Method and apparatus for denoising mechanical vibration signal, and medium and device

Abstract

The present invention provides a method and apparatus for denoising a mechanical vibration signal, and a medium and a device. The method comprises the following steps: using a generalized variational mode decomposition (GVMD) algorithm to process a vibration signal acquired in real time so as to obtain a plurality of intrinsic mode function sub-signals; clustering the plurality of intrinsic mode function sub-signals to divide the plurality of intrinsic mode function sub-signals into a plurality of categories; and assigning different Gaussian scale coefficients to different categories of intrinsic mode function sub-signals, and using a local means (NLM) algorithm to respectively filter the intrinsic mode function sub-signals and sum same, so as to obtain a denoised vibration signal. In the present invention, a vibration signal is decomposed into a plurality of intrinsic mode function sub-signals, different Gaussian scale coefficients are assigned to different categories of intrinsic mode function sub-signals, and the intrinsic mode function sub-signals are respectively filtered using the local means (NLM) algorithm, thereby improving the denoising effect.

Description

A method, device, medium and equipment for denoising mechanical vibration signals

[0001] This application claims priority to application number CN2023112410773, filed September 25, 2023, entitled “Method and device for denoising mechanical vibration signals,” filed by applicant. The entire contents of the aforementioned application are incorporated herein by reference in their entirety. Technical Field

[0002] The present invention belongs to the technical field of vibration signal processing, and in particular relates to a denoising processing method, device, medium and equipment for mechanical vibration signals.

[0003] Background of the Invention

[0004] Mechanical systems generate strong vibration signals during operation, which contain a wealth of useful information. For example, vibration signals can be measured at the rolling bearings of diesel engines, directly reflecting the engine's characteristic state and operating conditions. Therefore, industry researchers generally consider analyzing vibration signals collected from rolling bearings to accurately diagnose engine faults. However, bearing vibration signals measured in the field are easily drowned out by strong background noise. Therefore, to accurately extract bearing fault characteristics, it is necessary to effectively identify and filter out noise components in the measurement signal.

[0005] Currently, commonly used signal denoising methods include EMD (Empirical Mode Decomposition), VMD (Variational Mode Decomposition), and EWT (Empirical Wavelet Transform). Bearing vibration signals are nonlinear, non-stationary, and random signals, and their time-frequency characteristics vary randomly over time. Therefore, if these signal denoising methods are directly applied to rolling bearing noise, two problems may arise: first, modal aliasing may occur when processing vibration signals with a strong noise background, making it difficult to detect internal fault signature information; second, signal decomposition can easily distort the signal. Therefore, it is necessary to propose a more effective method to solve the signal denoising problem based on the characteristics of the measured rolling bearing vibration signal.

[0006] Summary of the Invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides a method, device, medium and equipment for denoising a mechanical vibration signal.

[0008] In order to achieve the above objectives, the present invention adopts the following technical solutions.

[0009] In a first aspect, the present invention provides a method for denoising a mechanical vibration signal, comprising the following steps:

[0010] The generalized variational mode decomposition algorithm GVMD is used to process the vibration signal acquired in real time to obtain multiple eigenmode function sub-signals;

[0011] Clustering the multiple intrinsic mode function sub-signals to divide the multiple intrinsic mode function sub-signals into a plurality of categories;

[0012] Different Gaussian scaling coefficients are assigned to different categories of intrinsic mode function sub-signals. The local mean algorithm (NLM) is used to filter each intrinsic mode function sub-signal separately and sum them to obtain the denoised vibration signal.

[0013] Furthermore, before performing GVMD processing on the vibration signal, the method further includes preprocessing the vibration signal as follows:

[0014] The vibration signal is represented as a time series X = {x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t = 1, 2, ..., N, and N is the number of data;

[0015] Perform Hampel filtering on X as follows: z(t) = |x(t)-median(X)| / (1.4826×median(Y))

[0016] Where z(t) is the filtered value of x(t), median(X) and median(Y) are the medians of X and Y, respectively, and Y = {|x(t) - median(X)|}, t = 1, 2, ..., N;

[0017] If z(t) is greater than the set threshold, then z(t) is abnormal data;

[0018] The abnormal data is replaced by a value obtained by interpolation calculation.

