Vehicle-mounted hydrogen system fault diagnosis method and device based on multi-vehicle data cooperation and small sample learning

Abstract

The invention discloses a vehicle-mounted hydrogen system fault diagnosis method and device based on multi-vehicle data cooperation and small sample learning, and belongs to the technical field of fault diagnosis. The method comprises the following steps: based on normal data of a normal vehicle-mounted hydrogen system and abnormal data of a fault vehicle-mounted hydrogen system, determining a target dimension strongly related to a fault in the vehicle-mounted hydrogen system; mixing the normal data and the abnormal data according to a set proportion to obtain a small sample data set; each sample in the small sample data set comprises sample data and a sample mark; the sample data is data corresponding to the target dimension, and the sample mark is a mark for judging whether a fault occurs or not; and training a neural network by using the small sample data set, and predicting data which is uploaded by the to-be-diagnosed vehicle and corresponds to the target dimension by using the trained neural network to obtain a fault diagnosis result. According to the method, the sensitivity degree of the neural network to fault features can be improved under the condition of small sample size, and the accuracy of fault diagnosis is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and in particular to a method and device for diagnosing faults of an on-board hydrogen system based on multi-vehicle data collaboration and small sample learning. Background Art

[0002] Hydrogen system failure is a failure mode unique to hydrogen fuel cell vehicles, including hydrogen system blockage, hydrogen leakage, etc. In related technologies, when the on-board hydrogen system fault diagnosis method relies on neural networks for fault prediction, the fault diagnosis accuracy is poor due to the small number of effective samples and the fact that abnormal features are easily masked by massive normal features.

[0003] Therefore, there is an urgent need to provide a new on-board hydrogen system fault diagnosis method to improve the accuracy of fault diagnosis. Summary of the invention

[0004] The present invention provides a method and device for diagnosing vehicle-mounted hydrogen system faults based on multi-vehicle data collaboration and small sample learning, which can solve the problem of poor fault diagnosis accuracy in related technologies. The technical solution is as follows:

[0005] On the one hand, a method for fault diagnosis of a vehicle-mounted hydrogen system based on multi-vehicle data collaboration and small sample learning is provided, the method comprising:

[0006] Based on the normal data of the normal onboard hydrogen system and the abnormal data of the faulty onboard hydrogen system, the target dimensions in the onboard hydrogen system that are strongly correlated with the fault are determined;

[0007] Normal data and abnormal data are mixed according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0008] The small sample data set is used to train a neural network, so that the trained neural network is used to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

[0009] On the other hand, a vehicle-mounted hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning is provided, the device comprising:

[0010] A determination unit, configured to determine a target dimension in the on-board hydrogen system that is strongly correlated with the fault based on normal data of the normal on-board hydrogen system and abnormal data of the faulty on-board hydrogen system;

[0011] An acquisition unit is used to mix normal data and abnormal data according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0012] The fault diagnosis unit is used to train a neural network using the small sample data set, so as to use the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

[0013] On the other hand, a computer device is provided, which includes a memory and a processor, the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the above-mentioned on-board hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning.

[0014] On the other hand, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned on-board hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning are implemented.

[0015] On the other hand, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for diagnosing faults of an on-board hydrogen system based on multi-vehicle data collaboration and small sample learning.

[0016] The technical solution provided by the present invention can at least bring the following beneficial effects:

[0017] The normal data and abnormal data in the multi-vehicle data are used to determine the target dimension that is strongly related to the fault in the vehicle hydrogen system, and then the normal data and abnormal data are mixed according to the set ratio to obtain a small sample data set. The sample data in the small sample data set is the data of the target dimension that is strongly related to the fault. In this way, the small sample data set is used to train the neural network, and a neural network that can accurately identify whether the hydrogen system is faulty can be trained without a large number of samples. It can be seen that this scheme uses the target dimension data that is strongly related to the fault of the vehicle hydrogen system to train the neural network, which can also improve the sensitivity of the neural network to the fault characteristics in the case of a small sample size, thereby improving the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of a vehicle-mounted hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning provided by an embodiment of the present invention;

