CN112733887A - Method for detecting fault of hub motor of electric vehicle driven by supervision data - Google Patents
Method for detecting fault of hub motor of electric vehicle driven by supervision data Download PDFInfo
- Publication number
- CN112733887A CN112733887A CN202011547614.3A CN202011547614A CN112733887A CN 112733887 A CN112733887 A CN 112733887A CN 202011547614 A CN202011547614 A CN 202011547614A CN 112733887 A CN112733887 A CN 112733887A
- Authority
- CN
- China
- Prior art keywords
- fault
- data set
- model
- data
- motor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Electric Propulsion And Braking For Vehicles (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a method for detecting the fault of an electric automobile hub motor driven by supervision data, which mainly comprises the following steps: 1. and (3) data set reorganization: collecting fault data, adding a motor fault label, and establishing a composite cross data set; 2. training and validation of the CDAE-SVC model: minimizing a loss function containing potential noise penalty terms and Jacobian normal mode penalty terms, optimizing CDAE model parameters, and cross-verifying an SVC classifier; 3. testing the faults of the hub motor: collecting various index data of the motor when the vehicle operates, inputting the data into a trained model, calculating posterior classification probability, judging whether the hub motor has a fault, and determining the fault category. The invention provides a supervised data driving method aiming at the technical problem that fault detection cannot be carried out by establishing an accurate vehicle model, and the supervised data driving method has the characteristics of rapidness, accuracy, strong robustness and the like.
Description
Technical Field
The invention relates to the field of fault detection of four-wheel independent drive hub motors of electric automobiles, in particular to a supervised fault detection technology based on a data drive method.
Background
The electric and intelligent automobile is the main trend of future automobile development, and in order to meet the requirements of an intelligent automobile wire-control chassis system of L3 and above on response speed, control potential and integration, the hub motor technology as one of automobile actuators is valued by extensive researchers due to excellent performance. The wheel hub motor technology is a drive-by-wire technology of new energy automobile, traditional mechanical structure power transmission chain has been eliminated, spread with the form of signal of telecommunication, response speed to control command is fast, secondly it has every wheel individual control, can realize the special advantage of independent moment distribution, the vehicle is indulged transverse control potentiality and is big, wheel hub motor technology installs the motor inside wheel hub in addition, multiple functions such as collection drive, the transmission, braking are as an organic whole, mechanical coupling connection has been reduced, the expanded space has been saved greatly, the requirement of intelligent car to the lightweight has been satisfied.
The running working conditions of the automobile are complex and changeable, the running environment of the wheels is severe and unusual, the hub motor is inevitably influenced by the external environment due to the specific structure, when one or more wheel motors fail, the motor cannot output expected torque (Roxie. distributed electrically driven vehicle system/driving force coordination and active fault-tolerant control [ D ]. Qinghua university, 2014.) so as to directly cause the reduction of the dynamic performance of the automobile and even the unbalance of the system, destroy the operation stability and the running safety of the automobile, seriously cause traffic accidents and not meet the requirement of the intelligent automobile on the safety. Therefore, the introduction of the in-wheel motor in the intelligent vehicle linear control Technology not only brings the advantages of the transverse and longitudinal control of the vehicle, but also makes the Fault detection of the in-wheel motor an important research field due to the characteristics of the independent driving of the wheels (Jinseong Park, Park Youngjin. optimal Input Design for Fault Identification of activated Electric groups [ J ]. IEEE Transactions on vehicle Technology,2016,65(4): 1912-.
Most of the traditional fault diagnosis methods are model-based fault diagnosis methods, and fault information can be clearly understood and completely separated on the basis of establishing an accurate model. However, the vehicle has very strong nonlinearity, the vehicle dynamics model, the tire model and the like have very strong nonlinear expressions on the input of a steering wheel of a driver and the input of an accelerator pedal and a brake pedal, most vehicle models (such as a complete vehicle seven-degree-of-freedom dynamics model) used for research cannot consider unmodeled dynamics (Liu Pan, research on fault-tolerant control strategies based on failure of intelligent vehicle brake actuators [ D ]. Hunan university, 2018.), in addition, the vehicle running environment is complex, the parameters of the vehicle and the parameters of a road surface are changeable, and the simplified vehicle equivalent model cannot completely and accurately describe the vehicle running characteristics, so the effect of the hub motor fault detection method based on the model developed on the basis is unsatisfactory, and the rapidity, the accuracy and the robustness of fault identification cannot be guaranteed.
