A method for processing vibration signals of port equipment

Abstract

The present invention relates to the technical field of signal processing, and discloses a vibration signal processing method for port equipment. By constructing a comprehensive function, based on the characteristics of the frequency scale, the second wavelet basis function, and the threshold, the concentration of energy distribution, the computational complexity of the second wavelet basis function under the frequency scale, the orthogonality value between the second wavelet basis functions under the frequency scale, the sparse value of the second wavelet coefficients under the frequency scale, and the signal-to-noise ratio of the reconstructed time-domain signal after threshold processing are comprehensively considered, realizing the selection of the optimized frequency scale, the optimized second wavelet basis function, and the optimized threshold. This innovative method gets rid of the limitations of single or optimized algorithms, fixed parameter settings, or predefined rules in the prior art, and significantly improves the accuracy and robustness of low-frequency vibration signal processing.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and particularly to a method for processing vibration signals of port equipment. Background Art

[0002] As a key hub of the international and domestic logistics chain, the safety and reliability of port infrastructure are of crucial importance. With the continuous expansion of port scale and the increase in complexity, the demand for health monitoring of port equipment (such as large cranes, wharf structures, ships, etc.) is growing. Vibration signals are one of the important indicators for evaluating the operating status and structural integrity of these devices. However, existing vibration sensors face many challenges in processing low-frequency vibration signals, especially in the application of complex port environments with multiple sea areas. Therefore, it is particularly urgent to develop a method that can effectively expand the measurement range of vibration sensors and accurately process low-frequency vibration signals.

[0003] Currently, the commonly used magnetoelectric vibration sensors perform well in the high-frequency band but have limited measurement capabilities in the low-frequency band. Moreover, port equipment usually involves large structures and heavy machinery, and their vibration frequencies are relatively low. Especially during startup, shutdown, or load changes, significant low-frequency vibrations will be generated. The environment is complex and affected by various factors such as temperature, humidity, salt spray, wind, and waves, which will further affect the acquisition and processing of low-frequency vibration signals. Most of the existing technologies for processing low-frequency signals remove background noise through filters to obtain denoised signals, then select wavelet functions to perform wavelet processing on the denoised signals to obtain wavelet coefficients, and obtain the processed time-domain signals through threshold processing. Finally, it is verified whether the signal-to-noise ratio of the time-domain signal is significantly improved. If the verification fails, the frequency, function, and threshold need to be reselected and the signal processing is carried out again. The existing selection of wavelet functions, frequency division, and thresholds is all achieved through single or optimized algorithms or relies on fixed parameter settings or predefined rules. Although these methods can achieve good results in certain specific scenarios, they mostly rely on single or optimized algorithms, or select wavelet functions, frequency division, and thresholds based on fixed parameter settings and predefined rules. The limitation of this method is that in some scenarios, signal processing cannot be completed quickly, and the frequency range, function, and threshold need to be reset, resulting in the inability to fully adapt to different types of low-frequency vibration signals and complex noise environments, and may not be able to quickly obtain the optimal processing effect in practical applications, with low signal processing efficiency. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for processing vibration signals of port equipment, including:

[0005] Step 1, collecting the original signal through a sensor;

[0006] Step 2, preprocess the original signal; perform initial frequency scale division on the preprocessed original signal to obtain sub-band signals of different frequencies; set a number of sampling points on the sub-band signals of different initial frequency scales, perform the first wavelet transform on each sampling point through the first wavelet basis function to obtain the first wavelet coefficients of each sampling point at different initial frequency scales; calculate the energy distribution at each initial frequency scale according to the first wavelet coefficients of each sampling point at different initial frequency scales, and obtain the concentration degree according to the energy distribution at each initial frequency scale;

[0007] Step 3, remove the background noise of the sub-band signal at each initial frequency scale through a filter to obtain a denoised signal, and record the filter length at each initial frequency scale; obtain the second wavelet basis function, select the second wavelet basis function to perform the second wavelet transform on each denoised signal at each initial frequency scale to obtain the second wavelet coefficients, calculate the orthogonality value of the current second wavelet basis function and other second wavelet basis functions at each initial frequency scale, and calculate the computational complexity of each second wavelet basis function at each initial frequency scale; obtain a threshold, perform threshold processing on the second wavelet coefficients through the threshold; input the threshold-processed second wavelet coefficients into the inverse wavelet transform to reconstruct the time-domain signal; calculate the signal-to-noise ratio according to the reconstructed time-domain signal and the original signal; obtain the sparsity value according to the second wavelet coefficients at each initial frequency scale;

[0008] Step 4, construct a comprehensive function through the sparsity value, signal-to-noise ratio, concentration degree, computational complexity, and orthogonality value, and select the optimized frequency scale from the initial frequency scales, select the optimized second wavelet basis function from the second wavelet basis functions, and select the optimized threshold from the thresholds according to the comprehensive function;

[0009] Step 5, determine the optimized sub-band signal according to the optimized frequency scale, remove the noise of the optimized sub-band signal through a filter to obtain an optimized denoised signal, perform the second wavelet transform on the optimized denoised signal through the optimized second wavelet basis function to obtain the optimized second wavelet coefficients, and perform threshold processing on the second wavelet coefficients according to the optimized threshold to obtain the optimized time-domain signal.