[0020] Figure 2 It is a structural diagram of a vehicle-mounted hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning provided by an embodiment of the present invention;

[0021] Figure 3 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0023] Please refer to Figure 1 , an embodiment of the present invention provides a vehicle-mounted hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning, the method comprising:

[0024] Step 100, based on normal data of a normal onboard hydrogen system and abnormal data of a faulty onboard hydrogen system, determining a target dimension in the onboard hydrogen system that is strongly correlated with the fault;

[0025] Step 102, normal data and abnormal data are mixed according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0026] Step 104 , using the small sample data set to train a neural network, so as to use the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

[0027] In the embodiment of the present invention, normal data and abnormal data in multi-vehicle data are used to determine the target dimension that is strongly related to the fault in the vehicle hydrogen system, and then the normal data and abnormal data are mixed according to a set ratio to obtain a small sample data set. The sample data in the small sample data set is the data of the target dimension that is strongly related to the fault. In this way, the small sample data set is used to train the neural network, and a neural network that can accurately identify whether the hydrogen system is faulty can be trained without a large number of samples. It can be seen that this scheme uses the target dimension data that is strongly related to the fault of the vehicle hydrogen system to train the neural network, and can also improve the sensitivity of the neural network to the fault characteristics in the case of a small sample size, thereby improving the accuracy of fault diagnosis.

[0028] Described below Figure 1 How the various steps are performed.

[0029] First, for step 100, based on normal data of a normal onboard hydrogen system and abnormal data of a faulty onboard hydrogen system, a target dimension in the onboard hydrogen system that is strongly correlated with the fault is determined.

[0030] In the embodiment of the present invention, the normal data of a normal on-board hydrogen system and the abnormal data of a faulty on-board hydrogen system can be collected from historical data, or can be obtained by conducting hydrogen system experiments. If data is obtained by conducting hydrogen system experiments, normal data can be obtained through normal working experiments of the on-board hydrogen system, and typical fault experiment projects can be carried out for the on-board hydrogen system to obtain abnormal data. Among them, the hydrogen system fault experiment projects that can be carried out include: hydrogen system overtemperature experiment, hydrogen leakage experiment, and hydrogen system blockage experiment.

[0031] In one embodiment of the present invention, the target dimension strongly related to the fault in the vehicle hydrogen system can be determined through the following steps 1000-1006:

[0032] Step 1000: Determine multiple dimensions involved in the onboard hydrogen system;

[0033] In one embodiment of the present invention, the multiple dimensions involved in the on-board hydrogen system may include multiple of the following: hydrogen system output flow, hydrogen system output pressure, hydrogen concentration, hydrogen bottle temperature, fuel cell voltage, fuel cell current, fuel cell temperature, fuel cell power and power battery power.

[0034] When conducting hydrogen system experiments, data for each of the above dimensions can be recorded, thereby obtaining normal data and abnormal data for the corresponding dimensions.

[0035] After obtaining the data of each of the above dimensions, in order to further improve the accuracy of determining the target dimension and thus improve the accuracy of the fault diagnosis results, the data of the above dimensions are used to expand the dimensions. After expansion, the following multiple dimensions (21 dimensions) involved in the on-board hydrogen system are obtained: hydrogen system output flow, change in hydrogen system output flow per unit time, total hydrogen system output, hydrogen system output pressure, change in hydrogen system output pressure per unit time, hydrogen concentration, change in hydrogen concentration per unit time, hydrogen bottle temperature, change in hydrogen bottle temperature per unit time, fuel cell voltage, change in fuel cell voltage per unit time, fuel cell current, change in fuel cell current per unit time, fuel cell temperature, change in fuel cell temperature per unit time, fuel cell power, change in fuel cell power per unit time, total fuel cell output work, power battery power, change in power battery power per unit time and total power battery output work.

