A method and system for diagnosing faults of LNG engine gas injection valves

Abstract

The present invention provides a method and system for diagnosing faults in LNG engine gas injection valves. The method comprises: acquiring a pressure fluctuation signal from a gas supply pipe, and dividing the acquired pressure fluctuation signal into a training signal and a test signal; filtering the pressure fluctuation signal using the SOA-CYCBD algorithm to obtain a pressure signal with enhanced impact components; processing the filtered pressure signal to obtain fault characteristics; and after training the fault feature vector, performing fault diagnosis and pattern recognition, and outputting the diagnosis results. The technical solution of the present invention is suitable for diagnosing faults in LNG engine gas injection valves in on-site industrial environments with strong noise interference. It can reduce noise interference, enhance periodic impact components, and improve the accuracy of LNG engine gas injection valve fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical fault diagnosis, and in particular to a method and system for diagnosing faults of a gas injection valve of an LNG engine. Background Art

[0002] The performance of the LNG engine's gas injection valve determines the quality of fuel atomization, thus affecting the entire combustion process. If the gas injection valve malfunctions, poor fuel atomization can lead to deteriorated combustion, resulting in reduced engine power, increased fuel consumption, black smoke, excessive emissions, difficulty starting, and even abnormal operation. Therefore, it is necessary to study and analyze the early signs of gas injection valve failure.

[0003] To extract early-stage fault signatures from mechanical systems, researchers have proposed numerous effective methods, including wavelet transforms, empirical mode decomposition, fuzzy theory, morphological filtering, and blind deconvolution (BD). The process by which the source signal from a gas injection valve fault is transmitted through a signal channel to the sensor can be viewed as a convolution between the source signal and the signal channel. Blind deconvolution theory, however, extracts fault pulses by finding a deconvolution filter that maximizes or minimizes the convolution target. Therefore, blind deconvolution offers unique advantages for processing gas injection valve fault signals. Since Ralph Wiggins proposed minimum entropy deconvolution (MED), deconvolution-based fault diagnosis methods have attracted considerable attention from experts and scholars, rapidly promoting their application in fault diagnosis. However, MED often extracts only one or a few pulses, which can lead to the loss of other impulses. To circumvent this problem, McDonald proposed the MCKD (Maximum Correlated Kurtosis Deconvolution) algorithm. MCKD uses deconvolution operations to highlight continuous impact pulses that are submerged in noise, thereby increasing the associated kurtosis value of the original signal. This makes it suitable for extracting continuous transient impacts from weak fault signals. However, although MCKD can extract periodic pulses, it can only extract a limited number of pulses, and the number of displacements greatly limits the filtering effect of MCKD. Therefore, McDonald proposed Multi-point Optimal Minimum Entropy Deconvolution Adjusted (MOMEDA). The process of solving the inverse filter in MOMEDA is a non-iterative process, which reduces the algorithm's running time. However, while MOMEDA reduces noise, it also significantly reduces the pulse amplitude in the signal.

[0004] Based on this, Marco Buzzoni proposed a new deconvolution method - Maximum Second-order Cyclostationarity Blind Deconvolution (CYCBD). CYCBD overcomes the defects of MED in recovering a single dominant pulse and MCKD in extracting only a limited number of pulses. It can extract continuous periodic pulses very well. Compared with MOMEDA, it can enhance the impact while extracting periodic shocks, and has good noise reduction performance. However, similar to MED, MCKD, and MOMEDA, it recovers the fault source by deconvolution by solving a finite-length filter, and the length of the filter has a great influence on its results. The SOA (Seagull optimization algorithm, SOA) algorithm is a new swarm intelligence optimization algorithm proposed by Gaurav Dhiman in 2019. The algorithm mainly simulates the migration of seagulls in nature and the aggressive behavior during migration. Compared with other optimization algorithms, the SOA algorithm can solve large-scale constraint problems and is highly competitive. Furthermore, due to its simple algorithm, it can obtain a global optimal solution while also achieving high search accuracy and efficiency, making it a promising candidate for automated parameter optimization. This patent utilizes the SOA algorithm to adaptively find the optimal CYCBD filter length. We use the optimal CYCBD filter to filter injector fault signals, reducing noise interference and highlighting the continuous impulse components that are often submerged in the noise.

