Method for fault diagnosis of diesel engine under sample imbalance, and apparatus, medium and device

Abstract

Disclosed in the present invention is a method for fault diagnosis of a diesel engine under sample imbalance. The method comprises: acquiring target vibration signals of a diesel engine, wherein the target vibration signals include target vibration signals under a normal state and target vibration signals under different fault states; performing wavelet-packet-based noise-assisted compressive sensing sample augmentation processing on the target vibration signals, in order to obtain augmented signal samples; performing a continuous wavelet transform on the augmented signal samples, in order to obtain time-frequency maps; on the basis of an edge detection algorithm, performing grayscaling on the time-frequency maps, calculating edge densities, and on the basis of states of the target vibration signals, constructing edge density vectors; and on the basis of the edge density vectors, constructing edge density threshold values for different states, calculating an edge density of a time-frequency map of a signal to be subjected to detection, and comparing the edge density with the edge density threshold values, in order to obtain the state of the diesel engine. The present invention improves the sample balance and the fault diagnosis capability of marine diesel engines, and is of great significance for ensuring the safe navigation of vessels, reducing the operating costs and extending the service life of devices.

Description

A method, apparatus, medium, and equipment for diagnosing diesel engine faults with imbalanced sample conditions. Technical Field

[0001] This invention belongs to the field of diesel engine diagnostic technology, and particularly relates to a method, apparatus, medium and equipment for diagnosing diesel engine faults with sample imbalance. Background Technology

[0002] Marine diesel engines, as the core equipment of ship propulsion systems, play a vital role in the shipping industry. They are widely used in various large vessels such as merchant ships, tankers, and cargo ships, providing powerful propulsion and stable power support. However, marine diesel engines operate for extended periods in harsh marine environments, facing high loads, high temperatures, and seawater corrosion, making them prone to various malfunctions. Therefore, fault diagnosis of marine diesel engines is necessary. However, fault diagnosis methods for marine diesel engines require extensive data to adjust the model. Since diesel engines spend a significant amount of time in normal operating conditions, the data on fault states is far less than that of normal states, causing a data sample imbalance. This leads to inaccurate fault model construction, making it difficult to detect faults in a timely manner, thus affecting navigation efficiency, and in severe cases, even causing serious equipment damage and safety accidents. Summary of the Invention

[0003] This invention proposes a method, apparatus, medium, and equipment for diagnosing diesel engine faults with sample imbalance, in order to solve the problems existing in the prior art.

[0004] To achieve the above objectives, the present invention provides a method for diagnosing diesel engine faults due to sample imbalance, comprising the following steps:

[0005] Acquire the target vibration signal of the diesel engine, which includes the target vibration signal under normal conditions and the target vibration signal under different fault conditions;

[0006] The target vibration signal is subjected to wavelet packet noise compression sensing sample enhancement processing to obtain enhanced signal samples;

[0007] Perform continuous wavelet transform on the enhanced signal samples to obtain a time-frequency diagram;

[0008] The time-frequency image is grayscaled according to the edge detection algorithm, and the edge density is calculated. An edge density vector is constructed based on the state of the target vibration signal.

[0009] Based on the edge density vector, edge density thresholds for different states are constructed. The edge density of the time-frequency map of the signal to be detected is calculated and compared with the edge density thresholds to obtain the state of the diesel engine.

[0010] Preferably, wavelet packet noise compression sensing sample enhancement includes:

[0011] The target vibration signal is decomposed by wavelet packet decomposition to obtain several sub-signals;

[0012] A sub-signal is encoded using compressed sensing coding, and a Gaussian mixture noise is constructed and superimposed on the encoded sub-signal.

[0013] The superimposed signal is decoded using compressed sensing decoding to obtain the enhanced component;

[0014] The enhanced component is then reconstructed with other unenhanced components.

[0015] Preferably, the wavelet packet decomposition expression is:

[0016] In the formula, x(t) represents the signal to be decomposed, J is the number of decomposition levels, and 2 j It is the number of sub-signals in each layer, d j,k It is the k-th sub-signal of the j-th layer, φ j,k It is the basis function of the k-th wavelet packet of the j-th layer.