Disclosure of Invention
The invention provides a supervised monitoring method based on a shrinkage noise reduction auto encoder (CDAE) and a Support Vector machine (SVC), aiming at the problem that fault detection cannot be carried out by establishing an accurate vehicle model, and the supervised monitoring method is applied to nonlinear fault detection of a hub motor technology.
The purpose of the invention is realized by the following technical scheme: a method for detecting the fault of an electric automobile hub motor driven by supervision data comprises the following steps:
(1) reconstructing a data set containing wheel hub motor fault information: acquiring input and output variable information with motor fault attributes, performing fault classification on training sets of different fault conditions, adding a fault classification label quantity to each sample, and establishing a composite working condition data set;
(2) training the CDAE-SVC model: forming a cross data set by using the data set formed in the step (1), minimizing a loss function containing potential noise penalty terms and Jacobian normal penalty terms by using the data set, optimizing CDAE model parameters, and cross-training an SVC classifier;
(3) testing the hub motor fault process: collecting various index data of the motor when the vehicle operates, processing a data set, inputting hidden layer variables into the trained CDAE-SVC model in the step (2), calculating posterior classification probability, judging whether the hub motor has a fault or not, and determining the fault category.
Further, the step (1) includes the sub-steps of:
(1.1) firstly defining a failure factor delta for describing the execution capacity of the actuator before and after failure, and then equivalently converting the failure factor into a reference standard capable of describing whether failure occurs or not according to the failure type of a specific hub motor;
(1.2) dividing the failure types of the automobile hub motors driven by four wheels independently into 6 modes according to the failure positions and quantity distribution of the hub motors, wherein the failure types of a single motor, two motors on the same side, two motors on different sides, three motors and four motors correspond to label quantities of 1, 2, 3, 4, 5 and 6 respectively;
and (1.3) acquiring input and output data of the vehicle during running under different failure types defined in the step (1.2), such as information of driving torque, braking torque, output torques of four hub motors and the like under a certain failure to form a data set with label quantity, combining the data sets under each type to form a total data set for training and testing, and selecting proper nodes to form a cross data set.
Further, the step (2) includes the sub-steps of:
(2.1) adding white Gaussian noise with the same dimension in the data set sample prepared in (1.3), namely adding a noise penalty term in the target loss function
Where f is the encoding layer function, g is the decoding layer function, and L is the square of the error between the two elements.
(2.2) adding another penalty term in the loss function, namely F norm square of Jacobian matrix of hidden layer vectors
Where lambda is a tuning parameter which is,
is a hidden layer vector pair input xiThe sum of the squares of the partial derivatives of (a);
(2.3) extracting the characterization vectors by adopting a CDAE model, inputting the characterization vectors into an SVC model for classification, and outputting label quantity to realize supervised process detection;
and (2.4) minimizing a loss function containing potential noise penalty terms and Jacobian norm penalty terms by adopting a random gradient descent method, optimizing CDAE model parameters, and cross-training the SVC classifier.
Further, the step (3) includes the sub-steps of:
(3.1) collecting sensor information such as steering wheel corners, driving torques, braking torques and output torques of four hub motors during vehicle running work to form a data set, and standardizing the data set in each dimension max-min;
(3.2) inputting the samples after the standardization treatment into the optimized CDAE model (2.4), and calculating hidden layer variables corresponding to the samples;
(3.3) inputting the hidden layer vector and the label into the optimized SVC model (2.4), and calculating the posterior classification probability corresponding to the sample;
(3.4) taking the maximum posterior classification probability in the corresponding model
And the corresponding label quantity is used as the fault category of the sample, whether the hub motor has a fault is judged, and the fault category is determined.
