Hydroelectric generating set vibration signal denoising method and system based on BPO-NLM

Abstract

The invention discloses a hydroelectric generating set vibration signal de-noising method and system based on BPO-NLM, and relates to the technical field of signal de-noising processing, and the method comprises the following steps: based on a non-local mean filtering algorithm NLM, through a Bayesian parameter optimization algorithm BPO, optimizing the NLM; and carrying out noise reduction processing on the collected vibration signals of the hydroelectric generating set by utilizing the optimized NLM. The processing efficiency and accuracy of the hydroelectric generating set vibration signals are improved, and powerful support is provided for fault diagnosis and state monitoring of the hydroelectric generating set.

Description

Technical Field

[0001] The present invention relates to the technical field of signal noise reduction processing, and in particular to a method and system for denoising vibration signals of a hydropower unit based on BPO-NLM. Background Art

[0002] The vibration signal of a hydropower unit is an important indicator for assessing the unit's operating status. However, the signal acquisition process is susceptible to interference from equipment operating noise, making it difficult for the sampled signal to truly reflect the unit's status. Therefore, denoising is a key step in obtaining accurate information. Traditional Fourier transforms are suitable for analyzing stationary linear signals, but the nonlinear and nonstationary nature of hydropower unit vibration signals requires a more appropriate denoising method.

[0003] The non-local means (NLM) algorithm is an emerging denoising algorithm, initially applied primarily to two-dimensional image noise removal, but it has also shown potential for one-dimensional signal denoising. The NLM algorithm estimates the true signal by calculating the weighted average of all similar blocks. Key parameters include the half-width of the target and similar structured blocks, the half-width of the search region, and the filter parameters, all of which have a significant impact on the denoising effect. However, the current NLM algorithm parameter settings still rely heavily on human experience and lack objectivity, limiting its further development and application.

[0004] The current hydropower unit vibration signal denoising process has the following major defects:

[0005] First, the traditional Fourier transform method is not very effective when processing nonlinear and non-stationary vibration signals, as it is more suitable for analyzing linear signals with stationary patterns. This makes it difficult for traditional methods to accurately extract the true information from the vibration signals of hydropower units.

[0006] Secondly, although wavelet analysis can process non-stationary signals, its parameter settings lack adaptability and require manual parameter adjustment, which increases the complexity and uncertainty of the operation. At the same time, the effectiveness of wavelet analysis is also limited by the rationality of parameter selection, which is not conducive to achieving efficient denoising.

[0007] Furthermore, while the non-local means (NLM) algorithm has shown potential for denoising one-dimensional signals, the setting of its key parameters still relies heavily on human experience. This makes it difficult to ensure the stability and consistency of the denoising effect of the NLM algorithm in practical applications, limiting its further development and application.

[0008] Specifically, the NLM algorithm's key parameters, such as the search domain half-width K, the building block half-width P, and the bandwidth parameter λ, have a crucial impact on denoising effectiveness. However, current research results lack objectivity, and parameter settings often rely on manual experience, making it difficult to achieve optimal denoising results in practical applications.

[0009] In summary, the current hydropower unit vibration signal denoising method has deficiencies in process and performance, and a more efficient, stable and adaptive denoising algorithm is needed to improve the accuracy and reliability of signal processing. Summary of the Invention

[0010] In order to solve the above problems, the purpose of the present invention is to provide an improved non-local mean (NLM) denoising algorithm, namely the BPO-NLM denoising algorithm, which aims to solve the problem that the parameter setting of the traditional NLM algorithm relies on human experience and the denoising effect is limited when processing the vibration signals of hydropower units.

[0011] In order to achieve the above technical objectives, the present application provides a method for denoising the vibration signal of a hydropower unit based on BPO-NLM, comprising the following steps:

[0012] Based on the non-local mean filter algorithm NLM, the NLM is optimized by the Bayesian parameter optimization algorithm BPO;

[0013] The optimized NLM is used to perform noise reduction on the collected vibration signals of the hydropower unit.

