Vehicle fault diagnosis method and system based on online transfer learning

Abstract

The invention discloses a vehicle fault diagnosis method and system based on online transfer learning. The method comprises the following steps: obtaining a source domain classifier; deploying the obtained source domain classifier in a cloud server, and initializing a target domain classifier and weights of the source domain classifier and the target domain classifier; acquiring operation data of a to-be-diagnosed vehicle; obtaining target domain data; a fault classification prediction result is obtained; updating the weights of the source domain classifier and the target domain classifier; the target domain classifier is updated; obtaining a final target domain classifier; and performing fault diagnosis by using the final target domain classifier. Through iterative updating of the online transfer learning architecture and the target domain classifier, the scheme of the invention can continuously adjust and optimize the fault diagnosis model according to the new target domain data sample. Therefore, the model can adapt to changes of different vehicles and fault conditions, and the self-adaptability is improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle fault diagnosis, and in particular to a vehicle fault diagnosis method and system based on online transfer learning. Background Art

[0002] With the rapid development of the economy and society, urban roads are becoming more and more complex, and vehicles are being used more and more widely. Traditional vehicle fault diagnosis methods can no longer meet users' requirements for real-time and adaptable fault diagnosis.

[0003] Traditional vehicle fault diagnosis methods include rule-based methods and machine learning-based methods, which have the following disadvantages: (1) Rule-based fault diagnosis methods need to work on the basis of a large number of rules and logics defined in advance. The writing of these rules and logics is a difficult and time-consuming task for complex vehicle systems. Because vehicle systems involve multiple components and subsystems, each fault may have different symptoms and impacts, so a large number of rules and logics need to be defined to cover various possible fault conditions. This requires not only professional domain knowledge and experience, but also a lot of time and effort to write and debug these rules and logics.

[0004] (2) Machine learning-based fault diagnosis methods require a large amount of labeled data to train the model. This data needs to contain the inputs and corresponding correct outputs under various fault conditions. However, obtaining sufficient labeled data is often a challenging task. First, collecting and labeling a large amount of fault data requires a lot of time and human resources. Second, for some rare fault types, it may be difficult to obtain enough data to train the model. In addition, due to the complexity and diversity of vehicle systems, traditional machine learning methods may not be able to adapt well to new fault types and scenarios. Therefore, in order to maintain the accuracy and adaptability of the model, regular offline training and updating are required, which also increases the complexity and cost of system maintenance.

[0005] In view of the above, it is necessary to improve the traditional vehicle fault diagnosis scheme. Summary of the invention

[0006] The purpose of the present invention is to provide a vehicle fault diagnosis method and system based on online transfer learning, so that the idea of ​​online transfer learning can be utilized to update and adapt the fault diagnosis model in real time in the cloud.

[0007] To achieve the purpose of the present invention, the technical solution provided by the present invention is as follows: First aspect The present invention provides a vehicle fault diagnosis method based on online transfer learning, comprising the following steps: Step S1: training existing vehicle fault samples to obtain a source domain classifier; Step S2: deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; Step S3: collecting the operating data of the vehicle to be diagnosed; Step S4: After preprocessing the running data of the vehicle to be diagnosed, a plurality of target domain data samples are obtained in sequence, and each target domain data sample is uploaded to the cloud server after being obtained; Step S5: input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain fault classification prediction results; Step S6: updating the weights of the source domain classifier and the target domain classifier; Step S7: Update the target domain classifier by constructing a loss function; Step S8: After the target domain classifier is updated according to the last target domain sample, a final target domain classifier is obtained; Step S9: Use the final target domain classifier to perform fault diagnosis.