The technical scheme of the invention is summarized as follows:
1. the invention makes a cross data set with label amount containing training and verifying functions. Firstly defining a hub motor failure factor delta, dividing the failure types of the hub motors of the four-wheel independent drive into 6 modes according to the failure positions and quantity distribution of the hub motors, respectively corresponding to different label quantities, selecting appropriate input and output variables, and forming a composite data set which can be used for model parameter optimization after standardization treatment;
2. in order to increase robustness of the model to characteristic noise, Gaussian white noise with the same dimensionality is added in an input sample, namely a noise penalty term is added in a target loss function, so that the extracted characteristic vector is more stable when the model processes a potential data sample with noise, meanwhile, another penalty term is added in the loss function, namely F norm square of a Jacobian matrix of a hidden layer vector, in the process of minimizing the loss function, the penalty term can enable a partial derivative of the hidden layer vector of the model to the input to tend to 0, the hidden layer vector of a self-encoder is insensitive to small amplitude change of the input vector, and the loss function of the established CDAE model comprises a potential noise penalty term and a Jacobian penalty term;
3. the invention develops a supervised fault detection method capable of improving the feature extraction capability of a CDAE model. Firstly, a CDAE model is adopted to extract a characterization vector, then the characterization vector is input into an SVC model for classification, a label amount is output, supervised process detection is realized, a random gradient descent method is adopted, a loss function containing potential noise punishment terms and Jacobian norm punishment terms is minimized, CDAE model parameters are optimized, and a cross data set is used for cross training of an SVC classifier
The method has the advantages that the method solves the problem of detecting the failure of the fault of the hub motor driven by the four wheels of the electric automobile independently, abandons the traditional fault detection method based on a model, provides a new method based on data driving, and can accurately describe the nonlinear characteristic of the automobile without establishing an accurate automobile model; a loss function containing potential noise punishment terms and Jacobian normal form punishment terms is provided, the provided hidden layer variable is more stable, and the robustness of the model to the characteristic noise is improved; a supervised fault detection strategy is provided, the label quantity of original data can be output, and the feature extraction capability of a CDAE model is improved. The method for detecting the motor fault of the hub of the electric automobile driven by the supervision data is superior to the existing method and has the characteristics of rapidness, accuracy, strong robustness and the like.
Drawings
FIG. 1 is a diagram illustrating input-output relationships between process variables in an established data set;
fig. 2 is a schematic diagram of failure types of in-wheel motors studied by the present invention, wherein (a) corresponds to a single motor failure type, (b) corresponds to two motor failure types on the same side, (c) corresponds to two motor failure types on different coaxial sides, (d) corresponds to two motor failure types on different coaxial sides, (e) corresponds to three motor failure types, and (f) corresponds to four motor failure types;
FIG. 3 is a schematic diagram of the structure of an AE model of an auto-encoder;
FIG. 4 is a flow chart of the CDAE-SVC method proposed by the present invention.
Detailed Description
The invention provides a method for detecting the fault of an electric automobile hub motor driven by supervision data, which comprises the following steps:
(1) reconstructing a data set containing wheel hub motor fault information: acquiring input and output variable information with motor fault attributes, performing fault classification on training sets of different fault conditions, adding a fault classification label quantity to each sample, and establishing a composite working condition data set; the method specifically comprises the following substeps:
(1.1) firstly defining a failure factor delta for describing the execution capacity of the actuator before and after failure, and then equivalently converting the failure factor into a reference standard capable of describing whether failure occurs or not according to the failure type of a specific hub motor;
in the research, the electrical fault of the hub motor is mainly considered, after the hub motor of the automobile fails, the motor cannot output torque according to expected torque, and a failure factor delta is defined for describing the execution capacity of the actuator before and after the failure:
the failure factor delta is used for describing the failure degree of the actuator, and the larger the delta is, the lighter the failure degree is; the smaller δ indicates a higher degree of failure. When the failure factor delta is 1, the actuator works normally; when the failure factor delta is 0, the actuator completely loses the execution capacity; when the failure factor 0< δ <1, the actuator is partially failed.
(1.2) in order to accurately position the failed motor and realize the function of fault classification, the specific failure degree is not counted, and whether a certain hub motor fails or not is focused. Based on the premise, according to the position and the quantity distribution of the motor failures, the failure types of the automobile hub motors driven by four wheels independently are divided into 6 modes, namely, a single motor failure mode, two motor failures on the same side, two motor failures on different sides and different shafts, three motor failures and four motor failures are shown in fig. 2 and respectively correspond to the label quantities 1, 2, 3, 4, 5 and 6, and the data quantity label quantity under the normal fault-free working condition is 0.