[0005] After filtering the fault signal, feature extraction is a key step in fault diagnosis. In recent years, numerous methods for measuring the complexity of nonlinear time series in mechanical dynamic systems have been proposed and applied to fault diagnosis, such as approximate entropy, sample entropy, fuzzy entropy, and permutation entropy. Permutation entropy (PE) quantifies dynamic changes based on the ordered patterns of time series structure. Due to its theoretical simplicity and fast computational speed, PE has been widely used in time series complexity analysis. However, the PE algorithm only utilizes the ordinal structure of the time series and ignores its amplitude information. Therefore, Bilal et al. proposed weighted permutation entropy (WPE) based on PE. However, WPE only considers the complexity of the time series at a single scale and ignores useful information at other scales. Therefore, Yin et al. combined WPE with multiscale entropy to propose multiscale weighted permutation entropy (MWPE). However, MWPE only considers the low-frequency components of the time series and ignores the high-frequency components. Based on this, this patent proposes a Hierarchical Weighted Permutation Entropy (HWPE) method that combines hierarchical analysis with weighted permutation entropy. This method not only considers the high-frequency and low-frequency components of the original sequence, but also improves anti-interference and sensitivity to signal bandwidth changes. The effectiveness and superiority of the proposed method are verified through analysis of simulated signals and experimental data.

[0006] Through the above analysis, the problems and defects of the existing technology are as follows:

[0007] 1. Although the currently commonly used weak fault diagnosis method can effectively enhance periodic pulses, the preset parameters of the algorithm itself have a decisive influence on the enhancement effect. Inaccurate selection of preset parameters leads to unsatisfactory enhancement effect and low accuracy of weak fault diagnosis.

[0008] 2. Existing fault feature extraction methods only consider the complexity of time series at low scales, ignoring useful information at other scales. They are prone to ignoring effective fault information, making it difficult to improve the accuracy of fault diagnosis.

[0009] The difficulty of solving the above problems and defects is:

[0010] 1. For different weak fault signals, how to adaptively use weak fault diagnosis methods to realize weak fault diagnosis and identification is an urgent problem to be solved.

[0011] 2. The existing multi-scale analysis method can only consider low-frequency components but not high-frequency components. Therefore, it is necessary to develop a new method that considers both low-frequency and high-frequency components.

[0012] The significance of solving the above problems and defects is:

[0013] The algorithm proposed in this patent is used to adaptively enhance weak impacts, reduce noise interference, highlight the continuous impact components submerged in the noise, analyze the enhanced fault signals from the perspectives of high frequency and low frequency, extract the effective fault characteristics of the LNG engine gas injection valve, improve the fault diagnosis and identification accuracy, and propose an LNG engine gas injection valve fault diagnosis method suitable for on-site industrial environments with strong noise interference. Summary of the Invention

[0014] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the weak fault diagnosis method commonly used in the prior art can effectively enhance periodic pulses, but the preset parameters of the algorithm itself have a decisive influence on the enhancement effect. Inaccurate selection of the preset parameters leads to unsatisfactory enhancement effect and low accuracy of weak fault diagnosis; the existing fault feature extraction method only considers the complexity of time series at low scales, ignores useful information at other scales, easily ignores effective fault information, and the fault diagnosis accuracy is difficult to improve, thereby providing an LNG engine gas injection valve fault diagnosis method and system suitable for LNG engine gas injection valve fault diagnosis in a field industrial environment with strong noise interference, which can reduce noise interference, enhance periodic impact components, and improve the LNG engine gas injection valve fault diagnosis accuracy.

[0015] In order to solve the above problems, the present invention provides a method for diagnosing faults of a gas injection valve of an LNG engine, comprising:

[0016] S1. Obtain a pressure fluctuation signal from the air supply pipe, and divide the acquired pressure fluctuation signal into a training signal and a test signal;

[0017] S2. Use the SOA-CYCBD algorithm to filter the pressure fluctuation signal to obtain a pressure signal with enhanced impact components;

[0018] S3. Process the filtered pressure signal to obtain fault characteristics;

[0019] S4. After training the fault feature vector, perform fault diagnosis and pattern recognition, and output the diagnosis results.

[0020] Optionally, the filtered pressure signal is processed in step S3, specifically comprising the following steps:

[0021] Perform hierarchical analysis on the filtered pressure signal;

[0022] Calculate the weighted permutation entropy of each level to form a fault feature subset.

[0023] Optionally, in step S4, the processed pressure signal is used as a fault feature vector for training, fault diagnosis and pattern recognition are performed, and a diagnosis result is output, which is specifically:

[0024] The hierarchical weighted permutation entropy of all training pressure signals as training samples is used as the feature vector to input the least squares support vector machine multi-classifier for training;

[0025] The trained least squares support vector machine multi-classifier is used to perform fault diagnosis and pattern recognition on the hierarchical weighted permutation entropy of the test signal samples, and the diagnosis results are output.