[0036] Step 1002: For each dimension, the following steps are performed: based on normal data of a normal vehicle hydrogen system and abnormal data of a faulty vehicle hydrogen system, a normal data set and an abnormal data set corresponding to the dimension are obtained; the normal data mean of the dimension is calculated using the normal data set; the characteristic deviation corresponding to the dimension is calculated using the abnormal data set and the normal data mean; and the characteristic risk priority value corresponding to the dimension is calculated using the characteristic deviation;

[0037] Normal data set X for a normal onboard hydrogen system norm and abnormal data set X of faulty onboard hydrogen system abnorm for:

[0038]

[0039]

[0040] in, is the normal data set of the i-th dimension, is the abnormal data set of the i-th dimension, M is the total number of target dimensions, and M is a positive integer, such as 21.

[0041] The mean of normal data in the i-th dimension for:

[0042]

[0043] Among them, L1 is the total amount of data in the normal data set of the i-th dimension, is the jth data in the normal data set corresponding to the i-th dimension.

[0044] The i-th dimension corresponds to the feature deviation D i for:

[0045]

[0046] Among them, L2 is the total amount of data in the abnormal data set of the i-th dimension, is the jth data in the abnormal data set corresponding to the i-th dimension.

[0047] The characteristic risk priority value RPN corresponding to the i-th dimension i for:

[0048] RPN i =S i ×O i ×D i

[0049] Among them, S i Represents the degree of influence of the feature of the i-th dimension on the fault, which is defined manually or evaluated by experts, and usually ranges from 1 to 10; i is the probability of abnormality of the feature of the i-th dimension;

[0050] Step 1004: Calculate the average of the characteristic risk priorities using the characteristic risk priority values ​​corresponding to each dimension, and use the average of the characteristic risk priorities as the risk threshold;

[0051] Risk threshold τ RPN for:

[0052]

[0053] Step 1006: Determine the dimension whose characteristic risk priority value is greater than the risk threshold as the target dimension that is strongly correlated with the fault.

[0054] If the characteristic risk priority value RPN of a certain dimension i >τ RPN , then the dimension is determined to be the target dimension that is strongly correlated with hydrogen system failure. Among them, the number of target dimensions is less than M.

[0055] Then, step 102 "mixing normal data and abnormal data according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to a target dimension, and the sample flag is a flag of whether a fault has occurred" and step 104 "training a neural network using the small sample data set, and using the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result" are explained at the same time.

[0056] In order to achieve accurate training of the neural network with a small sample data set, normal data and abnormal data can be mixed according to a set ratio, for example, the set ratio can be 1:100. Each sample in the sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag for whether a fault has occurred. Since the sample data is data corresponding to the target dimension, rather than data of all dimensions, the amount of data for network training can be reduced, and the sample data used is data of the target dimension that is strongly related to hydrogen system faults, thereby increasing the sensitivity of the neural network to fault characteristics.

[0057] In the embodiment of the present invention, the sample data is used as input and the sample mark is used as output to train the neural network, wherein the neural network may be a Transformer network.

[0058] After obtaining the trained neural network, the vehicle can upload data at a set frequency to input the data corresponding to the target dimension in the uploaded data into the neural network, and then output the fault diagnosis result.

[0059] In an embodiment of the present invention, the neural network is located at the server end, and the vehicle to be diagnosed can upload data at a set frequency so that the server end can perform fault diagnosis.

[0060] In one implementation, the vehicle to be diagnosed uploads data at different frequencies;

[0061] The multiple dimensions involved in the vehicle hydrogen system all have multiple fault detection thresholds corresponding to each other. When the vehicle to be diagnosed detects that the measured data of the vehicle hydrogen system exceeds the fault detection threshold of the corresponding dimension, the measured data is uploaded at a first frequency; otherwise, the measured data is uploaded at a second frequency; the first frequency is greater than the second frequency. The multiple dimensions are the dimensions before the expansion.

[0062] The fault detection threshold λ corresponding to each dimension can be calculated by the following formula:

[0063]

[0064] Among them, x norm is the mean of normal data under the corresponding dimension, x abnorm is the mean of abnormal data under the corresponding dimension.

[0065] Furthermore, in order to ensure the accuracy of the fault diagnosis results, the fault diagnosis results output by the neural network can be verified.