[0017] Preferably, the expression for encoding operations using compressed sensing coding is:

[0018] y=Φd j,k ;

[0019] In the formula, Φ represents the measurement matrix, and d j,k It is the k-th sub-signal of the j-th layer.

[0020] Preferably, the expression for the Gaussian mixed noise is:

[0021] In the formula, g represents the generated Gaussian mixture noise, z represents the number of Gaussian distributions, and w i Let η represent the weights of the i-th Gaussian distribution, and let μ represent the weights of the distribution. i With σ i The generated Gaussian noise, μ i Let σ represent the mean of the i-th Gaussian distribution. i Let represent the standard deviation of the i-th Gaussian distribution.

[0022] Preferably, the expression for the continuous wavelet transform is:

[0023] In the formula, x(t) is a one-dimensional time-domain signal, Ψ(*) is a wavelet basis function, t represents time, a is a scale parameter, and b is the transformation form of the translation parameter.

[0024] Preferably, calculating the edge density includes:

[0025] The time-frequency graph is converted to grayscale using a luminance method, and Gaussian filtering is applied to remove image noise.

[0026] Image edges are obtained by calculating image gradients using the Sobel operator;

[0027] Calculate the percentage of edge pixels to obtain the edge density.

[0028] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0029] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0030] The present invention also provides an electronic device, comprising: a memory and a processor; the memory for storing a program; and the processor for executing the program to implement the various steps of the method.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] This invention discloses a method for diagnosing diesel engine faults due to sample imbalance, comprising: acquiring a target vibration signal of the diesel engine, wherein the target vibration signal includes a target vibration signal under normal conditions and target vibration signals under different fault conditions; performing wavelet packet noise compression sensing sample enhancement processing on the target vibration signal to obtain enhanced signal samples; performing continuous wavelet transform on the enhanced signal samples to obtain a time-frequency map; converting the time-frequency map to grayscale according to an edge detection algorithm and calculating the edge density, constructing an edge density vector according to the state of the target vibration signal; constructing edge density thresholds for different states based on the edge density vector, calculating the edge density of the time-frequency map of the signal to be detected and comparing it with the edge density thresholds to obtain the state of the diesel engine. This invention improves the sample balance and fault diagnosis capabilities of marine diesel engines, which is of great significance for ensuring safe navigation of ships, reducing operating costs, and extending equipment service life. Attached Figure Description

[0033] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0034] Figure 1 is a flowchart of a method according to an embodiment of the present invention;

[0035] Figure 2 shows a normal sample image according to an embodiment of the present invention, wherein (a) is an image of the normal sample enhancement effect and (b) is the original image of the normal sample;

[0036] Figure 3 shows an abnormal sample image according to an embodiment of the present invention, where (a) is an image of the abnormal sample enhancement effect and (b) is the original image of the abnormal sample. Detailed Implementation

[0037] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0038] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0039] The following is an introduction to the technologies involved:

[0040] Wavelet packet decomposition is a technique for multi-scale and multi-frequency signal analysis. It recursively decomposes a signal into sub-signals containing different frequency components, thereby enabling in-depth extraction of local signal features. This decomposition method is highly flexible, allowing users to select different wavelet basis functions and decomposition depths as needed, making it particularly suitable for analyzing complex nonlinear and non-stationary signals. Wavelet packet decomposition has wide applications in signal processing, image analysis, audio analysis, and biomedical signal analysis, helping us to more accurately understand and process various signals by providing detailed frequency and time information.

[0041] Continuous wavelet transform (CWT) is a powerful signal analysis tool that reveals local features of signals, such as discontinuities, abrupt changes, and spikes, by analyzing signals at different scales and locations. This transform utilizes wavelet functions to probe signals, analogous to using magnifying glasses of varying magnifications to observe signal details. CWT is particularly well-suited for analyzing non-stationary signals, i.e., signals whose statistical properties change over time. It can accommodate the nonlinearity and non-stationarity of signals and increases analytical flexibility by selecting appropriate wavelet basis functions. Continuous wavelet transform has wide applications in signal processing, image analysis, seismology, and financial analysis, becoming a highly valuable analytical tool due to its ability to provide time-frequency information about signals.

