A fault diagnosis method, device and vehicle for an oil injector

Abstract

The present application relates to the technical field of fuel injector fault diagnosis, and particularly relates to a fault diagnosis method and device for a fuel injector and a vehicle. The method comprises: obtaining vibration data of a fuel injector to obtain to-be-identified data, wherein the to-be-identified data is vibration data different from working conditions of labeled data; using the to-be-identified data and the labeled data and based on an improved joint distribution adaptation, training a to-be-trained fault identification model to obtain a trained fault identification model; and inputting the to-be-identified data into the fault identification model to obtain a fault identification result. Through the labeled data, the to-be-identified data different from working conditions of the labeled data and the improved joint distribution adaptation, the to-be-trained fault identification model is trained, the to-be-trained fault identification model can be trained for different working conditions, and thus the accuracy of fault identification is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fuel injector fault diagnosis, and in particular to a fault diagnosis method and device for a fuel injector and a vehicle. BACKGROUND

[0002] Traditional fault diagnosis methods mainly rely on expert experience and simple signal processing techniques, and have the disadvantages of low diagnosis accuracy, poor real-time performance, weak adaptability to working condition changes, etc. In actual applications, high-pressure common rail fuel injectors often work under variable working conditions, including pressure changes, load changes, speed fluctuations, fuel quality differences, and environmental temperature changes, etc. These variable working conditions have a significant impact on the working state and fault characteristics of high-pressure common rail fuel injectors, increasing the complexity and difficulty of fault diagnosis. Therefore, it is particularly important to study fault diagnosis techniques for high-pressure common rail fuel injectors under variable working conditions.

[0003] In summary, there is a need to provide a fault diagnosis method, device and vehicle for a fuel injector that can improve the accuracy of fault diagnosis under different working conditions. SUMMARY

[0004] To solve the above problems, the present application provides a fault diagnosis method, device and vehicle for a fuel injector.

[0005] In one aspect, the present application provides a fault diagnosis method for a fuel injector, comprising the following steps:

[0006] Obtaining vibration data of the fuel injector to obtain to-be-identified data, wherein the to-be-identified data is vibration data different from the working condition of the labeled data;

[0007] Training a to-be-trained fault identification model using the to-be-identified data and the labeled data based on an improved joint distribution adaptation to obtain a trained fault identification model;

[0008] Inputting the to-be-identified data into the fault identification model to obtain a fault identification result.

[0009] Further, the fault diagnosis method for a fuel injector as described above, the training of the to-be-trained fault identification model using the to-be-identified data and the labeled data based on the improved joint distribution adaptation to obtain the trained fault identification model, comprises:

[0010] Inputting the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and respectively obtaining features of the to-be-identified data and the labeled data based on an attention mechanism to obtain first features and second features, wherein the first features are features of the to-be-identified data, and the second features are features of the labeled data.

[0011] inputting the first feature and the second feature into a failure prediction unit to obtain a first prediction result and a second prediction result, wherein the first prediction result is a prediction result obtained according to the first feature, and the second prediction result is a prediction result obtained according to the second feature;

[0012] inputting the first feature, the second feature, a label and a pseudo label into a domain adaptation unit, and determining a domain adaptation loss based on the improved joint distribution adaptation, wherein the label is a label of the labeled data, and the pseudo label is the first prediction result obtained by the failure prediction unit in the previous time when the first feature is predicted;

[0013] updating the to-be-trained model according to the domain adaptation loss and a classification loss to obtain a trained failure identification model.

[0014] Further, the failure diagnosis method for the fuel injector as described above, before the updating of the to-be-trained model according to the domain adaptation loss and the classification loss to obtain the trained failure identification model, further comprises:

[0015] determining the classification loss according to the second prediction result and the label.

[0016] Further, the failure diagnosis method for the fuel injector as described above, the inputting of the first feature, the second feature, a label and a pseudo label into a domain adaptation unit, and the determination of a domain adaptation loss based on the improved joint distribution adaptation, comprises:

[0017] determining the domain adaptation loss according to the first feature, the second feature, the label, the pseudo label and the improved joint distribution adaptation, and the calculation method is as follows:

[0018]

[0019] wherein IJDA represents the domain adaptation loss, s represents the labeled data, t represents the to-be-identified data, Y s represents a label data set including the label of the labeled data, Y t represents a pseudo label data set including the pseudo label of the to-be-identified data, represents a calculation data set feature, represents a first feature, represents a second feature, E(·) represents an expectation of data, represents an expectation of calculating a second feature edge probability distribution, represents an expectation of calculating a first feature edge probability distribution, represents the expectation of calculating the probability of the second feature occurring given that the class is c, represents the expectation of calculating the probability of the first feature occurring given that the class is c, P(X s represents the edge probability distribution of the data set of labeled data, P(Y s represents the class prior distribution of the label, P(Y t represents the class prior distribution of the pseudo label, c represents the data class, X s represents a first data set comprising the data to be identified, X t represents a second data set comprising the labeled data.

