Cabin component abnormity diagnosis method and system based on multi-dimensional signal analysis

Abstract

The invention belongs to the technical field of abnormity diagnosis, and discloses a cabin component abnormity diagnosis method and system based on multi-dimensional signal analysis. The method comprises the following steps: according to a plurality of collected historical multi-dimensional signals of a cabin, using a multi-dimensional fusion signal analysis algorithm to construct a cabin part abnormity diagnosis model, and using a reinforcement learning algorithm to construct a maintenance guidance model; acquiring real-time multi-dimensional signals of the cabin, and performing abnormality diagnosis on the preprocessed real-time multi-dimensional signals by using the cabin component abnormality diagnosis model to obtain a real-time cabin component abnormality diagnosis result; and by using the maintenance guidance model, performing maintenance guidance strategy generation on the real-time cabin component abnormality diagnosis result to obtain a real-time maintenance guidance strategy. According to the method, the problems of low accuracy, poor performability and poor practicability in the prior art are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of abnormality diagnosis, and in particular relates to a method and system for diagnosing abnormality of cabin components based on multi-dimensional signal analysis. Background Art

[0002] As the global demand for renewable energy continues to grow, wind power generation, as an important component of clean energy, is developing rapidly in terms of installed capacity and technology level. The large-scale and complex nature of wind turbines places higher demands on the reliability and operation and maintenance efficiency of the units. Wind turbines are usually installed in remote areas, such as offshore wind farms or mountainous areas, which makes maintenance costs high and maintenance cycles long. Once a key large component in the nacelle fails, it will not only affect the power generation efficiency, but may also lead to expensive maintenance costs and long production downtime.

[0003] The existing technology for abnormal diagnosis of cabin components has the following defects:

[0004] 1) They often rely on a single signal type (such as sound or vibration signals only), which may lead to low diagnostic accuracy in complex environments, and traditional signal processing methods may not be able to effectively extract complex features in the signal, further affecting the accuracy of diagnosis;

[0005] 2) The signal analysis and diagnosis process often requires a lot of manual operations by professional technicians, which is not only time-consuming but also prone to errors, resulting in poor executability;

[0006] 3) Lack of response mechanism. When an abnormality is detected, specific maintenance guidance strategies are often not provided, resulting in maintenance personnel being unable to quickly take correct maintenance measures, which is of poor practicality. Summary of the invention

[0007] In order to solve the problems of low accuracy, poor executability and poor practicality in the prior art, the present invention aims to provide a method and system for diagnosing abnormalities of cabin components based on multi-dimensional signal analysis.

[0008] The technical solution adopted by the present invention is:

[0009] A method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis comprises the following steps:

[0010] Based on the collected historical multi-dimensional signals of the cabin, a multi-dimensional fusion signal analysis algorithm is used to build a cabin component abnormality diagnosis model, and a reinforcement learning algorithm is used to build a maintenance guidance model;

[0011] Collect real-time multi-dimensional signals of the cabin, and use the cabin component abnormality diagnosis model to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signals to obtain real-time cabin component abnormality diagnosis results;

[0012] The maintenance guidance model is used to generate maintenance guidance strategies based on the real-time abnormal diagnosis results of cabin components to obtain real-time maintenance guidance strategies.

[0013] Furthermore, the historical / real-time multi-dimensional signal includes a historical / real-time sound signal and a historical / real-time vibration signal.

[0014] Furthermore, based on the collected historical multi-dimensional signals of the cabin, a multi-dimensional fusion signal analysis algorithm is used to construct a cabin component abnormality diagnosis model, and a reinforcement learning algorithm is used to construct a maintenance guidance model, including the following steps:

[0015] Collecting a number of historical multi-dimensional signals of the engine room under different operating conditions, and preprocessing the number of historical multi-dimensional signals to obtain a number of preprocessed historical multi-dimensional signals;

[0016] Performing STFT processing on the preprocessed historical sound signal of the preprocessed historical multidimensional signal to obtain a historical sound spectrogram, and performing STFT processing on the preprocessed historical vibration signal to obtain a historical vibration spectrogram;

[0017] Based on several pre-processed historical sound signals, corresponding historical sound spectrograms and historical vibration spectrograms, a multi-dimensional fusion signal analysis algorithm is used to construct a cabin component abnormality diagnosis model;

[0018] According to several historical cabin component abnormality diagnosis results generated by the cabin component abnormality diagnosis model, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated.

[0019] Furthermore, the abnormality diagnosis model of cabin components is constructed based on the logfBank-AAA-Attention-MLP algorithm.

[0020] Furthermore, the maintenance guidance model is constructed based on the DQN algorithm.

[0021] Furthermore, according to several historical abnormal diagnosis results of cabin components generated by the abnormal diagnosis model of cabin components, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated, including the following steps:

[0022] Generate questions based on the maintenance guidance strategy, define the simulation environment of the DQN algorithm, and build the agent and experience replay pool;

[0023] Convert the historical abnormal diagnosis results of the cabin components into a number of historical state parameters, and define the state space of the DQN algorithm based on the several historical state parameters;

[0024] Convert the preset maintenance guidance strategy into a number of historical action parameters, and define the action space of the DQN algorithm based on the historical action parameters;

[0025] Define the reward function of the DQN algorithm based on the influence of each action parameter in the action space;

[0026] Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space;

[0027] Based on the state space, action space and reward function, the deep Q network and intelligent agent are optimized and trained according to several historical abnormal diagnosis results of cabin components to build a maintenance guidance model;

[0028] A number of historical maintenance guidance strategy generation experiences are generated, and the number of historical maintenance guidance strategy generation experiences are stored in an experience replay pool.

[0029] Furthermore, real-time multi-dimensional signals of the cabin are collected, and the cabin component abnormality diagnosis model is used to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signals to obtain real-time cabin component abnormality diagnosis results, including the following steps:

[0030] Collecting real-time multi-dimensional signals from the engine room, and preprocessing the real-time multi-dimensional signals to obtain preprocessed real-time multi-dimensional signals;

[0031] Performing STFT processing on the preprocessed real-time sound signal of the preprocessed real-time multidimensional signal to obtain a real-time sound spectrogram, and performing STFT processing on the preprocessed real-time vibration signal to obtain a real-time vibration spectrogram;

[0032] The preprocessed real-time sound signal, the corresponding real-time sound spectrogram and the real-time vibration spectrogram are input into the abnormal diagnosis model of the cabin components to perform abnormal diagnosis and obtain the real-time abnormal diagnosis result of the cabin components.

[0033] Furthermore, the preprocessed real-time sound signal, the corresponding real-time sound spectrogram and the real-time vibration spectrogram are input into the abnormal diagnosis model of the cabin component to perform abnormal diagnosis and obtain the real-time abnormal diagnosis result of the cabin component, including the following steps:

[0034] Extracting the real-time acoustic analysis features of the preprocessed real-time sound signal, the first real-time spectrum features of the real-time sound spectrogram, and the second real-time spectrum features of the real-time vibration spectrogram;

[0035] According to a preset attention weight value, the real-time acoustic analysis feature, the first real-time spectrum feature, and the second real-time spectrum feature are feature spliced ​​to obtain a real-time splicing feature;

[0036] According to the real-time splicing features, abnormal diagnosis is performed to obtain real-time abnormal diagnosis results of cabin components.

[0037] Furthermore, the maintenance guidance model is used to generate a maintenance guidance strategy for the real-time abnormal diagnosis results of the cabin components, and the real-time maintenance guidance strategy is obtained, which includes the following steps:

[0038] Extracting a number of historical maintenance guidance strategy generation experiences from the experience replay pool, and updating the action space of the maintenance guidance model according to the number of historical maintenance guidance strategy generation experiences to obtain an updated action space;

[0039] According to the real-time abnormal diagnosis result of the cabin component, the state space of the maintenance guidance model is updated to obtain an updated state space;

[0040] According to the updated state space, the input layer of the deep Q network is updated, and according to the updated action space, the output layer of the deep Q network is updated to obtain an updated maintenance guidance model;

[0041] Using the updated maintenance guidance model, a real-time maintenance guidance strategy and corresponding real-time maintenance guidance strategy generation experience are generated, and the real-time maintenance guidance strategy generation experience is stored in an experience replay pool.

[0042] A cabin component abnormality diagnosis system based on multi-dimensional signal analysis is used to implement a cabin component abnormality diagnosis method. The system comprises a model building unit, an abnormality diagnosis unit and a maintenance guidance unit which are connected in sequence.

[0043] The beneficial effects of the present invention are:

[0044] The present invention discloses a method and system for diagnosing abnormalities of cabin components based on multi-dimensional signal analysis. The method adopts multi-dimensional signal fusion technology, and can simultaneously process and integrate various types of complex signals (sound signals, vibration signals and other signals), thereby improving the accuracy of diagnosis in complex environments; based on an intelligent feature extraction method, key features can be automatically extracted from multi-dimensional signals, reducing manual intervention, and significantly improving the efficiency and accuracy of feature extraction, and enhancing executability; a response mechanism is provided, and a maintenance guidance model constructed using a reinforcement learning algorithm can provide specific maintenance guidance strategies when an abnormality is detected, helping maintenance personnel to quickly take correct maintenance measures, thereby improving practicality; an efficient data processing flow is provided, which significantly improves the efficiency of processing a large amount of multi-dimensional signal data and meets the real-time requirements.

[0045] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flowchart of the method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis in the present invention.

[0047] Figure 2 It is a structural block diagram of the cabin component abnormality diagnosis system based on multi-dimensional signal analysis in the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0049] Embodiment 1:

[0050] like Figure 1 As shown, this embodiment provides a method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis, comprising the following steps:

[0051] S1: Based on the collected historical multi-dimensional signals of the cabin, a multi-dimensional fusion signal analysis algorithm is used to build a cabin component abnormality diagnosis model, and a reinforcement learning algorithm is used to build a maintenance guidance model, including the following steps:

[0052] S1-1: collecting a number of historical multi-dimensional signals of the cabin under different operating conditions, and preprocessing the number of historical multi-dimensional signals to obtain a number of preprocessed historical multi-dimensional signals;

[0053] The historical multi-dimensional signal includes a historical sound signal and a historical vibration signal;

[0054] S1-2: performing STFT processing on the preprocessed historical sound signal of the preprocessed historical multidimensional signal to obtain a historical sound spectrogram, and performing short-time Fourier transform (STFT) processing on the preprocessed historical vibration signal to obtain a historical vibration spectrogram;

[0055] S1-3: Based on several pre-processed historical sound signals, corresponding historical sound spectrograms and historical vibration spectrograms, a multi-dimensional fusion signal analysis algorithm is used to construct a cabin component abnormality diagnosis model;

[0056] The abnormality diagnosis model of the cabin components is constructed based on the logfBank-AAA-Attention-MLP algorithm, and the abnormality diagnosis model of the cabin components includes a first spectrum feature extraction model and a second spectrum feature extraction model both constructed based on the logfBank algorithm, an acoustic feature extraction module constructed based on the Acoustic Analysis Algorithms (AAA) algorithm, an attention weight module constructed based on the Attention mechanism, and a label prediction module constructed based on the Multilayer Perceptron (MLP) algorithm, the first spectrum feature extraction model, the second spectrum feature extraction model and the acoustic feature extraction module are all connected to the attention weight module, and the attention weight module is connected to the label prediction module;

[0057] The logfBank algorithm is similar to the Mel-Frequency Cepstrum (MFCC) algorithm. Both are based on the fBank feature sequence extraction results and then perform subsequent processing. The logfBank algorithm has a smaller amount of calculation than the MFCC algorithm, and the correlation between features is higher. The correlation between features can be better utilized by the model to improve the recognition accuracy and reduce the amount of calculation.

[0058] The AAA algorithm calculates the acoustic characteristics of the sound signal through the acoustic analysis function, which includes the volume calculation function, the pitch calculation function, the formant calculation function, and the duration calculation function;

[0059] The formula for the volume calculation function is:

[0060]

[0061] In the formula, E RMS is the volume calculation function; n is the sampling point indicator; N is the total number of sampling points; x[n] is the amplitude of the historical sound signal x(n) at sampling point n after preprocessing;

[0062] The formula for the pitch calculation function is:

[0063] E PITCH =X[n]+ε

[0064] In the formula, E PITCH is the pitch calculation function, generally using the automatic pitch detection algorithm; X[n] is the automatic pitch detection function of the preprocessed historical sound signal x(n) at sampling point n; ε is the pitch compensation value;

[0065] The formula for the resonance peak calculation function is:

[0066]

[0067] In the formula, is the predicted data; α i is the prediction coefficient; x(ni) is the historical sound signal after preprocessing; n is the sampling point indicator; i is the sampling point indicator, indicating the sampling point i before the current sampling point n; p is the prediction order; E FFT is the predicted data after fast Fourier transform; FFT[*] is the fast Fourier transform function; f k is the resonance peak frequency; max[*] is the maximum value retention function; W k is the resonance peak bandwidth; w[*] is the bandwidth calculation function;

[0068] The formula for the duration calculation function is:

[0069]

[0070] Where L is the duration calculation function; is the audio measurement duration at sampling point n; is the duration of the silence interval at sampling point n; β is the duration compensation value; n is the sampling point indication; N is the total number of sampling points;

[0071] S1-4: Based on several historical cabin component abnormality diagnosis results generated by the cabin component abnormality diagnosis model, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated;

[0072] The maintenance guidance model is constructed based on the Deep Q Network (DQN) algorithm, and the maintenance guidance model is provided with an intelligent agent, an experience replay pool, and a Deep Q Network;

[0073] Based on several historical abnormal diagnosis results of cabin components generated by the abnormal diagnosis model of cabin components, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated, including the following steps:

[0074] S1-4-1: Generate questions based on the maintenance guidance strategy, define the simulation environment of the DQN algorithm, and build the intelligent agent and experience replay pool;

[0075] S1-4-2: Convert the historical abnormal diagnosis results of the cabin components into a number of historical state parameters, and define the state space of the DQN algorithm based on the several historical state parameters;

[0076] S1-4-3: Convert the preset maintenance guidance strategy into a number of historical action parameters, and define the action space of the DQN algorithm based on the several historical action parameters;

[0077] S1-4-4: Define the reward function of the DQN algorithm based on the influence of each action parameter in the action space;

[0078] S1-4-5: Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space;

[0079] S1-4-6: Based on the state space, action space and reward function, and according to several historical abnormal diagnosis results of cabin components, the deep Q network and intelligent agent are optimized and trained to build a maintenance guidance model;

[0080] S1-4-7: Generate a number of historical maintenance guidance strategy generation experiences, and store the number of historical maintenance guidance strategy generation experiences in an experience playback pool;

[0081] S2: collecting real-time multi-dimensional signals of the cabin, and using the cabin component abnormality diagnosis model to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signals to obtain real-time cabin component abnormality diagnosis results, including the following steps:

[0082] S2-1: collecting real-time multi-dimensional signals of the cabin, and preprocessing the real-time multi-dimensional signals to obtain preprocessed real-time multi-dimensional signals;

[0083] Historical / real-time multi-dimensional signals include real-time sound signals and real-time vibration signals; pre-processing includes denoising, framing, pre-emphasis and windowing in sequence to improve data quality;

[0084] S2-2: performing STFT processing on the preprocessed real-time sound signal of the preprocessed real-time multidimensional signal to obtain a real-time sound spectrogram, and performing STFT processing on the preprocessed real-time vibration signal to obtain a real-time vibration spectrogram;

[0085] S2-3: inputting the preprocessed real-time sound signal, the corresponding real-time sound spectrogram and the real-time vibration spectrogram into the abnormality diagnosis model of the cabin component to perform abnormality diagnosis and obtain the real-time abnormality diagnosis result of the cabin component, including the following steps:

[0086] S2-3-1: Use the acoustic feature extraction module to extract the real-time acoustic analysis feature q = Q[E RMS ,E PITCH ,f k ,W k ,L], where Q[*] is the acoustic feature extraction function;

[0087] Using the first spectrum feature extraction model, extracting the first real-time spectrum feature of the real-time spectrogram includes the following steps:

[0088] A-1: Get the Mel-frequency cepstrum coefficients corresponding to the real-time spectrogram;

[0089] A-2: The formula for obtaining the Mel frequency cepstral coefficients is:

[0090]

[0091] Wherein, MFCCs is the Mel frequency cepstrum coefficient corresponding to the real-time spectrogram X(t,f); DCT[*] is the discrete cosine transform function; log(*) is the logarithmic function; |X(t,f)| is the modulus corresponding to the real-time spectrogram X(t,f); arg(X(t,f)) is the phase corresponding to the real-time spectrogram X(t,f); j is a complex number;

[0092] The formulas for first-order and second-order differences are:

[0093]

[0094] Where, MFCCs is the Mel frequency cepstrum coefficient corresponding to the real-time spectrogram X(t,f); d1 is the first-order difference; d2 is the second-order difference; i' is the difference indicator;

[0095] A-3: Integrate the Mel frequency cepstrum coefficients, the first order difference and the second order difference to obtain the corresponding first real-time spectrum feature h1;

[0096] Using a second frequency spectrum feature extraction model, extracting a second real-time frequency spectrum feature h2 of the real-time vibration spectrum;

[0097] S2-3-2: according to a preset attention weight value, the real-time acoustic analysis feature, the first real-time spectrum feature and the second real-time spectrum feature are feature spliced ​​to obtain a real-time splicing feature;

[0098] S2-3-3: Perform abnormal diagnosis based on the real-time splicing features to obtain real-time abnormal diagnosis results of cabin components;

[0099] S3: Using the maintenance guidance model, a maintenance guidance strategy is generated for the real-time abnormal diagnosis results of the cabin components to obtain a real-time maintenance guidance strategy, including the following steps:

[0100] S3-1: Extract several historical maintenance guidance strategy generation experiences from the experience playback pool, and update the action space of the maintenance guidance model based on the several historical maintenance guidance strategy generation experiences to obtain the updated action space A'=[a'1,...,a' j" ,...,a' I ], where a' j" is the updated j-th "action value", j" is the action indicator, and I is the total number of action space dimensions;

[0101] S3-2: Update the state space of the maintenance guidance model according to the real-time abnormal diagnosis results of the cabin components, and obtain the updated state space S'=[s'1,...,s' i" ,...,s' I ], where s' i" is the updated i'th state value, i' is the state indicator, and I' is the total number of state space dimensions;

[0102] S3-3: Update the input layer of the deep Q network according to the updated state space, and update the output layer of the deep Q network according to the updated action space to obtain an updated maintenance guidance model;

[0103] S3-4: using the updated maintenance guidance model, generating a real-time maintenance guidance strategy and corresponding real-time maintenance guidance strategy generation experience, and storing the real-time maintenance guidance strategy generation experience in an experience playback pool, including the following steps:

[0104] S3-4-1: Use the agent to control the updated deep Q network in the updated maintenance guidance model and output the updated action space A' = [a'1, ..., a' j' ,...,a' I ];

[0105] S3-4-2: Iteratively update the Q value according to the preset reward function to obtain an updated Q value until the number of iterations reaches the threshold;

[0106] The formula is:

[0107] Q(s' p' ,a' p' )=(1-α")·Q(s p' ,a p' )+α"·(R(s p' ,a p' ,s' p' )+γ·Q max (s p' ,a p' ))

[0108] In the formula, Q(s' p' ,a' p' ) is the updated state value s' p' and the updated action value a' p' The corresponding updated Q value; Q(s p' ,a p' ) is the state value s p' and action value a p' The corresponding predicted Q value; α" is the learning rate; Q max (s p' ,a p' ) is the highest predicted Q value; p' is the comprehensive indicator; γ is the update parameter;

[0109] S3-4-3: According to the greedy strategy, the possible action parameter corresponding to the highest updated Q value in each iteration is used as the execution action of the state parameter;

[0110] S3-4-4: Integrate the execution actions corresponding to all state parameters in the updated state space to obtain a real-time maintenance guidance strategy.