[0019] Furthermore, the Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows:

[0020] Determine the number of clusters K and extract the IMF sub-signals from the multiple eigenmode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, and j = 1, 2, ..., K;

[0021] Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows:

[0022] Where, d(IMF i (t),h j (t)) is the IMF i (t) to h j Euler distance of (t), i = 1, 2, ..., M, j = 1, 2, ..., K;

[0023] Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters to obtain K categories;

[0024] Calculate the average signal of all eigenmode function sub-signals in each category and use this average signal as the new cluster center;

[0025] Iterate by repeatedly executing the above steps. If J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is:

[0026] In the formula, min() means finding the minimum value.

[0027] Furthermore, K=3, and the Kmeans clustering algorithm is used to divide multiple intrinsic mode function sub-signals into three categories. The main signal component of the first category is the vibration signal, the main signal component of the second category is the mixed signal of the vibration signal and the noise signal, and the main signal component of the third category is the noise signal.

[0028] Furthermore, the three categories of identification methods include:

[0029] Calculate the maximum amplitude value of the eigenmode function sub-signal of each category;

[0030] Sort the three categories in descending order according to the maximum amplitude value, from front to back they are the first category, the second category and the third category.

[0031] Furthermore, the method of filtering the intrinsic mode function sub-signal IMF using the local mean algorithm NLM includes:

[0032] The weighted coefficient of IMF(i) at the i-th data collection moment is calculated as follows:

[0033] Where ω(i,j) is the weighted coefficient, i.e., the similarity of similar blocks centered on i and j, δ kis the Gaussian scaling coefficient corresponding to the intrinsic mode function sub-signal of the k-th category, k = 1, 2, 3, δ1 < δ2 < δ3, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered on i;

[0034] Calculate the IMF of the filtered signal as follows: S (i)

[0035] Where, Ω i is the search window centered at i.

[0036] Furthermore, δ1=1, δ2=4, δ3=8.

[0037] In a second aspect, the present invention provides a denoising device for a mechanical vibration signal, comprising:

[0038] The signal decomposition module is used to process the vibration signal acquired in real time using the generalized variational mode decomposition algorithm GVMD to obtain multiple eigenmode function sub-signals;

[0039] a signal classification module, configured to cluster the plurality of intrinsic mode function sub-signals and divide the plurality of intrinsic mode function sub-signals into a plurality of categories;

[0040] The NLM processing module is used to assign different Gaussian scaling coefficients to different categories of intrinsic mode function sub-signals, and uses the local mean algorithm NLM to filter each intrinsic mode function sub-signal separately and sum them to obtain the denoised vibration signal.

[0041] Furthermore, the device further includes a preprocessing module for preprocessing the vibration signal in the following manner:

[0042] The vibration signal is represented as a time series X = {x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t = 1, 2, ..., N, and N is the number of data;

[0043] Perform Hampel filtering on X as follows: z(t) = |x(t)-median(X)| / (1.4826×median(Y))

[0044] Where z(t) is the filtered value of x(t), median(X) and median(Y) are the medians of X and Y, respectively, and Y = {|x(t) - median(X)|}, t = 1, 2, ..., N;

[0045] If z(t) is greater than the set threshold, then z(t) is abnormal data;

[0046] The abnormal data is replaced by a value obtained by interpolation calculation.