[0066] Specifically, the method of using the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed to obtain the fault diagnosis result may include the following steps A1-A6:

[0067] A1: Obtaining an initial fault diagnosis result output by the trained neural network for data corresponding to the target dimension uploaded by the vehicle to be diagnosed;

[0068] A2: The data to be diagnosed uploaded by the vehicle to be diagnosed corresponding to the target dimension is added as a sample to a pre-generated outlier data set; the outlier data set is generated by mixing normal data and abnormal data in a set ratio in advance; each sample in the outlier data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0069] The set ratio may also be 1: 100. It should be noted that the outlier data set may directly use the small sample data set, but considering that the sample data in the small sample data set may be continuous, while the sample data in the outlier data set needs to calculate the outlier degree, the sample data in the outlier data set may be the mean of the data segment, thereby ensuring the accuracy of the fault detection result when the outlier degree is used to determine the fault detection result.

[0070] A3: For each sample in the outlier data set, the outlier degree is calculated by taking K nearest neighbors; K is a positive integer;

[0071]

[0072] Among them, OD i is the outlier degree of the i-th data in the outlier data set, K represents the number of neighboring points, d() represents the Euclidean distance calculation, and x i represents the point whose outlier degree is calculated, x j Represents x i Neighboring points around the outlier, N K (x i ) represents x i The set of neighboring points of .

[0073] A4: Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance;

[0074] Outlier mean μ OD for:

[0075]

[0076] The outlier variance σ is:

[0077]

[0078] N is the total number of samples in the outlier data set;

[0079] Outlier threshold OD thresh For: OD thresh =μ OD +α·σ

[0080] Among them, α is the adjustment coefficient.

[0081] A5: Determine the outlier diagnosis result according to the outlier degree of the data to be diagnosed and the outlier degree threshold;

[0082] If the outlier degree of the data to be diagnosed is greater than the outlier degree threshold, the outlier diagnosis result is a hydrogen system failure;

[0083] If the outlier degree of the data to be diagnosed is not greater than the outlier degree threshold, the outlier diagnosis result is that the hydrogen system is normal.

[0084] A6: Determine the fault diagnosis result according to the initial fault diagnosis result and the outlier diagnosis result.

[0085] In one implementation, if the initial fault diagnosis result and the outlier diagnosis result are the same, the same result is used as the fault diagnosis result; otherwise, the fault diagnosis result is re-determined.

[0086] Among them, re-determining the fault diagnosis result can be achieved by continuing the diagnosis through subsequent data.

[0087] In the embodiment of the present invention, in order to improve the accuracy of the outlier diagnosis result, the value of K is predetermined in the following manner:

[0088] S1: Determine the initial value of K to execute S2;

[0089] S2: Calculate the outlier degree of each sample in the outlier data set by taking K nearest neighbor points;

[0090] S3: Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance;

[0091] S4: Perform fault diagnosis on each sample in the outlier data set using the outlier degree threshold, and compare the fault diagnosis result with the sample flag to determine whether the fault diagnosis accuracy reaches the set value based on the comparison result; if not, readjust the value of K, and continue to execute S2 using the adjusted value of K; if so, determine the current value of K as the final value of K.

[0092] In addition, in order to improve the accuracy of the outlier diagnosis result, the adjustment coefficient may be adjusted according to the value adjustment method of K, so as to determine the value of the adjustment coefficient.

[0093] In this way, the embodiment of the present invention uses target dimension data that is strongly correlated with the on-board hydrogen system fault to train the neural network, which can improve the sensitivity of the neural network to fault characteristics even in the case of a small sample size, thereby improving the accuracy of fault diagnosis.

[0094] Please refer to Figure 2 The embodiment of the present invention provides a vehicle-mounted hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning, the device comprising:

[0095] A determination unit 200, configured to determine a target dimension in the onboard hydrogen system that is strongly correlated with the fault based on normal data of the normal onboard hydrogen system and abnormal data of the faulty onboard hydrogen system;

[0096] The acquisition unit 202 is used to mix the normal data and the abnormal data according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0097] The fault diagnosis unit 204 is used to train a neural network using the small sample data set, so as to use the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

[0098] In one embodiment of the present invention, the determining unit is specifically configured to perform the following operations:

[0099] Identify the multiple dimensions involved in onboard hydrogen systems;