[0014] Preferably, when optimizing the NLM, the decisive parameters of the non-local mean filtering algorithm NLM are optimized by using a Bayesian parameter optimization algorithm BPO.

[0015] Preferably, when optimizing the decisive parameters, the decisive parameters include: filter bandwidth parameter, parameter K and parameter P, wherein parameter K affects the computational amount and computational time of the NLM algorithm, and parameter P affects the number of similar structural blocks found during the algorithm operation.

[0016] Preferably, when optimizing the decisive parameters, an objective function of a Bayesian parameter optimization algorithm BPO is constructed based on the spectral kurtosis value, the peak signal-to-noise ratio and the mean square error, so as to optimize the decisive parameters.

[0017] Preferably, when the optimized NLM is used for noise reduction processing, a vibration signal of the hydropower unit consisting of a real vibration signal and external interference noise is obtained, and the weighted average of all similar blocks is calculated using the optimized NLM to estimate the real signal.

[0018] The present invention discloses a hydropower unit vibration signal denoising system based on BPO-NLM, which is used in the above-mentioned hydropower unit vibration signal denoising method based on BPO-NLM, comprising:

[0019] Data acquisition module, used to collect vibration signals of hydropower units consisting of real vibration signals and external interference noise;

[0020] The noise reduction processing module is used to perform noise reduction processing on the vibration signal of the hydropower unit through the optimized NLM, wherein the NLM is optimized based on the non-local mean filter algorithm NLM through the Bayesian parameter optimization algorithm BPO.

[0021] Preferably, the noise reduction processing module is further used to optimize the decisive parameters of the non-local mean filtering algorithm NLM through the Bayesian parameter optimization algorithm BPO, wherein the decisive parameters include: filter bandwidth parameter, parameter K and parameter P, parameter K affects the calculation amount and calculation time of the NLM algorithm, and parameter P affects the number of similar structural blocks found during the operation of the algorithm.

[0022] Preferably, the noise reduction processing module is further used to construct an objective function of the Bayesian parameter optimization algorithm BPO based on the spectral kurtosis value, peak signal-to-noise ratio and mean square error, so as to optimize the decisive parameters, wherein the optimized NLM is used to calculate the weighted average of all similar blocks to estimate the true signal.

[0023] The present invention discloses the following technical effects:

[0024] The present invention introduces a Bayesian parameter optimization method, which adaptively determines the key parameters of the NLM algorithm by constructing an optimization objective function (combining spectral kurtosis and peak signal-to-noise ratio). This method not only avoids the interference of human experience, but also improves the denoising efficiency and accuracy of the algorithm.

[0025] Compared with the traditional NLM algorithm, the BPO-NLM denoising algorithm can significantly reduce the root mean square error (RMSE) of the denoised signal when processing hydropower unit vibration signals with different signal-to-noise ratios, while improving the signal-to-noise ratio (SNR) of the signal, thereby more effectively enhancing the effective components of the signal and suppressing the noise components. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 It is a schematic flow chart of the method described in the present invention;

[0028] Figure 2 is the analog signal of the present invention, wherein (a) is a noise-free analog signal, and (b) is a noisy analog signal;

[0029] Figure 3is a comparison between the noise-free / noisy signal and the noisy denoised signal of the present invention;

[0030] Figure 4 is a comparison between NLM denoised signals using different algorithm parameters according to the present invention;

[0031] Figure 5 It is a comparison of the denoising effectiveness between the traditional NLM and BPO-NLM (NLM based on Bayesian parameter optimization) described in the present invention;

[0032] Figure 6 This is a comparison of the BPO-NLM denoising effects under different signal-to-noise ratio conditions described in the present invention;

[0033] Figure 7 is the root mean square error and signal-to-noise ratio of the denoised signal under different signal-to-noise ratios described in the present invention;

[0034] Figure 8 It is the optimal parameter of the BPO-NLM algorithm under different signal-to-noise ratio conditions described in the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0036] like Figures 1-8 As shown, the present invention provides an improved non-local means (NLM) denoising algorithm, namely the BPO-NLM denoising algorithm, which specifically includes the following contents:

[0037] A method for preprocessing and denoising vibration signals of hydropower units based on NLM algorithm.