[0008] Second aspect The present invention provides a vehicle fault diagnosis system based on online transfer learning, comprising the following units: a source domain classifier construction unit, a cloud deployment and initialization unit, a collection unit, a target domain data sample acquisition unit, a fault classification prediction unit, a weight updating unit, a classifier updating unit, a final target domain classifier acquisition unit, and a diagnosis unit; The source domain classifier construction unit is used to train the existing vehicle fault samples to obtain a source domain classifier; The cloud deployment and initialization unit is used to deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; The collecting unit is used to collect the operating data of the vehicle to be diagnosed; The target domain data sample acquisition unit is used to pre-process the operation data of the vehicle to be diagnosed, and sequentially obtain multiple target domain data samples, and upload each target domain data sample to the cloud server after obtaining it; The fault classification prediction unit is used to input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain a fault classification prediction result; The weight updating unit is used to update the weights of the source domain classifier and the target domain classifier; The classifier updating unit is used to update the target domain classifier by constructing a loss function; The final target domain classifier obtaining unit is used to obtain the final target domain classifier after updating the target domain classifier according to the last target domain sample; The diagnosis unit is used to perform fault diagnosis using the final target domain classifier.

[0009] Compared with the prior art, the present invention has the following beneficial effects: (1) By building an online transfer learning architecture in the cloud and initializing the target domain classifier and related weight parameters, the solution of the present invention has the ability to process large-scale data and multiple vehicles and has good scalability.

[0010] (2) Through the online transfer learning solution, we can make full use of existing data and models, accelerate the training of fault diagnosis models, improve the speed and accuracy of fault diagnosis, and reduce false positives and negative negatives.

[0011] (3) Through the online transfer learning architecture and iterative updates of the target domain classifier, the solution of the present invention can continuously adjust and optimize the fault diagnosis model according to new target domain data samples. This enables the model to adapt to changes in different vehicles and fault conditions, improving its adaptability.

[0012] (4) By using existing fault samples and cloud servers for online transfer learning, the cost of re-collecting and labeling data is avoided. This saves time and resources and builds a reliable fault diagnosis system more quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings 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.

[0015] It should be noted that the present invention proposes a vehicle fault diagnosis solution based on online transfer learning. By utilizing the idea of ​​online transfer learning, the present invention can update and adapt the fault diagnosis model in real time in the cloud, thereby overcoming the limitations of traditional methods. Compared with traditional methods, the method of the present invention does not require a large number of rules and logics to be defined in advance, nor does it require a large amount of labeled data to train the model. It applies previously learned knowledge to new fault diagnosis tasks through online learning and iterative updates, and continuously updates for each new sample, thereby improving the accuracy, real-time and scalability of the model. Therefore, the method of the present invention has higher adaptability and generalization capabilities, and can better adapt to new fault types and scenarios.

[0016] like Figure 1 As shown, an embodiment of the present invention provides a vehicle fault diagnosis method based on online transfer learning, comprising the following steps: Step S1: training existing vehicle fault samples to obtain a source domain classifier; In step S1, the source domain classifier is used Indicates that the source domain classifier The type of is support vector machine classifier, Bayesian classifier, or deep learning based classifier.

[0017] It should be noted that the existing vehicle fault samples are used as source domain data samples. The existing vehicle fault samples can be measured by noise sensors arranged at the inner and outer ears of the driver and the inner and outer ears of the right rear passenger.

[0018] Step S2: deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; It should be noted that the source domain classifier is not only used for online transfer learning, but can also directly participate in the diagnosis of existing vehicle fault samples.

[0019] In step S2, the target domain classifier , initialized to ,in, Indicates The target domain classifier when target domain samples arrive, ∈[1, t ]; and The weights are: , initialized to ,in and For the When target domain samples arrive, and The weight of .

[0020] Step S3: collecting the operating data of the vehicle to be diagnosed; It should be noted that by arranging relevant sensors at relevant test points of the vehicle to be diagnosed, operating data is collected when the vehicle is running; the sensor types include acceleration sensors, noise sensors, etc., and the collection points are flexibly selected according to the target fault, including the cab, front and rear axles, etc.

[0021] Step S4: After preprocessing the running data of the vehicle to be diagnosed, a plurality of target domain data samples are obtained in sequence, and each target domain data sample is uploaded to the cloud server after being obtained.