[0047] Furthermore, the Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows:

[0048] Determine the number of clusters K and extract the IMF sub-signals from the multiple eigenmode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, and j = 1, 2, ..., K;

[0049] Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows:

[0050] Where, d(IMF i (t),h j (t)) is the IMF i (t) to h j Euler distance of (t), i = 1, 2, ..., M, j = 1, 2, ..., K;

[0051] Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters to obtain K categories;

[0052] Calculate the average signal of all eigenmode function sub-signals in each category and use this average signal as the new cluster center;

[0053] Iterate by repeatedly executing the above steps. If J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is:

[0054] In the formula, min() means finding the minimum value.

[0055] In a third aspect, the present invention provides a readable medium, wherein the readable medium includes an execution instruction. When a processor of an electronic device executes the execution instruction, the electronic device executes the method as described above.

[0056] In a fourth aspect, the present invention provides an electronic device comprising a processor and a memory storing execution instructions. When the processor executes the execution instructions stored in the memory, the processor executes the method as described above.

[0057] Compared with the prior art, the present invention has the following beneficial effects.

[0058] The present invention processes a vibration signal acquired in real time using a generalized variational modal decomposition algorithm to obtain a plurality of intrinsic mode function sub-signals, clusters the plurality of intrinsic mode function sub-signals, divides them into a plurality of categories, assigns different Gaussian scaling coefficients to the intrinsic mode function sub-signals of different categories, and uses a local mean algorithm (NLM) to filter each intrinsic mode function sub-signal separately and sum them to obtain a denoised vibration signal, thereby achieving denoising of the vibration signal. The present invention decomposes the vibration signal into a plurality of intrinsic mode function sub-signals, assigns different Gaussian scaling coefficients to the intrinsic mode function sub-signals of different categories, and uses a local mean algorithm (NLM) to filter each intrinsic mode function sub-signal separately, thereby significantly improving the denoising effect.

[0059] BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the existing technical solutions, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0061] FIG1 is a flow chart of a method for denoising a mechanical vibration signal according to an embodiment of the present invention.

[0062] Figure 2 is a schematic diagram of the vibration signal waveform, and the small squares in the figure are the identified abnormal data.

[0063] FIG3 is a schematic diagram of multiple IMF waveforms decomposed from a vibration signal using GVMD.

[0064] FIG4 is a block diagram of a denoising apparatus for a mechanical vibration signal according to an embodiment of the present invention.

[0065] FIG5 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention.

[0066] Modes for Carrying Out the Invention

[0067] In order to make the technical objectives, technical solutions and beneficial effects of the present invention clearer, the specific embodiments of the present invention are clearly and completely described below in conjunction with the accompanying drawings. The specific embodiments described are only part of the embodiments of the present invention, rather than all the embodiments. Based on the specific embodiments of the present invention, other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of the present invention.

[0068] FIG1 is a flow chart of a method for denoising a mechanical vibration signal according to an embodiment of the present invention, comprising the following steps:

[0069] Step 101: Processing the vibration signal acquired in real time using the generalized variational mode decomposition algorithm GVMD to obtain multiple eigenmode function sub-signals;

[0070] Step 102: clustering the plurality of intrinsic mode function sub-signals to divide the plurality of intrinsic mode function sub-signals into a plurality of categories;

[0071] Step 103 : assigning different Gaussian scaling coefficients to different types of intrinsic mode function sub-signals, and using the local mean algorithm (NLM) to filter and sum each intrinsic mode function sub-signal to obtain a denoised vibration signal.

[0072] In this embodiment, step 101 is mainly used to decompose the vibration signal into multiple eigenmode function sub-signals. In this embodiment, the vibration signal is processed by using the GVMD algorithm to obtain multiple eigenmode function sub-signals.