[0020] Further, the fault diagnosis method for the fuel injector as described above, the updating the to-be-trained model according to the domain adaptation loss and the classification loss to obtain the trained fault identification model, comprising:

[0021] updating the feature extraction unit, the fault prediction unit and the domain adaptation unit according to the domain adaptation loss and the classification loss;

[0022] continuing to perform the step of inputting the to-be-identified data and the labeled data into the feature extraction unit in the to-be-trained fault identification model, and obtaining the features of the to-be-identified data and the labeled data based on the attention mechanism to obtain the first feature and the second feature until a preset training threshold is reached, or the domain adaptation loss and / or the classification loss is less than or equal to a preset loss threshold, to obtain the trained fault identification model comprising the feature extraction unit and the fault prediction unit.

[0023] Further, the fault diagnosis method for the fuel injector as described above, the obtaining the features of the to-be-identified data and the labeled data based on the attention mechanism to obtain the first feature and the second feature, comprising:

[0024] extracting a plurality of frequency features of the to-be-identified data and the labeled data based on a convolution block attention module to obtain a plurality of first frequency features corresponding to the to-be-identified data and a plurality of second frequency features corresponding to the labeled data;

[0025] merging the plurality of first frequency features and then extracting features again to obtain the first feature, and merging the plurality of second frequency features and then extracting features again to obtain the second feature.

[0026] Further, the fault diagnosis method for the fuel injector as described above, the obtaining the vibration data of the fuel injector to obtain the to-be-identified data, comprising:

[0027] obtaining the vibration data of the fuel injector;

[0028] The working condition of the vibration data is judged to determine the to-be-identified data.

[0029] In a second aspect, the application provides a fault diagnosis device for an oil injector, comprising a data acquisition module and a fault identification module.

[0030] The data acquisition module is configured to acquire vibration data of the oil injector to obtain to-be-identified data.

[0031] The fault identification module is configured to use the to-be-identified data and labeled data and based on an improved joint distribution adaptation to train a to-be-trained fault identification model to obtain a trained fault identification model; and input the to-be-identified data into the fault identification model to obtain a fault identification result.

[0032] Further, the fault diagnosis device for an oil injector described above, the fault identification module comprises a feature extraction unit, a fault prediction unit, a domain adaptation unit and a training unit.

[0033] The feature extraction unit is configured to input the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and respectively acquire features of the to-be-identified data and the labeled data based on an attention mechanism to obtain first features and second features, wherein the first features are features of the to-be-identified data, and the second features are features of the labeled data.

[0034] The fault prediction unit is configured to input the first features and the second features into a fault prediction unit to obtain first prediction results and second prediction results, wherein the first prediction results are prediction results obtained according to the first features, and the second prediction results are prediction results obtained according to the second features.

[0035] The domain adaptation unit is configured to input the first features, the second features, labels and pseudo labels into a domain adaptation unit, and determine a domain adaptation loss based on the improved joint distribution adaptation, wherein the labels are labels of the labeled data, and the pseudo labels are the first prediction results obtained by the fault prediction unit in the previous time for predicting the first features.

[0036] The training unit is configured to update the to-be-trained model according to the domain adaptation loss and a classification loss to obtain the trained fault identification model.

[0037] In a third aspect, the application provides a vehicle comprising a vehicle terminal, wherein the vehicle terminal is configured to execute the fault diagnosis method for an oil injector according to any one of the first aspect.

[0038] The application has the advantages that the training of the to-be-trained fault identification model is performed through the labeled data, the to-be-identified data different from the working conditions of the labeled data, and the improved joint distribution adaptation, the to-be-trained fault identification model can be trained for different working conditions, and the accuracy of fault identification is improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of preferred implementations, and are not meant to limit the present application. Furthermore, the same reference numerals are used throughout the various drawings to denote the same or similar parts. In the drawings:

[0040] Figure 1 is a schematic diagram of a fault diagnosis method for an oil injector provided by the application;

[0041] Figure 2 is a schematic diagram of a flow of a fault diagnosis method for an oil injector provided by the application;