[0022] It should be noted that each time a target domain data sample is collected and processed, the target domain data sample is uploaded.

[0023] It should be noted that the preprocessing includes data cleaning, data trimming, data labeling, etc., all of which are performed using well-known technologies and will not be described in detail here.

[0024] Step S5: input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain fault classification prediction results; Among them, in step S5, the target domain sample is ,in Expressed as The target domain samples that arrive, ∈[1, t ]; for , classified by the following formula: ; in, for The fault classification prediction results are is the prediction function, T Represents a matrix transpose operation 。

[0025] Step S6: updating the weights of the source domain classifier and the target domain classifier; Among them, according to The fault classification prediction results of target domain samples are calculated by the following formula: and Weight and To update, get the updated weight and : ; in, , They are and The evaluation function of .

[0026] Step S7: Update the target domain classifier by constructing a loss function; In step S7, the loss function is ;if , then the target domain classifier Updated to ,in , C is the initial trade-off coefficient, For the The target domain samples that have arrived Fault classification label; if , then the target domain classifier .

[0027] Step S8: After the target domain classifier is updated according to the last target domain sample, a final target domain classifier is obtained; Step S9: Use the final target domain classifier to perform fault diagnosis.

[0028] It should be noted that both the existing vehicle fault samples and the target domain samples include data ontology and fault classification labels.

[0029] Accordingly, the present embodiment provides a vehicle fault diagnosis system based on online transfer learning, comprising the following units: a source domain classifier construction unit, a cloud deployment and initialization unit, a collection unit, a target domain data sample acquisition unit, a fault classification prediction unit, a weight updating unit, a classifier updating unit, a final target domain classifier acquisition unit, and a diagnosis unit; The source domain classifier construction unit is used to train the existing vehicle fault samples to obtain a source domain classifier; The cloud deployment and initialization unit is used to deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; The collecting unit is used to collect the operating data of the vehicle to be diagnosed; The target domain data sample acquisition unit is used to pre-process the operation data of the vehicle to be diagnosed, and sequentially obtain multiple target domain data samples, and upload each target domain data sample to the cloud server after obtaining it; The fault classification prediction unit is used to input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain a fault classification prediction result; The weight updating unit is used to update the weights of the source domain classifier and the target domain classifier; The classifier updating unit is used to update the target domain classifier by constructing a loss function; The final target domain classifier obtaining unit is used to obtain the final target domain classifier after updating the target domain classifier according to the last target domain sample; The diagnosis unit is used to perform fault diagnosis using the final target domain classifier.

[0030] Among them, in the source domain classifier construction unit, the source domain classifier is used Indicates that the source domain classifier The type of is support vector machine classifier, Bayesian classifier, or deep learning based classifier.

[0031] Among them, in the cloud deployment and initialization unit, the type of the target domain classifier is a SVM classifier, a Bayesian classifier or a classifier based on deep learning.

[0032] It should be noted that the solution of the present invention is applicable to various vehicle-related fields. For example, automobile manufacturers can apply the method to fault diagnosis of new models to improve vehicle quality and reliability. Maintenance service providers can use the method to quickly and accurately diagnose and repair vehicle faults and improve service efficiency. Vehicle operating companies can use the method to monitor the operating status of vehicles, promptly discover and solve potential fault problems, and ensure the safety and reliability of vehicles.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate and explain the present invention, and are not intended to limit the present invention to the scope of the described embodiments. In addition, those skilled in the art can understand that the present invention is not limited to the above embodiments, and more variations and modifications can be made according to the teachings of the present invention, and these variations and modifications all fall within the scope of the protection claimed by the present invention.

Claims

1. A vehicle fault diagnosis method based on online transfer learning, It is characterized in that The steps include: Step S1: training existing vehicle fault samples to obtain a source domain classifier; Step S2: deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; Step S3: collecting the operating data of the vehicle to be diagnosed; Step S4: After preprocessing the running data of the vehicle to be diagnosed, a plurality of target domain data samples are obtained in sequence, and each target domain data sample is uploaded to the cloud server after being obtained; Step S5: input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain fault classification prediction results; Step S6: updating the weights of the source domain classifier and the target domain classifier; Step S7: Update the target domain classifier by constructing a loss function; Step S8: After the target domain classifier is updated according to the last target domain sample, a final target domain classifier is obtained; Step S9: Use the final target domain classifier to perform fault diagnosis.

2. A vehicle fault diagnosis method based on online transfer learning according to claim 1, It is characterized in that In step S1, the source domain classifier is used Indicates that the source domain classifier The type of is support vector machine classifier, Bayesian classifier, or deep learning based classifier.

3. A vehicle fault diagnosis method based on online transfer learning according to claim 2, It is characterized in that In step S2, the type of the target domain classifier is a SVM classifier, a Bayesian classifier, or a classifier based on deep learning.

4. A vehicle fault diagnosis method based on online transfer learning according to claim 3, It is characterized in that In step S2, the target domain classifier , initialized to ,in, Indicates The target domain classifier when target domain samples arrive, ∈[1, t ]; and The weights are: , initialized to ,in and For the When target domain samples arrive, and The weight of .

5. A vehicle fault diagnosis method based on online transfer learning according to claim 4, It is characterized in that In step S5, the target domain sample is ,in Expressed as The target domain samples that arrive, ∈[1, t ]; for , classified by the following formula: ； in, for The fault classification prediction results are is the prediction function, T Represents a matrix transpose operation 。 6. A vehicle fault diagnosis method based on online transfer learning according to claim 5, It is characterized in that According to The fault classification prediction results of target domain samples are calculated by the following formula: and Weight and To update, get the updated weight and : ； in, , They are and The evaluation function of .

7. A vehicle fault diagnosis method based on online transfer learning according to claim 6, It is characterized in that In step S7, the loss function is ;if , then the target domain classifier Updated to ,in , C is the initial trade-off coefficient, For the The target domain samples that have arrived Fault classification label; if , then the target domain classifier .

8. A vehicle fault diagnosis system based on online transfer learning, It is characterized in that It includes the following units: source domain classifier construction unit, cloud deployment and initialization unit, acquisition unit, target domain data sample acquisition unit, fault classification prediction unit, weight update unit, classifier update unit, final target domain classifier acquisition unit and diagnosis unit; The source domain classifier construction unit is used to train the existing vehicle fault samples to obtain a source domain classifier; The cloud deployment and initialization unit is used to deploy the obtained source domain classifier on the cloud server, initialize the target domain classifier and the weights of the source domain classifier and the target domain classifier; The collecting unit is used to collect the operating data of the vehicle to be diagnosed; The target domain data sample acquisition unit is used to pre-process the operation data of the vehicle to be diagnosed, and sequentially obtain multiple target domain data samples, and upload each target domain data sample to the cloud server after obtaining it; The fault classification prediction unit is used to input the received target domain data samples into the source domain classifier and the target domain classifier for classification to obtain a fault classification prediction result; The weight updating unit is used to update the weights of the source domain classifier and the target domain classifier; The classifier updating unit is used to update the target domain classifier by constructing a loss function; The final target domain classifier obtaining unit is used to obtain the final target domain classifier after updating the target domain classifier according to the last target domain sample; The diagnosis unit is used to perform fault diagnosis using the final target domain classifier.

9. A vehicle fault diagnosis system based on online transfer learning according to claim 8, It is characterized in that In the source domain classifier construction unit, the source domain classifier is used Indicates that the source domain classifier The type of is support vector machine classifier, Bayesian classifier, or deep learning based classifier.

10. A vehicle fault diagnosis system based on online transfer learning according to claim 9, It is characterized in that In the cloud deployment and initialization unit, the type of the target domain classifier is a SVM classifier, a Bayesian classifier, or a classifier based on deep learning.

Images