[0073] The GVMD algorithm constructs a constrained optimization problem for each sub-signal and incorporates a fixed-frequency decomposition solution to solve the constructed optimization problem, thereby completing multi-scale fixed-frequency decomposition in the frequency domain. Therefore, the GVMD algorithm mainly consists of two parts: constructing the constrained optimization problem and solving the constrained optimization problem. First, assuming that the vibration signal X is composed of M sub-signals, the optimization problem constructed by GVMD can be expressed as:

[0074] In the formula is the partial derivative with respect to time; t is time; δ(*) is the Dirac function; u k is the kth sub-signal; ω k Corresponding to the center frequency of the kth sub-signal. First, the above constrained optimization problem is transformed into an unconstrained optimization problem using the multiplier method:

[0075] Where, α k represents the scale parameter; λ k is the Lagrange multiplier. Then the alternating direction method of the multiplier is introduced to solve the above unconstrained optimization problem, thereby decomposing the main problem into three sub-problems: sub-signal u k The minimization problem, the center frequency ω k Minimization problem, Lagrange multiplier λ k The specific formula is as follows:

[0076] Where the parameter τ is the dual lifting step size and n is the iteration step.

[0077] Finally, all three sub-problems are solved by k 、ω k and λ k Perform iterative updates to solve. The resulting sub-signals and center frequencies are:

[0078] Where, and are the Fourier transforms of x, u, and λ, respectively. The real part of the inverse Fourier transform is the sub-signal u k , which is also the intrinsic mode function IMF. Figure 3 is a schematic diagram of the IMF obtained by decomposing the vibration signal using the GVMD method.

[0079] In this embodiment, step 102 is primarily used to classify multiple intrinsic mode function sub-signals. The purpose of classifying the intrinsic mode function sub-signals is to group sub-signals with similar characteristics into the same category and to filter sub-signals of different categories in a targeted manner to achieve ideal filtering and denoising effects. Numerous classification algorithms are available in the prior art, and this embodiment does not limit the specific classification algorithm. A specific classification algorithm will be described in subsequent embodiments.

[0080] In this embodiment, step 103 primarily filters the different categories of intrinsic mode function sub-signals. This embodiment employs the local mean (NLM) algorithm to filter each intrinsic mode function sub-signal separately. The NLM algorithm can effectively reduce noise while minimizing signal distortion. To achieve ideal filtering and denoising effects, this embodiment assigns different Gaussian scaling coefficients to different categories of sub-signals. Each sub-signal is filtered and summed to obtain a denoised vibration signal.

[0081] As an optional embodiment, before performing GVMD processing on the vibration signal, the method further includes preprocessing the vibration signal as follows:

[0082] The vibration signal is represented as a time series X = {x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t = 1, 2, ..., N, and N is the number of data;

[0083] Perform Hampel filtering on X as follows: z(t) = |x(t)-median(X)| / (1.4826×median(Y)) (1)

[0084] Where z(t) is the filtered value of x(t), median(X) and median(Y) are the medians of X and Y, respectively, and Y = {|x(t) - median(X)|}, t = 1, 2, ..., N;

[0085] If z(t) is greater than the set threshold, then z(t) is abnormal data;

[0086] The abnormal data is replaced by a value obtained by interpolation calculation.

[0087] This embodiment provides a technical solution for preprocessing vibration signals. During the on-site collection of vibration signals, measurement errors and interference signals are likely to occur, resulting in abnormal data points in the signal, as shown in Figure 2. This embodiment uses the Hampel filter method to process the vibration signal and identify abnormal data points. As a mean filter, the Hampel filter can design a probability distribution model for known data, and abnormal data points will be identified because they differ greatly from the expectations of the probability model. The calculation formula for Hampel filtering the vibration signal is as shown in formula (1). By comparing the filtered data with the set threshold, if the value of a data point exceeds the set threshold, the data point is considered to be an abnormal data point. In order to eliminate the adverse effects of abnormal data points, they are eliminated and replaced by the values ​​obtained by the interpolation method. There are many existing interpolation algorithms, the simplest of which is the linear interpolation method. In order to obtain higher accuracy, the fourth-order polynomial fitting interpolation of the least squares method can be used.