[0042] Figure 3 is a schematic diagram of a feature extraction unit of a fault diagnosis method for an oil injector provided by the application;

[0043] Figure 4 is a schematic diagram of another fault diagnosis method for an oil injector provided by the application;

[0044] Figure 5 is a schematic diagram of a fault diagnosis device for an oil injector provided by the application;

[0045] Figure 6 is a schematic diagram of another fault diagnosis device for an oil injector provided by the application. DETAILED DESCRIPTION

[0046] Exemplary embodiments of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0047] The existing fault identification method still faces challenges such as insufficient feature extraction and insufficient model generalization ability when dealing with complex working condition changes and diversified fault modes. Especially under variable working condition conditions, the traditional model often has difficulty in adapting to the change of fault features under different working conditions, resulting in a significant reduction in diagnosis accuracy and robustness.

[0048] To solve the above problems, the embodiments of the present application propose a fault diagnosis method, device and vehicle for an injector. By training a fault recognition model to be trained with labeled data, data to be identified with working conditions different from those of the labeled data, and an improved joint distribution adaptation, the fault recognition model to be trained can be trained for different working conditions, thereby improving the accuracy and robustness of fault recognition.

[0049] In the first aspect, according to the embodiment of the present application, a fault diagnosis method for a fuel injector is proposed, such as Figure 1 As shown, the following steps are included:

[0050] S101, acquiring vibration data of the injector to obtain data to be identified, wherein the data to be identified is vibration data of a different working condition from the labeled data;

[0051] S102, using the data to be identified and the labeled data and adaptively training the fault identification model to be trained based on the improved joint distribution to obtain a trained fault identification model;

[0052] S103: Input the data to be identified into the fault identification model to obtain a fault identification result.

[0053] like Figure 2 As shown, the vibration data of the injector is obtained to obtain the data to be identified, including: obtaining the vibration data of the injector; and performing working condition judgment on the vibration data to determine the data to be identified.

[0054] Among them, the working conditions include the pressure during operation (working pressure). The vibration data can be classified according to the working pressure at which the vibration data is located. When the working pressure of the vibration data does not belong to the working pressure of the labeled data, the vibration data is used as data to be identified. That is, assuming that the working pressure of the labeled data includes 70 MPa and 90 MPa, if the working pressure of the vibration data is 70 MPa, then the vibration data is not used as data to be identified. If the working pressure of the vibration data is 80 MPa, then the vibration data is used as data to be identified.

[0055] like Figure 2 As shown, data is collected through high-precision vibration sensors installed in key parts of the high-pressure common rail injector equipment. The vibration sensors are used to capture the vibration signals of the injector equipment under different working conditions in real time and convert the vibration signals into electrical signals for output.

[0056] like Figure 2As shown, the electrical signal of the vibration sensor can then be received by a data acquisition board card, which has multi-channel high-speed acquisition function and can perform high-precision digital processing on the vibration signal. After real-time filtering processing and combined with adaptive noise reduction technology, external interference noise and low-frequency interference or other noise sources that may be introduced during the acquisition process are effectively removed, ensuring high-quality output of the acquired signal and improving the clarity and accuracy of the signal. The acquisition board card can also dynamically adjust the noise reduction strategy, optimize the noise reduction effect according to the vibration signal characteristics under different working conditions, and improve the reliability and stability of the data. The board supports high-resolution digital processing to ensure that the details of the vibration signal are preserved with high fidelity. During this process, the board automatically adjusts the filter parameters for the current moment using the filter parameters obtained at the previous moment, and then filters the collected digital signal using the filter for the current moment to achieve real-time noise reduction, providing more accurate signal input for subsequent data analysis and fault diagnosis. The digitized vibration signal is transmitted to the data division module, such as the upper computer, as vibration data through a data transmission interface.

[0057] As shown in Figure 2 The data division module includes a data processing unit and a storage unit. The data processing unit performs real-time analysis on the vibration signal (vibration data) according to the working conditions of the high-pressure common rail fuel injector, divides the vibration data into source domain data containing fault categories (labeled data) and target domain data not containing fault categories (to be identified data) according to the working conditions, and saves the divided data to the storage unit for further analysis and fault identification (fault diagnosis).

[0058] Before updating the to-be-trained model according to the domain adaptation loss and the classification loss to obtain the trained fault identification model, it further includes: determining the classification loss according to the second prediction result and the label.

[0059] Among them, the classification loss includes cross-entropy loss. The classification loss can be determined by a cross-entropy loss function according to the second prediction result and the label.

