A diesel engine vibration signal feature extraction method and system based on fuzzy entropy

Abstract

The application provides a diesel engine vibration signal feature extraction method and system based on fuzzy entropy, which comprises the following steps: reconstructing a phase space of a diesel engine vibration signal based on a given embedding dimension m and a time delay to generate a reconstructed vector; calculating a Chebyshev distance based on the reconstructed vector and any other reconstructed vector, inputting the Chebyshev distance into an exponential fuzzy membership function to calculate a single-point similarity; calculating the average similarity of the reconstructed vector and all other reconstructed vectors based on the single-point similarity; iteratively calculating the average similarity of m+1-dimensional vectors to output a fuzzy entropy scalar; and using the fuzzy entropy scalar to represent the complexity of the diesel engine vibration signal. The application enhances the quantification ability of signal complexity and irregularity by establishing a fuzzy entropy model, and has higher noise resistance and stability compared with a traditional RMS feature.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to a method and system for extracting features from diesel engine vibration signals based on fuzzy entropy. Background Technology

[0002] As a critical power source, the monitoring of diesel engine operating status is of great significance for ensuring the reliability of major equipment. Under complex operating conditions, diesel engine vibration signals exhibit significant nonlinear and non-stationary characteristics, making it difficult for traditional time-domain and frequency-domain feature extraction methods to effectively characterize their dynamic behavior. This technical challenge has spurred research into feature extraction methods based on nonlinear dynamics, among which entropy features have attracted considerable attention due to their ability to quantify system complexity and disorder.

[0003] Currently, the application of entropy analysis methods in diesel engine fault diagnosis faces three main challenges: First, single-scale entropy is difficult to comprehensively characterize the dynamic characteristics of multi-scale coupling, especially under non-stationary operating conditions such as variable speed. Second, the sensitivity of traditional sample entropy to short data sequences leads to insufficient stability in early and weak fault diagnosis. Third, existing multi-scale entropy methods have a contradiction between computational efficiency and feature discrimination, affecting their engineering practicality.

[0004] Scholars have made some progress in entropy feature extraction. One research team proposed a composite multi-scale entropy algorithm, which improves feature stability by refining the coarsening process, but it is still insufficient for capturing high-speed impact features. Other scholars have combined fuzzy entropy with recursive quantitative analysis to enhance feature noise resistance, but its computational complexity still needs optimization. It is worth noting that most existing methods are based on single sensor signals and fail to fully utilize the complementary information from multi-source heterogeneous data. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a method and system for extracting diesel engine vibration signal features based on fuzzy entropy. By establishing a fuzzy entropy model, the method enhances the quantification capability for signal complexity and irregularity, exhibiting higher noise resistance and stability compared to traditional RMS features.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for extracting features from diesel engine vibration signals based on fuzzy entropy, comprising: Based on a given embedding dimension m and time delay, the phase space of the diesel engine vibration signal is reconstructed to generate a reconstruction vector. Based on reconstruction vector Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into the exponential fuzzy membership function to calculate the single-point similarity; Calculate the reconstructed vector based on single-point similarity. Average similarity with all other reconstructed vectors; The average similarity of the m+1 dimensional vectors is calculated iteratively, and the fuzzy entropy scalar is output; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal.

[0007] Secondly, the present invention provides a diesel engine vibration signal feature extraction system based on fuzzy entropy, comprising: The phase space reconstruction module is used to reconstruct the phase space of the diesel engine vibration signal based on a given embedding dimension m and time delay, and generate a reconstruction vector. The single-point similarity calculation module is used to calculate similarity based on the reconstructed vector. Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into the exponential fuzzy membership function to calculate the single-point similarity; The average similarity calculation module is used to calculate the reconstructed vector based on single-point similarity. Average similarity with all other reconstructed vectors; The fuzzy entropy scalar output module is used to iteratively calculate the average similarity of m+1 dimensional vectors and output a fuzzy entropy scalar; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal.

[0008] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for extracting diesel engine vibration signals based on fuzzy entropy as described in the first aspect.