[0088] As an optional embodiment, a Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows:

[0089] Determine the number of clusters K and extract the IMF sub-signals from the multiple eigenmode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, and j = 1, 2, ..., K;

[0090] Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows:

[0091] Where, d(IMF i (t),h j (t)) is the IMF i (t) to h jEuler distance of (t), i = 1, 2, ..., M, j = 1, 2, ..., K;

[0092] Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters to obtain K categories;

[0093] Calculate the average signal of all eigenmode function sub-signals in each category and use this average signal as the new cluster center;

[0094] Iterate by repeatedly executing the above steps. If J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is:

[0095] In the formula, min() means finding the minimum value.

[0096] This embodiment provides a specific classification algorithm. This embodiment uses the Kmeans clustering algorithm to cluster multiple intrinsic mode function sub-signals. The Kmeans clustering algorithm is the most commonly used clustering algorithm and is also an unsupervised learning algorithm. The core idea of ​​the Kmeans algorithm is: first determine the number of clusters K, randomly select K initial cluster centers C from the data set, and then calculate the number of clusters K. i , calculate the remaining data and the cluster center C i Euclidean distance, find the cluster center C closest to the data i , and assign the data to the cluster center C i The corresponding cluster is then calculated. The average value of the data objects in each cluster is used as the new cluster center, and the next iteration is performed until the cluster center no longer changes or the maximum number of iterations is reached. This embodiment differs from the general clustering operation in that the cluster center is not a data point, but all the data points contained in a sub-signal. Of course, the classification object is not a single data point, but rather a sub-signal as a whole.

[0097] As an optional embodiment, K=3, and the Kmeans clustering algorithm is used to divide multiple intrinsic mode function sub-signals into three categories, the main signal component of the first category is the vibration signal, the main signal component of the second category is the mixed signal of the vibration signal and the noise signal, and the main signal component of the third category is the noise signal.

[0098] This embodiment limits the number of clusters K. This embodiment sets K to 3. After performing the Kmeans clustering algorithm, three categories of intrinsic mode function sub-signals are obtained. The three categories are basically classified according to signal energy or intensity: the first category is the sub-signal with the strongest signal, whose main signal component is the vibration signal; the second category is the sub-signal with a relatively strong signal, whose main signal component is a mixed signal of the vibration signal and the noise signal; and the third category is the sub-signal with the weakest signal, whose main signal component is the noise signal.

[0099] As an optional embodiment, the three categories of identification methods include:

[0100] Calculate the maximum amplitude value of the eigenmode function sub-signal of each category;

[0101] Sort the three categories in descending order according to the maximum amplitude value, from front to back they are the first category, the second category and the third category.

[0102] This embodiment provides a method for identifying three categories. Although the previous embodiment has obtained three categories, the signal components of each category are unknown, that is, it is unknown which category is the first category, which category is the second category, and which category is the third category. This embodiment calculates the maximum signal amplitude value of all sub-signals in each category and sorts them by size, from large to small, into the first category, the second category, and the third category. Of course, comparison can also be made based on the signal mean of each category cluster center. In short, the three categories are distinguished based on signal strength.

[0103] As an optional embodiment, a method for filtering the intrinsic mode function sub-signal IMF using a local mean algorithm NLM includes:

[0104] The weighted coefficient of IMF(i) at the i-th data collection moment is calculated as follows:

[0105] Where ω(i,j) is the weighted coefficient, i.e., the similarity of similar blocks centered on i and j, δ k is the Gaussian scaling coefficient corresponding to the intrinsic mode function sub-signal of the k-th category, k = 1, 2, 3, δ1 < δ2 < δ3, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered on i;

[0106] Calculate the IMF of the filtered signal as follows: S (i)

[0107] Where, Ω i is the search window centered at i.