[0060] As shown in Figure 2As shown, the training of the to-be-trained fault identification model based on the improved joint distribution adaptation using the to-be-identified data and the labeled data and based on the improved joint distribution adaptation training of the to-be-trained fault identification model to obtain the trained fault identification model includes: inputting the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and respectively obtaining features of the to-be-identified data and the labeled data based on an attention mechanism to obtain first features and second features, wherein the first features are features of the to-be-identified data, and the second features are features of the labeled data; inputting the first features and the second features into a fault prediction unit to obtain first prediction results and second prediction results, wherein the first prediction results are prediction results obtained according to the first features, and the second prediction results are prediction results obtained according to the second features; inputting the first features, the second features, a label, and a pseudo label into a domain adaptation unit, and determining a domain adaptation loss based on the improved joint distribution adaptation, wherein the label is a label of the labeled data, and the pseudo label is the first prediction result obtained by the fault prediction unit in the previous time for predicting the first features; updating the to-be-trained model according to the domain adaptation loss and a classification loss to obtain the trained fault identification model.

[0061] In the case of predicting the first features and the second features for the first time (the first time), since the first prediction result of the previous time before the first time is not obtained, that is, there is no pseudo label of the previous time corresponding to the first prediction result obtained for the first time, in the case of predicting the first features and the second features for the first time, after obtaining the first prediction result and the second prediction result, the step of obtaining the vibration data of the fuel injector to obtain the to-be-identified data is continued. That is, in the case of predicting the first features and the second features for the first time, the classification loss is not determined according to the second prediction result and the label, and the to-be-trained model is not updated according to the domain adaptation loss and the classification loss.

[0062] Updating the to-be-trained model according to the domain adaptation loss and the classification loss includes updating the feature extraction unit, the fault prediction unit, and the domain adaptation unit according to the domain adaptation loss and the classification loss.

[0063] Respectively obtaining the features of the to-be-identified data and the labeled data based on the attention mechanism to obtain the first features and the second features includes: respectively extracting a plurality of frequency features of the to-be-identified data and the labeled data based on a convolution block attention module (CBAM) to obtain a plurality of first frequency features corresponding to the to-be-identified data and a plurality of second frequency features corresponding to the labeled data; after merging the plurality of first frequency features, the features are extracted again to obtain the first features, and after merging the plurality of second frequency features, the features are extracted again to obtain the second features.

[0064] The embodiment of the present application extracts features by constructing a multi-attention convolutional network as a feature extraction unit, expands the receptive field by using multi-channel convolution, and constructs a one-dimensional multi-scale convolutional neural network model for feature extraction. Since a wider convolution kernel (such as any one of 1x64 to 1x128 width) can capture a larger range of features, it can be used to extract low-frequency features in the vibration signal, while a smaller (narrow) convolution kernel (such as 1x3) can be used to capture local details and high-frequency features in the vibration signal. The feature extraction unit of the embodiment of the present application includes at least 3 channels, wherein each channel includes 2 convolution blocks, and after all channels and before pooling, 1 convolution block is further included.

[0065] As shown in Figure 3 , the feature extraction unit includes 3 channels and 7 convolution blocks. Among them, the first channel Ch1 includes the first convolution block 11 and the first second convolution block 12; the second channel Ch2 includes the second convolution block 21 and the second second convolution block 22; the third channel Ch3 includes the third convolution block 31 and the third second convolution block 32; after the 3 channels, the fourth convolution block 40 is further included. Each convolution block in turn includes a one-dimensional convolution layer 51, a batch normalization layer 52, a convolution attention layer 53, and a nonlinear activation layer 54. The convolution attention layer 53 includes a convolution block attention module for dynamically focusing on the key parts of the features to improve the network's feature expression ability for the vibration signal (vibration data). Attention mechanism is a way to simulate the human attention allocation process, and is usually applied in neural networks and machine learning models. In deep learning, attention mechanism can make the model focus on specific parts of the input data to better process information. By adding a convolution attention layer 53 with attention mechanism after the one-dimensional convolution layer 51, channel attention and spatial attention can be jointly applied to further improve the feature extraction capability.

[0066] As shown in Figure 3 , in the embodiment of the present application, at least one of the convolution kernels of the first convolution block 11, the second convolution block 21 and the third convolution block 31 is a smaller (narrow) convolution kernel (such as 1x3), and the other two are different from each other. Two of the wider convolution kernels (such as any one of 1x64 to 1x128 width). Different frequencies of features in the to-be-identified data or labeled data are extracted through the 3 channels, and then the features of different frequencies are spliced and further extracted by the fourth convolution block. After the fourth convolution block 40, an adaptive average pooling layer (such as a one-dimensional adaptive average pooling layer) 55 and a fully connected layer 56 are further included, wherein the fully connected layer 56 is used to flatten the received features to obtain the first feature or the second feature.

