GENERATION OF REFERENCE VALUE FOR VEHICLE FAILURE DIAGNOSIS

• US 20110196572A1
• Filed: 09/09/2009
• Published: 08/11/2011
• Est. Priority Date: 10/10/2008
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: Aihara, Jiro, Tsuchikiri, Hideyuki, Sato, Yasushi
• Granted Patent: US 9,043,079 B2
• Time in Patent Office: Days
• US Class Current: 701/35
• CPC Class Codes: B60R 25/00 Fittings or systems for preventing or indicating unauthorised use or theft of vehicles locks for vehicles E05B77/00-E05B85/00
• CPC Class Codes: G07C 5/0808 Diagnosing performance data testing of vehicles G01M17/00; testing of electrical installation on vehicles G01R31/005
• CPC Class Codes: G07C 5/085 using electronic data carriers

Abstract

Data during normal driving is generated using travel data during the occurrence of a failure or the like, the travel data being accumulated into ordinary vehicles traveling in a city or other places on a daily basis. A failure diagnosis of a vehicle is performed by comparing a reference value with the time sequence electronic control unit (ECU) data of a plurality of driving parameters during the occurrence of a failure, the data being stored into a storage device in the ECU of the vehicle when a failure of the vehicle has occurred. The time sequence ECU data obtained from a large number of vehicles are sequentially accumulated and stored and the numerical vectors of the accumulated time sequence ECU data are generated. The numerical vectors are clustered and classified into a plurality of clusters according to the characteristics. In each of the plurality of clusters, a range of high occurrence rate values is obtained for the values of the respective driving parameters. The obtained range of high occurrence rate values is then stored as a range of the normal values of the driving parameters, and the range of the normal values is set as the reference value for a failure diagnosis.

Claims

1. A reference value generating apparatus for failure diagnosis in which time sequence ECU data for a plurality of driving parameters during the occurrence of a failure stored in a storage device of an electronic control unit (ECU) of a vehicle is compared with reference values when the failure of the vehicle occurs, to make a failure diagnosis of the vehicle, comprising:

an accumulation unit for sequentially accumulating and storing the time sequence ECU data obtained from a large number of vehicles;

means for generating numerical vectors of the time sequence ECU data accumulated in the accumulation unit;

clustering means for clustering the numerical vectors and classifying the clustered numerical vectors into a plurality of clusters according to the characteristics;

an occurrence rate calculating means for finding a range of high occurrence rate values for the parameter values of the driving parameters in the plurality of clusters; and

means for storing the range of high occurrence rate values obtained in the occurrence rate calculating means, as a range of normal values of the driving parameters,

wherein the range of normal values is set as a reference value for failure diagnosis.

2. The reference value generating apparatus according to claim 1, wherein the time sequence ECU data is data on a plurality of driving parameters including at least one of engine revolution speed, vehicle speed and cooling water temperature at a predetermined time before a failure of a vehicle occurs.

3. The reference value generating apparatus according to claim 2, wherein the means for generating numerical vectors divides the time sequence ECU data into a plurality of time zones per record, calculates the characteristic amounts of the driving parameters per time zone, and generates numerical vectors of the characteristic amounts.

4. The reference value generating apparatus according to claim 3, wherein the clustering means, mainly for a plurality of first core vectors having a different initial value, calculates degrees of approximation between the numerical vectors for the time zones and the plurality of core vectors to cause each numerical vector to belong to a cluster having a main core vector with the highest degree of approximation, and generates the plurality of first clusters.

5. The reference value generating apparatus according to claim 4,

wherein a) an average of the characteristic amounts of a plurality of numerical vectors belonging to the cluster is taken to replace a numerical value of the main core vector in the cluster with the average,

b) the obtained core vector is used to calculate degrees of approximation between the numerical vectors for the time zones and a plurality of core vectors to cause each numerical vector to belong to the cluster having the main core vector with the highest degree of approximation.

6. The reference value generating apparatus according to claim 5.

Wherein c) the a) and the b) processings are repeated until the centers of the core vectors converge or until a preset number of times of trial ends.

7. The reference value generating apparatus according to claim 6, wherein a processing of causing numerical vectors to belong to the plurality of first clusters is performed on one or a plurality of predetermined driving parameters, and performs secondary clustering on a plurality of second clusters for the plurality of first clusters about a larger number of driving parameters in terms of the numerical vectors belonging to the plurality of obtained first clusters.

8. The reference value generating apparatus according to claim 7, wherein an occurrence rate of the driving parameter for the numerical vector belonging to the cluster obtaMed in the secondary clustering is calculated and the range of high occurrence rate numerical values is set as a reference value of each driving parameter.

9. A method for generating a reference value for failure diagnosis which compares time sequence ECU data for a plurality of driVing parameters during the occurrence of a failure stored in a storage device of an electronic control unit (ECU) of a vehicle when the failure of the vehicle occurs with a reference value to make a failure diagnosis of the vehicle, comprising:

sequentially accumulating and storing the time sequence ECU data obtained from a large number of vehicles;

generating numerical vectors of the stored time sequence ECU data;

clustering the numerical vectors and classifyin g the clustered numerical vectors into a plurality of clusters according to the characteristics;

finding a range of high occiuTence rate values for the parameter values of the driving parameters for the plurality of clusters; and

storing the range of high occurrence rate values obtained in the occurrence rate calculating means as a range of normal values of the driving parameters,

wherein the range of normal values is set as a reference value for failure diagnosis.

10. The method for generating a reference value according to claim 9, wherein the time sequence ECU data is data relating to a plurality of driving parameters including at least one of engine revolution speed, vehicle speed and cooling water at a predetermined time before a failure of a vehicle occurs.

11. The method for generating a reference value according to claim 10, wherein the numerical vectors are generated by dividing the time sequence ECU data into a plurality of time zones per record, calculating the characteristic amounts of the driving parameters per time zone and generating numerical vectors of the characteristic amounts.

12. The method for generating a reference value according to claim 11, wherein the clustering, mainly for a plurality of first core vectors having a different initial value, calculates degrees of approximation between the numerical vectors of the time zones and the plurality of core vectors to cause each numerical vector to belong to a cluster having the main core vector with the highest degree of approximation, and generates the plurality of first clusters.

13. The method for generating a reference value according to claim 12,

wherein a) taking an average of the characteristic amounts of a plurality of numerical vectors belonging to the cluster, and replacing the numerical values of the main core vectors in the cluster with the average,

b) using the obtained core vectors to calculate degrees of approximation between the numerical vectors for the time zones and a plurality of core vectors, thereby causing each numerical vector to belong to a cluster having the math vector with the highest degree of approximation.

14. The method for generating a reference value according to claim 13,

wherein c) the a) and the b) processings are repeated until the centers of the core vectors converge or until a preset number of times of trial ends.

15. The method for generating a reference value according to claim 14, wherein a processing of causing in merical vectors to belong to the plurality of first clusters is performed on one or a plurality of predetermined driving parameters and performs secondary clustering on a plurality of second clusters for the plurality of first clusters about a larger number of driving parameters in terms of the numerical vectors belonging to the obtained plurality of first clusters.

16. The method for generating a reference value according to claim 15, wherein an occurrence rate of the driving parameter for a numerical vector belonging to the cluster obtained by the secondary clustering is calculated, and a range of the high occurrence rate values is set as a reference value of a respective driving parameter.