[0108] This embodiment provides a technical solution for filtering the eigenmode function sub-signal using the NLM algorithm. i The weighted mean of the internal IMF sub-signals is used to filter the IMF sub-signals. The calculation formulas are shown in (5) and (6). Search window Ω i The larger the width, the better the denoising effect, but the longer the calculation time. ω(i,j) is the weighting coefficient, and the calculation formula is shown in formula (4). The calculation process of ω(i,j) can be understood as: comparing the areas around similar blocks centered on i and j. The higher the similarity, the larger ω(i,j). δ k is the Gaussian scaling coefficient, which affects the smoothness of the denoised signal. P is the coefficient variable that affects ω(i,j), affecting the number of similar blocks. λ represents the search step size, that is, the step size of the search window movement. Δ is the search block centered on i. Ω i , δ k Both δ and P will affect the results of NLM, but it can be clearly seen from formula (4) that the Gaussian scaling coefficient δ k The influence of is greater than the other two parameters. In order to obtain a good filtering effect, this embodiment uses Gaussian scaling coefficients of different sizes for different signal components, δ k This is the Gaussian scaling coefficient corresponding to the k-th category IMF sub-signal, where k = 1, 2, and 3. Since the signal energies of the first, second, and third categories decrease in sequence, δ1 < δ2 < δ3. The Gaussian scaling coefficient of the third category is the largest, which can most effectively suppress the noise signal.

[0109] As an optional embodiment, δ1=1, δ2=4, and δ3=8.

[0110] This embodiment provides a set of specific values ​​of Gaussian scaling coefficients for three categories. δ1, δ2, and δ3 are 1, 4, and 8 respectively, satisfying δ1<δ2<δ3. It is worth noting that this embodiment only provides a preferred implementation method and does not exclude or deny other feasible implementation methods, such as δ k Different from the value in this embodiment.

[0111] FIG4 is a schematic diagram of a denoising device for a mechanical vibration signal according to an embodiment of the present invention, the device comprising:

[0112] The signal decomposition module 11 is used to process the vibration signal acquired in real time using the generalized variational mode decomposition algorithm GVMD to obtain multiple eigenmode function sub-signals;

[0113] a signal classification module 12, configured to cluster the plurality of intrinsic mode function sub-signals and divide the plurality of intrinsic mode function sub-signals into a plurality of categories;

[0114] The NLM processing module 13 is used to assign different Gaussian scaling coefficients to different types of intrinsic mode function sub-signals, and use the local mean algorithm NLM to filter and sum each intrinsic mode function sub-signal to obtain a denoised vibration signal.

[0115] The device of this embodiment can be used to implement the technical solution of the method embodiment shown in Figure 1. Its implementation principle and technical effects are similar and will not be described in detail here. The same is true for the following embodiments and will not be described in detail.

[0116] As an optional embodiment, the device further includes a preprocessing module for preprocessing the vibration signal in the following manner:

[0117] The vibration signal is represented as a time series X = {x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t = 1, 2, ..., N, and N is the number of data;

[0118] Perform Hampel filtering on X as follows: z(t) = |x(t)-median(X)| / (1.4826×median(Y)) (1)

[0119] Where z(t) is the filtered value of x(t), median(X) and median(Y) are the medians of X and Y, respectively, and Y = {|x(t) - median(X)|}, t = 1, 2, ..., N;

[0120] If z(t) is greater than the set threshold, then z(t) is abnormal data;

[0121] The abnormal data is replaced by a value obtained by interpolation calculation.

[0122] As an optional embodiment, a Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows:

[0123] Determine the number of clusters K and extract the IMF sub-signals from the multiple eigenmode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, and j = 1, 2, ..., K;

[0124] Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows:

[0125] Where, d(IMF i (t),h j (t)) is the IMF i (t) to h j Euler distance of (t), i = 1, 2, ..., M, j = 1, 2, ..., K;

[0126] Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters to obtain K categories;

[0127] Calculate the average signal of all eigenmode function sub-signals in each category and use this average signal as the new cluster center;

[0128] Iterate by repeatedly executing the above steps. If J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is:

[0129] In the formula, min() means finding the minimum value.

[0130] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

[0131] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Among them, the memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage, etc. Of course, the electronic device may also include hardware required for other services.