[0067] The embodiment of the application proposes a high-pressure common rail fuel injector feature extraction method based on multi-attention convolution under variable working conditions to reduce the confusion between different faults of the target domain data (to-be-identified data), and extract more comprehensive features under complex conditions of variable working conditions.

[0068] The obtained labeled data and to-be-identified data are input into the feature extraction unit with the multi-attention convolution network for feature extraction to obtain features with more characteristics. Then, the joint distribution adaptation is improved by the Bayes theorem, and the conditional distribution is calculated using the pseudo label to make up for the deficiency in the joint distribution adaptation, so as to improve the accuracy of fault identification.

[0069] As shown in Figure 4 , the first feature, the second feature, the label and the pseudo label are input into the domain adaptation unit, and the domain adaptation loss is determined based on the improved joint distribution adaptation, including: determining the domain adaptation loss according to the first feature, the second feature, the label, the pseudo label and the improved joint distribution adaptation, and the calculation method is as follows:

[0070]

[0071] wherein, IJDA represents the domain adaptation loss, s represents the labeled data, t represents the to-be-identified data, Y s represents a label data set including labels of the labeled data, Y t represents a pseudo label data set including pseudo labels of the to-be-identified data, represents the calculation of the data set feature, represents the first feature, represents the second feature, and E(·) represents the expectation of the data, represents the calculation of the expectation of the second feature edge probability distribution, represents the calculation of the expectation of the first feature edge probability distribution, and P(X|Y=c) represents the calculation of the class conditional probability distribution, represents the calculation of the expectation of the second feature occurrence probability under the condition that the class is c, represents the calculation of the expectation of the first feature occurrence probability under the condition that the class is c, and P(X s represents the calculation of the edge probability distribution of the data set of the labeled data, P(Y s =c) represents the class prior distribution of the label, and P(Y t =c) represents the class prior distribution of the pseudo label, c represents the data class, X s represents a first data set including the to-be-identified data, X t represents a second data set including the labeled data.

[0072] As shown in Figure 4As shown, the to-be-trained model is updated according to the domain adaptation loss and the classification loss to obtain a trained fault identification model, including: updating the feature extraction unit, the fault prediction unit and the domain adaptation unit according to the domain adaptation loss and the classification loss; continuing to perform the steps of inputting the to-be-identified data and the labeled data into the feature extraction unit in the to-be-trained fault identification model, and obtaining the features of the to-be-identified data and the labeled data based on the attention mechanism to obtain the first feature and the second feature until a preset training threshold is reached, or the domain adaptation loss and / or the classification loss is less than or equal to a preset loss threshold, to obtain the trained fault identification model including the feature extraction unit and the fault prediction unit.

[0073] wherein the trained multi-attention convolutional network is output as the fault identification model when a training round requirement (a preset training threshold) is reached or the domain adaptation loss and / or the classification loss is less than or equal to a preset loss threshold, otherwise the training is continued.

[0074] Then, the obtained to-be-identified data is input into the trained fault identification model to output the fault diagnosis result under the variable working condition.

[0075] Next, the improved joint distribution adaptation determination of the domain adaptation loss of the embodiment of the present application is further described.

[0076] First, it is assumed that X s represents a source domain data set, that is, a labeled data set including labeled data, Y s represents a source domain data label set, that is, a label data set including labels of labeled data, X t represents a target domain data set, that is, a to-be-identified data set including to-be-identified data, Y t represents a target domain data label set, that is, a pseudo label data set including pseudo labels of to-be-identified data, s represents the data distribution of the source domain, that is, the labeled data, and t represents the data distribution of the target domain, that is, the to-be-identified data. The objective function of joint distribution adaptation (JDA) is defined as:

[0077]

[0078] wherein, represents the calculation of the feature of the data set, represents the calculation of the feature of the source domain data, represents the calculation of the feature of the target domain data, c represents the data category, P(X) represents the calculation of the edge probability distribution of the data set X, P(X s ) represents the calculation of the edge probability distribution of the source domain data set, E(·) is the expectation of the data, represents the calculation of the expectation of the edge probability distribution of the feature of the source domain data, represents the expectation of computing the edge probability distribution of the feature of the target domain data, P(X|Y=c) represents the computation of the class conditional probability distribution, represents the expectation of computing the probability of the feature of the source domain data in the case of class c, represents the expectation of computing the probability of the feature of the target domain data in the case of class c.

[0079] Secondly, in order to overcome the influence of the poor final migration effect caused by the approximation of the conditional probability distribution by using the class conditional probability distribution in the calculation of the conditional distribution adaptation (CDA) in formula (1), an improved conditional probability distribution adaptation is proposed to adjust the conditional probability distribution in the two domains (source domain and target domain). By using the Bayes theorem, the calculation of the conditional probability distribution is expressed by the class conditional probability distribution, and the expression is as follows:

[0080]

[0081] wherein P(Y=c) represents the class prior distribution, which can be calculated by the following formula:

[0082]

[0083] wherein c represents the class label value, represents the size of the cth class in the source domain (labeled data), represents the size of the cth class in the target domain (to-be-identified data), and thus the following formula exists.

[0084]

[0085] The Bayes theorem is an important theorem in probability theory, which describes the probability of another related event occurring under the condition that a certain event has occurred. The theorem uses the relationship between prior probability and posterior probability to calculate the posterior probability by updating the prior probability. In machine learning and statistical inference, the Bayes theorem is widely used in parameter estimation, classification problems and model prediction.

[0086] In the embodiments of the present application, the working conditions corresponding to the data (labeled data) that have labels (true labels) are all regarded as the source domain, and the working conditions corresponding to the vibration data that do not have true labels are regarded as the target domain. Therefore, when the real-time acquired vibration data is the same as the working condition of the source domain data (i.e. the labeled data with labels) in the existing (original) database, this data can be used, and only the vibration data obtained under different working conditions is regarded as the target domain data (to-be-identified data).

[0087] After that, the improved conditional probability distribution adaptation using formula (2) can be expressed by the following formula:

[0088]

[0089] Finally, by improving the conditional probability distribution adaptation, the final improved joint distribution adaptation (IJDA) can be defined as:

[0090]

[0091] where IJDA represents the domain adaptation loss, s represents the labeled data, t represents the data to be identified, Y s represents the labeled data set including the label of the labeled data, Y t represents the pseudo-label data set including the pseudo-label of the data to be identified, represents the calculation of the feature of the data set, represents the first feature, represents the second feature, E(·) represents the expectation of the data, represents the expectation of calculating the edge probability distribution of the second feature, represents the expectation of calculating the edge probability distribution of the first feature, P(X|Y=c) represents the calculation of the class conditional probability distribution, represents the expectation of calculating the probability of the occurrence of the second feature under the condition that the class is c, represents the expectation of calculating the probability of the occurrence of the first feature under the condition that the class is c, P(X s represents the calculation of the edge probability distribution of the data set of the labeled data, P(Y s represents the class prior distribution of the label, P(Y t represents the class prior distribution of the pseudo-label, c represents the data class, X s represents the first data set including the data to be identified, X t represents the second data set including the labeled data.

[0092] The embodiments of the present application use pseudo-labels to replace real labels to calculate conditional probability distribution based on Bayes theorem to make up for the poor transfer effect caused by using approximate calculation in the principle of JDA, and propose an improved joint distribution adaptation method for a fault diagnosis method of a high-pressure common rail fuel injector under variable working conditions.

[0093] In a second aspect, the embodiments of the present application also propose a fault diagnosis device for a fuel injector, as shown in Figure 5 , comprising a data acquisition module 100 and a fault identification module 200;

[0094] The data acquisition module 100 is configured to acquire vibration data of the fuel injector to obtain the data to be identified.

[0095] The fault identification module 200 is used to use the data to be identified and the labeled data and to adaptively train the fault identification model to be trained based on the improved joint distribution to obtain a trained fault identification model; and input the data to be identified into the fault identification model to obtain a fault identification result.

[0096] The fault identification module includes: feature extraction unit, fault prediction unit, domain adaptation unit and training unit;

[0097] a feature extraction unit, configured to input the data to be identified and the labeled data into the feature extraction unit of the fault identification model to be trained, and obtain features of the data to be identified and the labeled data based on an attention mechanism, respectively, to obtain a first feature and a second feature, wherein the first feature is a feature of the data to be identified and the second feature is a feature of the labeled data;

[0098] a fault prediction unit, configured to input the first feature and the second feature into the fault prediction unit to obtain a first prediction result and a second prediction result, wherein the first prediction result is a prediction result obtained based on the first feature, and the second prediction result is a prediction result obtained based on the second feature;

[0099] a domain adaptation unit, configured to input the first feature, the second feature, the label, and the pseudo-label into the domain adaptation unit, and determine a domain adaptation loss based on the improved joint distribution adaptation, wherein the label is a label of the labeled data, and the pseudo-label is a first prediction result obtained by the fault prediction unit in a previous prediction of the first feature;

[0100] The training unit is used to update the model to be trained according to the domain adaptation loss and the classification loss to obtain a trained fault recognition model.

[0101] like Figure 6 As shown, the data acquisition module includes a vibration sensing unit and a data acquisition and transmission unit. The embodiment of the present application also includes a data partitioning module, which is used to perform working condition judgment on the vibration data and determine the data to be identified. The data partitioning module may include a host computer system. Specifically, the vibration sensing unit adopts a high-precision vibration sensor, which is fixedly installed at a preset position of the high-pressure common rail injector to collect the vibration data of the injector and convert the vibration signal into an electrical signal. Among them, there can be multiple preset positions, which can be key positions or positions that are easy to install. The data acquisition and transmission unit includes a multi-channel high-speed acquisition board for receiving the electrical signal of the vibration sensor and transmitting the vibration data to the data partitioning module through digital processing. The acquisition board has a high sampling rate and anti-interference ability, which can ensure the accuracy and real-time nature of the data.

[0102] The data partitioning module, such as the host computer system, includes an industrial computer equipped with a data partitioning algorithm. The industrial computer divides the vibration data into labeled data (source domain data) containing fault categories and data to be identified (target domain data) that does not contain fault categories according to different working conditions (such as different working pressures).

[0103] The labeled data includes a stored labeled data set. The labeled data input to the fault identification module may only use the labeled data of the stored labeled data set.

[0104] The fault identification module includes a feature extraction unit of a multi-attention convolutional network, which inputs the labeled data and the data to be identified into the constructed multi-attention convolutional network to extract the vibration data features as the first feature and the second feature. The domain adaptation unit improves the joint distribution adaptation based on the Bayesian principle to perform feature alignment and determine the domain adaptation loss; the fault prediction unit uses the cross entropy function to determine the classification loss; the training unit updates the iterative fault identification module according to the domain adaptation loss and the classification loss until the training round requirement is met. After the training is completed, only the feature extraction unit and the fault prediction unit can be retained in the fault identification module. The fault prediction unit in the fault identification module performs fault identification and classification and outputs the fault diagnosis result (fault identification result), thereby realizing the fault diagnosis of the high-pressure common rail injector under variable working conditions. Among them, in addition to being able to identify normal and faulty conditions, the implementation method of the present application can also identify specific fault types, such as wear, blockage, etc.

[0105] In a third aspect, according to an embodiment of the present application, a vehicle is further proposed, comprising an on-board terminal, the on-board terminal being used to execute a fault diagnosis method for a fuel injector as described in any one of the first aspects.

[0106] In the method of the present application, by training the fault recognition model to be trained with labeled data, data to be identified with different working conditions from the labeled data, and improved joint distribution adaptation, the fault recognition model to be trained can be trained for different working conditions, thereby improving the accuracy and robustness of fault identification of the high-pressure common rail injector under variable working conditions.

[0107] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A failure diagnosis method for an oil jet, characterized by, The method comprises: obtaining vibration data of an oil injector to obtain to-be-identified data, wherein the to-be-identified data is vibration data different from working conditions of labeled data; training a to-be-trained fault identification model based on improved joint distribution adaptation using the to-be-identified data and the labeled data to obtain a trained fault identification model; inputting the to-be-identified data into the fault identification model to obtain a fault identification result; the step of training the to-be-trained fault identification model based on improved joint distribution adaptation using the to-be-identified data and the labeled data to obtain a trained fault identification model comprises: inputting the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and obtaining features of the to-be-identified data and the labeled data based on an attention mechanism to obtain first features and second features, wherein the first features are features of the to-be-identified data, and the second features are features of the labeled data; inputting the first features and the second features into a fault prediction unit to obtain first prediction results and second prediction results, wherein the first prediction results are prediction results obtained according to the first features, and the second prediction results are prediction results obtained according to the second features; inputting the first features, the second features, labels, and pseudo labels into a domain adaptation unit, and determining a domain adaptation loss based on the improved joint distribution adaptation, wherein the labels are labels of the labeled data, and the pseudo labels are the first prediction results obtained by the fault prediction unit in the last time of predicting the first features; updating the to-be-trained model according to the domain adaptation loss and a classification loss to obtain a trained fault identification model.

2. The failure diagnosis method according to claim 1, characterized by, Before the step of updating the to-be-trained model according to the domain adaptation loss and the classification loss to obtain a trained fault identification model, the method further comprises: determining the classification loss according to the second prediction results and the labels.