[0026] Optionally, the SOA-CYCBD in step S2 uses the Seagull optimization algorithm to seek the optimal values ​​of the CYCBD parameter filter length L and the fault period T with the harmonic significance index HSI as the objective function. The calculation of HSI is as follows:

[0027]

[0028] Where ω is the frequency, F(ω) represents the amplitude at frequency ω in the envelope spectrum, Q is the highest-order harmonic used in the calculation, and Q = 5 is usually used. N(ω) is the amount of noise around frequency ω, which can be calculated using a sliding average filter, and the harmonic amplitudes must be removed beforehand. q is a constant ranging from 1 to Q. p(qω) = F(ω) / N(ω) is the amplitude ratio, which indicates the significance of a specific frequency component.

[0029] Optionally, the performing hierarchical analysis on the filtered pressure signal is specifically performing hierarchical analysis on the pressure signal with a signal length of N:

[0030] Based on the vector [γ 1, γ 2, ...,γ n ], define the node components of each layer of time series u(i) as follows:

[0031]

[0032] is the average operator, γ n = 0 or 1, Q0(u) and Q1(u) are as follows:

[0033]

[0034] Where N = 2 n , n is a positive integer. The length of operator Q0(u) and operator Q1(u) is 2 n-1 ; k is the number of decomposition layers, uk,e is the node component of the k-layer decomposition of the time series u, e is a positive integer, and an n-dimensional vector [γ 1, γ 2, ...,γ n ]∈{0,1}, then the integer e can be expressed as:

[0035]

[0036] In the formula, the vector corresponding to the positive integer e is [γ1,γ 2, ...,γ n ].

[0037] Optionally, calculate the weighted permutation entropy of each level. The entropy calculation results are as follows:

[0038]

[0039] HWPE=E(u k,e ,m,d)=[E1,E2,...,E e ] T (7)

[0040] Where m is the embedding dimension, d is the time delay, T is the matrix transpose, E1, E2, ..., E e is the weighted permutation entropy value.

[0041] The present invention also provides a LNG engine gas injection valve fault diagnosis system, comprising:

[0042] A signal acquisition system is used to acquire the pressure fluctuation signal of the air supply pipe and divide the acquired pressure fluctuation signal into a training signal and a test signal;

[0043] A filtering system for filtering the pressure fluctuation signal using the SOA-CYCBD algorithm to obtain a pressure signal with enhanced impact components;

[0044] A signal processing system is used to process the filtered pressure signal and obtain fault characteristics;

[0045] The fault analysis system, after training the fault feature vector, performs fault diagnosis and pattern recognition and outputs the diagnosis results.

[0046] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program implements the above-mentioned LNG engine gas injection valve fault diagnosis method.

[0047] The present invention has the following advantages:

[0048] First, SOA-CYCBD is effectively used to adaptively filter the pressure signal of the LNG engine gas injection valve to enhance the periodic impact component and reduce noise interference;

[0049] Second, the hierarchical weighted permutation entropy is used to comprehensively and accurately reflect the fault information of the gas supply pipe pressure signal, which is suitable for completing the fault diagnosis of the gas injection valve of the LNG engine in a strong noise environment, improving the fault diagnosis rate of the gas injection valve and reducing the misdiagnosis rate and missed diagnosis rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 This is a flow chart of the LNG engine gas injection valve fault diagnosis algorithm based on SOA-CYCBD and hierarchical weighted permutation entropy of the present invention;

[0052] Figure 2 It is a flow chart of the SOA-CYCBD optimization algorithm of the present invention;

[0053] Figure 3 This is a diagram of the fault diagnosis result of the LNG engine gas injection valve based on SOA-CYCBD and hierarchical weighted permutation entropy of the present invention;

[0054] Figure 4 This is a physical picture of the whole machine test bench of the present invention;

[0055] Figure 5 This is a physical diagram of the sensor installation position of the present invention;

[0056] Figure 6 This is a diagram showing the vibration signal of a high-pressure gas supply pipeline of a typical fault of the present invention;

[0057] Figure 7 This is one of the flow charts of the steps of the LNG engine gas injection valve fault diagnosis method of the present invention;

[0058] Figure 8 This is the second step flow chart of the LNG engine gas injection valve fault diagnosis method of the present invention;

[0059] Figure 9 It is a connection diagram of the LNG engine gas injection valve fault diagnosis system of the present invention. DETAILED DESCRIPTION

[0060] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] Application Overview

[0062] While commonly used weak fault diagnosis methods can effectively enhance periodic pulses, the algorithm's preset parameters have a decisive influence on the enhancement effect. Improperly selected preset parameters can lead to suboptimal enhancement and low weak fault diagnosis accuracy. Existing fault feature extraction methods only consider the complexity of time series at low scales, ignoring useful information at other scales. This can easily overlook valid fault information, making it difficult to improve fault diagnosis accuracy.

[0063] Exemplary LNG engine gas injection valve fault diagnosis method

[0064] Figure 7 The flowchart of a method for diagnosing a fault of a gas injection valve of an LNG engine is shown. The method for diagnosing a fault of a gas injection valve of an LNG engine includes:

[0065] S1. Obtain a pressure fluctuation signal from the air supply pipe, and divide the acquired pressure fluctuation signal into a training signal and a test signal;

[0066] S2. Use the SOA-CYCBD algorithm to filter the pressure fluctuation signal to obtain a pressure signal with enhanced impact components;

[0067] S3. Process the filtered pressure signal to obtain fault characteristics;

[0068] S4. Train the fault feature vector, perform fault diagnosis and pattern recognition, and output the diagnosis results.

[0069] The first stage of the above-mentioned LNG engine gas injection valve fault diagnosis method is to obtain a pressure fluctuation signal using a pressure sensor installed on the gas supply pipe and preliminarily establish a sample set of training signals and test signals; the second stage is to filter the pressure fluctuation signal using the SOA-CYCBD algorithm to obtain a pressure signal with enhanced impact components; the third stage is to calculate the hierarchical weighted permutation entropy of the filtered pressure signal and use the hierarchical weighted permutation entropy as the pressure signal fault feature; the fourth stage is to use the hierarchical weighted permutation entropy of all training samples as the feature vector to input the least squares support vector machine multi-classifier for training; the fifth stage is to use the trained least squares support vector machine multi-classifier to perform fault diagnosis and pattern recognition on the hierarchical weighted permutation entropy of the test sample and output the diagnosis result.

[0070] By installing a pressure sensor on the gas supply pipe to measure the pressure state parameters of the gas supply pipe, accurate data measurement can be achieved and the accuracy of fault diagnosis can be improved.

[0071] Therefore, the acquired pressure fluctuation signal is used to form a training signal sample set and a test signal sample set.

[0072] Specifically, in step S2, the pressure signal is adaptively filtered using the SOA-CYCBD algorithm to obtain a filtered signal with enhanced impact components. The SOA-CYCBD optimization algorithm process is as follows: Figure 2 As shown, the migratory behavior of seagulls is modeled as:

[0073] D=|C+M| (8)

[0074] Where D represents the distance between the search agent and the best adaptation agent, and C is defined as:

[0075] C=A×P(t) (9)

[0076] Where P(t) is the position of the seagull in the tth generation, and the parameter A represents the movement behavior of the search agent in the given search space, which is used to avoid collisions between seagulls. At the same time, A is a parameter that decreases linearly with the progress of iterations.

[0077] A=a-(t×(a / MAX iteration )) (10)

[0078] Where a is a constant that controls the frequency of using variable A, and t represents the number of generations of the current iteration. M is defined as

[0079] M=B×(Pbs(t)-P(t)) (11)

[0080] Where M represents the direction of the best position, Pbs(t) is the best seagull of generation t, and B is a random parameter responsible for balancing global and local search, calculated as follows:

[0081] B=2×A 2 ×rd (12)

[0082] where rd is a random parameter between [0,1]. After arriving at a new location, the seagulls attack their prey using a spiral motion. The seagulls’ attack behavior is modeled as

[0083] P(t)=(D×x×y×z)+Pbs(t) (13)

[0084] in,

[0085] x=r×cos(θ) (14)

[0086] y=r×sin(θ) (15)

[0087] z=r×θ (16)

[0088] r=u×e θv (17)

[0089] Here, u and v are constants, e is the basis of natural logarithms, and θ is a random number between [0 and 2π]. The Seagull Optimizer begins with a randomly generated population. During iterations, the search agents update their positions based on the best search agent. To smoothly transition between global and local search, the variable B is used. As a result, the Seagull Optimizer has excellent global and local search capabilities.

[0090] Specifically, if Figure 8 As shown, the LNG engine gas injection valve fault diagnosis system method also includes:

[0091] The filtered pressure signal is processed in step S3, specifically comprising the following steps:

[0092] Step S301: performing hierarchical analysis on the filtered pressure signal;

[0093] Given a time series {u(i), i = 1, 2, ..., N} of length N, define the averaging operators Q0(u) and Q1(u) as follows:

[0094]

[0095] Where N = 2 n , n is a positive integer. The length of operator Q0(u) and operator Q1(u) is 2 n-1 According to the average operators Q0(u) and Q1(u), the original sequence can be reconstructed as

[0096] u={(Q0(u) j +Q1(u) j ),(Q0(u) j -Q1(u) j )},j=0,1,2,...,2 n-1

[0097] When j = 0 or j = 1, define the matrix Q j The operator is as follows

[0098]

[0099] Step S302: Calculate the weighted permutation entropy of each level to form a fault feature subset:

[0100] Step 2: Construct an n-dimensional vector [γ 1, γ 2, ...,γ n ]∈{0,1}, then the integer e can be expressed as

[0101]

[0102] In the formula, the vector corresponding to the positive integer e is [γ 1, γ 2, ...,γ n ].

[0103] Step 3: Based on the vector [γ 1, γ 2, ...,γ n ], define the node components of each layer of time series u(i) as follows

[0104]

[0105] Where k represents the kth layer in hierarchical segmentation, and the low-frequency and high-frequency parts of the original time series u(i) at the k+1th layer are represented by u k,0 and u k,1 express.

[0106] Step 4: For the time series u k,e Perform phase space reconstruction to obtain a series of subsequences

[0107]

[0108] Where m is the embedding dimension, T is the time delay, K is the number of reconstructed components, and K = N-(m-1)T, N is the length of the time series. Arrange the elements in each reconstructed subsequence in ascending order according to their numerical values. After arranging each reconstructed subsequence in ascending order, a set of symbol sequences π can be obtained. i =[k1,k2,…,k m ].

[0109] Calculate the weight value ω of each subsequence i ;

[0110]

[0111] Any subsequence The feature information is represented by the weight value ω i and the permutation pattern π i For this time series U, there are K kinds of arrangement patterns, and each arrangement pattern π i The weighted probability value of is;

[0112]

[0113] According to the definition of information entropy, calculate the weighted permutation entropy WPE value of time series U;

[0114]

[0115] Hierarchical discrete entropy can be expressed as:

[0116] HWPE=E(u k,e ,m,d)=[E1,E2,...,E e ] T .

[0117] S4. Use the hierarchical weighted permutation entropy of all training samples as the feature vector and input it into the least squares support vector machine multi-classifier for training.

[0118] S5. Use the trained least squares support vector machine multi-classifier to perform fault diagnosis and pattern recognition on the hierarchical weighted permutation entropy of the test samples and output the diagnosis results. The classification results are as follows: Figure 3 shown.

[0119] like Figure 9 As shown, the LNG engine gas injection valve fault diagnosis system of the present invention achieves the reduction of noise interference, enhancement of periodic impact components, and improvement of LNG engine gas injection valve fault diagnosis accuracy by successively acquiring voltage fluctuation signals, voltage filtering, fault feature analysis and training, fault diagnosis and output results.

[0120] Exemplary computer-readable storage media

[0121] The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned LNG engine gas injection valve fault diagnosis method is implemented.

[0122] The above-mentioned computer-readable storage medium can implement the steps of the LNG engine gas injection valve fault diagnosis method in any of the above-mentioned embodiments and can achieve the same technical effects. Therefore, it has all the beneficial effects of any of the above-mentioned embodiments and will not be repeated here.

[0123] The computer-readable storage medium in this embodiment is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0124] According to the above description, this application has the following advantages:

[0125] 1. Use the algorithm proposed in this patent to adaptively enhance weak impacts, reduce noise interference, and highlight the continuous impact components submerged in the noise.

[0126] 2. Analyze the enhanced fault signal from the perspective of high frequency and low frequency, extract the effective fault characteristics of the LNG engine gas injection valve, and improve the fault diagnosis and identification accuracy.

[0127] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0128] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0129] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0130] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A method for diagnosing faults of a gas injection valve of an LNG engine, characterized in that: include: Step S1, obtaining a pressure fluctuation signal of the air supply pipe, and dividing the collected pressure fluctuation signal into a training signal and a test signal; Step S2, using the SOA-CYCBD algorithm to filter the pressure fluctuation signal to obtain a pressure signal with enhanced impact components; Step S3, processing the filtered pressure signal to obtain fault characteristics; Step S4: After training the fault feature vector, perform fault diagnosis and pattern recognition, and output the diagnosis result; The filtered pressure signal is processed in step S3, specifically comprising the following steps: Perform hierarchical analysis on the filtered pressure signal; Calculate the weighted permutation entropy of each level to form a subset of fault features; The SOA-CYCBD in step S2 uses the Seagull optimization algorithm to find the optimal values ​​of the CYCBD parameter filter length L and the fault period T with the harmonic significance index HSI as the objective function. The calculation of HSI is as follows: Where ω is the frequency, F(ω) represents the amplitude at frequency ω in the envelope spectrum, Q is the maximum harmonic used in the calculation, usually Q = 5, N(ω) is the amount of noise around frequency ω, which is calculated using a sliding average filter and the harmonic amplitude needs to be removed in advance; q is a constant with a value range of [1, Q]; p(qω) = F(ω) / N(ω) is the amplitude ratio, which is used to indicate the significance of a specific frequency component.

2. The LNG engine gas injection valve fault diagnosis method according to claim 1, characterized in that: After the fault feature vector is trained in step S4, fault diagnosis and pattern recognition are performed, and the diagnosis results are output, which are specifically: The hierarchical weighted permutation entropy of all training pressure signals as training samples is used as the feature vector to input the least squares support vector machine multi-classifier for training; The trained least squares support vector machine multi-classifier is used to perform fault diagnosis and pattern recognition on the hierarchical weighted permutation entropy of the test signal samples, and the diagnosis results are output.

3. The LNG engine gas injection valve fault diagnosis method according to claim 2, characterized in that: The hierarchical analysis of the filtered pressure signal is specifically performed by performing hierarchical analysis on the pressure signal with a signal length of N. Based on the vector [γ1,γ2,...,γ n ], define the node components of each layer of time series u(i) as follows: is the average operator, γ n = 0 or 1, Q0(u) and Q1(u) are as follows: Where N = 2 n , n is a positive integer, the length of operator Q0(u) and operator Q1(u) is 2 n-1 ,k is the number of decomposition layers, u k,e is the node component of the k-layer decomposition of the time series u, e is a positive integer, and an n-dimensional vector [γ1,γ2,...,γ n ]∈{0,1}, then the integer e can be expressed as: In the formula, the vector corresponding to the positive integer e is [γ1,γ 2, ...,γ n ].

4. The LNG engine gas injection valve fault diagnosis method according to claim 3, characterized in that: Calculate the weighted permutation entropy of each level. The entropy calculation results are as follows: HWPE=E(u k,e ,m,d)=[E1,E2,...,E e ] T (6) Where m is the embedding dimension, d is the time delay, T is the matrix transpose, E1, E2, ..., E e is the weighted permutation entropy value.

5. A LNG engine gas injection valve fault diagnosis system, characterized in that: include: A signal acquisition system is used to acquire the pressure fluctuation signal of the air supply pipe and divide the acquired pressure fluctuation signal into a training signal and a test signal; The filtering system is used to filter the pressure fluctuation signal using the SOA-CYCBD algorithm to obtain a pressure signal with enhanced impact components. SOA-CYCBD uses the Seagull optimization algorithm and the harmonic significance index HSI as the objective function to find the optimal values ​​of the CYCBD parameter filter length L and the fault period T. The calculation of HSI is as follows: Where ω is the frequency, F(ω) represents the amplitude at frequency ω in the envelope spectrum, Q is the maximum harmonic used in the calculation, usually Q = 5, N(ω) is the amount of noise around frequency ω, which is calculated using a sliding average filter and the harmonic amplitude needs to be removed in advance; q is a constant with a value range of [1, Q]; p(qω) = F(ω) / N(ω) is the amplitude ratio, which is used to indicate the significance of a specific frequency component; The signal processing system is used to process the filtered pressure signal, and the processed pressure signal is used as a fault feature. The processing of the filtered pressure signal specifically includes the following steps: performing hierarchical analysis on the filtered pressure signal; calculating the weighted permutation entropy of each level to form a fault feature subset; The fault analysis system is used to use the processed pressure signal as a fault feature vector for training, perform fault diagnosis and pattern recognition, and output the diagnosis results.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the LNG engine gas injection valve fault diagnosis method according to any one of claims 1 to 4.

Images