(1.3) acquiring input and output data of the vehicle during running under different failure types defined in (1.2), such as information of driving torque, braking torque, output torques of four hub motors and the like under a certain failure to form a data set with label quantity, combining the data sets under each type to form a total data set for training and verification, selecting proper nodes to form a cross data set
The vehicle model has high nonlinear characteristics, the whole vehicle model, the tire model and the hub motor model which are included in the vehicle model present complex mapping relationships, the model parameters to be considered are huge and time-varying, a state observer is required to be established to estimate state quantity, and unmodeled dynamic errors are inevitably added in the process of arranging the observer, so that the fault detection based on the model method is inaccurate. Although the model-based method has great disadvantages, the input and output variables used by the model-based method are the best choices for characterizing the non-linear characteristics of the vehicle model, so the data set selects the same process variables as the model-based method, and therefore the data set contains the variables of steering wheel angle, driving torque, braking torque, output torque of four in-wheel motors, and longitudinal speed of the vehicle, which are related as shown in fig. 1.
(2) Training the CDAE-SVC model: forming a cross data set by using the data set formed in the step (1), minimizing a loss function containing a potential noise penalty term and a jacobian penalty term by using the data set, optimizing CDAE model parameters, and cross-verifying an SVC classifier; the method specifically comprises the following substeps:
(2.1) adding white Gaussian noise with the same dimension in the data set sample prepared in (1.3), namely adding a noise penalty term in the target loss function
(2.2) adding another penalty term in the loss function, namely F norm square of Jacobian matrix of hidden layer vectors
The most basic self-encoder (AE) is a special three-layer neural network, the structure of which is shown in fig. 3. The input layer and the output layer of the self-encoder have the same dimension, and the structure of the self-encoder determines the good information retention capacity of the self-encoder, so that the self-encoder can be used for feature extraction. The single hidden layer self-encoder comprises two mapping functions, namely an encoding layer function f and a decoding layer function g. f will sample
Nonlinear projection from input vector space to feature vector space hi＝f(xi) Obtaining
And g is then
Projecting back into the input vector space to obtain a reconstruction of the sample input
The loss function of the auto-encoder (AE) consists only of the sum of the squares of the reconstruction error:
the proposed CDAE model contains a potential noise penalty term and a jacobian penalty term, so the established CDAE model loss function formula is as follows:
wherein the content of the first and second substances,
is the hidden layer vector hiFor input xiThe sum of the squares of the partial derivatives of (c), hi＝f(Wxi+ b), W is the projection matrix, b is the bias term, and λ is the tuning parameter used to tune the effect of the jacobian penalty term on the overall loss function.
If the activation function f (-) is sigmoid (), Jacobian form
The following formula can be rewritten:
wherein is IpIs a p-dimensional real number vector space RpAll 1 vectors, hi∈Rp，W∈Rp×m。
If the reconstruction error squared is taken as the main loss function, the loss function of the CDAE is rewritten as:
where · is the sign of the vector inner product, λ is the conditioning parameter, W is the projection matrix, hiIs a hidden layer vector.
(2.3) extracting the characterization vectors by adopting a CDAE model, inputting the characterization vectors into an SVC model for classification, and outputting label quantity to realize supervised process detection;
and (2.4) minimizing a loss function containing potential noise penalty terms and Jacobian norm penalty terms by adopting a random gradient descent method, optimizing CDAE model parameters, and cross-verifying an SVC classifier to obtain the optimal value of each parameter.
(3) Testing the hub motor fault process: collecting various index data of the motor when the vehicle operates, processing a data set, inputting hidden layer variables into the trained CDAE-SVC model in the step (2), calculating posterior classification probability, judging whether the hub motor has a fault or not, and determining the fault category. The method specifically comprises the following substeps:
(3.1) collecting sensor information such as steering wheel corners, driving torques, braking torques, output torques of four hub motors, longitudinal speeds of vehicles and the like when the vehicles run to form a data set, standardizing the data set in each dimension max-min, and recording the value of each dimension max and min;
(3.2) inputting the samples after the standardization treatment into the optimized CDAE model (2.4), and calculating hidden layer variables corresponding to the samples;
(3.3) inputting the hidden layer vector and the label into the optimized SVC model (2.4), and calculating the posterior classification probability corresponding to the sample
(3.4) taking the maximum posterior classification probability in the corresponding model
And the corresponding label quantity is used as the fault category of the sample, whether the hub motor has a fault is judged, and the fault category is determined.
Claims (4)
1. A method for detecting the fault of an electric automobile hub motor driven by supervision data is characterized by comprising the following steps:
(1) reconstructing a data set containing wheel hub motor fault information: the method comprises the steps of collecting input and output variable information with motor fault attributes, carrying out fault category division on training sets of different fault conditions, adding a fault category label quantity to each sample, and establishing a composite working condition data set.