[0132] The processor, network interface, and memory can be interconnected via an internal bus, such as an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. These buses can be categorized as address buses, data buses, and control buses. For ease of illustration, FIG5 shows only one bidirectional arrow, but this does not imply that there is only one bus or only one type of bus.

[0133] Memory is used to store execution instructions. Specifically, execution instructions are computer programs that can be executed. Memory can include internal memory and non-volatile memory, and provides execution instructions and data to the processor.

[0134] In one possible implementation, a processor reads corresponding execution instructions from a non-volatile memory into an internal memory and then executes them. Alternatively, the processor may obtain corresponding execution instructions from another device to logically form a blood vessel segmentation device. The processor executes the execution instructions stored in the memory to implement the mechanical vibration signal denoising method provided in any embodiment of the present application.

[0135] The denoising method for mechanical vibration signals provided in the embodiment shown in FIG1 of the present application can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0136] The steps of the method disclosed in the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method. Industrial Applicability

[0137] The present invention provides a method, device, medium, and equipment for denoising mechanical vibration signals. By decomposing the vibration signal into multiple intrinsic mode function sub-signals, assigning different Gaussian scaling coefficients to different types of intrinsic mode function sub-signals, and filtering each intrinsic mode function sub-signal using a local mean (NLM) algorithm, the denoising effect is significantly improved. This method can be combined with computer technology to form signal processing devices and equipment for mass production, enabling rapid application in systems or scenarios with high demands for vibration signal analysis.

Claims

1. A method for denoising a mechanical vibration signal, characterized in that: The following steps are involved: The generalized variational mode decomposition algorithm GVMD is used to process the vibration signal acquired in real time to obtain multiple eigenmode function sub-signals; Clustering the multiple intrinsic mode function sub-signals to divide the multiple intrinsic mode function sub-signals into a plurality of categories; Different Gaussian scaling coefficients are assigned to different types of intrinsic mode function sub-signals, and the local mean algorithm (NLM) is used to filter and sum each intrinsic mode function sub-signal to obtain the denoised vibration signal.

2. The method for denoising a mechanical vibration signal according to claim 1, characterized in that: Before performing GVMD processing on the vibration signal, the method further includes preprocessing the vibration signal as follows: The vibration signal is represented as a time series X={x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t=1, 2, ..., N, and N is the number of data; Perform Hampel filtering on X as follows: z(t)=|x(t)-median(X)| / (1.4826×median(Y)) Where z(t) is the value after filtering x(t), median(X) and median(Y) are the medians of X and Y respectively, Y = {|x(t) - median(X)|}, t = 1, 2, ..., N; If z(t) is greater than the set threshold, then z(t) is abnormal data; The abnormal data is replaced by a numerical value obtained by interpolation calculation.

3. The method for denoising a mechanical vibration signal according to claim 1, characterized in that: The Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows: Determine the number of clusters K, and extract the IMF sub-signals from the multiple intrinsic mode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, j = 1, 2, ..., K; Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows: Where, d(IMF i (t),h j (t)) is IMF i (t) to h j (t), i = 1, 2, ..., M, j = 1, 2, ..., K; Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters, and K categories are obtained; Calculate the average signal of all intrinsic mode function sub-signals in each category, and use this average signal as the new cluster center; By repeatedly executing the above steps, if J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is: In the formula, min() means finding the minimum value.

4. The method for denoising a mechanical vibration signal according to claim 3, characterized in that: K=3, and the Kmeans clustering algorithm is used to divide multiple intrinsic mode function sub-signals into three categories. The main signal component of the first category is the vibration signal, the main signal component of the second category is the mixed signal of the vibration signal and the noise signal, and the main signal component of the third category is the noise signal.

5. The method for denoising a mechanical vibration signal according to claim 4, characterized in that: The 3 categories of identification methods include: Calculate the maximum amplitude value of each category of eigenmode function sub-signal; Sort the three categories in descending order according to the maximum amplitude value, from the front to the back, they are the first category, the second category, and the third category.

