Diesel engine lubricating oil system cross-working-condition fault diagnosis method and system based on deep transfer learning

Abstract

The invention provides a diesel engine lubricating oil system cross-working-condition fault diagnosis method and system based on deep transfer learning. The method comprises the following steps that physical signals of different fault types of a diesel engine lubricating oil system under different loads are collected; segmenting the physical signal by adopting a sliding window method, establishing a fault label for each segment of segmented physical signal based on a fault type, and obtaining an initial data set containing the fault labels; and on the basis of the initial data set and a diesel engine fault diagnosis model combined with deep transfer learning, diesel engine cross-working-condition fault diagnosis is carried out, and a fault diagnosis result is obtained. According to the method, data feature hierarchical fusion is carried out based on the one-dimensional convolutional neural network, feature reconstruction of structured data is realized, and the data feature representation capability is enhanced. And the generalization ability of the fault diagnosis model is improved through model fine tuning, and accurate fault identification under cross-working conditions is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of marine diesel engine fault diagnosis, and specifically relates to a cross-operating condition fault diagnosis method and system for a diesel engine lubricating oil system based on deep transfer learning. Background Art

[0002] Marine diesel engines are the core power source of ships, and their performance is directly related to the safety of ships at sea. However, diesel engines are susceptible to problems such as component corrosion, pipeline blockage, and poor lubrication under complex and changeable working conditions, resulting in performance degradation and increased failure rates. Traditional fault diagnosis methods rely on sensor monitoring and engineer experience, but with the growth of industrial big data, data-driven diagnosis methods have become a research hotspot. However, due to the diversity of diesel engine operating conditions, the generalization ability of traditional data-driven models under cross-operating conditions is limited, and it is difficult to meet the needs of practical applications. Cross-operating fault diagnosis refers to fault diagnosis under different operating conditions, specifically using data obtained under one operating condition (source domain) to identify faults under another different operating condition (target domain). In the context of diesel engine fault diagnosis, this process involves model training through source domain data (e.g., data obtained under 25% load), and then applying the model to perform fault diagnosis under different operating conditions (e.g., 50% load). This method is of great significance in practical applications, because the operating conditions of diesel engines may lead to different failure modes under different loads, so it is very critical to perform effective fault diagnosis under different operating conditions.

[0003] Although the existing technology provides a new solution for diesel engine fault diagnosis, it still has the following two limitations: (1) The model provided by the existing technology is trained and tested under specific data distribution and operating conditions. Its generalization ability may be limited when facing unprecedented operating conditions or changes in data distribution; (2) The existing technology mainly uses diesel engine vibration signals collected in experiments as the data source, while most signals in actual applications are presented in the form of thermodynamic signals such as temperature and pressure. Although the model performance has been verified in some cases, it still has significant limitations in actual industrial applications. This is mainly reflected in the fact that the vibration signal is different from the temperature and pressure signals that are more common in actual industry. The data type may lead to reduced applicability and generalization ability of the model; secondly, the collection and processing of vibration signals may be more challenging in complex industrial environments, and data acquisition is difficult and costly. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a cross-operating condition fault diagnosis method and system for a diesel engine lubricating oil system based on deep transfer learning, which improves the generalization ability of the fault diagnosis model through model fine-tuning and realizes accurate fault identification under cross-operating conditions.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A cross-operating condition fault diagnosis method for a diesel engine lubricating oil system based on deep transfer learning comprises the following steps:

[0007] Collect physical signals of different fault types of diesel engine lubricating oil system under different loads;

[0008] The physical signal is segmented by a sliding window method, and a fault label is established for each segment of the physical signal based on the fault type, so as to obtain an initial data set including the fault label;

[0009] Based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning, cross-operating condition fault diagnosis of the diesel engine is performed to obtain a fault diagnosis result.

[0010] Preferably, the fault types include five fault modes: oil filter blockage, oil cooler fouling, oil shortage, oil leakage and bypass valve leakage;

[0011] The physical signals include the pressure after the lubricating oil pump, the temperature after the lubricating oil pump, the pressure before the first lubricating oil filter, the temperature before the first lubricating oil filter, the pressure before the second lubricating oil filter, the temperature before the second lubricating oil filter, the pressure before the third lubricating oil filter, the temperature before the third lubricating oil filter, the pressure after the lubricating oil filter, the temperature after the lubricating oil filter, the pressure of the lubricating oil main oil channel, the temperature of the lubricating oil main oil channel and the cooling water outlet temperature of the lubricating oil cooler.

[0012] Preferably, the method of segmenting the physical signal using a sliding window method is:

[0013] Based on the working cycle and fault characteristics of the diesel engine lubricating oil system, the size of the sliding window is obtained;

[0014] Based on the preset sliding step size and the sliding window, the time series of the physical signal is segmented so that each window generates a subsequence containing the data features of the physical signal.

[0015] Preferably, the method for constructing a diesel engine fault diagnosis model is:

[0016] Based on the transfer learning strategy, the initial data set is divided into a source domain data set and a target domain data set;

[0017] Based on the comprehensive data feature importance calculation method, the relevance and classification contribution of the data features of the source domain data set are integrated to perform data feature classification and obtain classified features; wherein the classified features include important features and less important features;

[0018] The important features are modeled using 5 convolutional layers to obtain data detail features;

[0019] Four convolutional layers are used to model the sub-important features to obtain the broad features of the data;

[0020] splicing and reconstructing the data detail features and the data broad features to obtain fusion features;

[0021] Extracting fault features and time series features of the fused features, and performing feature classification based on the MLP model to obtain a pre-trained model for cross-operating condition fault diagnosis of diesel engines;

[0022] The pre-trained model is fine-tuned based on the target domain data set to obtain the diesel engine fault diagnosis model.

[0023] Preferably, the method for obtaining the classification characteristics is:

[0024] Based on the principal component analysis method, the weight of the correlation degree of data features is calculated;

[0025] The MDA method based on random forest is used to calculate the weight of the contribution degree of data feature classification;

[0026] Sum the weights of the data feature classification contribution degrees to obtain the total classification contribution degree of the corresponding data feature in the corresponding fault type;

[0027] Sum the weights of the correlation degree of the data features to obtain the sum of the correlation degree between the data features;

[0028] The sum of the classification contribution degrees and the sum of the correlation degrees are superimposed and normalized to obtain the percentage of comprehensive feature importance;

[0029] The importance percentages of the comprehensive features are sorted to obtain the graded features.

[0030] The present invention also provides a diesel engine lubricating oil system cross-operating condition fault diagnosis system based on deep transfer learning, which is used to implement the method, including:

[0031] A signal acquisition module, used to collect physical signals of different fault types of the diesel engine lubricating oil system under different loads;

[0032] A data set construction module, used to segment the physical signal using a sliding window method, and to establish a fault label for each segmented physical signal based on the fault type, so as to obtain an initial data set containing the fault label;

[0033] The fault diagnosis module is used to perform cross-operating condition fault diagnosis of the diesel engine based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning to obtain a fault diagnosis result.

[0034] Preferably, the data set construction module includes a signal segmentation unit, which is used to segment the physical signal using a sliding window method; the signal segmentation unit includes:

[0035] The window size definition subunit is used to obtain the size of the sliding window based on the working cycle and fault characteristics of the diesel engine lubricating oil system;

[0036] The subsequence generation subunit is used to segment the time series of the physical signal based on a preset sliding step and the sliding window, so that each window generates a subsequence containing the data features of the physical signal.

[0037] Preferably, the fault diagnosis module includes a model building unit, which is used to build a diesel engine fault diagnosis model based on the initial data set combined with deep transfer learning; the model building unit includes:

[0038] A data set division subunit, used for dividing the initial data set into a source domain data set and a target domain data set based on a transfer learning strategy;

[0039] A data feature classification subunit is used to classify data features based on a comprehensive data feature importance calculation method, integrating the relevance and classification contribution of data features of source domain data sets, and obtaining classified features; wherein the classified features include important features and less important features;

[0040] A feature modeling subunit is used to model the important features using 5 convolution layers to obtain data detail features; and to model the less important features using 4 convolution layers to obtain data broad features;

[0041] A feature fusion and reconstruction subunit is used to splice and reconstruct the data detail features and the data broad features to obtain fusion features;

[0042] A pre-training model construction sub-unit is used to extract the fault features and time series features of the fusion features, and perform feature classification based on the MLP model to obtain a pre-training model for cross-operating condition fault diagnosis of diesel engines;

[0043] The model fine-tuning subunit is used to fine-tune the pre-trained model based on the target domain data set to obtain the diesel engine fault diagnosis model.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. Through hierarchical feature fusion and deep transfer learning, accurate fault identification under cross-operating conditions is achieved, overcoming the shortcomings of traditional methods in terms of poor generalization ability under cross-operating conditions.