(2) Training and validation of the CDAE-SVC model: and (3) forming a cross data set by using the data set established in the step 1, minimizing a loss function containing a potential noise penalty term and a jacobian penalty term by using the data set, optimizing CDAE model parameters, and cross-verifying the SVC classifier.
(3) Testing the faults of the hub motor: and (3) when the vehicle runs and works, acquiring various index data of the hub motor to be detected, processing a data set, inputting hidden layer variables into the CDAE-SVC model trained in the step (2), calculating posterior classification probability, and determining fault classification.
2. The method for detecting the fault of the hub motor of the electric vehicle driven by the supervised data as recited in claim 1, wherein the step (1) comprises the following substeps:
(1.1) firstly defining a failure factor delta for describing the executing capacity of the hub motor before and after failure, and then equivalently converting the failure factor into a reference standard capable of describing whether failure occurs or not according to the specific failure type of the hub motor.
(1.2) according to the position and the quantity distribution of the failure of the hub motors, the failure types of the hub motors of the four-wheel independent drive automobile are divided into 6 modes, namely, single motor failure, two motor failures on the same side, two motor failures on different sides and different shafts, three motor failures and four motor failures correspond to the label quantities of 1, 2, 3, 4, 5 and 6 respectively.
And (1.3) acquiring input and output data of the vehicle during running under different failure types defined in the step (1.2), wherein the input and output data comprise information such as driving torque, braking torque, output torques of four hub motors and the like under a certain failure, forming a data set with corresponding label quantity, combining the data sets under each type to form a total data set for training and testing, and selecting proper nodes to form a cross data set.
3. The method for detecting the failure of the hub motor of the electric vehicle driven by the supervised data as recited in claim 1, wherein the step (2) comprises the following substeps:
(2.1) adding white Gaussian noise with the same dimension in the data set sample prepared in the step (1.3), namely adding a noise penalty term in the target loss function
(2.2) adding another penalty term in the loss function, namely F norm square of Jacobian matrix of hidden layer vectors
And (2.3) extracting the characterization vectors by adopting a CDAE model, inputting the characterization vectors into an SVC model for classification, outputting the label quantity, and realizing supervised process detection.
And (2.4) minimizing a loss function containing potential noise penalty terms and Jacobian norm penalty terms by adopting a random gradient descent method, optimizing CDAE model parameters, and cross-verifying the SVC classifier.
4. The method for detecting the fault of the hub motor of the electric vehicle driven by the supervised data as recited in claim 1, wherein the step (3) comprises the following substeps:
(3.1) collecting sensor information such as steering wheel corners, driving torques, braking torques and output torques of four hub motors during vehicle running work to form a data set, standardizing the data set in each dimension max-min, and recording each dimension max and min value;
(3.2) inputting the samples after the standardization treatment into the optimized CDAE model (2.4), and calculating hidden layer variables corresponding to the samples;
(3.3) inputting the hidden layer vector and the label into the optimized SVC model (2.4), and calculating the posterior classification probability corresponding to the sample;
(3.4) taking the maximum posterior classification probability in the corresponding model
And the corresponding label quantity is used as the fault category of the sample, whether the hub motor has a fault is judged, and the fault category is determined.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011547614.3A CN112733887A (en) | 2020-12-24 | 2020-12-24 | Method for detecting fault of hub motor of electric vehicle driven by supervision data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011547614.3A CN112733887A (en) | 2020-12-24 | 2020-12-24 | Method for detecting fault of hub motor of electric vehicle driven by supervision data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112733887A true CN112733887A (en) | 2021-04-30 |
Family
ID=75605175
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011547614.3A Pending CN112733887A (en) | 2020-12-24 | 2020-12-24 | Method for detecting fault of hub motor of electric vehicle driven by supervision data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112733887A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116989730A (en) * | 2023-09-27 | 2023-11-03 | 郯城鸿顺机动车检测有限公司 | Hub roundness detection equipment for automobile detection |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108248598A (en) * | 2018-01-08 | 2018-07-06 | 武汉理工大学 | A kind of hybrid electric vehicle driven by wheel hub Failure Control system and method |