6. The method for denoising a mechanical vibration signal according to claim 5, characterized in that: The method of filtering the intrinsic mode function sub-signal IMF using the local mean algorithm NLM includes: The weighted coefficient of IMF(i) at the i-th data collection moment is calculated as follows: In the formula, ω(i,j) is the weighting coefficient, that is, the similarity of similar blocks centered on i and j, δ k For the kth class The Gaussian scaling coefficients corresponding to other eigenmode function sub-signals are k=1, 2, 3, δ1<δ2<δ3, P is the influence coefficient related to the number of similar blocks, λ is the search step size, and Δ is the search block centered on i; The IMF of the filtered signal is calculated as follows: S (i) In the formula, Ω i is the search window centered at i.

7. The method for denoising a mechanical vibration signal according to claim 6, characterized in that: δ1=1, δ2=4, δ3=8.

8. A denoising device for a mechanical vibration signal, characterized in that: include: The signal decomposition module is used to process the vibration signal acquired in real time by using the generalized variational mode decomposition algorithm GVMD to obtain multiple eigenmode function sub-signals; A signal classification module, used for clustering the multiple intrinsic mode function sub-signals and dividing the multiple intrinsic mode function sub-signals into several categories; The NLM processing module is used to assign different Gaussian scaling coefficients to different categories of intrinsic mode function sub-signals, and use the local mean algorithm NLM to filter and sum each intrinsic mode function sub-signal to obtain a denoised vibration signal.

9. The denoising device for mechanical vibration signals according to claim 8, characterized in that: The device also includes a preprocessing module, which is used to preprocess the vibration signal in the following way: The vibration signal is represented as a time series X={x(t)}, where x(t) is the vibration signal amplitude at the t-th data collection moment, t=1, 2, ..., N, and N is the number of data; Perform Hampel filtering on X as follows: z(t)=|x(t)-median(X)| / (1.4826×median(Y)) Where z(t) is the value after filtering x(t), median(X) and median(Y) are the medians of X and Y respectively, Y = {|x(t) - median(X)|}, t = 1, 2, ..., N; If z(t) is greater than the set threshold, then z(t) is abnormal data; The abnormal data is replaced by a numerical value obtained by interpolation calculation.

10. The denoising device for mechanical vibration signals according to claim 8, characterized in that: The Kmeans clustering algorithm is used to cluster the multiple intrinsic mode function sub-signals, and the method is as follows: Determine the number of clusters K, and extract the IMF sub-signals from the multiple intrinsic mode functions i (t) randomly selects K eigenmode function sub-signals as the center h of the initial sample cluster j (t), where t is the data acquisition time, t = 1, 2, ..., N, N is the number of data, i = 1, 2, ..., M, M is the number of eigenmode function sub-signals, j = 1, 2, ..., K; Calculate each IMF i (t) to each cluster center h j The Euler distance of (t) is as follows: Where, d(IMF i (t),h j (t)) is IMF i (t) to h j (t), i = 1, 2, ..., M, j = 1, 2, ..., K; Based on the nearest Euler distance criterion, the M intrinsic mode function sub-signals are respectively assigned to the centers of K clusters, and K categories are obtained; Calculate the average signal of all intrinsic mode function sub-signals in each category, and use this average signal as the new cluster center; By repeatedly executing the above steps, if J SSE If the value is less than the set threshold, the iteration is stopped and the final K categories are obtained. SSE The calculation formula is: In the formula, min() means finding the minimum value.

11. A readable medium, characterized in that: The readable medium includes execution instructions. When a processor of an electronic device executes the execution instructions, the electronic device executes the method according to any one of claims 1 to 7.

12. An electronic device, characterized in that: The electronic device includes a processor and a memory storing execution instructions. When the processor executes the execution instructions stored in the memory, the processor executes the method according to any one of claims 1 to 7.

Images