[0038] The non-local mean filtering algorithm exploits the presence of numerous similar structures in an image and achieves image denoising by performing a weighted average of these similar structures. This is why the NLM algorithm is widely used for denoising two-dimensional images. However, these similar characteristics also exist in one-dimensional signals, and the NLM algorithm has been successfully applied to the processing of one-dimensional rolling bearing vibration signals. Therefore, the NLM algorithm can also be used to process hydropower unit vibration signals.

[0039] The present invention uses the NLM algorithm to pre-process and reduce noise on the vibration signal of the hydropower unit to facilitate subsequent signal feature extraction. The principle of the NLM algorithm is as follows:

[0040] Assume that the actual noisy hydropower unit vibration acquisition signal y is the superposition of the real vibration signal u and the external interference noise n, that is:

[0041] y=u+n (1)

[0042] The NLM algorithm calculates the weighted average of all similar blocks to estimate the true signal u*(s), that is:

[0043]

[0044] Where: u*(s) represents the true signal value estimated at position s; s represents the position index of a pixel or signal point; D(s) represents the search window centered on pixel s; t is the position index of another pixel within the search window D(s); y(t) represents the noisy signal value actually collected at position t; w(s, t) is the weight function, which represents the similarity weight between pixel t and pixel s; Z(s) is the normalization factor, also known as the partition function.

[0045]

[0046] Where: ω(s, t) represents the weight, which refers to the similarity between two search blocks centered on t and s, and must satisfy 0≤ω(s, t) / Z(s)≤1 and ∑ t The basic condition of ω(s,t)=1. Among them, λ is the filter bandwidth parameter, which affects the smoothness of the denoised signal; Δ is the search block centered on t, and K is half the length of the Δ area. The parameter K affects the computational complexity and computational time of the NLM algorithm; L Δ =2P+1 is the neighborhood block centered on s, and the parameter P affects the number of similar structure blocks found during the algorithm operation.

[0047] Parameters λ, K, and P are the decisive parameters of the NLM algorithm, which greatly affect the denoising effect of the algorithm on one-dimensional signals. However, the setting of these parameters still depends largely on human experience.

[0048] Bayesian parameter optimization method: Bayesian optimization is used to optimize hyperparameters in machine learning models. Its essence is to estimate the optimal value of a function based on existing sampling points when the function equation is unknown. The problem expression considering extreme values ​​is

[0049]

[0050] The decision function f(x) is in the range Rd In internal optimization, x represents the decision vector in d-dimensional space. Bayesian optimization only needs to specify the objective function to be optimized (a generalized function that only needs to specify input and output), and updates the posterior distribution of the objective function by continuously adding sample points. Bayesian optimization has unique advantages over traditional network global search and random search. Bayesian uses Gaussian process fitting to optimize the objective function f(x), while traditional algorithms require that the objective function is a known mathematical model, and the calculation is simple and does not involve human intervention. Many practical problems do not meet these prerequisites, resulting in weak adaptability of traditional optimization algorithms. However, Bayesian optimization has no rigid requirements for the objective function f(x) and takes into account the continuous update of the prior information in the previous step, while traditional methods do not consider previous parameter information. When dealing with non-convex problems, traditional methods are prone to fall into local optimal solutions, while Bayesian optimization is still very effective for non-convex problems and can obtain global optimal solutions.

[0051] Bayesian optimization consists of two parts: a Gaussian process and an acquisition function. The Gaussian process is used to model an optimization function of unknown form. After obtaining the posterior probability of the function through the Gaussian process, the acquisition function samples new points based on certain metrics of this posterior probability. These new points are added to the observed data as a reference for the next calculation, resulting in a more accurate posterior probability.

[0052] Compared to other hyperparameter optimization algorithms, such as grid search, random search, genetic algorithms, and particle swarm optimization, Bayesian optimization requires fewer initial sample points and has higher optimization efficiency, making it more suitable for model hyperparameter adjustment scenarios. Parameter selection has always been a research focus in NLM denoising algorithms, but current research results lack objectivity and rely heavily on manual experience, which has hindered the further development and application of NLM algorithms.

[0053] To better improve the NLM denoising algorithm and achieve effective denoising in high-noise environments, the present invention proposes incorporating Bayesian optimization into the algorithm's key parameter settings, seeking optimal solutions for the search domain half-width K, the building block half-width P, and the bandwidth parameter λ. Bayesian optimization hyperparameters require a given objective function, and the choice of the objective function directly determines the parameter selection results, thereby influencing the algorithm's denoising performance. The current work proposes combining spectral kurtosis and peak signal-to-noise ratio (PSNR) as the objective function for Bayesian optimization. Kurtosis is often used in fault diagnosis of rotating machinery, but as an overall metric, kurtosis cannot reflect the impact of changes in characteristic signal components. To overcome the shortcomings of kurtosis, the prior art has used the concept of spectral kurtosis and further studied it, providing a detailed explanation, including its definition, algorithm flow, and application conditions. Spectral kurtosis is sensitive to periodic transient pulse signals caused by faults in rolling bearing vibration signals, and the PSNR is a commonly used metric for evaluating denoising effectiveness. Therefore, the present invention combines the two as the objective function for Bayesian optimization hyperparameters, effectively suppressing noise components in the original signal and enhancing the useful signal components. The optimization objective function is

[0054] -αmax(KYRT)-βPSNR+ηMEAN (5)

[0055] Where KYRT represents the spectral kurtosis value, PSNR is the peak signal-to-noise ratio, MEAN is the mean square error, α is the coefficient multiplied by max(KYRT), β is the coefficient multiplied by PSNR (peak signal-to-noise ratio), and η is the coefficient multiplied by MEAN (mean).

[0056] The experimental results show that this method is more effective than using spectral kurtosis or peak signal-to-noise ratio as the objective function alone. The flowchart of introducing Bayesian optimization into the non-local means denoising algorithm is shown in the attached figure. Figure 1 shown.

[0057] The present invention proposes a BPO-NLM denoising algorithm, which uses Bayesian parameter optimization to adaptively set algorithm parameters, avoiding the interference of human experience. Compared with the traditional NLM algorithm that relies on human experience, the denoising effect of BPO-NLM is significantly improved. Under different signal-to-noise ratios (SNRs), the root mean square error (RMSE) of the signal after BPO-NLM denoising is much smaller than that of the signal after traditional NLM algorithm denoising. At the same time, the signal-to-noise ratio (SNR) of the signal after BPO-NLM denoising is higher, that is, the effective component of the signal is enhanced, while the noise component of the signal is suppressed. The key points mainly include the following aspects:

[0058] (1) The denoising effect of BPO-NLM is much better than that of traditional NLM, and its parameters are obtained through optimization rather than human experience;

[0059] (2) After adopting the optimal parameters, the root mean square error (RMSE) of the denoised signal will be significantly reduced, and the signal-to-noise ratio (SNR) will be significantly increased;

[0060] Optimization method validation:

[0061] The swing signal of a hydroelectric generator set is an important monitoring input indicator. To verify the application value of the NLM algorithm based on Bayesian parameter optimization in denoising vibration signal processing, this patent selected the swing signal of a hydroelectric generator set for simulation analysis. The swing of a hydroelectric generator set is mainly affected by mechanical excitation and hydraulic excitation. Mechanical excitation is generally based on medium frequency (1, 2, or 3 times the rotation frequency), while hydraulic excitation is mainly based on low frequency (0.20 to 45 times the rotation frequency). Therefore, a simulation signal is constructed as shown below:

[0062]

[0063] Where A i＝1,2,3,4,5,6 =20, 4.5, 2.55, 1.5, 0.4 and 0.3um, f i＝1,2,3,4,5,6 = 1.25, 1.25×2, 1.25×3, 1.25×4, 1.25×0.2 and 1.25×0.3. The sampling frequency is set to 1000. Gaussian white noise with a signal-to-noise ratio of 5dB is superimposed on the original signal without noise. The analog signal without noise is as follows Figure 2 As shown in (a).

[0064] In order to evaluate the effectiveness of NLM based on Bayesian parameter optimization in vibration signal denoising, the root mean square error RMSE and signal-to-noise ratio SNR are defined.

[0065]

[0066] Where N is the number of sampling points, x i is the original signal without noise, y i The smaller the RMS error and the larger the SNR, the better the algorithm is at denoising vibration signals.

[0067] Figure 3The effects of noise-contaminated and noise-free signals were compared with those of signals processed using non-local means denoising, with the algorithm parameters set to λ = 0.3σ (σ represents the standard error of the noisy signal), K = 20, and P = 12. Compared to the noisy signal, the non-local means algorithm (NLM) performed significantly better in denoising vibration signals. While the original signal contaminated by noise was noticeably distorted, the denoised signal was smoother and closely resembled the noise-free signal. Furthermore, the differences between the noise-free, noisy, and denoised signals were amplified. Although the non-local means algorithm (NLM) performed satisfactorily in denoising vibration signals, the differences between the noise-free and denoised signals cannot be ignored, which is closely related to the algorithm parameters.

[0068] To verify the influence of parameter settings on the denoising effect, the present invention adopts three different parameter combinations: λ=0.3σ, K=20 / 40 / 80 and P=12. Figure 4 This comparison shows the signal effects of NLM (non-local means) denoising using different algorithm parameters. Clearly, the denoising effect of the NLM algorithm is significantly affected by the different algorithm parameters. Furthermore, to quantitatively evaluate the denoising effect, the root mean square error and signal-to-noise ratio (SNR) corresponding to the three algorithm parameters were calculated. See Table 1 for detailed data.

[0069] As the parameter K (i.e., the half-width of the search domain) increases, the root mean square error (RMS) of the denoised signal increases, while the signal-to-noise ratio (SNR) decreases, gradually deteriorating the denoising effect. Among the three algorithm parameters, the combination of λ = 0.3σ, K = 20, and P = 12 performs best, achieving the lowest RMS error and the highest SNR. Parameter settings significantly influence denoising effectiveness, and finding the optimal algorithm parameters remains a key challenge.

[0070] Table 1 RMSE and SNR of three algorithm parameters

[0071]

[0072] Bayesian optimization provides a solution for adaptively setting the parameters of nonlinear mapping (NLM) that does not rely on expert experience. Figure 5The denoising performance of the traditional non-local means algorithm (NLM) was compared with that of the non-local means algorithm based on Bayesian parameter optimization (BPO-NLM). Compared to the traditional NLM, whose parameters depend on user experience, the BPO-NLM significantly improves denoising performance. It is important to note that the parameters of the traditional NLM algorithm were λ = 0.3σ, K = 20, and P = 12, which have demonstrated excellent denoising performance in previous studies. The optimal parameters obtained using Bayesian parameter optimization (BPO-NLM) are λ = 1.4632, K = 20, and P = 50. Clearly, the denoised signal processed by BPO-NLM is closer to the original noise-free signal and is smoother than the signal denoised by the traditional NLM algorithm. Table 2 lists the root mean square error (RMSE) and signal-to-noise ratio (SNR) of the denoised signal to quantitatively compare the denoising performance of the traditional NLM algorithm and the BPO-NLM algorithm. Compared with the traditional non-local means algorithm, the root mean square error of the signal after denoising by the BPO-NLM algorithm is reduced by 21.9%, the signal-to-noise ratio is improved by 5.3%, and the denoising effect is significantly improved.

[0073] Table 2 RMSE and SNR of three algorithm parameters

[0074] Traditional NLM denoising signal BPO-NLM denoised signal Relative error (%) RMSE 0.1273 0.0994 21.9 Signal-to-noise ratio (SNR) 41.238 43.439 5.3

[0075] Furthermore, we investigate the denoising effect of BPO-NLM under different signal-to-noise ratios (SNRs). Figure 6 The denoising performance of BPO-NLM under different SNR conditions is demonstrated. Regardless of the SNR value, the signal processed by BPO-NLM is closer to the original signal without noise, that is, the error between the denoised signal and the original signal is smaller. Therefore, the BPO-NLM denoising algorithm is an effective signal for general vibrations. At the same time, as the SNR increases, the original signal is more seriously affected by the distortion of environmental noise, and the signal denoised by the BPO-NLM algorithm is smoother than the traditional NLM method. In addition, the root mean square error and SNR indicators of the denoised signal under different SNRs are calculated, and the results are shown in Figure 2. Figure 7 As shown. Clearly, compared to traditional NLM, the root mean square error of the signal after BPO-NLM processing is significantly reduced, which means that the denoised signal is closer to the original noise-free signal. At the same time, as the signal-to-noise ratio increases, the root mean square error of the signal denoised by the BPO-NLM algorithm decreases rapidly. The signal-to-noise ratio of the BPO-NLM denoised signal is always higher than that of the traditional NLM denoising result, indicating that this method effectively enhances the effective components of the signal while suppressing the noise component. When the signal-to-noise ratio is low, the denoising effect of the BPO-NLM algorithm is particularly improved - when the input signal-to-noise ratio is 20, the signal-to-noise ratio of the denoised signal can be close to 55.

[0076] Finally, we investigate how the algorithm parameters vary with the signal-to-noise ratio (SNR). Figure 8 The optimal parameter configuration of the BPO-NLM algorithm under different signal-to-noise ratio (SNR) conditions is demonstrated. As mentioned above, the decisive parameters of the non-local mean denoising algorithm are the bandwidth λ, the search domain half-width K and the structural block half-width P, which will significantly affect the noise reduction effect of NLM. Therefore, how to set these parameters is crucial for vibration signal denoising. The present invention introduces a Bayesian parameter optimization method to adaptively set these parameters, which can significantly improve the denoising effect of the non-local mean algorithm. The bandwidth λ is insensitive to the signal-to-noise ratio, that is, under different signal-to-noise ratios (SNRs), the change of λ is very small, so λ=1.463 is suitable for most noisy signals. Under low signal-to-noise ratio (SNR) conditions, the half-width K of the search domain remains constant; as the signal-to-noise ratio (SNR) increases, the K value decreases sharply, which means that a narrow search domain is more suitable for high signal-to-noise ratio (SNR) signals. Interestingly, the dependence of the half-width of the structural block P on the signal-to-noise ratio (SNR) is non-monotonic. Under low SNR conditions, the P value increases rapidly as the SNR increases; however, as the SNR continues to increase, the P value drops back to its previous level. This indicates that narrow building blocks are recommended for extremely low and high SNR environments, while wide building blocks are the optimal choice for signals with medium SNRs.

[0077] Compared with the traditional NLM algorithm, the BPO-NLM denoising algorithm of the present invention shows significant effects and advantages in denoising the vibration signals of hydropower units.

[0078] (1) The BPO-NLM algorithm achieves adaptive setting of key parameters of the NLM algorithm by introducing the Bayesian parameter optimization method, avoiding the interference of human experience, thereby improving the accuracy and stability of denoising. Experimental results show that under different signal-to-noise ratios (SNRs), the root mean square error (RMSE) of the signal denoised by BPO-NLM is much smaller than that of the signal denoised by the traditional NLM algorithm, which means that the BPO-NLM algorithm can more effectively remove noise and retain the true information of the signal.

[0079] (2) The BPO-NLM algorithm also performs well in improving signal quality. The signal-to-noise ratio (SNR) of the denoised signal is significantly improved, that is, the effective components of the signal are enhanced, while the noise components are effectively suppressed. This advantage enables the BPO-NLM algorithm to play a better role in subsequent signal feature extraction and fault diagnosis, improving the accuracy and reliability of diagnosis.

[0080] (3) The BPO-NLM algorithm also has the advantages of high computational efficiency and simple parameter optimization. Compared with traditional global search and random search methods, the Bayesian optimization method requires fewer initial sample points and can find the global optimal solution more quickly. This makes the BPO-NLM algorithm more efficient and practical in practical applications.

[0081] In summary, the BPO-NLM denoising algorithm of the present invention has significant denoising effects and advantages in processing vibration signals of hydropower units, which not only improves the accuracy and stability of the signal, but also improves the reliability and efficiency of subsequent fault diagnosis.

[0082] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0084] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A hydropower unit vibration signal denoising method based on BPO-NLM is characterized by: The following steps are involved: Based on the non-local mean filter algorithm NLM, the NLM is optimized by the Bayesian parameter optimization algorithm BPO; The optimized NLM is used to perform noise reduction on the collected vibration signals of the hydropower unit.

2. The method for denoising vibration signals of a hydropower unit based on BPO-NLM according to claim 1, characterized in that: When optimizing NLM, the decisive parameters of the non-local mean filtering algorithm NLM are optimized through the Bayesian parameter optimization algorithm BPO.

3. The method for denoising vibration signals of a hydropower unit based on BPO-NLM according to claim 2, characterized in that: When optimizing the decisive parameters, the decisive parameters include: filter bandwidth parameter, parameter K and parameter P, wherein the parameter K affects the computational complexity and computational time of the NLM algorithm, and the parameter P affects the number of similar structural blocks found during the algorithm operation.

4. The method for denoising vibration signals of a hydropower unit based on BPO-NLM according to claim 3, characterized in that: When optimizing the decisive parameters, the objective function of the Bayesian parameter optimization algorithm BPO is constructed according to the spectral kurtosis value, peak signal-to-noise ratio and mean square error to optimize the decisive parameters.

5. The method for denoising vibration signals of a hydropower unit based on BPO-NLM according to claim 4, characterized in that: When using the optimized NLM for noise reduction processing, the vibration signal of the hydropower unit consisting of the real vibration signal and external interference noise is obtained, and the weighted average of all similar blocks is calculated using the optimized NLM to estimate the real signal.

6. A hydropower unit vibration signal denoising system based on BPO-NLM, used to implement the hydropower unit vibration signal denoising method based on BPO-NLM according to claim 1, characterized in that: include: Data acquisition module, used to collect vibration signals of hydropower units consisting of real vibration signals and external interference noise; The noise reduction processing module is used to perform noise reduction processing on the vibration signal of the hydropower unit through the optimized NLM, wherein the NLM is optimized based on the non-local mean filter algorithm NLM through the Bayesian parameter optimization algorithm BPO.

7. The hydropower unit vibration signal denoising system based on BPO-NLM according to claim 6, characterized in that: The noise reduction processing module is further used to optimize the decisive parameters of the non-local mean filtering algorithm NLM through the Bayesian parameter optimization algorithm BPO, wherein the decisive parameters include: filter bandwidth parameter, parameter K and parameter P, the parameter K affects the calculation amount and calculation time of the NLM algorithm, and the parameter P affects the number of similar structural blocks found during the algorithm operation.

8. The hydropower unit vibration signal denoising system based on BPO-NLM according to claim 7, characterized in that: The noise reduction processing module is further used to construct an objective function of the Bayesian parameter optimization algorithm BPO based on the spectral kurtosis value, peak signal-to-noise ratio and mean square error, so as to optimize the decisive parameters, wherein the optimized NLM is used to calculate the weighted average of all similar blocks to estimate the real signal.

Images