| US20190124045A1 (en) * | 2017-10-24 | 2019-04-25 | Nec Laboratories America, Inc. | Density estimation network for unsupervised anomaly detection |
| CN110207997A (en) * | 2019-07-24 | 2019-09-06 | 中国人民解放军国防科技大学 | Liquid rocket engine fault detection method based on convolution self-encoder |
| CN110481338A (en) * | 2019-07-23 | 2019-11-22 | 武汉理工大学 | A kind of hub motor vehicle disablement control method and entire car controller |
| CN111428788A (en) * | 2020-03-24 | 2020-07-17 | 西安交通大学 | Deep learning-based multi-fault diagnosis method and system for steam turbine generator set rotor |
| CN112116029A (en) * | 2020-09-25 | 2020-12-22 | 天津工业大学 | Intelligent fault diagnosis method for gearbox with multi-scale structure and characteristic fusion |
-
2020
- 2020-12-24 CN CN202011547614.3A patent/CN112733887A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190124045A1 (en) * | 2017-10-24 | 2019-04-25 | Nec Laboratories America, Inc. | Density estimation network for unsupervised anomaly detection |
| CN108248598A (en) * | 2018-01-08 | 2018-07-06 | 武汉理工大学 | A kind of hybrid electric vehicle driven by wheel hub Failure Control system and method |
| CN110481338A (en) * | 2019-07-23 | 2019-11-22 | 武汉理工大学 | A kind of hub motor vehicle disablement control method and entire car controller |
| CN110207997A (en) * | 2019-07-24 | 2019-09-06 | 中国人民解放军国防科技大学 | Liquid rocket engine fault detection method based on convolution self-encoder |
| CN111428788A (en) * | 2020-03-24 | 2020-07-17 | 西安交通大学 | Deep learning-based multi-fault diagnosis method and system for steam turbine generator set rotor |
| CN112116029A (en) * | 2020-09-25 | 2020-12-22 | 天津工业大学 | Intelligent fault diagnosis method for gearbox with multi-scale structure and characteristic fusion |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116989730A (en) * | 2023-09-27 | 2023-11-03 | 郯城鸿顺机动车检测有限公司 | Hub roundness detection equipment for automobile detection |
| CN116989730B (en) * | 2023-09-27 | 2024-01-02 | 郯城鸿顺机动车检测有限公司 | Hub roundness detection equipment for automobile detection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105741637B (en) | Four-wheel hub motor electric car automated steering control method | |
| CN106662501B (en) | System and method for analyzing the efficiency of vehicle | |
| JP7053147B2 (en) | Systems and methods for analyzing the energy efficiency of automobiles, especially automobile equipment | |
| US20210171050A1 (en) | System for evaluating vehicle performance | |
| CN107247824A (en) | Consider the car mass road grade combined estimation method of brake and influence of turning | |
| CN111775925B (en) | Working mode decision method and device for power split hybrid electric vehicle | |
| CN113359457B (en) | High-dimensional dynamic model resolving device and method for intelligent vehicle chassis area controller | |
| CN110091914B (en) | Distributed automobile multi-working-condition identification differential steering method and system | |
| CN109357900B (en) | Method and system for determining test system of mechanical electronic equipment | |
| CN107554356A (en) | The moment of torsion distribution method and device of electric automobile | |
| Zhao et al. | Distributed electric powertrain test bench with dynamic load controlled by neuron PI speed-tracking method | |
| CN112733887A (en) | Method for detecting fault of hub motor of electric vehicle driven by supervision data | |
| CN111198501B (en) | Method for determining fuel equivalent factor by RBF neural network | |
| CN111746304B (en) | Distributed driving electric vehicle driving system control method based on failure state | |
| Joshi | Powertrain and chassis hardware-in-the-loop (HIL) simulation of autonomous vehicle platform | |
| Wang et al. | Vehicle stability control strategy based on recognition of driver turning intention for dual-motor drive electric vehicle | |
| CN109900399A (en) | A kind of wheel side/hub drive system test platform | |
| CN113177291B (en) | Electric vehicle efficacy analysis method and system based on cloud data platform | |
| US20210342500A1 (en) | Systems and methods for vehicle modeling | |
| CN112883500A (en) | Intelligent vehicle system early function safety assessment method based on fault injection | |
| CN114739554B (en) | Test experiment method and evaluation method for torsional strength of four-wheel drive train | |
| CN115421473A (en) | Simulation verification method and system for automobile driving function | |
| CN112810602B (en) | Intelligent line control chassis personalized stability control method and system based on clustering | |
| CN114997748A (en) | New energy automobile operation safety risk prediction method and system based on model fusion | |
| CN110018678B (en) | Fault diagnosis method for networked automobile control system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |