milling process into multiple IMF components, calculate the sample entropy of each IMF component and classify and reorganize them to obtain the reconstructed signal and high-complexity components. Step S02: Using sample entropy as a fitness function to measure signal complexity, guide the RIME algorithm to automatically find the parameter combination that optimizes VMD decomposition. Step S03: Use the optimized parameter combination to perform VMD secondary decomposition on the high-complexity component, calculate the PCC of the decomposed IMF component and the original signal, and select the effective IMF component according to the correlation coefficient weight. Step S04: Superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

2. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 1, characterized in that, The specific steps of step S01 are as follows: Step S011: Collect raw signals during the milling process using a vibration acceleration sensor and force gauge installed on the machine tool spindle or worktable. ; Step S012: Use the CEEMDAN algorithm to process the original signal. Adaptive decomposition is performed, a process that involves iteratively adding adaptive white noise and extracting modes. The core iterative steps are as follows: ， ， in, No. The signal to be decomposed after adding noise. Noise figure Indicates the first First-order modal operator, It is Gaussian white noise. This represents the first-order intrinsic mode function of the extracted signal; iterative execution continues until the termination condition is met, ultimately obtaining a series of intrinsic mode function components. ; Step S013: Calculate the sample entropy of each IMF component. The simplified formula for calculating sample entropy is: ， in, and Each under similar tolerance Below, the length is and The number of template pairs; using the quartile method, all IMFs are divided into three categories: high complexity components ( ), medium complexity components ( ) and low-complexity components ( ); Step S014: Superimpose the classified low-complexity components and medium-complexity components to form the reconstructed signal after the first purification. Meanwhile, the separated high-complexity components For subsequent processing.

3. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 2, characterized in that, The steps for classifying all IMFs into three categories using the quartile method are as follows: 1) The components described in step S012 Convert to Reconstructed to dimension: ； 2) Matrix The Column and number The columns are denoted as vectors. and The distance between the two is denoted as ,calculate : ， In the formula: ; 3) Let For a threshold value, Calculate the threshold condition. The number of them is denoted as And then with Comparison is recorded as : ； 4) Calculation average : ； 5) Increase the dimension to And repeat steps 1)-3), to get: ； 6) Sequence-like for: 。 4. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 1, characterized in that, The specific steps of step S02 are as follows: Step S021: Parameter initialization, setting the population size and maximum number of iterations for the RIME algorithm. VMD parameters and upper and lower bounds of the search and And randomly initialize the positions of the frost particles. ,in ; Step S022: For each particle position The sample entropy value of the corresponding IMF component after VMD decomposition is calculated as the fitness value. Step S023: For each particle, generate a random number. ,like If the fitness value is less than the normalized fitness value of the current particle, then a local perturbation is applied to enhance the algorithm's fine-grained search capability. The formula for local fine-grained search is: ； Step S024: Compare the updated fitness values ​​of each particle and update the global optimal solution. That is, finding the minimum fitness value Parameter combinations; If the maximum number of iterations is reached Or it meets the convergence condition Then the optimal parameter combination will be output. Otherwise, continue iterating.

5. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 1, characterized in that, The specific steps of step S03 are as follows: Step S031: Combine the obtained optimal parameters. Input VMD algorithm for highly complex signals Perform precise decomposition; Step S032: During the initialization phase, set the convergence tolerance to... Maximum number of iterations and initialize modal components. Center frequency and Lagrange multipliers ; Step S033: During the iterative update process, at the... In the next iteration, the three variables are updated sequentially: Modal components are updated according to the formula: ， The center frequency is obtained through calculation: ， The Lagrange multipliers are then updated accordingly: ； Step S034: After each iteration, check the convergence condition. Check if the conditions are met; if so, stop the iteration. Encapsulate the above process into a VMD module and set the parameter values ​​in step S032. Step S035: In order to effectively identify the feature information related to the tool wear state, the correlation coefficient index of each modal component obtained by decomposition in step S034 is quantitatively analyzed.

6. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 5, characterized in that, The PCC index calculation for each modal component is used to evaluate the correlation strength between each modal component and the original signal. The calculation steps are as follows: First, calculate the covariance: ， Then calculate the standard deviation of the modal components from the original signal: ， Then PCC is: ， Modal components can be filtered based on PCC: when When the value is greater than 0.5, the component is considered to be strongly correlated with the original signal and contains the main signal characteristics; when the value is less than 0.3... When the correlation coefficient is ≤0.5, it indicates moderate correlation, containing some signal characteristics; when When the correlation coefficient is ≤0.3, it is a weak correlation, and the main component is noise.

7. The adaptive noise reduction method for milling cutter condition monitoring signals according to claim 1, characterized in that, The specific steps of step S04 are as follows: Step S041: Linearly superimpose the selected effective IMF components, i.e., select those with PCC values ​​greater than 0.

5. The components, combined with the reconstructed signal obtained from CEEMDAN decomposition, generate a preliminary denoised signal: ， in, For the set of indices of the valid components, For the first One effective modal component; Step S042: Calculate the signal-to-noise ratio between the original signal and the reconstructed signal. ， In the formula: and These represent the power of the signal and the power of the noise, respectively. Indicates power, For signal length, The first of the original signals One sampling point; Step S043: Calculate the root mean square error: ， In the formula: The original signal, The signal after denoising. The signal length; Step S044: Calculate the normalized correlation coefficient: ， A comprehensive evaluation of the denoised signal is performed using multiple indicators.

8. An adaptive noise reduction device for milling cutter condition monitoring signals, characterized in that, The device implements the method as described in any one of claims 1 to 7, comprising: Signal decomposition module: Used to decompose the original complex signal during milling into multiple IMF components using the CEEMDAN algorithm, calculate the sample entropy of each IMF component and classify and reassemble it to obtain the reconstructed signal and high-complexity components. Parameter optimization module: Used to guide the RIME algorithm to automatically find the parameter combination that optimizes the VMD decomposition effect, using sample entropy as a fitness function to measure signal complexity. Component selection module: Used to perform VMD secondary decomposition on highly complex components using the optimized parameter combination, calculate the PCC of the decomposed IMF components and the original signal, and select effective IMF components based on the correlation coefficient weight. Component reconstruction module: used to superimpose and reconstruct the selected effective IMF components to generate a denoised signal with a high signal-to-noise ratio, which serves as the feature input signal for end mill status monitoring.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 7.

Images