Multiple algorithm event discrimination method

• US 7,269,483 B2
• Filed: 08/19/2004
• Issued: 09/11/2007
• Est. Priority Date: 08/19/2004
• Status: Active Grant

• Current Assignee: Aptiv Technologies AG (Aptiv PLC)
• Original Assignee: Delphi Technologies, Inc. (BorgWarner Incorporated)
• Inventors: Schubert, Peter J., Aeschliman, Chad M., Wiles, Benjamin C.
• Primary Examiner(s): NGUYEN, TAN QUANG
• Application Number:
  US10/921,722
• Publication Number:
  US 20060041336A1
• Time in Patent Office: 1,118 Days
• Field of Search: 701/1, 701/38, 701/45, 180/271, 180/282, 280/756, 280/728.1, 280/5.502, 340/440
• US Class Current: 701/1
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Abstract

An event discrimination methodology executes multiple versions of the same or different event discrimination algorithms and logically or arithmetically combines their outputs to distinguish between specified events and non-events. One given algorithm is repeatedly executed with different sets of calibration data, or alternately, a number of different algorithms are executed. In cases where the algorithm results are arithmetically combined, the weights accorded to each algorithm result are dynamically adjusted based on driver input or vehicle dynamic behavior data to accord highest weight to the algorithm(s) calibrated to identify events associated with the detected driver input or vehicle dynamic behavior.

Claims

1. A method of discriminating an occurrence of a specified event in a system based on measured or estimated system parameter data pertinent to said event, the method comprising the steps of:

developing a plurality of different calibrations for one event discrimination algorithm for detecting said different patterns of parameter data that are characteristic of the occurrence of said specified event;

successively executing said one event discrimination algorithm with said different calibrations to produce a plurality of event discrimination outputs; and

combining said plurality of event discrimination outputs to form a single output that discriminates the occurrence of said specified event.

2. The method of claim 1, wherein the step of combining said event discrimination outputs includes the step of:

combining said event discrimination outputs with a logical-OR function.

3. A method of discriminating an occurrence of a specified event in a system based on measured or estimated system parameter data pertinent to said event, the method comprising the steps of:

executing a number of event discrimination algorithms individually calibrated to detect different patterns of parameter data that are characteristic of the occurrence of said specified event, each algorithm producing an event discrimination output;

applying calibrated weights to each of said event discrimination outputs;

combining the weighted event discrimination outputs to form a combined output; and

forming a single output based on said combined output.

4. The method of claim 3, where said different patterns of parameter data are characteristic of occurrences of said specified event due to different dynamic conditions of said system, and the method includes the steps of:

monitoring system data to detect at least one of said different dynamic conditions of said system;

identifying one or more of the event discrimination algorithms that are calibrated to detect a pattern of parameter data that is characteristic of occurrences of said specified event due to the detected dynamic condition; and

according higher weight to the event discrimination output of the identified event discrimination algorithm when combining the event discrimination outputs produced by said algorithms.

5. The method of claim 4, including the steps of:

storing a set of calibrated weights to be applied to each of said event discrimination outputs for arithmetically combining said event discrimination outputs; and

adjusting said calibrated weights in response to detection of said at least one the different dynamic condition of said system so as to increase the weight applied to the event discrimination output of the identified event discrimination algorithm.

6. The method of claim 3, where the specified event is a rollover of a vehicle operated by a driver, and said different patterns of parameter data are characteristic of occurrences of said specified event due to different actions of said driver, the method including the steps of:

monitoring system data to detect at least one of said different actions of said driver;

identifying one or more of the event discrimination algorithms that are calibrated to detect a pattern of parameter data that is characteristic of occurrences of said specified event due to the at least one detected action of the driver; and

according higher weight to the event discrimination output of the identified event discrimination algorithm when combining the event discrimination outputs produced by said algorithm.

7. The method of claim 6, including the steps of:

storing a set of calibrated weights to be applied to each of said event discrimination outputs for arithmetically combining said event discrimination outputs; and

adjusting said calibrated weights in response to detection of said at least one action of said driver so as to increase the weight applied to the event discrimination output of the identified event discrimination algorithm.