[0100] For each dimension, the following are performed: based on the normal data of the normal vehicle hydrogen system and the abnormal data of the faulty vehicle hydrogen system, a normal data set and an abnormal data set corresponding to the dimension are obtained; the normal data mean of the dimension is calculated using the normal data set; the characteristic deviation corresponding to the dimension is calculated using the abnormal data set and the normal data mean; the characteristic risk priority value corresponding to the dimension is calculated using the characteristic deviation;

[0101] Using the characteristic risk priority value corresponding to each dimension, calculate the average characteristic risk priority, and use the average characteristic risk priority as the risk threshold;

[0102] The dimension whose characteristic risk priority value is greater than the risk threshold is determined as the target dimension that is strongly correlated with the fault.

[0103] In one embodiment of the present invention, the vehicle to be diagnosed uploads data at different frequencies;

[0104] The multiple dimensions involved in the on-board hydrogen system all have multiple fault detection thresholds corresponding to each other. When the vehicle to be diagnosed detects that the measured data of the on-board hydrogen system exceeds the fault detection threshold of the corresponding dimension, the measured data is uploaded at a first frequency; otherwise, the measured data is uploaded at a second frequency; the first frequency is greater than the second frequency.

[0105] In one embodiment of the present invention, the fault diagnosis unit is specifically configured to perform the following operations:

[0106] Obtaining an initial fault diagnosis result output by the trained neural network for data corresponding to the target dimension uploaded by the vehicle to be diagnosed;

[0107] The data to be diagnosed uploaded by the vehicle to be diagnosed and corresponding to the target dimension are added as samples to a pre-generated outlier data set; the outlier data set is generated by pre-mixing normal data and abnormal data according to a set ratio; each sample in the outlier data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs;

[0108] For each sample in the outlier data set, the outlier degree is calculated by taking K nearest neighbor points; K is a positive integer;

[0109] Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance;

[0110] Determine the outlier diagnosis result according to the outlier degree of the data to be diagnosed and the outlier degree threshold;

[0111] The fault diagnosis result is determined according to the initial fault diagnosis result and the outlier diagnosis result.

[0112] In one embodiment of the present invention, the value of K is predetermined in the following manner:

[0113] S1: Determine the initial value of K to execute S2;

[0114] S2: Calculate the outlier degree of each sample in the outlier data set by taking K nearest neighbor points;

[0115] S3: Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance;

[0116] S4: Perform fault diagnosis on each sample in the outlier data set using the outlier degree threshold, and compare the fault diagnosis result with the sample flag to determine whether the fault diagnosis accuracy reaches the set value based on the comparison result; if not, readjust the value of K, and continue to execute S2 using the adjusted value of K; if so, determine the current value of K as the final value of K.

[0117] In one embodiment of the present invention, when the fault diagnosis unit determines the fault diagnosis result based on the initial fault diagnosis result and the outlier diagnosis result, it specifically includes: if the initial fault diagnosis result and the outlier diagnosis result are the same, then the same result is used as the fault diagnosis result; otherwise, the fault diagnosis result is re-determined.

[0118] It should be noted that the on-board hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the on-board hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning provided in the above embodiment and the on-board hydrogen system fault diagnosis method embodiment based on multi-vehicle data collaboration and small sample learning belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0119] The embodiment of the present application also provides a computer device, please refer to Figure 3 The computer device includes a processor and a memory, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the on-board hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning provided by the above-mentioned method embodiments.

[0120] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the on-board hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning provided by the above-mentioned method embodiments.

[0121] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes any of the on-board hydrogen system fault diagnosis methods based on multi-vehicle data collaboration and small sample learning described in the above embodiments.

