Monitoring and Identifying Sensor Failure in an Electric Drive System

Abstract

A method for monitoring and identifying sensor faults in an electric drive system of a vehicle includes collecting corresponding data using sensors in the electric drive system, inputting the collected data to an already-established sensor fault mode identification model, and determining whether a fault mode exists and a fault mode type based on the collected data using the sensor fault mode identification model. The method quickly determines the fault mode caused by a sensor fault in the electric drive system, and the fault mode type of the sensor fault.

Claims

1. A method for monitoring and identifying sensor faults in an electric drive system of a vehicle, comprising: collecting corresponding data using sensors in the electric drive system of the vehicle; inputting the collected data to an already-established sensor fault mode identification model; and determining whether a fault mode exists and a fault mode type based on the collected data using the sensor fault mode identification model.

2. The method for monitoring and identifying sensor faults according to claim 1, further comprising: after determining that the fault mode exists and the fault mode type, inputting fault data in a database, the fault data corresponding to the fault mode and/or the fault mode type; or after determining that the fault mode does not exist, inputting normal data in the database, the normal data indicating that the fault mode does not exist.

3. The method for monitoring and identifying sensor faults according to claim 2, further comprising: optimizing and updating the sensor fault mode identification model based on the fault data and/or the normal data.

4. The method for monitoring and identifying sensor faults according to claim 1, further comprising: after determining that the fault mode exists and the fault mode type, running a corresponding compensation algorithm program stored in the electric drive system or a controller of the vehicle, to compensate for data inaccuracy caused by a faulty sensor of the sensors in the electric drive system.

5. The method for monitoring and identifying sensor faults according to claim 1, wherein the sensor fault mode identification model is established based on initial data by a machine learning method.

6. The method for monitoring and identifying sensor faults according to claim 1, further comprising: performing the method using (i) a vehicle network monitoring system comprising corresponding electric drive systems in each of a plurality of vehicles including the vehicle, and (ii) a cloud computing system configured to communicate wirelessly with each electric drive system or a controller of each vehicle of the plurality of vehicles, wherein the cloud computing system comprises a database for storing the collected data, and a central processing unit for performing calculations and communication.

7. The method for monitoring and identifying sensor faults according to claim 2, wherein the sensor fault mode identification model is based on:
S(normal, Flt.sub.1, Flt.sub.2, . . . , Flt.sub.n)=Func(X) wherein X is an input variable corresponding to variation of the electric drive system with a time domain or a frequency domain, wherein Func(X) is the sensor fault mode identification model established based on initial data, and wherein S corresponds to the fault data and/or the normal data.

8. The method for monitoring and identifying sensor faults according to claim 7, wherein: X includes a parameter X(t) corresponding variation of the electric drive system with the time domain, and is one or more of a motor power, a motor rotation speed, a motor torque, a DC voltage from a battery pack, a current measurement value, a position sensor measurement value, a stator temperature, an insulated gate bipolar transistor temperature, a direct-axis current, a quadrature-axis current, a direct-axis voltage, and a quadrature-axis voltage.

9. The method for monitoring and identifying sensor faults according to claim 7, wherein: X includes a parameter X(f) corresponding to variation of the electric drive system with the frequency domain, and is one or more of harmonic information, standard deviation, kurtosis, or skewness acquired from a time domain variable by frequency domain calculation or statistical calculation.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In the drawings:

[0023] FIG. 1 shows schematically a typical electric drive system in simplified form.

[0024] FIG. 2 shows schematically a vehicle network monitoring system.

[0025] FIG. 3 is a schematic drawing explaining the method for monitoring and identifying sensor faults in an electric drive system according to the present invention.

[0026] FIG. 4 is a flow chart showing a preferred embodiment of the method for monitoring and identifying sensor faults in an electric drive system according to the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0027] Preferred embodiments of the present invention are described in detail below in conjunction with examples. Those skilled in the art will understand that these exemplary embodiments do not mean that any limitation is applied to the present invention.

[0028] FIG. 2 shows schematically a vehicle network monitoring system. As shown in FIG. 2, the vehicle network monitoring system comprises an electric drive system in each of multiple vehicles 6, and a cloud 7 capable of communicating wirelessly with a vehicle controller in each vehicle 6. The vehicle controller is in electrical communication with the electric drive system in the corresponding vehicle, in order to monitor the electric drive system. Data can be transmitted between the vehicle controller and the cloud. Of course, the cloud 7 may also be in direct electrical communication with, and communicate wirelessly with, the electric drive system in each vehicle 6. A typical cloud in the vehicle network monitoring system comprises a database 8 for storing data from all of the vehicles, and a central processing unit 9 for performing all necessary calculation and communication.

[0029] The method for monitoring and identifying sensor faults in an electric drive system according to the present invention uses the vehicle network monitoring system as a basic platform. FIG. 3 is a schematic drawing explaining the method for monitoring and identifying sensor faults in an electric drive system according to the present invention. The cloud collects all necessary data from each of the multiple interconnected vehicles and stores same in the database 8. Data for monitoring and identifying sensor faults in an electric drive system includes but is not limited to: current sensor measurement values, position sensor measurement values, battery pack DC voltage, stator temperature, insulated gate bipolar transistor (IGBT) temperature, motor rotation speed, motor torque, motor power, etc. Various data collected on each occasion from each of the multiple interconnected vehicles are called a data set 10; the data in the data set is divided into normal data 11 and fault data 12, and stored in the database 8 as historical data. Normal data is sensor measurement values when vehicle operation is normal; fault data is sensor measurement values when vehicle operation is abnormal, i.e. when a fault has occurred. For example, when a vehicle experiences vibration or a fault such as insufficient motive power, a gain error or offset error might occur in the measurement value of the current sensor, or a gain error or offset error might occur in the measurement value of the position sensor. When a gain error occurs in the measurement value of the current sensor, this will cause torque ripple, and the frequency of this type of torque ripple is generally double the frequency of the output current of the inverter; when an offset error occurs in the measurement value of the current sensor, this will also cause torque ripple, and the frequency of this type of torque ripple is generally the same as the frequency of the output current of the inverter. When a gain error or offset error occurs in the measurement value of the position sensor, in both cases the result will be that the output torque of the motor is lower than the desired torque. Thus, it is possible to determine possible sensor fault types according to the vehicle fault situation, e.g. what type of fault has occurred in which sensor. The measurement value of the sensor in which the fault has occurred can then be subjected to compensation, in order to eliminate the inaccuracy thereof, by means of a compensation algorithm program stored in the inverter or vehicle controller. Initial fault data (historical data) can be acquired online by a fault detection method of the vehicle itself when a fault occurs during vehicle operation (e.g. in the case of a fault which occurs in the position sensor), or acquired offline by vehicle maintenance personnel removing the sensor.

[0030] The central processing unit 9 of the cloud then performs data processing by means of a machine learning method 13 on the basis of the initial historical data acquired, in order to establish a sensor fault mode identification model 14. There are many machine learning methods that are capable of performing these functions, e.g. artificial neural networks, clustering, similarity and metric learning, etc., in order to establish the sensor fault mode identification model. It is possible to identify or determine various fault modes and normal modes by means of the sensor fault mode identification model. An input of the sensor fault mode identification model is a new data set 15 collected from each vehicle, and an output of the sensor fault mode identification model is whether a fault mode exists 16, and a fault mode type 17. Information relating to fault mode type is then transmitted to the inverter or vehicle controller, in order to run a compensation algorithm program 18 at the electric drive system side, in order to subject the measurement value inaccuracy of the faulty sensor to compensation. The new data set 15 collected from each vehicle is also simultaneously transmitted to the database 8 as historical data, to further perfect and update the database 8.

[0031] For a vehicle that joined the vehicle network monitoring system at an early stage, it might be necessary to perform initial historical fault data acquisition offline, but it is not necessary to perform initial historical fault data acquisition offline for a vehicle that joined the vehicle network monitoring system after a fault identification model was established; instead, data thereof is inputted directly to the fault identification model to perform identification and simultaneously inputted to the database 8 to further perfect and update the database 8. With the aid of Big Data that is continuously collected from the various vehicles, it is possible to continuously perfect the sensor fault mode identification model.

[0032] More specifically, the sensor fault mode identification model is established on the basis of the following formula:)

S(normal, Flt.sub.1, Flt.sub.2, . . . , Flt.sub.n)=Funct(X)

[0033] wherein X is an input variable that reflects the variation of the vehicle's electric drive system with the time domain or frequency domain, and is collected and stored in the database of the cloud. When X is a parameter X(t) reflecting variation of the electric drive system with the time domain, X(t)=[P, n.sub.motor, T.sub.c, U.sub.dc, I.sub.ac, 0, T.sub.c, T.sub.igbt, I.sub.d, I.sub.q, U.sub.d, U.sub.q . . . ], wherein P is the motor power P, n.sub.motor is the motor rotation speed, T.sub.c is the motor torque, U.sub.dc is the DC voltage from the battery pack, I.sub.ac is the current measurement value, θ is the position sensor measurement value, T.sub.s is the stator temperature, T.sub.igbt is the IGBT temperature, I.sub.d is the direct-axis current, I.sub.q is the quadrature-axis current, U.sub.d is the direct-axis voltage, U.sub.q is the quadrature-axis voltage . . . . X(t) may be any one or more of these parameters.