[0010] Further, selecting the optimized frequency scale from the initial frequency scales, selecting the optimized second wavelet basis function from the second wavelet basis functions, and selecting the optimized threshold from the thresholds according to the comprehensive function includes the following steps:

[0011] Step 41, calculate the concentration degree of the energy distribution of the sub-band signal at each initial frequency scale; take the initial frequency scale with the largest concentration degree as the initial optimized frequency scale;

[0012] Step 42, preset candidate second wavelet basis functions and candidate thresholds;

[0013] Step 43, select a candidate second wavelet basis function to perform a second wavelet transform on each denoised signal at each initial optimized frequency scale to obtain candidate second wavelet coefficients;

[0014] Step 44, calculate the orthogonality value between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimized frequency scale;

[0015] Step 45, calculate the computational complexity of each candidate wavelet basis function at each initial optimized frequency scale;

[0016] Step 46, perform threshold processing on each candidate second wavelet coefficient through a candidate threshold; input the candidate second wavelet coefficients after threshold processing into the inverse wavelet transform to reconstruct the candidate time-domain signal; calculate the candidate signal-to-noise ratio based on the reconstructed candidate time-domain signal and the original signal; obtain the candidate sparsity value based on the candidate second wavelet coefficients at each initial optimized frequency scale;

[0017] Step 47, calculate the comprehensive function value based on the orthogonality value between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimized frequency scale, the computational complexity of each candidate second wavelet basis function at each initial optimized frequency scale, the candidate sparsity value obtained from the candidate second wavelet coefficients at each initial optimized frequency scale, the candidate signal-to-noise ratio calculated based on the reconstructed candidate time-domain signal and the original signal, and the concentration degree of the sub-band signal energy distribution at each initial optimized frequency scale. Take the candidate second wavelet basis function with the maximum comprehensive function value as the optimized second wavelet basis function, and take the candidate threshold with the maximum comprehensive function value as the optimized threshold;

[0018] Step 48, based on the optimized second wavelet basis function and the optimized threshold, recalculate the comprehensive function value at each initial frequency scale, and take the initial frequency scale with the maximum comprehensive function value as the optimized frequency scale.

[0019] Further, the calculation process of the orthogonality value of each candidate wavelet function at the initial optimized frequency scale is as follows: calculate the correlation between each candidate second wavelet basis function and other candidate second wavelet basis functions at the initial optimized frequency scale to obtain a correlation matrix; calculate the maximum singular value and the minimum singular value of the correlation matrix, and obtain the condition number of the correlation matrix based on the maximum singular value and the minimum singular value, and take the condition number as the orthogonality value; construct an evaluation threshold for the orthogonality value, and divide the orthogonality value into a reward term and a penalty term according to the evaluation threshold.

[0020] Further, the computational complexity of each candidate wavelet basis function at each initial optimized frequency scale is:

[0021] ;

[0022] In the formula, represents the computational complexity of the g-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the signal length at the initial optimized frequency scale c, represents the filter length at the initial optimized frequency scale c, where N is the length of the original signal.

[0023] Furthermore, the orthogonality value of each candidate wavelet function at the initial optimized frequency scale is:

[0024] ;

[0025] wherein, represents the orthogonality value of the correlation matrix A between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the maximum singular value of the correlation matrix between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the minimum singular value of the correlation matrix between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, and Z represents the total number of candidate second wavelet basis functions at the initial optimized frequency scale c.

[0026] Furthermore, the synthesis function is:

[0027]

[0028] wherein, represents the synthesis function comprehensively considering the initial frequency scale d, the second wavelet basis function , the threshold , represents the concentration degree of the energy distribution on the i-th initial frequency scale d, represents the transfer function, represents the orthogonality value of the j-th second wavelet basis function, represents the computational complexity of the j-th second wavelet basis function, represents the signal-to-noise ratio of the reconstructed time-domain signal and the original signal after being processed by the k-th threshold, represents the sparsity value of the optimized second wavelet coefficients after being processed by the k-th threshold, , , respectively represent the adjustment coefficients, .

[0029] Furthermore, the transfer function is:

[0030] ;

[0031] wherein, a conversion value representing the orthogonality value of the j-th second wavelet basis function, z represents a positive number, and z represents an evaluation threshold, representing a reward term, representing a penalty term.

[0032] The embodiments of the present invention have the following technical effects:

[0033] By constructing a comprehensive function, based on the characteristics of the frequency scale, the second wavelet basis function, and the threshold, the present invention comprehensively considers the concentration of energy distribution, the computational complexity of the second wavelet basis function under the frequency scale, the orthogonality value between the second wavelet basis functions under the frequency scale, the sparse value of the second wavelet coefficients under the frequency scale, and the signal-to-noise ratio of the reconstructed time-domain signal after threshold processing, and realizes the selection of the optimized frequency scale, the optimized second wavelet basis function, and the optimized threshold. This innovative method gets rid of the limitations of single or optimized algorithms, fixed parameter settings, or predefined rules in the prior art, and significantly improves the accuracy and robustness of low-frequency vibration signal processing. First of all, through multi-dimensional comprehensive optimization, the present invention can more comprehensively capture the multi-scale characteristics of the signal and ensure the best denoising effect in different application scenarios. Secondly, the highly adaptive optimization method enables the method to flexibly handle different types and complexities of signals. Especially in complex environments such as port equipment, affected by various external factors such as temperature, humidity, salt spray, and wind and waves, it can still maintain stable denoising performance. In addition, the flexible selection of wavelet basis functions and the dynamically adjusted frequency division enable the method to adapt to different types of low-frequency vibration signals. Whether it is a high-frequency signal, a low-frequency signal, or a non-stationary signal, it can find the most suitable processing solution. Moreover, the search for the global optimal solution ensures the best denoising effect throughout the signal processing process, rather than just local optimization. Finally, by introducing the computational complexity as one of the optimization objectives, the present invention can reasonably control the consumption of computing resources while ensuring the denoising effect, and is applicable to real-time processing and large-scale data processing scenarios. In summary, the present invention not only improves the processing accuracy and robustness of low-frequency vibration signals, but also has stronger adaptability and flexibility, and can achieve efficient and stable signal denoising in complex environments, providing strong technical support for the health monitoring and fault prediction of port equipment.

[0034] The present invention significantly improves the accuracy and robustness of low-frequency vibration signal processing through a systematic multi-stage optimization strategy. First, the present invention takes the initial frequency scale with the highest concentration as the initial optimization frequency scale. This selection is based on the concentration of energy distribution, which can ensure that the initially selected frequency scale can best capture the main components of the signal, thus providing a solid foundation for subsequent optimization. Then, at the initial optimization frequency scale, the computational complexity, orthogonality value, sparsity value of the second wavelet basis function, and the signal-to-noise ratio of the reconstructed time-domain signal after threshold processing are calculated respectively. By constructing a comprehensive function and combining these multi-dimensional parameters, the second wavelet basis function corresponding to the maximum comprehensive function value is selected as the optimized second wavelet basis function, and the threshold corresponding to the maximum comprehensive function value is selected as the optimized threshold. This optimization method based on the comprehensive function not only considers the denoising effect but also takes into account the computational complexity and orthogonality, ensuring the efficiency and stability of the algorithm in different application scenarios.

[0035] Subsequently, the present invention further recalculates the comprehensive function value at each initial frequency scale based on the optimized second wavelet basis function and the optimized threshold, and finally selects the initial frequency scale with the maximum comprehensive function value as the optimized frequency scale. This process ensures the best denoising effect throughout the signal processing process through global optimization, rather than just a local optimal solution. This method avoids the limitations of relying on single or fixed parameter settings in the prior art and can flexibly handle different types and complexities of signals in complex environments. Especially in the case of port equipment affected by various external factors such as temperature, humidity, salt spray, and wind and waves, it can still maintain stable denoising performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0037] Figure 1 It is a flowchart of a method for processing vibration signals of port equipment provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope protected by the present invention.

[0039] Figure 1 It is a flowchart of a method for processing vibration signals of port equipment provided by an embodiment of the present invention. Refer to Figure 1 , specifically including:

[0040] Step 1, collect the original signal through a sensor.

[0041] The sensor is preferably a magnetoelectric vibration sensor for low-frequency vibration measurement, such as a magnetoelectric vibration velocity sensor with the model LDT05-06C. This sensor has the following characteristics:

[0042] Frequency range: 0.5 Hz to 500 Hz, which can cover the common low-frequency vibration range of port equipment.

[0043] Sensitivity: 40 mV / (mm / s), ensuring that sufficient output voltage can still be provided in the low-frequency band.

[0044] Operating temperature range: -40°C to +125°C, adapting to temperature changes in the port environment.

[0045] Protection level: IP67, with good waterproof and dustproof performance, suitable for the complex environment of the port.

[0046] Anti-interference ability: Built-in electromagnetic shielding to reduce the influence of external electromagnetic interference on the signal.

[0047] Installation position of the sensor: According to the specific structure of the port equipment, select key parts to install the sensor. For example, install the sensor at parts such as the boom, base, and transmission shaft of the crane that are prone to low-frequency vibration. Ensure that the sensor is in close contact with the object to be measured to avoid signal distortion caused by looseness.

[0048] Step 2, preprocess the original signal.

[0049] The preprocessing includes:

[0050] Remove the DC component: Use a high-pass filter (such as a cut-off frequency of 0.1 Hz) to remove the DC component in the signal to ensure the accuracy of subsequent processing.

[0051] Normalization processing: Normalize the signal amplitude to the interval [-1, 1] for subsequent wavelet transform and other signal processing operations.

[0052] Remove high-frequency noise: Use a low-pass filter (such as a cut-off frequency of 500 Hz) to remove the high-frequency noise in the signal and retain the low-frequency vibration components.

[0053] Perform an initial frequency scale division on the preprocessed original signal to obtain sub-band signals of different frequencies. Set a number of sampling points for the sub-band signals of different initial frequency scales, and perform a first wavelet transform on each sampling point through the first wavelet basis function (Db4 wavelet) to obtain the first wavelet coefficients of each sampling point under different initial frequency scales; according to the first wavelet coefficients of each sampling point under different initial frequency scales, calculate the energy distribution under each initial frequency scale, and obtain the concentration degree according to the energy distribution under each initial frequency scale.

[0054] The initial frequency scales are: , ,..., , where is the lowest frequency scale, is the spurious high frequency. Each frequency scale corresponds to a sub-band signal , where a represents the frequency scale and b represents the translation parameter.

[0055] The first wavelet coefficients: ;

[0056] represents the first wavelet coefficient of the q-th sampling point under the i-th initial frequency scale d, x represents the original signal, represents the first wavelet basis function under the i-th initial frequency scale d.

[0057] The energy distribution under each initial frequency scale is:

[0058] , represents the energy distribution under the i-th initial frequency scale d, and H is the total number of initial frequency scales.

[0059] The concentration degree represents the degree of concentration of the energy distribution of the signal at a certain scale. Usually, the higher the energy concentration degree, the more the main energy of the signal is concentrated on fewer wavelet coefficients, which is helpful for subsequent denoising and feature extraction.

[0060] ;

[0061] represents the concentration degree of the energy distribution on the i-th initial frequency scale d, represents the variance of the energy distribution on the i-th initial frequency scale d, represents the mean of the energy distribution on the i-th initial frequency scale d.

[0062] The concentration degree obtained based on variance and mean can effectively reflect the distribution of signal energy. Even in the case of high noise, it can accurately identify the main components of the signal. Compared with other complex energy distribution indicators, the energy concentration based on variance is simple to calculate and easy to implement. It is applicable to multi-scale wavelet transform and can evaluate the energy concentration degree of the signal at different scales, helping to select the optimal optimization scale.

[0063] Step 3: Remove the background noise of the sub-band signals at each initial frequency scale through a filter to obtain the denoised signals, and record the filter length at each initial frequency scale; obtain the second wavelet basis function, select the second wavelet basis function to perform the second wavelet transform on each denoised signal at each initial frequency scale to obtain the second wavelet coefficients, calculate the orthogonality values between the current second wavelet basis function and other second wavelet basis functions at each initial frequency scale, and calculate the computational complexity of each second wavelet basis function at each initial frequency scale; obtain the threshold, and perform threshold processing on the second wavelet coefficients through the threshold; input the threshold-processed second wavelet coefficients into the inverse wavelet transform to reconstruct the time-domain signal; calculate the signal-to-noise ratio according to the reconstructed time-domain signal and the original signal; obtain the sparsity value according to the second wavelet coefficients at each initial frequency scale.

[0064] Among them, the calculation processes of the orthogonality value and the computational complexity are the same as the calculation logic in step 47, and will not be elaborated in detail in this embodiment. The signal-to-noise ratio is calculated through the prior art.

[0065] The sparsity value is calculated by the ratio of the second wavelet coefficients at each initial frequency scale to the maximum second wavelet coefficient at each initial frequency scale.

[0066] In this embodiment, the calculation of the sparsity value is mainly used to evaluate the sparsity of the second wavelet coefficients after threshold processing. The higher the sparsity, the more the main energy of the signal is concentrated on a few non-zero coefficients after threshold processing, which helps to remove noise and retain the key features of the signal. By introducing the sparsity value as part of the comprehensive optimization function, the present invention can ensure that the sparse structure of the signal is better retained while guaranteeing the denoising effect, thereby improving the quality of the reconstructed time-domain signal. By calculating the sparsity value, the processing flow of the low-frequency vibration signal is further optimized. Specifically, the introduction of the sparsity value enables the algorithm to better retain the sparse structure of the signal, reduce the influence of noise on the signal, and ensure that the reconstructed time-domain signal has higher clarity and reliability. Combining other parameters such as energy concentration degree, computational complexity, orthogonality value, and signal-to-noise ratio, the present invention constructs a multi-dimensional comprehensive optimization function to achieve the global optimal selection of the frequency scale, the second wavelet basis function, and the threshold, significantly improving the accuracy and robustness of signal processing.

[0067] Step 4: Construct a comprehensive function based on the sparsity value, signal-to-noise ratio, concentration, computational complexity, and orthogonality value. Select the optimized frequency scale from the initial frequency scales, the optimized second wavelet basis function from the second wavelet basis functions, and the optimized threshold from the thresholds according to the comprehensive function.

[0068] Based on the signal characteristics, this embodiment has made multiple innovative improvements, significantly enhancing the accuracy, robustness, and computational efficiency of low-frequency vibration signal processing. First, in the selection of the frequency scale, the present invention introduces the concentration of energy distribution as a key evaluation index. By calculating the energy concentration at each initial frequency scale, the scale with the highest concentration is selected as the preliminary optimized frequency scale. This improvement ensures that the selected scale can most effectively capture the main components of the signal. Especially in the low-frequency vibration signals commonly found in port equipment, it can more accurately identify the key vibration characteristics. Second, for the selection of the second wavelet basis function, the present invention not only considers its computational complexity at a specific frequency scale but also introduces the orthogonality value as an evaluation criterion. The introduction of computational complexity enables the algorithm to reasonably control the consumption of computational resources while ensuring the denoising effect, making it suitable for real-time processing and large-scale data processing scenarios; the introduction of the orthogonality value ensures the independence between different wavelet basis functions, avoiding the introduction of redundant information and further improving the accuracy of signal processing. Third, for the selection of the threshold, the present invention comprehensively considers two key parameters, the sparsity value and the signal-to-noise ratio. The introduction of the sparsity value enables the algorithm to better retain the sparse structure of the signal and reduce the influence of noise on the signal; the introduction of the signal-to-noise ratio ensures that the reconstructed time-domain signal after threshold processing has higher clarity and reliability. Finally, by constructing a comprehensive function, the parameters in the above multiple dimensions (energy concentration, computational complexity, orthogonality value, sparsity value, and signal-to-noise ratio) are organically combined to achieve global optimization of multiple objectives. This comprehensive optimization strategy not only overcomes the limitations of relying on single or fixed parameter settings in the prior art but also can flexibly handle different types and complexities of signals in complex environments, ensuring the best denoising effect in various application scenarios. In summary, through multi-dimensional comprehensive optimization, this embodiment significantly enhances the accuracy, robustness, and computational efficiency of low-frequency vibration signal processing, providing strong technical support for the health monitoring and fault prediction of port equipment.

[0069] The process of selecting the optimized frequency scale, the optimized second wavelet basis function, and the optimized threshold through the comprehensive function is as follows:

[0070] Step 41: Calculate the concentration of the sub-band signal energy distribution at each initial frequency scale; take the initial frequency scale with the maximum concentration as the initial optimized frequency scale.

[0071] Step 42: Preset candidate second wavelet basis functions and candidate thresholds.

[0072] Step 43: Select a candidate second wavelet basis function to perform a second wavelet transform on each denoised signal at each initial optimized frequency scale to obtain candidate second wavelet coefficients.

[0073] Step 44: Calculate the orthogonality values between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimized frequency scale.

[0074] Step 45: Calculate the computational complexity of each candidate wavelet basis function at each initial optimized frequency scale.

[0075] Step 46: Threshold each candidate second wavelet coefficient with a candidate threshold; input the thresholded candidate second wavelet coefficients into the inverse wavelet transform to reconstruct a candidate time-domain signal; calculate the candidate signal-to-noise ratio based on the reconstructed candidate time-domain signal and the original signal; obtain candidate sparsity values based on the candidate second wavelet coefficients at each initial optimized frequency scale.

[0076] Step 47: Calculate a comprehensive function value based on the orthogonality values between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimized frequency scale, the computational complexity of each candidate second wavelet basis function at each initial optimized frequency scale, the candidate sparsity values obtained from the candidate second wavelet coefficients at each initial optimized frequency scale, the candidate signal-to-noise ratio calculated from the reconstructed candidate time-domain signal and the original signal, and the concentration degree of the sub-band signal energy distribution at each initial optimized frequency scale. Take the candidate second wavelet basis function with the maximum comprehensive function value as the optimized second wavelet basis function, and take the candidate threshold with the maximum comprehensive function value as the optimized threshold.

[0077] Calculate the correlation between each candidate second wavelet basis function and other candidate second wavelet basis functions at the initial optimized frequency scale to obtain a correlation matrix; calculate the maximum singular value and the minimum singular value of the correlation matrix, and obtain the condition number of the correlation matrix based on the maximum singular value and the minimum singular value. Take the condition number as the orthogonality value; construct an evaluation threshold for the orthogonality value, and divide the orthogonality value into a reward term and a penalty term according to the evaluation threshold.

[0078] ;

[0079] In the formula, represents the orthogonality value of the correlation matrix A between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the maximum singular value of the correlation matrix between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the minimum singular value of the correlation matrix between the h-th candidate second wavelet basis function and the o-th candidate second wavelet basis function at the initial optimized frequency scale c, and Z represents the total number of candidate second wavelet basis functions at the initial optimized frequency scale c.

[0080] Obtain the signal length of the denoised signal at the initial optimized frequency scale and the filter length at the initial optimized frequency scale. According to the signal length and the filter length, calculate the computational complexity of each candidate second wavelet basis function.

[0081] ;

[0082] In the formula, represents the computational complexity of the g-th candidate second wavelet basis function at the initial optimized frequency scale c, represents the signal length of the initial optimized frequency scale c, represents the filter length at the initial optimized frequency scale c, N is the length of the original signal, represents the candidate second wavelet basis function.

[0083] ;

[0084] In the formula, represents the comprehensive function considering the initial frequency scale d, the second wavelet basis function , the threshold , represents the concentration degree of the energy distribution on the i-th initial frequency scale d, represents the conversion function, represents the orthogonality value of the j-th second wavelet basis function, represents the computational complexity of the j-th second wavelet basis function, represents the signal-to-noise ratio of the reconstructed time-domain signal and the original signal after the k-th threshold processing, represents the sparsity value of the optimized second wavelet coefficients after the k-th threshold processing, , , respectively represent the adjustment coefficients, .

[0085] Furthermore, ;

[0086] In the formula, represents the conversion value of the orthogonality value of the j-th second wavelet basis function, z represents a positive number, z > 2, Y represents the evaluation threshold, represents the reward term, represents the penalty term.

[0087] In signal processing, the orthogonality value is a crucial metric that measures the independence and non-interference among different wavelet basis functions. The higher the orthogonality value, the lower the correlation among the wavelet basis functions, which helps improve the accuracy and robustness of signal processing. When constructing the synthesis function, the consideration of the orthogonality value is divided into a reward term and a penalty term.

[0088] When , the reward term is . As increases, the reward term gradually decreases but always remains greater than 0. This indicates that wavelet basis functions with higher orthogonality will receive higher rewards and thus higher weights in the synthesis function.

[0089] When , the penalty term is . As decreases, the penalty term gradually increases but always remains less than 0. This shows that wavelet basis functions with lower orthogonality will receive lower weights and may even be excluded.

[0090] Step 48: Based on the optimized second wavelet basis function and the optimized threshold, recalculate the synthesis function value at each initial frequency scale, and take the initial frequency scale with the maximum synthesis function value as the optimized frequency scale.

[0091] The optimized second wavelet basis function can better match the time-frequency characteristics of the signal, while the optimized threshold can retain the key features of the signal while removing noise. On this basis, recalculate the synthesis function value at each initial frequency scale, and combine parameters in multiple dimensions such as energy concentration, computational complexity, orthogonality value, sparsity value, and signal-to-noise ratio to ensure a more comprehensive and accurate evaluation at each frequency scale. Finally, selecting the initial frequency scale with the maximum synthesis function value as the optimized frequency scale not only ensures the global optimality of signal processing but also enables flexible response to different types and complexities of signals in complex environments, ensuring the best denoising effect in various application scenarios.

[0092] Step 5: Determine the optimized subband signal according to the optimized frequency scale, perform denoising on the optimized subband signal through a filter to obtain the optimized denoised signal, perform the second wavelet transform on the optimized denoised signal through the optimized second wavelet basis function to obtain the optimized second wavelet coefficients, and perform threshold processing on the second wavelet coefficients according to the optimized threshold to obtain the optimized time-domain signal.

[0093] The processing principle of this step is the same as that of the prior art, except that the optimized frequency scale, the optimized denoised signal, and the optimized threshold are selected.

[0094] This processing method significantly improves the efficiency and reliability of low-frequency vibration signal processing, completely getting rid of the problem of repeated verification required in the prior art after obtaining the time-domain signal. Traditional methods often rely on fixed parameter settings or empirical values to determine the frequency scale, filter type, wavelet basis function, and threshold. This results in the need for multiple trials and verifications when processing different types of signals to ensure the accuracy of the final result. However, through the global optimization strategy in this application, these key parameters are dynamically adjusted during the processing to ensure that each step is based on the optimal solution. Specifically, optimizing the selection of the frequency scale enables the main components of the signal to be captured more accurately, avoiding information loss caused by inappropriate frequency selection; optimizing the denoised signal effectively removes noise through an adaptive filter while retaining the key features of the signal; optimizing the application of the threshold ensures that in the processing of the coefficients after the second wavelet transform, noise can be effectively suppressed without over-smoothening the signal. Therefore, the entire processing process is not only more efficient, but also a high-quality optimized time-domain signal can be obtained once without subsequent repeated verification, greatly saving time and computing resources.

[0095] It should be noted that the terms used in this invention are only for describing specific embodiments and do not limit the scope of this application. As shown in the specification of this invention, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. The term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method or device including a series of elements not only includes those elements but also includes other elements not explicitly listed, or also includes elements inherent to such a process, method or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of other identical elements in the process, method or device including the said element.

[0096] It should also be noted that the orientation or positional relationship indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing this invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to this invention. Unless otherwise clearly specified and limited, terms such as "installed", "connected", "connected to" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this invention can be understood according to specific circumstances.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for processing vibration signals of port equipment, characterized in that: include: Step 1, collecting original signals through sensors; Step 2, preprocessing the original signal; Performing initial frequency scale division on the preprocessed original signal to obtain sub-band signals of different frequencies; setting a number of sampling points on the sub-band signals of different initial frequency scales, performing the first wavelet transform on each sampling point through the first wavelet basis function, and obtaining the first wavelet coefficient of each sampling point at different initial frequency scales; calculating the energy distribution at each initial frequency scale according to the first wavelet coefficient of each sampling point at different initial frequency scales, and obtaining the concentration according to the energy distribution at each initial frequency scale; Step 3, removing the background noise of the subband signal at each initial frequency scale through a filter to obtain a denoised signal, and recording the filter length at each initial frequency scale; obtaining a second wavelet basis function, selecting a second wavelet basis function to perform a second wavelet transform on each denoised signal at each initial frequency scale to obtain a second wavelet coefficient, calculating the orthogonal value of the current second wavelet basis function and other second wavelet basis functions at each initial frequency scale, and calculating the computational complexity of each second wavelet basis function at each initial frequency scale; Obtain a threshold value, and perform threshold processing on the second wavelet coefficient by using the threshold value; The second wavelet coefficient after threshold processing is input into the inverse wavelet transform to reconstruct the time domain signal; the signal-to-noise ratio is calculated based on the reconstructed time domain signal and the original signal; the sparse value is obtained based on the second wavelet coefficient at each initial frequency scale; The sparse value is obtained by calculating the ratio of the second wavelet coefficient at each initial frequency scale to the maximum second wavelet coefficient at the scale; Step 4, constructing a comprehensive function through sparse value, signal-to-noise ratio, concentration, computational complexity, and orthogonal value, and selecting an optimized frequency scale from the initial frequency scale, selecting an optimized second wavelet basis function from the second wavelet basis function, and selecting an optimized threshold from the threshold according to the comprehensive function; Step 5, determine the optimized subband signal according to the optimized frequency scale, denoise the optimized subband signal through a filter to obtain an optimized denoised signal, perform a second wavelet transform on the optimized denoised signal by optimizing the second wavelet basis function to obtain an optimized second wavelet coefficient, perform threshold processing on the second wavelet coefficient according to the optimized threshold to obtain an optimized time domain signal.

2. A method for processing vibration signals of port equipment according to claim 1, characterized in that: The method comprises the following steps: selecting an optimized frequency scale from an initial frequency scale according to a comprehensive function, selecting an optimized second wavelet basis function from a second wavelet basis function, and selecting an optimized threshold value from a threshold value. Step 41, calculating the concentration of the subband signal energy distribution at each initial frequency scale; taking the initial frequency scale with the largest concentration as the initial optimized frequency scale; Step 42, presetting a candidate second wavelet basis function and a candidate threshold; Step 43, selecting a candidate second wavelet basis function to perform a second wavelet transform on each denoised signal at each initial optimized frequency scale to obtain a candidate second wavelet coefficient; Step 44, calculating the orthogonality value between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimized frequency scale; Step 45, calculating the computational complexity of each candidate wavelet basis function at each initial optimized frequency scale; Step 46, threshold processing is performed on each candidate second wavelet coefficient by using a candidate threshold; the candidate second wavelet coefficient after threshold processing is input into an inverse wavelet transform to reconstruct a candidate time domain signal; a candidate signal-to-noise ratio is calculated based on the reconstructed candidate time domain signal and the original signal; a candidate sparse value is obtained based on the candidate second wavelet coefficient at each initial optimized frequency scale; Step 47, calculating the comprehensive function value according to the orthogonal value between the current candidate wavelet basis function and other candidate wavelet basis functions at each initial optimization frequency scale, the computational complexity of each candidate second wavelet basis function at each initial optimization frequency scale, the candidate sparse value obtained from the candidate second wavelet coefficients at each initial optimization frequency scale, the candidate signal-to-noise ratio calculated between the reconstructed candidate time domain signal and the original signal, and the concentration of the subband signal energy distribution at each initial optimization frequency scale, taking the candidate second wavelet basis function with the maximum comprehensive function value as the optimized second wavelet basis function, and taking the candidate threshold value with the maximum comprehensive function value as the optimized threshold value; Step 48, based on optimizing the second wavelet basis function and optimizing the threshold, recalculate the comprehensive function value at each initial frequency scale, and use the initial frequency scale with the maximum comprehensive function value as the optimized frequency scale.

3. A method for processing vibration signals of port equipment according to claim 2, characterized in that: The calculation process of the orthogonal value of each candidate wavelet function at the initial optimized frequency scale is as follows: calculating the correlation between each candidate second wavelet basis function and other candidate second wavelet basis functions at the initial optimized frequency scale to obtain a correlation matrix; Calculate the maximum singular value and the minimum singular value of the correlation matrix, obtain the condition number of the correlation matrix according to the maximum singular value and the minimum singular value, and use the condition number as the orthogonal value; An evaluation threshold is constructed for the orthogonal value, and the orthogonal value is divided into reward items and penalty items according to the evaluation threshold.

4. A method for processing vibration signals of port equipment according to claim 3, characterized in that: The computational complexity of each candidate second wavelet basis function at the initial optimization frequency scale includes: obtaining the signal length of the de-noised signal at the initial optimization frequency scale and the filter length at the initial optimization frequency scale, and calculating the computational complexity of each candidate second wavelet basis function according to the signal length and the filter length.

5. A method for processing vibration signals of port equipment according to claim 4, characterized in that: The computational complexity of each candidate wavelet basis function at each initial optimized frequency scale is: ； In the formula, represents the computational complexity of the g-th candidate second wavelet basis function under the initial optimization frequency scale c, represents the signal length of the initial optimized frequency scale c, represents the filter length at the initial optimized frequency scale c, and N is the length of the original signal.

6. A method for processing vibration signals of port equipment according to claim 3, characterized in that: The orthogonal value of each candidate wavelet function at the initial optimized frequency scale is: ； In the formula, represents the orthogonal value of the correlation matrix A between the hth candidate second wavelet basis function and the oth candidate second wavelet basis function under the initial optimization frequency scale c, represents the maximum singular value of the correlation matrix between the hth candidate second wavelet basis function and the oth candidate second wavelet basis function at the initial optimized frequency scale c, represents the minimum singular value of the correlation matrix between the hth candidate second wavelet basis function and the oth candidate second wavelet basis function at the initial optimization frequency scale c, and Z represents the total number of candidate second wavelet basis functions at the initial optimization frequency scale c.

7. A method for processing vibration signals of port equipment according to claim 1, characterized in that: The comprehensive function is: ； In the formula, Represents a comprehensive consideration of the initial frequency scale d, the second wavelet basis function , Threshold The comprehensive function of represents the concentration of energy distribution on the i-th initial frequency scale d, represents the conversion function, represents the orthogonal value of the jth second wavelet basis function, represents the computational complexity of the jth second wavelet basis function, represents the signal-to-noise ratio of the time domain signal reconstructed after the kth threshold processing and the original signal, represents the sparse value of the optimized second wavelet coefficient after the kth threshold processing, , , Represent the adjustment coefficients, .

8. A method for processing vibration signals of port equipment according to claim 7, characterized in that: The conversion function is: ； In the formula, represents the conversion value of the orthogonal value of the jth second wavelet basis function, z represents a positive number, z represents the evaluation threshold, Represents the reward item, Represents a penalty term.

Images