3. The failure diagnosis method according to claim 1, characterized by, the step of inputting the first features, the second features, labels, and pseudo labels into a domain adaptation unit, and determining a domain adaptation loss based on the improved joint distribution adaptation comprises: determining a domain adaptation loss according to the first features, the second features, the labels, the pseudo labels, and the improved joint distribution adaptation, and the calculation method is as follows: wherein IJDA denotes the domain adaptation loss, s denotes the labeled data, t denotes the to-be-identified data, Y s denotes a label data set comprising the label of the labeled data, t denotes a pseudo-label data set comprising the pseudo-label of the to-be-identified data, denotes calculating a data set feature, denotes a first feature, denotes a second feature, E(·) denotes an expectation of data, denotes calculating an expectation of a second feature edge probability distribution, denotes calculating an expectation of a first feature edge probability distribution, denotes calculating an expectation of a probability of the second feature occurring in the case of a class being c, denotes calculating an expectation of a probability of the first feature occurring in the case of a class being c, P(X s ) denotes calculating an edge probability distribution of a data set of the labeled data, P(Y s =c) denotes a class prior distribution of the label, P(Y t =c) denotes a class prior distribution of the pseudo-label, c denotes a data class, X s denotes a first data set comprising the to-be-identified data, X t denotes a second data set comprising the labeled data.

4. The failure diagnosis method according to claim 1, characterized by, the step of updating the to-be-trained model according to the domain adaptation loss and the classification loss to obtain a trained fault identification model comprises: updating the feature extraction unit, the fault prediction unit, and the domain adaptation unit according to the domain adaptation loss and the classification loss; continuing to perform the steps of inputting the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and obtaining features of the to-be-identified data and the labeled data based on an attention mechanism to obtain first features and second features until a preset training threshold is reached, or the domain adaptation loss and / or the classification loss is less than or equal to a preset loss threshold, to obtain a trained fault identification model comprising the feature extraction unit and the fault prediction unit.

5. The failure diagnosis method according to claim 1, characterized by, The attention mechanism-based feature acquisition separately obtains features of the to-be-identified data and the labeled data, to obtain first features and second features, including: The convolution block attention module-based frequency feature extraction separately extracts multiple frequency features of the to-be-identified data and the labeled data, to obtain multiple first frequency features corresponding to the to-be-identified data and multiple second frequency features corresponding to the labeled data; The first frequency features are merged and then re-extracted to obtain first features, and the second frequency features are merged and then re-extracted to obtain second features.

6. The failure diagnosis method according to claim 1, characterized by, The vibration data of the fuel injector is obtained to obtain to-be-identified data, including: The vibration data of the fuel injector is obtained; The vibration data is subjected to working condition judgment to determine the to-be-identified data.

7. A failure diagnosing device for an oil jet, characterized by comprising: It includes: A data acquisition module and a fault identification module; The data acquisition module is configured to obtain vibration data of a fuel injector to obtain to-be-identified data; The fault identification module is configured to use the to-be-identified data and labeled data and based on an improved joint distribution adaptation to train a to-be-trained fault identification model to obtain a trained fault identification model; The to-be-identified data is input into the fault identification model to obtain a fault identification result; The fault identification module includes a feature extraction unit, a fault prediction unit, a domain adaptation unit, and a training unit; The feature extraction unit is configured to input the to-be-identified data and the labeled data into a feature extraction unit in the to-be-trained fault identification model, and based on an attention mechanism, features of the to-be-identified data and the labeled data are separately obtained to obtain first features and second features, wherein the first features are features of the to-be-identified data, and the second features are features of the labeled data; The fault prediction unit is configured to input the first features and the second features into a fault prediction unit to obtain first prediction results and second prediction results, wherein the first prediction results are prediction results obtained according to the first features, and the second prediction results are prediction results obtained according to the second features; The domain adaptation unit is configured to input the first features, the second features, labels, and pseudo labels into a domain adaptation unit, and determine a domain adaptation loss based on the improved joint distribution adaptation, wherein the labels are labels of the labeled data, and the pseudo labels are the first prediction results obtained by the fault prediction unit in the previous time of predicting the first features; The training unit is configured to update the to-be-trained model according to the domain adaptation loss and a classification loss to obtain the trained fault identification model.

8. A vehicle characterized by comprising: It includes a vehicle terminal, and the vehicle terminal is configured to execute a fault diagnosis method for a fuel injector according to any one of claims 1 to 6.

Images