[0042] The Sobel operator identifies edges by calculating the gradient magnitude at each pixel in an image. This operator uses two 3x3 convolutional kernels: one to detect horizontal brightness changes and the other to detect vertical brightness changes. Using the gradients in these two directions, the Sobel operator can determine the intensity and direction of the edge. This method is particularly effective for horizontal and vertical edges in images, but its ability to detect diagonal edges is weaker. The Sobel operator is widely used in practical applications due to its simplicity and effectiveness, especially when fast and relatively accurate edge detection is required.

[0043] Example 1

[0044] As shown in Figure 1, this embodiment provides a method for diagnosing diesel engine faults due to sample imbalance, including the following steps:

[0045] Acquire the target vibration signal of the diesel engine, which includes the target vibration signal under normal conditions and the target vibration signal under different fault conditions;

[0046] The target vibration signal is subjected to wavelet packet noise compression sensing sample enhancement processing to obtain the enhanced signal sample;

[0047] Continuous wavelet transform is performed on the enhanced signal samples to obtain the time-frequency diagram;

[0048] The time-frequency image is converted to grayscale based on the edge detection algorithm, and the edge density is calculated. An edge density vector is constructed based on the state of the target vibration signal.

[0049] Edge density thresholds for different states are constructed based on edge density vectors. The edge density of the time-frequency map of the signal to be detected is calculated and compared with the edge density thresholds to obtain the state of the diesel engine.

[0050] The specific steps are as follows:

[0051] 1. Wavelet packet noise compression sensing sample enhancement:

[0052] The vibration signals S collected by the accelerometer under different states have an imbalance in data samples because the normal state time of the equipment is often much longer than the abnormal state time. This imbalance can lead to the data evaluation index being distorted and biased towards the majority type, thus affecting the accuracy. To solve the problem of sample imbalance, the wavelet packet noise compression sensing sample enhancement method is used to enhance the data samples and obtain balanced data samples. The specific implementation method is as follows: a sample signal is obtained by wavelet packet decomposition into n sub-signals. One sub-signal is selected and encoded using compression sensing. A Gaussian mixed noise is constructed and superimposed on the encoded signal. The superimposed signal is then decoded using compression sensing to obtain the enhanced component. Finally, this component is reconstructed with other components to obtain a signal similar to but different from the original signal. The above operation is repeated n times to complete the enhancement of a sample signal. The wavelet packet decomposition method is shown in Equation (1).

[0053] In equation (1), x(t) represents the signal to be decomposed, and J represents the number of decomposition layers, 2. j It is the number of sub-signals in each layer, d j,k It is the k-th sub-signal of the j-th layer. φ j,k It is the basis function of the k-th wavelet packet of the j-th layer, which is the form of the mother wavelet after scaling and shifting. Multiple sub-signals can be obtained through equation (1), and one of the sub-signals d is selected in the process. j,k Compressed sensing encoding is performed, and compressed sensing assumes that signal d j,k It is sparse, and furthermore, by measuring the matrix Φ, the sub-signals d are... j,k Compressed sampling is performed, as shown in equation (2):

[0054] y=Φd j,k (2)

[0055] In the above process, a randomly generated matrix Φ is used for sampling to generate a measurement value y, thereby encoding the signal and sub-signal d. j,k Effective reconstruction is achieved. Based on this encoding, Gaussian mixture noise (with different standard deviations) is added to the signal to achieve differentiation of the sample enhancement signal. The mathematical expression of Gaussian mixture noise is shown in Equation (4). In order to ensure the consistency of signal quality after adding noise and control the same signal-to-noise ratio, it is necessary to calculate the standard deviation of the Gaussian mixture noise signal, as shown in Equation (3).

[0056] In the formula, σ represents the calculated standard deviation of Gaussian noise, and N represents the signal d. j,kThe length of . g represents the generated Gaussian noise, z represents the number of Gaussian distributions, w i μ represents the weight of the i-th Gaussian distribution. i Let σ represent the mean of the i-th Gaussian distribution. i Let represent the standard deviation of the i-th Gaussian distribution. A new signal y is obtained by superimposing the compressed sensing coded signal with Gaussian mixture noise. noisy Based on this, a compressed sensing decoding operation is performed on the signal. The purpose of the decoding operation is to obtain sub-signals that are similar but not identical. The process involves reconstructing the original signal x through an optimization problem. The optimization process is shown in (5).

[0057] The optimization objective of this process is to find an x ​​that minimizes equation (5), where λ is the regularization parameter used to control sparsity, ||·||2 represents the L2 norm, which is the square root of the sum of squares, and ||·||1 represents the L1 norm, which is the sum of absolute values.

[0058] According to equation (1), the processed sub-signal and atomic signal are replaced, and then the signal is reconstructed to obtain a signal sample that is similar to but not the same as the original sample, thereby enhancing the faulty sample. Figures 2 and 3 are schematic diagrams of normal and abnormal samples generated by the sample enhancement method.

[0059] 2. Method for drawing time-frequency graphs:

[0060] Continuous wavelet transform is an effective tool for simultaneously analyzing signals in the time and frequency domains, possessing the advantages of time-frequency localization and multi-scale analysis. It is mainly achieved by performing a continuous inner product on a one-dimensional signal using equation (6), utilizing wavelet basis functions of different scales. To further extract fault feature information from the signal, continuous wavelet transform is used to generate a time-frequency diagram.

[0061] In the formula, y(t) is a one-dimensional time-domain signal, Ψ(*) is the wavelet basis function, a is the scaling parameter, and b is the transformation form of the translation parameter. In the continuous wavelet time-frequency plot, the energy distribution in different frequency ranges can reflect different fault characteristics. Therefore, the continuous wavelet time-frequency plot can be used to extract fault features through the energy distribution in different frequency ranges and the energy concentration area.

[0062] 3. Simple diagnostic method for edge detection:

[0063] In two-dimensional images, by identifying regions with significant brightness changes, these regions correspond to important features in the image. Therefore, edge detection can be used to diagnose faulty two-dimensional images. To achieve edge detection, the image first needs to be converted to grayscale. Image grayscale conversion can be achieved using the brightness method, as shown in equation (7).gray (x,y)=w R ×R(x,y)+w G ×G(x,y)+w B ×B(x,y) (7)

[0064] Where I gray Given a grayscale image, R(x,y), G(x,y), and I(x,y) represent the pixel values ​​for red, green, and blue, respectively. R w G w B Three different weights are used. Based on this, Gaussian filtering is used to smooth the image and remove noise, reducing interference factors in the detection process. The specific method is shown in Equation (8).

[0065] Where ρ represents the standard deviation of the Gaussian kernel, controlling the smoothing degree, and the smoothed image is obtained through the above process. The smoothed signal is then used to perform image gradient calculation using the Sobel operator to calculate the gradient of the image in the x and y directions. The gradient calculation process is shown in equations (9) and (10), where the magnitude and direction of the gradient are calculated as shown in equations (11) and (12). θ(x,y)=atan2(G y (x,y),G x (x,y)) (12)

[0066] Among them G x For the horizontal gradient, G y The gradient is the vertical gradient, representing the edge strength at a point in the image, and θ is the gradient angle. To simplify calculations, the gradient direction is quantized to values ​​such as 0°, 45°, 90°, and 135°. In different gradient directions, if the gradient value of the current pixel is not a local maximum along the gradient direction, it is set to 0; otherwise, the pixel is retained, thus refining the edge. A dual-threshold detection method is used to distinguish between three different types: strong edges, weak edges, and non-edges. A higher threshold T is set... high and low threshold T low The mathematical expression for dual threshold detection is shown in (13).

[0067] After the above processing, in order to measure the density of edge information in the image signal and thus determine the number of fault components, the edge density is used to calculate the proportion of edge pixels. The specific implementation formula is shown in Equation (14).

[0068] Where W and H are the width and height of the image, respectively, and W×H equals the total number of pixels in the image. The edge density information of the fault time-frequency map under different states is calculated using the above method, and its standard deviation σ is calculated.edge and mean μ edge According to the Gaussian distribution principle, 99.7% of the data is concentrated within the range of μ ± 3σ. Therefore, the Gaussian distribution principle can be used to calculate the mean σ of the edge density of multiple images under different conditions. edge With standard deviation μ edge Construct different state thresholds, as shown in Equation (15).

[0069] This embodiment improves the sample balance and fault diagnosis capabilities of marine diesel engines, which is of great significance for ensuring safe navigation of ships, reducing operating costs, and extending equipment service life.

[0070] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0071] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method.

[0072] This embodiment also provides an electronic device, including: a memory and a processor; the memory is used to store a program; the processor is used to execute the program to implement the various steps of the method.

[0073] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for diagnosing diesel engine faults due to sample imbalance, characterized in that, Includes the following steps: Acquire the target vibration signal of the diesel engine, which includes the target vibration signal under normal conditions and the target vibration signal under different fault conditions; The target vibration signal is subjected to wavelet packet noise compression sensing sample enhancement processing to obtain enhanced signal samples; Perform continuous wavelet transform on the enhanced signal samples to obtain a time-frequency diagram; The time-frequency image is grayscaled according to the edge detection algorithm, and the edge density is calculated. An edge density vector is constructed based on the state of the target vibration signal. Based on the edge density vector, edge density thresholds for different states are constructed. The edge density of the time-frequency map of the signal to be detected is calculated and compared with the edge density thresholds to obtain the state of the diesel engine.

2. The method according to claim 1, characterized in that, Wavelet packet noise compression sensing sample enhancement includes: The target vibration signal is decomposed by wavelet packet decomposition to obtain several sub-signals; A sub-signal is encoded using compressed sensing coding, and a Gaussian mixture noise is constructed and superimposed on the encoded sub-signal. The superimposed signal is decoded using compressed sensing decoding to obtain the enhanced component; The enhanced component is then reconstructed with other unenhanced components.

3. The method according to claim 2, characterized in that, The wavelet packet decomposition expression is: In the formula, x(t) represents the signal to be decomposed, J is the number of decomposition levels, and 2 j It is the number of sub-signals in each layer, d j,k It is the k-th sub-signal of the j-th layer, φ j,k It is the basis function of the k-th wavelet packet of the j-th layer.

4. The method according to claim 2, characterized in that, The expression for encoding operations using compressed sensing coding is: y = Φd j,k ; In the formula, Φ represents the measurement matrix, and d j,k It is the k-th sub-signal of the j-th layer.

5. The method according to claim 2, characterized in that, The expression for the Gaussian mixed noise is: In the formula, g represents the generated Gaussian mixture noise, z represents the number of Gaussian distributions, and w i Let η represent the weights of the i-th Gaussian distribution, and let μ represent the weights of the distribution. i With σ i The generated Gaussian noise, μ i Let σ represent the mean of the i-th Gaussian distribution. i Let represent the standard deviation of the i-th Gaussian distribution.

6. The method according to claim 1, characterized in that, The expression for the continuous wavelet transform is: In the formula, x(t) is a one-dimensional time-domain signal, Ψ(*) is a wavelet basis function, t represents time, a is a scale parameter, and b is the transformation form of the translation parameter.

7. The method according to claim 1, characterized in that, Calculating edge density includes: The time-frequency graph is converted to grayscale using a luminance method, and Gaussian filtering is applied to remove image noise. Image edges are obtained by calculating image gradients using the Sobel operator; Calculate the percentage of edge pixels to obtain the edge density.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-7.

10. An electronic device, characterized in that, include: Memory and processor; The memory is used to store a program; the processor is used to execute the program to implement the steps of the method as described in any one of claims 1-7.

Images