[0111] Embodiment 2:

[0112] like Figure 2 As shown, this embodiment provides a cabin component abnormality diagnosis system based on multi-dimensional signal analysis, which is used to implement a cabin component abnormality diagnosis method. The system includes a model building unit, an abnormality diagnosis unit, and a maintenance guidance unit connected in sequence;

[0113] A model building unit, used to build a cabin component abnormality diagnosis model based on a number of collected historical multi-dimensional signals of the cabin using a multi-dimensional fusion signal analysis algorithm, and to build a maintenance guidance model using a reinforcement learning algorithm;

[0114] The abnormality diagnosis unit is used to collect the real-time multi-dimensional signals of the cabin, and use the cabin component abnormality diagnosis model to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signals to obtain real-time cabin component abnormality diagnosis results;

[0115] The maintenance guidance unit is used to generate a maintenance guidance strategy for the real-time abnormal diagnosis result of the cabin component by using the maintenance guidance model to obtain a real-time maintenance guidance strategy.

[0116] The present invention discloses a method and system for diagnosing abnormalities of cabin components based on multi-dimensional signal analysis. The method adopts multi-dimensional signal fusion technology, and can simultaneously process and integrate various types of complex signals (sound signals, vibration signals and other signals), thereby improving the accuracy of diagnosis in complex environments; based on an intelligent feature extraction method, key features can be automatically extracted from multi-dimensional signals, reducing manual intervention, and significantly improving the efficiency and accuracy of feature extraction, and enhancing executability; a response mechanism is provided, and a maintenance guidance model constructed using a reinforcement learning algorithm can provide specific maintenance guidance strategies when an abnormality is detected, helping maintenance personnel to quickly take correct maintenance measures, thereby improving practicality; an efficient data processing flow is provided, which significantly improves the efficiency of processing a large amount of multi-dimensional signal data and meets the real-time requirements.

[0117] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis, characterized in that: The steps include: Based on the collected historical multi-dimensional signals of the cabin, a multi-dimensional fusion signal analysis algorithm is used to build a cabin component abnormality diagnosis model, and a reinforcement learning algorithm is used to build a maintenance guidance model; Collect real-time multi-dimensional signals of the cabin, and use the cabin component abnormality diagnosis model to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signals to obtain real-time cabin component abnormality diagnosis results; The maintenance guidance model is used to generate maintenance guidance strategies based on the real-time abnormal diagnosis results of cabin components to obtain real-time maintenance guidance strategies.

2. The method for diagnosing abnormality of cabin components based on multidimensional signal analysis according to claim 1, characterized in that: The historical / real-time multi-dimensional signals include historical / real-time sound signals and historical / real-time vibration signals.

3. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 2, characterized in that: Based on the collected historical multi-dimensional signals of the cabin, a multi-dimensional fusion signal analysis algorithm is used to build a cabin component abnormality diagnosis model, and a reinforcement learning algorithm is used to build a maintenance guidance model, including the following steps: Collecting a number of historical multi-dimensional signals of the engine room under different operating conditions, and preprocessing the number of historical multi-dimensional signals to obtain a number of preprocessed historical multi-dimensional signals; Performing STFT processing on the preprocessed historical sound signal of the preprocessed historical multidimensional signal to obtain a historical sound spectrogram, and performing STFT processing on the preprocessed historical vibration signal to obtain a historical vibration spectrogram; Based on several pre-processed historical sound signals, corresponding historical sound spectrograms and historical vibration spectrograms, a multi-dimensional fusion signal analysis algorithm is used to construct a cabin component abnormality diagnosis model; According to several historical cabin component abnormality diagnosis results generated by the cabin component abnormality diagnosis model, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated.

4. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 3, characterized in that: The cabin component abnormality diagnosis model is constructed based on the logfBank-AAA-Attention-MLP algorithm.

5. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 3, characterized in that: The maintenance guidance model is constructed based on the DQN algorithm.

6. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 5, characterized in that: Based on several historical abnormal diagnosis results of cabin components generated by the abnormal diagnosis model of cabin components, a maintenance guidance model is constructed using a reinforcement learning algorithm, and several maintenance guidance strategy generation experiences are generated, including the following steps: Generate questions based on the maintenance guidance strategy, define the simulation environment of the DQN algorithm, and build the intelligent agent and experience replay pool; Convert the historical abnormal diagnosis results of the cabin components into a number of historical state parameters, and define the state space of the DQN algorithm based on the several historical state parameters; Convert the preset maintenance guidance strategy into a number of historical action parameters, and define the action space of the DQN algorithm based on the historical action parameters; Define the reward function of the DQN algorithm based on the influence of each action parameter in the action space; Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space; Based on the state space, action space and reward function, the deep Q network and intelligent agent are optimized and trained according to several historical abnormal diagnosis results of cabin components to build a maintenance guidance model; A number of historical maintenance guidance strategy generation experiences are generated, and the number of historical maintenance guidance strategy generation experiences are stored in an experience replay pool.

7. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 4, characterized in that: The real-time multi-dimensional signal of the cabin is collected, and the cabin component abnormality diagnosis model is used to perform abnormality diagnosis on the pre-processed real-time multi-dimensional signal to obtain the real-time cabin component abnormality diagnosis result, including the following steps: Collecting real-time multi-dimensional signals from the engine room, and preprocessing the real-time multi-dimensional signals to obtain preprocessed real-time multi-dimensional signals; Performing STFT processing on the preprocessed real-time sound signal of the preprocessed real-time multidimensional signal to obtain a real-time sound spectrogram, and performing STFT processing on the preprocessed real-time vibration signal to obtain a real-time vibration spectrogram; The preprocessed real-time sound signal, the corresponding real-time sound spectrogram and the real-time vibration spectrogram are input into the abnormal diagnosis model of the cabin components to perform abnormal diagnosis and obtain the real-time abnormal diagnosis result of the cabin components.

8. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 7, characterized in that: The preprocessed real-time sound signal, the corresponding real-time sound spectrogram and the real-time vibration spectrogram are input into the abnormal diagnosis model of the cabin component to perform abnormal diagnosis and obtain the real-time abnormal diagnosis result of the cabin component, including the following steps: Extracting the real-time acoustic analysis features of the preprocessed real-time sound signal, the first real-time spectrum features of the real-time sound spectrogram, and the second real-time spectrum features of the real-time vibration spectrogram; According to a preset attention weight value, the real-time acoustic analysis feature, the first real-time spectrum feature, and the second real-time spectrum feature are feature spliced ​​to obtain a real-time splicing feature; According to the real-time splicing features, abnormal diagnosis is performed to obtain real-time abnormal diagnosis results of cabin components.

9. The method for diagnosing abnormality of cabin components based on multi-dimensional signal analysis according to claim 6, characterized in that: Using the maintenance guidance model, a maintenance guidance strategy is generated for the real-time abnormal diagnosis results of the cabin components, and the real-time maintenance guidance strategy is obtained, including the following steps: Extracting a number of historical maintenance guidance strategy generation experiences from the experience replay pool, and updating the action space of the maintenance guidance model according to the number of historical maintenance guidance strategy generation experiences to obtain an updated action space; According to the real-time abnormal diagnosis result of the cabin component, the state space of the maintenance guidance model is updated to obtain an updated state space; According to the updated state space, the input layer of the deep Q network is updated, and according to the updated action space, the output layer of the deep Q network is updated to obtain an updated maintenance guidance model; Using the updated maintenance guidance model, a real-time maintenance guidance strategy and corresponding real-time maintenance guidance strategy generation experience are generated, and the real-time maintenance guidance strategy generation experience is stored in an experience replay pool.

10. A cabin component abnormality diagnosis system based on multi-dimensional signal analysis, used to implement the cabin component abnormality diagnosis method according to any one of claims 1 to 9, characterized in that: The system comprises a model building unit, an abnormality diagnosis unit and a maintenance guidance unit which are connected in sequence.

Images