[0046] 2. Through experimental verification, the proposed fine-tuning method has a significant impact on the fault diagnosis effect, and the fine-tuning strategy significantly improves the migration performance of the model.

[0047] 3. Through hierarchical feature processing and convolutional layer reconstruction, the computational complexity is reduced, while the risk of model overfitting is avoided and the computational efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 This is a flow chart of a cross-operating-condition fault diagnosis method for a diesel engine lubricating oil system based on deep transfer learning according to an embodiment of the present invention;

[0050] Figure 2 This is a structural diagram of a pre-training model in an embodiment of the present invention;

[0051] Figure 3 This is an explanatory diagram of the model fine-tuning method in an embodiment of the present invention;

[0052] Figure 4 4 is a comparison chart of the model diagnosis results in the embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Embodiment 1

[0056] like Figure 1 As shown, a cross-operating condition fault diagnosis method for a diesel engine lubricating oil system based on deep transfer learning comprises the following steps:

[0057] S1: Collect physical signals of different fault types of diesel engine lubricating oil system under different loads.

[0058] A further implementation manner is that the fault types include five fault modes: lubricating oil filter blockage F1, lubricating oil cooler fouling F2, lubricating oil shortage F3, lubricating oil leakage F4 and bypass valve leakage F5.

[0059] In this embodiment, based on the pressure and temperature sensors arranged in the diesel engine lubricating oil system, thermodynamic signals and dynamic signals of different types of faults are collected at 25%, 50% and 75% load states respectively.

[0060] The physical signals include the pressure after the lubricating oil pump, the temperature after the lubricating oil pump, the pressure before the first lubricating oil filter, the temperature before the first lubricating oil filter, the pressure before the second lubricating oil filter, the temperature before the second lubricating oil filter, the pressure before the third lubricating oil filter, the temperature before the third lubricating oil filter, the pressure after the lubricating oil filter, the temperature after the lubricating oil filter, the pressure of the lubricating oil main oil channel, the temperature of the lubricating oil main oil channel and the cooling water outlet temperature of the lubricating oil cooler.

[0061] S2: Use the sliding window method to segment the physical signal, and establish a fault label for each segment of the physical signal based on the fault type to obtain an initial data set containing fault labels.

[0062] A further implementation method is to use a sliding window method to segment the physical signal:

[0063] Based on the working cycle and fault characteristics of the diesel engine lubricating oil system, the size of the sliding window is obtained; in this embodiment, the length of the sliding window is 150.

[0064] Based on the preset sliding step and sliding window, the time series of the physical signal is segmented so that each window generates a subsequence containing the data features of the physical signal. An appropriate sliding step is selected to balance the granularity of data segmentation and the computational complexity. In this embodiment, the step size is 15.

[0065] S3: Based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning, perform diesel engine cross-operating condition fault diagnosis and obtain fault diagnosis results.

[0066] A further implementation method is that the method for constructing a diesel engine fault diagnosis model is:

[0067] Based on the transfer learning strategy, the initial data set is divided into a source domain data set and a target domain data set. In this embodiment, the source domain data set is the time series data and corresponding fault labels under the load condition of diesel engine A; the target domain data set is the time series data and corresponding fault labels under the load condition of diesel engine B.

[0068] Based on the comprehensive data feature importance calculation method (CCMFI), the correlation degree and classification contribution degree of the data features of the source domain dataset are integrated to classify the data features and obtain the graded features; among them, the graded features include important features (SF) and less important features (LSF).

[0069] A further implementation method is that the method for obtaining the classification features is:

[0070] Based on the principal component analysis method, the weight of the correlation degree of data features is calculated;

[0071] In this embodiment, principal component analysis (PCA) is used to evaluate the correlation of data features, and the calculation expression is:

[0072]

[0073] Where Cov(X centered ) is the covariance matrix calculation, λ i is the eigenvalue, and υ is the corresponding eigenvector. The data feature correlation weight calculated by EVR, where p is the number of data features and e is the number of principal components. In this embodiment, the number of principal components is selected as 6, and the calculation results of the correlation weight of data features under different fault conditions are shown in Table 1:

[0074] Table 1

[0075]

[0076] Based on the MDA method of random forest, the weight of the contribution degree of data feature classification is calculated; specifically, the calculation expression is:

[0077]

[0078] Among them, MDA(X) represents the decrease in model accuracy after feature disruption, and also reflects the classification contribution weight of the feature. Acc baseline represents the classification accuracy using the original validation set, Denotes the classification accuracy of the sample set after feature exchange, D val represents the original validation set, Represents the sample set after feature exchange.

[0079] Specifically, 90% of the source domain data is used as the training set, and 10% of the source domain data is used as the validation set. When constructing each decision tree, Random Forest uses a method called Bootstrap sampling, which is to randomly select multiple sub-sample sets from the original training set with replacement. There will definitely be samples that are not selected. This part of the data is called out-of-bag data. Randomly exchanging feature values ​​means that for each feature, the out-of-bag data of each tree is traversed and the value of the feature in the sample is randomly exchanged.

[0080] In this embodiment, the classification contribution weight calculation results of data features under different fault states are shown in Table 2:

[0081] Table 2

[0082]

[0083] The weights of the data feature classification contribution are summed to obtain the sum of the classification contribution of the corresponding data features in the corresponding fault type; the weights of the data feature correlation are summed to obtain the sum of the correlations between the data features; the sum of the classification contribution and the sum of the correlation are superimposed and normalized to obtain the percentage of comprehensive feature importance; specifically, CCMFI takes the above two indicators into consideration, and the calculation process is as follows:

[0084]

[0085] Where k represents the classification category, m represents the number of data features, Find the total contribution of feature X in k categories. Calculate the sum of the feature correlations between feature X and other feature variables, add the last two items together, and normalize them to get the percentage of comprehensive feature importance.

[0086] In this embodiment, the calculation results of the comprehensive feature importance percentage are shown in Table 3:

[0087] Table 3

[0088]

[0089] The importance percentages of the comprehensive features are sorted to obtain hierarchical features. In this embodiment, six important features (SF) and seven less important features (LSF) are obtained according to the importance percentages of the comprehensive features, forming a feature hierarchy.

[0090] The SF combination is the temperature of the main oil gallery, the temperature before the first oil filter, the temperature after the oil pump, the temperature before the third oil filter, the temperature after the oil filter and the temperature before the second oil filter.

[0091] The LSF combination is the lubricating oil main oil channel pressure, the lubricating oil cooler cooling water outlet temperature, the pressure before the first lubricating oil filter, the pressure before the third lubricating oil filter, the pressure after the lubricating oil pump, the pressure before the second lubricating oil filter and the pressure after the lubricating oil filter.

[0092] In this embodiment, convolutional layers of different depths are used to model the data features in different scales, complete feature reconstruction and fusion, and 5 convolutional layers are used to model important features to obtain data detail features; 4 convolutional layers are used to model less important features to obtain data extensive features. The method of multi-scale feature learning is used to capture data features. By increasing the number of channels layer by layer, the model can extract features at different levels and enhance the ability to identify important features in the data; after feature amplification, feature reduction is performed to compress information, reduce redundancy, and improve computational efficiency. Finally, combined with the activation function, nonlinear transformation can be introduced, so that the model can learn more complex mapping relationships.

[0093] Specifically, two base network classes for feature extraction are built using the Pytorch framework. One base network class has a 5-layer one-dimensional convolutional neural network, and Relu is used as the activation function to extract the features of SF. The other base network class has a 4-layer one-dimensional convolutional neural network to extract the features of LSF. Finally, Concat is used to complete the fusion of the two parts of data features. The specific parameters are shown in Table 4.

[0094] In this embodiment, the convolution layer feature extraction and convolution calculation process are as follows:

[0095]

[0096] in represents the i-th channel output in the l-1th layer, indicating The output of the cth channel in layer l, which has a convolution kernel represents bias, * represents convolution calculation, and f(·) represents activation function.

[0097] The Trasforms module is mainly used for feature fusion and realizes the fusion of tensors through the Concat function.

[0098] The detailed features and extensive features of the data are spliced ​​and reconstructed to obtain fusion features;

[0099] Table 4

[0100]

[0101] Data fusion involves integrating data features from different sources to enhance the completeness and accuracy of information. In this embodiment, two one-dimensional convolutional neural networks are used to extract the features of important and less important parameters respectively, and finally the two feature vectors are combined by concatenation. This method can improve the performance of the model.

[0102] Extract the fault features and time series features of the fusion features, and perform feature classification based on the MLP model to obtain a pre-trained model for cross-operating condition fault diagnosis of a diesel engine; in this embodiment, the pre-trained model for cross-operating condition fault diagnosis of a diesel engine consists of two core parts: a feature extraction module and a classifier module, such as Figure 2 As shown in Figure 2, the feature extraction is divided into two submodules: CNN module and time series module, while the classifier module adopts MLP model.

[0103] In a specific embodiment, a pre-trained network is built using the Pytorch framework. The convolutional neural network (CNN) module includes four convolutional computing blocks for capturing general features of the data (fault features), of which the first three convolutional computing blocks are used for underlying feature extraction. The basic structure of the convolutional computing module includes a convolutional layer, batch normalization, activation function, and pooling layer. The timing module uses a gated recurrent unit (GRU) module to extract timing features.

[0104] The timing module mainly uses the GRU module, and the calculation process is as follows:

[0105]

[0106] Among them, Z t and r t They represent the switch states of the reset gate and the update gate, respectively. t ' is a candidate hidden state, is the GRU output.

[0107] The specific parameters are shown in Table 5.

[0108] Table 5

[0109]

[0110]

[0111] The pre-trained model is fine-tuned based on the target domain dataset and is fine-tuned to the target working conditions to adapt to the working conditions and fault types of the target domain, thus obtaining a diesel engine fault diagnosis model.

[0112] Specifically, the fine-tuning method is as follows Figure 3 As shown, it includes freezing the parameters of the first three convolution blocks in the CNN layer, and fine-tuning CNNBlock4, GRU and classifier to achieve the best migration effect. The parameters of the first three convolution blocks in the CNN layer come from the parameters of the pre-training process. These parameters represent the diagnostic knowledge of the source domain dataset. CNNBlock4, GRU and classifier are fine-tuned using target domain data to allow the model to learn diagnostic knowledge applicable to the target domain dataset.

[0113] The data of the target domain test data set is input into the fine-tuned diesel engine fault diagnosis model, and the diesel engine cross-operating condition fault diagnosis is performed to obtain the fault type.

[0114] Specifically, it involves six transfer learning tasks in the initial data set of the diesel engine lubricating oil system, marked as Task1 to Task6. Taking Task1 as an example, Task 1 uses data under 25% load as source domain data and data under 50% load as target domain data for fine-tuning and testing.

[0115] In this embodiment, Task 1: 25% load → 50% load migration task; Task 2: 25% load → 75% load migration task; Task 3: 50% load → 25% load migration task; Task 4: 50% load → 75% load migration task; Task 5: 75% load → 25% load migration task; Task 6: 75% load → 50% load migration task.

[0116] In this embodiment, the target domain dataset is divided into a training set, a validation set, and a test set in a ratio of 7:1.5:1.5, which means that 70% of the dataset is designated for training, 15% for model validation, and the remaining 15% for testing.

[0117] In this embodiment, the accuracy rate is used as a performance indicator to evaluate the diagnostic performance of the proposed method. The accuracy rate refers to the proportion of the number of correctly classified diesel engine data samples to the total number of samples. The calculation formula is as follows:

[0118] Accuracy=(TP+TN) / (TP+TN+FN+FP)

[0119] Specifically, TP represents true positive examples, which are actually the number of samples in normal state and predicted by the model as normal state; TN represents true negative examples, which are actually faulty samples and predicted by the model as correct faulty samples; FN represents false positive examples, which are actually faulty samples but predicted by the model as normal state; FP represents false negative examples, which are actually the number of samples in normal state but predicted by the model as faulty samples.

[0120] In this embodiment, the diesel engine cross-operating condition fault diagnosis model based on model fine-tuning has good accuracy for diesel engine fault diagnosis under cross-operating conditions. The performance in the six migration tasks is as follows: Figure 4 As shown in the figure, in the six migration tasks, the diagnostic accuracy is guaranteed to be above 88.06%, and the average accuracy can reach 93.99%, which has good generalization.

[0121] Embodiment 2

[0122] The present invention also provides a diesel engine lubricating oil system cross-operating condition fault diagnosis system based on deep transfer learning, which is used to implement the method, including:

[0123] A signal acquisition module, used to collect physical signals of different fault types of the diesel engine lubricating oil system under different loads;

[0124] A data set construction module is used to segment the physical signal using a sliding window method, and to establish a fault label for each segmented physical signal based on the fault type, so as to obtain an initial data set containing the fault label;

[0125] The fault diagnosis module is used to perform cross-operating condition fault diagnosis of diesel engines based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning to obtain fault diagnosis results.

[0126] A further implementation method is that the data set construction module includes a signal segmentation unit for segmenting the physical signal using a sliding window method; the signal segmentation unit includes:

[0127] The window size definition subunit is used to obtain the size of the sliding window based on the working cycle and fault characteristics of the diesel engine lubricating oil system;

[0128] The subsequence generation subunit is used to segment the time series of the physical signal based on a preset sliding step and a sliding window, so that each window generates a subsequence containing the data features of the physical signal.

[0129] A further implementation method is that the fault diagnosis module includes a model building unit, which is used to build a diesel engine fault diagnosis model based on the initial data set combined with deep transfer learning; the model building unit includes:

[0130] A data set division subunit is used to divide the initial data set into a source domain data set and a target domain data set based on a transfer learning strategy;

[0131] The data feature classification subunit is used to classify data features based on a comprehensive data feature importance calculation method, integrating the relevance and classification contribution of data features of the source domain data set, and obtaining classified features; wherein the classified features include important features and less important features;

[0132] The feature modeling subunit is used to model important features using 5 convolutional layers to obtain detailed features of the data; and to model less important features using 4 convolutional layers to obtain broad features of the data;

[0133] The feature fusion and reconstruction subunit is used to splice and reconstruct the data detail features and the data broad features to obtain fusion features;

[0134] The pre-training model construction sub-unit is used to extract the fault features and time series features of the fusion features, and perform feature classification based on the MLP model to obtain a pre-training model for cross-operating condition fault diagnosis of diesel engines;

[0135] The model fine-tuning subunit is used to fine-tune the pre-trained model based on the target domain dataset to obtain a diesel engine fault diagnosis model.

[0136] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A cross-operating fault diagnosis method for diesel engine lubricating oil system based on deep transfer learning, characterized in that: The following steps are involved: Collect physical signals of different fault types of diesel engine lubricating oil system under different loads; The physical signal is segmented by a sliding window method, and a fault label is established for each segment of the physical signal based on the fault type, so as to obtain an initial data set including the fault label; Based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning, cross-operating condition fault diagnosis of the diesel engine is performed to obtain a fault diagnosis result.

2. The cross-operating condition fault diagnosis method for diesel engine lubricating oil system based on deep transfer learning according to claim 1 is characterized in that: The fault types include five fault modes: oil filter blockage, oil cooler fouling, oil shortage, oil leakage and bypass valve leakage; The physical signals include the pressure after the lubricating oil pump, the temperature after the lubricating oil pump, the pressure before the first lubricating oil filter, the temperature before the first lubricating oil filter, the pressure before the second lubricating oil filter, the temperature before the second lubricating oil filter, the pressure before the third lubricating oil filter, the temperature before the third lubricating oil filter, the pressure after the lubricating oil filter, the temperature after the lubricating oil filter, the pressure of the lubricating oil main oil channel, the temperature of the lubricating oil main oil channel and the cooling water outlet temperature of the lubricating oil cooler.

3. The cross-operating condition fault diagnosis method for diesel engine lubricating oil system based on deep transfer learning according to claim 1 is characterized in that: The method of segmenting the physical signal using the sliding window method is as follows: Based on the working cycle and fault characteristics of the diesel engine lubricating oil system, the size of the sliding window is obtained; Based on the preset sliding step size and the sliding window, the time series of the physical signal is segmented so that each window generates a subsequence containing the data features of the physical signal.

4. The cross-operating condition fault diagnosis method for diesel engine lubricating oil system based on deep transfer learning according to claim 3 is characterized in that: The method of constructing a diesel engine fault diagnosis model is: Based on the transfer learning strategy, the initial data set is divided into a source domain data set and a target domain data set; Based on the comprehensive data feature importance calculation method, the relevance and classification contribution of the data features of the source domain data set are integrated to perform data feature classification and obtain classified features; wherein the classified features include important features and less important features; The important features are modeled using 5 convolutional layers to obtain data detail features; Four convolutional layers are used to model the sub-important features to obtain the broad features of the data; splicing and reconstructing the data detail features and the data broad features to obtain fusion features; Extracting fault features and time series features of the fused features, and performing feature classification based on the MLP model to obtain a pre-trained model for cross-operating condition fault diagnosis of diesel engines; The pre-trained model is fine-tuned based on the target domain data set to obtain the diesel engine fault diagnosis model.

5. The cross-operating condition fault diagnosis method for diesel engine lubricating oil system based on deep transfer learning according to claim 4 is characterized in that: The method for obtaining the classification features is: Based on the principal component analysis method, the weight of the correlation degree of data features is calculated; The MDA method based on random forest is used to calculate the weight of the contribution degree of data feature classification; Sum the weights of the data feature classification contribution degrees to obtain the total classification contribution degree of the corresponding data feature in the corresponding fault type; Sum the weights of the correlation degree of the data features to obtain the sum of the correlation degree between the data features; The sum of the classification contribution degrees and the sum of the correlation degrees are superimposed and normalized to obtain the percentage of comprehensive feature importance; The importance percentages of the comprehensive features are sorted to obtain the graded features.

6. A diesel engine lubricating oil system cross-operating condition fault diagnosis system based on deep transfer learning, used to implement the method described in any one of claims 1 to 5, characterized in that: include: A signal acquisition module, used to collect physical signals of different fault types of the diesel engine lubricating oil system under different loads; A data set construction module, used to segment the physical signal using a sliding window method, and to establish a fault label for each segmented physical signal based on the fault type, so as to obtain an initial data set containing the fault label; The fault diagnosis module is used to perform cross-operating condition fault diagnosis of the diesel engine based on the initial data set and the diesel engine fault diagnosis model combined with deep transfer learning to obtain a fault diagnosis result.

7. The diesel engine lubricating oil system cross-operating condition fault diagnosis system based on deep transfer learning according to claim 6 is characterized in that: The data set construction module includes a signal segmentation unit, which is used to segment the physical signal using a sliding window method; the signal segmentation unit includes: The window size definition subunit is used to obtain the size of the sliding window based on the working cycle and fault characteristics of the diesel engine lubricating oil system; The subsequence generation subunit is used to segment the time series of the physical signal based on a preset sliding step and the sliding window, so that each window generates a subsequence containing the data features of the physical signal.

8. The diesel engine lubricating oil system cross-operating condition fault diagnosis system based on deep transfer learning according to claim 7 is characterized in that: The fault diagnosis module includes a model building unit, which is used to build a diesel engine fault diagnosis model based on the initial data set combined with deep transfer learning; the model building unit includes: A data set division subunit, used for dividing the initial data set into a source domain data set and a target domain data set based on a transfer learning strategy; A data feature classification subunit is used to classify data features based on a comprehensive data feature importance calculation method, integrating the relevance and classification contribution of data features of source domain data sets, and obtaining classified features; wherein the classified features include important features and less important features; A feature modeling subunit is used to model the important features using 5 convolution layers to obtain data detail features; and to model the less important features using 4 convolution layers to obtain data broad features; A feature fusion and reconstruction subunit is used to splice and reconstruct the data detail features and the data broad features to obtain fusion features; A pre-training model construction sub-unit is used to extract the fault features and time series features of the fusion features, and perform feature classification based on the MLP model to obtain a pre-training model for cross-operating condition fault diagnosis of diesel engines; The model fine-tuning subunit is used to fine-tune the pre-trained model based on the target domain data set to obtain the diesel engine fault diagnosis model.

Images