[0122] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0123] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0124] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0125] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A vehicle-mounted hydrogen system fault diagnosis method based on multi-vehicle data collaboration and small sample learning, characterized in that: The method comprises: Based on the normal data of the normal onboard hydrogen system and the abnormal data of the faulty onboard hydrogen system, the target dimensions in the onboard hydrogen system that are strongly correlated with the fault are determined; Normal data and abnormal data are mixed according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs; The small sample data set is used to train a neural network, so that the trained neural network is used to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

2. The method according to claim 1, characterized in that The determining of target dimensions strongly related to the fault in the on-board hydrogen system based on the normal data of the normal on-board hydrogen system and the abnormal data of the faulty on-board hydrogen system includes: Identify the multiple dimensions involved in onboard hydrogen systems; For each dimension, the following are performed: based on the normal data of the normal vehicle hydrogen system and the abnormal data of the faulty vehicle hydrogen system, a normal data set and an abnormal data set corresponding to the dimension are obtained; the normal data mean of the dimension is calculated using the normal data set; the characteristic deviation corresponding to the dimension is calculated using the abnormal data set and the normal data mean; the characteristic risk priority value corresponding to the dimension is calculated using the characteristic deviation; Using the characteristic risk priority value corresponding to each dimension, calculate the average characteristic risk priority, and use the average characteristic risk priority as the risk threshold; The dimension whose characteristic risk priority value is greater than the risk threshold is determined as the target dimension that is strongly correlated with the fault.

3. The method according to claim 1, characterized in that The vehicles to be diagnosed upload data at different frequencies; The multiple dimensions involved in the on-board hydrogen system all have multiple fault detection thresholds corresponding to each other. When the vehicle to be diagnosed detects that the measured data of the on-board hydrogen system exceeds the fault detection threshold of the corresponding dimension, the measured data is uploaded at a first frequency; otherwise, the measured data is uploaded at a second frequency; The first frequency is greater than the second frequency.

4. The method according to any one of claims 1 to 3, characterized in that: The trained neural network is used to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed to obtain a fault diagnosis result, including: Obtaining an initial fault diagnosis result output by the trained neural network for data corresponding to the target dimension uploaded by the vehicle to be diagnosed; The data to be diagnosed uploaded by the vehicle to be diagnosed and corresponding to the target dimension are added as samples to a pre-generated outlier data set; the outlier data set is generated by pre-mixing normal data and abnormal data according to a set ratio; each sample in the outlier data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs; For each sample in the outlier data set, the outlier degree is calculated by taking K nearest neighbor points; K is a positive integer; Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance; Determine the outlier diagnosis result according to the outlier degree of the data to be diagnosed and the outlier degree threshold; The fault diagnosis result is determined according to the initial fault diagnosis result and the outlier diagnosis result.

5. The method according to claim 4, characterized in that The value of K is predetermined as follows: S1: Determine the initial value of K to execute S2; S2: Calculate the outlier degree of each sample in the outlier data set by taking K nearest neighbor points; S3: Calculate the outlier degree mean and outlier degree variance of multiple samples in the outlier data set, and calculate the outlier degree threshold according to the outlier degree mean and the outlier degree variance; S4: Perform fault diagnosis on each sample in the outlier data set using the outlier degree threshold, and compare the fault diagnosis result with the sample flag to determine whether the fault diagnosis accuracy reaches the set value based on the comparison result; if not, readjust the value of K, and continue to execute S2 using the adjusted value of K; if so, determine the current value of K as the final value of K.

6. The method according to claim 4, characterized in that Determining the fault diagnosis result according to the initial fault diagnosis result and the outlier diagnosis result includes: If the initial fault diagnosis result and the outlier diagnosis result are the same, the same result is used as the fault diagnosis result; otherwise, the fault diagnosis result is re-determined.

7. A vehicle-mounted hydrogen system fault diagnosis device based on multi-vehicle data collaboration and small sample learning, characterized in that: The device comprises: A determination unit, configured to determine a target dimension in the on-board hydrogen system that is strongly correlated with the fault based on normal data of the normal on-board hydrogen system and abnormal data of the faulty on-board hydrogen system; An acquisition unit is used to mix normal data and abnormal data according to a set ratio to obtain a small sample data set; each sample in the small sample data set includes sample data and a sample flag; the sample data is data corresponding to the target dimension, and the sample flag is a flag indicating whether a fault occurs; The fault diagnosis unit is used to train a neural network using the small sample data set, so as to use the trained neural network to predict the data corresponding to the target dimension uploaded by the vehicle to be diagnosed, so as to obtain a fault diagnosis result.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-6.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The method comprises a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images