[0034] When X is a parameter X(f) reflecting variation of the electric drive system with the frequency domain, X(f) is harmonic information, standard deviation, kurtosis, skewness, etc., acquired from a time domain variable by frequency domain calculation or statistical calculation.

[0035] Func( ) is the sensor fault mode identification model established on the basis of historical data stored in the vehicle network monitoring system or stored in the cloud. As stated above, the sensor fault mode identification model may be established by any suitable machine learning method, such as an artificial neural network, cluster analysis, etc. S(normal, Flt.sub.1, Flt.sub.2, . . . , Flt.sub.n) is an output of the sensor fault mode identification model, and represents a mode corresponding to normal data, a fault type 1 (Flt.sub.1), a fault type 2 (Flt.sub.2), . . . , a fault type n (Flt.sub.n). For example, in the case of a normal situation, a fault caused by a current sensor measurement value gain error, a fault caused by a current sensor measurement value offset error, a fault caused by a position sensor measurement value gain error, a fault caused by a position sensor measurement value offset error, etc., the output of S(normal, Flt.sub.1, Flt.sub.2, . . . , Flt.sub.n) may use the numbers 000, 001, 010, 100, 011, 110, 101, 111, etc. to represent the corresponding fault modes.

[0036] FIG. 4 is a flow chart showing a preferred embodiment of the method for monitoring and identifying sensor faults in an electric drive system according to the present invention. As shown in FIG. 4, the method for monitoring and identifying sensor faults in an electric drive system according to the present invention starts at step 20; the starting step may start at a set time. In step 21, various corresponding necessary data are collected by means of various sensors on a vehicle such as a position sensor, current sensor, voltage sensor and temperature sensor, and transmitted to a central processing unit in a cloud. In step 22, the various collected data are applied to an already-established sensor fault mode identification model to perform fault mode type identification. In step 23, a determination is made as to whether a fault exists. If a fault indeed exists, in step 25, corresponding data is added to fault data of a database. In step 26, a compensation algorithm program in an inverter or a vehicle controller is activated according to a known fault mode. It should be understood that steps 25 and 26 may be performed simultaneously, or step 26 may be performed before step 25. If there is no fault, in step 27, corresponding data is added to normal data of the database. After step 26 or 27, the method for monitoring and identifying sensor faults in an electric drive system according to the present invention returns to step 21, and the above steps are repeated. If a set operating time has been reached, the method for monitoring and identifying sensor faults in an electric drive system according to the present invention ends at step 28.

[0037] According to the present invention, once a fault mode has been identified, the central processing unit will communicate with the corresponding electric drive system, and if necessary activate the compensation algorithm program stored in the electric drive system or vehicle controller with a suitable parameter, to compensate for a sensor error. Thus, optimal performance of the electric drive system can be maintained even when a fault occurs in a sensor.

[0038] Vehicles that join the vehicle network monitoring system should have identical electric drive systems. If the number of vehicles joining the vehicle network monitoring system is greater, the amount of data acquired will be greater, and this will be of greater help in perfecting the sensor fault mode identification model, and therefore more favourable for increasing the accuracy with which sensor faults in the electric drive system are monitored and identified.

[0039] Although, in the above preferred embodiments, the sensor fault mode identification model is established on the basis of data from the database of the cloud of the vehicle network monitoring system, it should be understood that it is also possible to establish the sensor fault mode identification model on the basis of historical data of a single electric drive system, and the sensor fault mode identification model can then be applied to new operating data, in order to monitor and identify newly developed sensor faults in the same electric drive system.

[0040] The present invention has been described in detail above in conjunction with specific preferred embodiments. Obviously, the embodiments described above and displayed in the drawings are exemplary, and should not limit the present invention. Those skilled in the art should understand that various amendments and alterations may be made without departing from the spirit of the present invention, and such amendments and alterations will not depart from the scope of protection of the present invention.