[0009] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the method for extracting diesel engine vibration signal features based on fuzzy entropy described in the first aspect.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention enhances the quantification capability of diesel engine vibration signals by constructing a fuzzy entropy model, exhibiting superior noise resistance and stability compared to traditional RMS features. Phase space reconstruction maps one-dimensional signals to a high-dimensional space, uncovering hidden dynamic characteristics and providing rich information for feature extraction. Chebyshev distance combined with an exponential fuzzy membership function calculates single-point similarity, avoiding absolute boundary definition problems and flexibly handling noise and short time series. The fuzzy entropy scalar accurately characterizes signal complexity by statistically analyzing a large number of vector similarities, and its variation pattern is consistent across different time periods. Experiments show that in processing vibration signals of cylinder A1, gearbox axial, and base z-axis, this method achieves standard deviation and variance values ​​smaller than RMS, exhibiting low data dispersion and high stability, providing an effective means for diesel engine operating status monitoring and fault diagnosis.

[0011] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0012] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0013] Figure 1 A flowchart illustrating the main steps of a diesel engine vibration signal feature extraction method based on fuzzy entropy, as provided in this embodiment of the invention. Figure 2 The following are schematic diagrams of the time-domain waveform and spectrum of the vibration signal of cylinder A1 provided in the embodiments of the present invention: (a) is the time-domain waveform of the signal of cylinder A1; (b) is the spectrum of the signal of cylinder A1. Figure 3 This is a schematic diagram comparing the eigenvalues / mean values ​​of the A1 cylinder vibration signal using fuzzy entropy (FE) and root mean square (RMS) in an embodiment of the present invention. Figure 4 A schematic diagram comparing the sliding standard deviation / mean of fuzzy entropy FE and root mean square RMS for the vibration signal of cylinder A1 provided in this embodiment of the invention; Figure 5 The following are schematic diagrams of the time-domain waveform and spectrum of the axial vibration signal of the gear transmission box provided in the embodiments of the present invention: (a) is a time-domain waveform of the axial signal of the gear transmission box; (b) is a spectrum of the axial signal of the gear transmission box. Figure 6 A schematic diagram comparing the eigenvalues / mean values ​​of fuzzy entropy (FE) and root mean square (RMS) of the axial vibration signal of the gear transmission box provided in this embodiment of the invention. Figure 7A schematic diagram comparing the sliding standard deviation / mean of fuzzy entropy (FE) and root mean square (RMS) of the axial vibration signal of the gear transmission box provided in this embodiment of the invention. Figure 8 The following are schematic diagrams of the time-domain waveform and spectrum of the base z-axis vibration signal provided in the embodiments of the present invention: (a) is the time-domain waveform of the base z-axis signal; (b) is the spectrum of the base z-axis signal. Figure 9 A schematic diagram comparing the eigenvalues / mean values ​​of the base z-axis vibration signal provided in this embodiment of the invention using fuzzy entropy (FE) and root mean square (RMS). Figure 10 A schematic diagram comparing the sliding standard deviation / mean of the z-axis vibration signal of the base provided in the embodiments of the present invention using fuzzy entropy (FE) and root mean square (RMS). Detailed Implementation

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Example 1 Combination Figure 1 This embodiment provides a detailed description of a diesel engine vibration signal feature extraction method based on fuzzy entropy.

[0016] During the operation of a diesel engine, the vibration of its various components (such as cylinders, gearboxes, and bases) generates dynamic electrical signals. These signals can be collected by accelerometers or vibration sensors placed in key parts of the diesel engine.

[0017] Taking a 16-cylinder diesel engine as an example, piezoelectric vibration sensors are installed on the surface of the cylinder block of its A1 cylinder, the axial end face of the gear transmission box, and the z-axis direction of the base, respectively. The sensors convert mechanical vibration into voltage signals, which are then converted from analog to digital by the data acquisition system at a certain sampling frequency (such as 10kHz) to obtain discrete digital signal sequences.

[0018] During the acquisition process, to ensure signal integrity, the duration of each acquisition is usually set (e.g., 10 seconds), thus obtaining an original time series of length N (N = sampling frequency × acquisition time, e.g., N=100000 when sampling at 10kHz).

[0019] The acquired signals need to be pre-filtered first. By using a bandpass filter to remove environmental noise and irrelevant frequency band interference, the characteristic frequency band of diesel engine vibration (such as 100-5000Hz) is retained. Finally, time series data of length N can be obtained for subsequent analysis, providing the original input for subsequent phase space reconstruction and fuzzy entropy calculation.

[0020] Traditional methods for feature extraction from diesel engine vibration signals have significant limitations in handling nonlinear and non-stationary signals. While phase space reconstruction can map a one-dimensional time series to a higher-dimensional space to uncover dynamic characteristics, more effective methods are still needed for subsequent feature quantization. Therefore, fuzzy entropy is introduced. It can quantify the regularity and complexity of signals from a new perspective, better adapting to the nonlinear and non-stationary characteristics of diesel engine vibration signals. This provides a new approach to overcoming the limitations of traditional methods, thus enabling more accurate characterization of signal features.

[0021] S1. Based on the given embedding dimension m and time delay, the phase space of the diesel engine vibration signal is reconstructed to generate a reconstruction vector.

[0022] Specifically, for lengths of diesel engine vibration signal time series Given the embedding dimension and time delay Perform phase space reconstruction and construct the reconstruction vector: ; In the formula, m represents the dimension of the eigenvectors in the phase space, mapping a one-dimensional time series to a high-dimensional phase space, which helps to uncover the dynamic characteristics of the series; time delay The value is 1; Input diesel engine vibration signal time series The length.

[0023] Wherein, embedding dimension Used to determine the basic dimensions of the reconstructed phase space, its core function is to unfold a one-dimensional time series in a higher-dimensional space in order to fully reveal the nonlinear features contained in the original dynamic system that are folded or hidden in the one-dimensional signal.

[0024] It is important to note that choosing the appropriate Value is crucial. If If the value is too small, the reconstructed phase space cannot fully accommodate the dynamic structure of the original system, causing the trajectories of different states to intersect, which will affect the correct assessment of the system's complexity. The value should not be too large. Too large a value... This would unnecessarily increase computational cost. More importantly, for time series of finite length, excessively high dimensionality leads to extremely sparse data points in high-dimensional space, reducing the statistical significance of distances between vectors and resulting in unstable and unreliable calculated fuzzy entropy values. Those skilled in the art can determine the appropriate dimensionality by comprehensively experimenting, analyzing the dynamic characteristics of the sequence, and balancing the impact of computational cost and data sparsity. This embodiment is not limited here.

[0025] Phase space reconstruction is used to map a one-dimensional diesel engine vibration signal time series to a high-dimensional phase space. This breaks the limitations of the signal in one-dimensional space and more comprehensively reveals the dynamic characteristics of the signal. This embodiment performs phase space reconstruction based on the diesel engine vibration signal time series, which can reveal the hidden nonlinear structure and dynamic evolution law in the signal in a high-dimensional space, providing richer feature information for subsequent fuzzy entropy calculation. This is achieved by appropriately selecting the embedding dimension *m* and the time delay. This ensures that the reconstructed phase space accurately reflects the dynamic characteristics of the original signal, making the extracted features more representative and laying a good foundation for diesel engine operating status monitoring and fault diagnosis.

[0026] S2, Based on Reconstructed Vectors Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into an exponential fuzzy membership function to calculate the single-point similarity.

[0027] Specifically, reconstructing vectors and The distance is defined as the maximum absolute difference between its corresponding components, i.e., the Chebyshev distance: ; in, Represents the reconstructed vector and The Chebyshev distance is calculated as follows: k is an integer index variable of a vector component in the Chebyshev distance calculation, used to iterate through all vector components from 0 to m-1 to ensure that the differences at all corresponding positions are compared; j is used to iterate through the set of reconstructed vectors and, together with i, selects different reconstructed vectors. The index parameter.

[0028] This embodiment calculates single-point similarity using Chebyshev distance and an exponential fuzzy membership function. This calculation method fully considers the maximum absolute difference between the components of the vectors, enabling it to more sensitively capture the differences between vectors. Simultaneously, the introduction of the fuzzy membership function makes the calculation of single-point similarity not a simple binary judgment, but rather fuzzy and continuous, allowing for a more nuanced description of the degree of similarity between vectors. This is highly advantageous for processing noise and uncertainty in diesel engine vibration signals, improving the accuracy and anti-interference capability of feature extraction, and making the extracted features more reflective of the signal's essence.

[0029] Furthermore, a fuzzy membership function is introduced. To characterize the similarity between vectors, an exponential function is used: ; in, It is a fuzzy threshold. This represents the gradient of the membership function.

[0030] In this embodiment, an exponential membership function is constructed to provide a more flexible way to calculate the similarity between vectors. The exponential function allows for rapid adjustment of the membership value based on the Chebyshev distance. When the distance is small, the membership value quickly approaches 1, indicating high vector similarity; when the distance is large, the membership value quickly approaches 0, indicating low vector similarity. This rapid response effectively highlights key features in the signal and suppresses the influence of noise. Furthermore, by adjusting parameters such as the fuzzy threshold, the gradient of the membership function can be optimized according to the actual signal characteristics, making the similarity calculation more consistent with the characteristics of diesel engine vibration signals and further improving the feature extraction effect.

[0031] S3. Calculate the reconstructed vector based on single-point similarity. The average similarity with all other reconstructed vectors.

[0032] For each Calculate the fuzzy similarity between it and other vectors and take the average: ; Traditional methods for calculating similarity often employ absolute boundary definitions, which have significant drawbacks when processing signals. When faced with noise interference or short time series, absolute boundaries make similarity calculations rigid and unable to accurately reflect the true relationships between samples. Fuzzy membership functions, however, differ. They use an exponential function to continuously calculate single-point similarity based on Chebyshev distance, using membership values ​​to characterize the degree of similarity between vectors. This breaks the limitations of absolute boundaries, allowing for more flexible handling of various complex situations. In noisy environments, its sensitivity to distance can be adjusted through gradient parameters, reducing the impact of noise on similarity calculations. For short time series, it avoids feature distortion caused by small data volumes and strict absolute boundary definitions, making similarity calculations more robust and improving the stability and reliability of feature extraction.

[0033] S4. Iteratively calculate the average similarity of the m+1 dimensional vectors and output the fuzzy entropy scalar; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal.

[0034] The fuzzy entropy proposed in this embodiment is a method for measuring the regularity of time series, defined as the negative natural logarithm of the fuzzy similarity between samples that meet certain conditions.

[0035] Specifically, through construction Repeat the above steps for the vector to obtain the result. The ratio on the right-hand side of the equation essentially represents a conditional probability, that is, the probability that two vectors, if similar in m-dimensional space, will still be similar in m+1-dimensional space. -ln is the negative natural logarithm, defined as the negative natural logarithm of the ratio of the average similarity of vectors in m+1-dimensional space to the average similarity of vectors in m-dimensional space. The formula for calculating fuzzy entropy is as follows: ; The complexity of diesel engine vibration signals is reflected in the irregularity and diversity of their underlying dynamic behavior.

[0036] Fuzzy entropy maps signals to a high-dimensional space through phase space reconstruction, capturing the dynamic characteristics of signals from a multi-dimensional perspective. During the calculation, an exponential fuzzy membership function is used to calculate the similarity between vectors. This method can finely characterize the degree of similarity between different patterns in the signal. When the signal complexity is high, its internal pattern differences are large, the similarity between vectors is low, and the fuzzy entropy scalar is large; conversely, when the signal complexity is low, the patterns are relatively regular, the similarity is high, and the fuzzy entropy scalar is small. Moreover, the fuzzy entropy scalar is derived through statistical calculation of a large number of vector similarities, and can comprehensively reflect the overall irregularity and dynamic change of the signal. Therefore, the fuzzy entropy scalar is used to characterize the complexity of diesel engine vibration signals.

[0037] Example 1 To verify the effectiveness of this embodiment, taking the vibration signal of a 16-cylinder diesel engine as an example, the fuzzy entropy feature extraction algorithm is used to extract features from the vibration signals of cylinder A1, the axial direction of the gear transmission box, and the z-axis of the base; the RMS of the signal samples is used as the comparison feature of the fuzzy entropy.

[0038] Among them, cylinder A1, the axial direction of the gearbox, and the z-axis of the base are typical representatives of a diesel engine system. Cylinder A1, as the core working unit, reflects key operating conditions such as combustion and piston movement; the axial direction of the gearbox is related to power transmission stability, gear meshing, and shaft vibration; the z-axis of the base reflects the overall machine foundation vibration, involving equipment installation and structural rigidity. Furthermore, these three correspond to key subsystems of the diesel engine's core working, power transmission, and foundation support, respectively, and are most sensitive to faults such as cylinder block anomalies, transmission component failures, and structural loosening. Selecting them allows for comprehensive verification of the algorithm's ability to extract vibration characteristics from different functional modules, as well as its effectiveness in fault feature capture scenarios.

[0039] (1) Cylinder A1 The time-domain waveform and spectrum of the vibration signal of cylinder A1 are as follows: Figure 2 As shown: from Figure 2 As can be seen, the vibration signal of cylinder A1 is quite complex, and it still has strong energy in the mid-to-high frequency range.

[0040] The fuzzy entropy (hereinafter referred to as FE), traditional root mean square (hereinafter referred to as RMS) eigenvalues, and the dimensionalized results of the moving standard deviation (ratio to the mean) proposed in this embodiment are as follows: Figure 3 , 4 As shown, the ratios of eigenvalues ​​and moving standard deviations to the mean indicate that the fluctuation range of FE is extremely low, all falling between 0.9 and 1.1, and the standard deviation curve shows a stable trend with no significant fluctuations, which are significantly smaller than those of RMS. This suggests that the fuzzy entropy exhibits a consistent pattern of change across different time periods, demonstrating good stability.

[0041] As shown in Table 1, the standard deviation and variance of FE reflect the absolute dispersion of the data. Both of these values ​​are less than RMS, proving the consistency of fuzzy entropy under experimental conditions. The fluctuation range of FE is also less than RMS, representing a lower degree of dispersion. Furthermore, the coefficient of variation (COP), which measures relative dispersion, is much less than 0.1 for FE, with a COP of only 0.0345 in the table. This indicates that the fluctuation of the fuzzy entropy data accounts for only 3.45% of its mean, and is less than the COP coefficient of variation for RMS, demonstrating the method's very high stability. Therefore, fuzzy entropy exhibits better anti-interference capability and data stability than RMS when processing diesel engine vibration signals.

[0042] Table 1 Comparison of standard deviation, variance, mean, fluctuation range and coefficient of variation of A1 cylinder signal FE and RMS;

[0043] (2) Axial gearbox The time-domain waveform and spectrum of the axial vibration signal of the gearbox are as follows: Figure 5 As shown, the low frequency has significant harmonic components, making it quite complex.

[0044] The eigenvalues ​​of FE and RMS, and the results of the moving standard deviation after unification of dimensions (ratio to the mean) are as follows: Figure 6 , 7 As shown, the ratio of eigenvalues ​​to the mean indicates that the fluctuation range of FE is extremely low, roughly between 0.95 and 1.05, and the standard deviation curve shows a stable trend with no significant fluctuations. In contrast, the fluctuation of RMS is significantly greater than that of FE, a fact also verified by the ratio plot of the moving standard deviation to the mean. This suggests that the fuzzy entropy exhibits a consistent pattern of change across different time periods, demonstrating good stability.

[0045] The fuzzy entropy stability index of the axial vibration signal samples of the gearbox is shown in Table 2: the standard deviation and variance of FE are both less than RMS, proving the consistency of fuzzy entropy under experimental conditions. The fluctuation amplitude of FE is also less than that of RMS, representing a lower degree of dispersion. Furthermore, the coefficient of variation of FE, which measures relative dispersion, is much less than 1 and less than that of RMS, indicating that this method has very high stability. Therefore, fuzzy entropy has better anti-interference ability and data stability than RMS when processing diesel engine vibration signals.

[0046] Table 2 Standard deviation, variance, mean, fluctuation range, and coefficient of variation of axial signals FE and RMS of gear transmission box;

[0047] (3) Z-axis of the base The time-domain waveform and spectrum of the vibration signal of the base z-axis are as follows: Figure 8 As shown: The signal is quite complex, especially with strong energy in the mid-to-low frequency range.

[0048] Table 3 Standard deviation, variance, mean, fluctuation amplitude, and coefficient of variation of the z-axis vibration signals FE and RMS of the base;

[0049] The eigenvalues ​​of FE and RMS, and the results of the moving standard deviation after unification of dimensions (ratio to the mean) are as follows: Figure 9 , 10 As shown, the ratio of eigenvalues ​​to the mean indicates that the fluctuation range of FE is extremely low, roughly between 0.97 and 1.03, and the standard deviation curve shows a stable trend with no significant fluctuations. In contrast, the fluctuation of RMS is significantly greater than that of FE, a finding also verified by the ratio plot of the moving standard deviation to the mean. This suggests that the fuzzy entropy exhibits a consistent pattern of change across different time periods, demonstrating good stability.

[0050] As shown in Tables 1-3, the standard deviation and variance of FE reflect the absolute dispersion of the data. Both values ​​are less than RMS, demonstrating the consistency of fuzzy entropy under experimental conditions. The fluctuation range of FE is also less than RMS, representing a lower degree of dispersion. Furthermore, the coefficient of variation (COP), which measures relative dispersion, is much less than 0.1 for FE. Taking the fuzzy entropy of the base z-axis as an example, the COP in Table 3 is only 0.0134. This indicates that the fluctuation of the fuzzy entropy data accounts for only 1.34% of its mean, and is less than the COP coefficient of variation for RMS, indicating that this method has very high stability. Therefore, fuzzy entropy has better anti-interference ability and data stability than RMS when processing diesel engine vibration signals.

[0051] also, Figure 3 , 4The eigenvalues ​​and the ratio of the moving standard deviation to the mean in values ​​6, 7, 9, and 10 show that the fluctuation range of FE is extremely low, the standard deviation curve shows a stable trend with no significant fluctuations, and the fluctuation of the moving standard deviation is significantly smaller than that of RMS. This indicates that the fuzzy entropy changes in a consistent manner over different time periods, exhibiting good stability.

[0052] This specific embodiment addresses the limitations of existing entropy analysis methods in diesel engine fault diagnosis, including limited single-scale entropy representation capabilities, insufficient stability of traditional sample entropy, the trade-off between computational efficiency and feature discriminative power in multi-scale entropy calculations, and underutilization of multi-source heterogeneous data. It proposes a diesel engine vibration signal feature extraction method based on fuzzy entropy. This method enhances the quantification capability for signal complexity and irregularity through phase space reconstruction and fuzzy entropy calculation. In handling multi-scale coupled dynamic characteristics, fuzzy entropy can characterize signals from a more comprehensive perspective; for short data sequences, its stability is superior to traditional sample entropy; and it achieves a better balance between computational efficiency and feature discriminative power. Furthermore, this method provides a new approach for the fusion and application of multi-source heterogeneous data, improving the accuracy and engineering practicality of diesel engine fault diagnosis, effectively solving the problems existing in current technologies, and promoting the development of diesel engine vibration signal feature extraction technology.

[0053] Example 2 This embodiment provides a diesel engine vibration signal feature extraction system based on fuzzy entropy, including: The phase space reconstruction module is used to reconstruct the phase space of the diesel engine vibration signal based on a given embedding dimension m and time delay, and generate a reconstruction vector. The single-point similarity calculation module is used to calculate similarity based on the reconstructed vector. Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into the exponential fuzzy membership function to calculate the single-point similarity; The average similarity calculation module is used to calculate the reconstructed vector based on single-point similarity. Average similarity with all other reconstructed vectors; The fuzzy entropy scalar output module is used to iteratively calculate the average similarity of m+1 dimensional vectors and output a fuzzy entropy scalar; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal.

[0054] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the diesel engine vibration signal feature extraction method based on fuzzy entropy as described in Embodiment 1 above.

[0055] Example 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the diesel engine vibration signal feature extraction method based on fuzzy entropy as described in Embodiment 1 above.

[0056] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for extracting features from diesel engine vibration signals based on fuzzy entropy, characterized in that, include: Based on a given embedding dimension m and time delay, the phase space of the diesel engine vibration signal is reconstructed to generate a reconstruction vector. Based on reconstruction vector Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into the exponential fuzzy membership function to calculate the single-point similarity; Calculate the reconstructed vector based on single-point similarity. The average similarity with all other reconstructed vectors; specifically: ； in, It is single-point similarity. Represents the reconstructed vector and Chebyshev distance, It is the fuzzy threshold; m represents the dimension of the feature vector in the phase space. Input diesel engine vibration signal time series The length of i; j is used to traverse the set of reconstructed vectors and, in conjunction with i, select different reconstructed vectors. The index parameter; The average similarity of m+1 dimensional vectors is calculated iteratively, and a fuzzy entropy scalar is output; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal; specifically: ； in, , These represent the average similarity in m dimensions and m+1 dimensions, respectively.

2. The diesel engine vibration signal feature extraction method based on fuzzy entropy as described in claim 1, characterized in that, The step of reconstructing the phase space of the diesel engine vibration signal based on a given embedding dimension m and time delay to generate a reconstruction vector specifically includes: For length of diesel engine vibration signal time series Given the embedding dimension and time delay Perform phase space reconstruction and construct the reconstruction vector: ； Where m represents the dimension of the eigenvectors in the phase space, mapping a one-dimensional time series to a high-dimensional phase space; time delay The value is 1.

3. The diesel engine vibration signal feature extraction method based on fuzzy entropy as described in claim 1, characterized in that, The reconstruction vector-based The Chebyshev distance is calculated with any other reconstructed vector, and the Chebyshev distance is input into an exponential fuzzy membership function to calculate the single-point similarity, specifically including: Reconstructed vector and The distance is defined as the maximum absolute difference between its corresponding components, i.e., the Chebyshev distance: ； in, Represents the reconstructed vector and The Chebyshev distance is calculated as follows: k is an integer index variable for a vector component in the Chebyshev distance calculation, used to iterate through all vector components from 0 to m-1; j is used to iterate through the set of reconstructed vectors and, together with i, selects different reconstructed vectors. The index parameter; The Chebyshev distance Calculate single-point similarity using an exponential fuzzy membership function: ； in, It is a fuzzy threshold. This represents the gradient of the membership function.

4. A diesel engine vibration signal feature extraction system based on fuzzy entropy, characterized in that, include: The phase space reconstruction module is used to reconstruct the phase space of the diesel engine vibration signal based on a given embedding dimension m and time delay, and generate a reconstruction vector. The single-point similarity calculation module is used to calculate similarity based on the reconstructed vector. Calculate the Chebyshev distance with any other reconstructed vector, and input the Chebyshev distance into the exponential fuzzy membership function to calculate the single-point similarity; The average similarity calculation module is used to calculate the reconstructed vector based on single-point similarity. The average similarity with all other reconstructed vectors; specifically: ； in, It is single-point similarity. Represents the reconstructed vector and Chebyshev distance, It is the fuzzy threshold; m represents the dimension of the feature vector in the phase space. Input diesel engine vibration signal time series The length of i; j is used to traverse the set of reconstructed vectors and, in conjunction with i, select different reconstructed vectors. The index parameter; The fuzzy entropy scalar output module is used to iteratively calculate the average similarity of m+1 dimensional vectors and output a fuzzy entropy scalar; the fuzzy entropy scalar is used to characterize the complexity of the diesel engine vibration signal; specifically: ； in, , These represent the average similarity in m dimensions and m+1 dimensions, respectively.

5. The diesel engine vibration signal feature extraction system based on fuzzy entropy as described in claim 4, characterized in that, The step of reconstructing the phase space of the diesel engine vibration signal based on a given embedding dimension m and time delay to generate a reconstruction vector specifically includes: For length of diesel engine vibration signal time series Given the embedding dimension and time delay To reconstruct the phase space, construct a vector sequence: ； Where m represents the dimension of the eigenvectors in the phase space, mapping a one-dimensional time series to a high-dimensional phase space; time delay The value is 1.

6. The diesel engine vibration signal feature extraction system based on fuzzy entropy as described in claim 4, characterized in that, The reconstruction vector-based The Chebyshev distance is calculated with any other reconstructed vector, and the Chebyshev distance is input into an exponential fuzzy membership function to calculate the single-point similarity, specifically including: Reconstructed vector and The distance is defined as the maximum absolute difference between its corresponding components, i.e., the Chebyshev distance: ； in, Represents the reconstructed vector and The Chebyshev distance is calculated as follows: k is an integer index variable for a vector component in the Chebyshev distance calculation, used to iterate through all vector components from 0 to m-1; j is used to iterate through the set of reconstructed vectors and, together with i, selects different reconstructed vectors. The index parameter; The Chebyshev distance Calculate single-point similarity using an exponential fuzzy membership function: ； in, It is a fuzzy threshold. This represents the gradient of the membership function.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the diesel engine vibration signal feature extraction method based on fuzzy entropy as described in any one of claims 1-3.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the diesel engine vibration signal feature extraction method based on fuzzy entropy as described in any one of claims 1-3.

Images