Diagnostic system using pattern recognition for electronic automotive control systems

• US 5,041,976 A
• Filed: 08/28/1989
• Issued: 08/20/1991
• Est. Priority Date: 05/18/1989
• Status: Expired due to Fees

• Current Assignee: Ford Motor Company
• Original Assignee: Ford Motor Company
• Inventors: Marko, Kenneth A., James, John V., Dosdall, James M.
• Primary Examiner(s): Lall, Parshotam S.
• Assistant Examiner(s): Zanelli, Michael
• Application Number:
  US07/399,683
• Time in Patent Office: 722 Days
• Field of Search: 364/424.01, 364/424.03, 364/431.02, 364/550, 364/551.01, 364/575, 364/578, 364/431.03, 364/424.04, 364/431.11, 364/513, 73/117.2, 73/117.3, 73/116, 307/201
• US Class Current: 701/29
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: F02D 41/22 Safety or indicating devices for abnormal conditions in air/fuel ratio feedback systems F02D41/1495, in electric control linkage F02D11/107, in purge control systems F02M25/0809
• CPC Class Codes: G01R 31/007 using microprocessors or computers
• CPC Class Codes: Y02T 10/40 Engine management systems
• CPC Class Codes: Y10S 706/911 Nonmedical diagnostics

Abstract

A system is disclosed for diagnosing faults in electronic control systems wherein a large volume of information is exchanged between the electronic control processor and a mechanical system under its control. The data is acquired such that parameter vectors describing the system operation are formed. The vectors are provided to a pattern recognition system such as a neural network for classification according to the operating condition of the electronically controlled system. For diagnosis of electronically controlled engine operation, the parameters included in the vectors correspond to individual firing events occurring in the engine operating under a predetermined condition. The diagnostic system can be implemented as a service tool in an automotive service bay or can be implemented within the on-board electronic control system itself.

Claims

1. A method for identifying selected conditions in an electronically controlled automotive system comprising the steps of:

conducting a plurality of system operating trials using a test system, each trial including at least one respective one of a plurality of known selected conditions;

collecting system operating data during said operating trials corresponding to a plurality of system parameters, said operating data including signals passing to and from an electronic control;

inputting said system parameters and corresponding known system fault identities as training vectors to a trainable pattern recognition system;

recursively adjusting said pattern recognition system to converge to a configuration of said pattern recognition system for matching input vectors with said system fault identities;

conducting a system operating trial using a system to be diagnosed;

collecting system diagnostic data corresponding to said plurality of system parameters; and

inputting said system parameters based on said diagnostic data as input vectors to said pattern recognition system to identify said selected conditions.

2. The method of claim 1 wherein said pattern recognition system performs a neural network simulation.

3. A method for identifying selected conditions in a combustion engine comprising the steps of:

conducting a plurality of engine operating trials using a test engine, each trial including at least one respective one of a plurality of known selected conditions;

collecting engine operating data during said operating trials corresponding to a plurality of engine parameters, said parameters including individual firing events;

inputting said engine parameters and condition codes as training vectors to a trainable pattern recognition system, each condition code identifying one of said selected conditions;

recursively adjusting said pattern recognition system to converge to a configuration of said pattern recognition system for matching input vectors with said condition codes;

conducting an engine operating trial using an engine to be diagnosed;

collecting engine diagnostic data corresponding to said plurality of engine parameters; and

inputting said engine parameters based on said diagnostic data as input vectors to said adjusted pattern recognition system to generate at least one condition code corresponding to one of said selected conditions present in said engine to be diagnosed.

4. The method of claim 3 wherein said selected conditions include a proper running condition of said combustion engine and conditions corresponding to fault conditions causing improper operation of said combustion engine.

5. The method of claim 3 further comprising the steps of:

calculating at least one of said engine parameters in said training vectors from said engine operating data; and

calculating at least one of said engine parameters in said input vectors from said diagnostic data.

6. The method of claim 3 wherein said test engine and said engine to be diagnosed include an electronic engine control and wherein said engine parameters are derived from input signals to and output signals from said electronic engine control.

7. The method of claim 6 wherein said input signals and said output signals include both analog and digital representations.

8. The method of claim 3 wherein said parameters include engine conditions averaged over a predetermined time period.

9. The method of claim 8 wherein said predetermined time period is one engine cycle.

10. The method of claim 8 wherein said averaged conditions include idle speed control duty cycle.

11. The method of claim 8 wherein said averaged conditions include engine speed.

12. The method of claim 8 wherein said averaged conditions include engine acceleration.

13. The method of claim 8 wherein said averaged conditions include throttle position.

14. The method of claim 3 wherein said parameters include engine conditions determined once per cylinder firing.

15. The method of claim 14 wherein said conditions determined once per engine cycle include air charge temperature.

16. The method of claim 14 wherein said conditions determined once per engine cycle include engine coolant temperature.

17. The method of claim 3 wherein said pattern recognition system is comprised of a microcomputer programmed to perform a neural network simulation.

18. The method of claim 3 wherein said pattern recognition system is comprised of a microcomputer programmed to perform a backpropagation analysis.

19. The method of claim 3 wherein said firing events include spark advance for each engine cylinder.

20. The method of claim 3 wherein said firing events include spark duration for each engine cylinder.

21. The method of claim 3 wherein said firing events include manifold absolute pressure.

22. The method of claim 3 wherein said firing events include engine position.

23. The method of claim 3 wherein said firing events include deviation of acceleration of a single cylinder from average acceleration over one engine cycle.

24. The method of claim 3 wherein said firing events include exhaust gas oxygen content.

25. The method of claim 3 wherein said firing events include exhaust gas recirculating pressure.

26. The method of claim 3 wherein said firing events include each fuel injector pulse width.

27. The method of claim 3 wherein said firing events include each fuel injector pulse timing.

28. The method of claim 3 wherein each of said vectors includes engine parameters corresponding to each cylinder firing within a single engine cycle.

29. Apparatus for identifying possible fault conditions in a combustion engine comprising:

monitor means for collecting a plurality of time-referenced values corresponding to selected engine parameters during operation of said engine, said parameters including individual firing event information;

processing means coupled to said monitor means for processing said values to derive said engine parameters and organizing said engine parameters as vectors;

memory means coupled to said processing means for storing said vectors; and

pattern recognition means coupled to said memory means for classifying said stored vectors according to said possible fault conditions.

30. The apparatus of claim 29 wherein said pattern recognition means includes a data base derived from a plurality of test vectors corresponding to said selected engine parameters collected during operation of a test engine for a plurality of trials, said trials including a plurality of known engine faults.

31. The apparatus of claim 29 wherein said time-referenced values include digital values and analog values, said digital values being monitored upon occurrence of a transition therein, and said analog values being monitored at predetermined intervals.

32. The apparatus of claim 29 wherein said pattern recognition means is comprised of a neural network.

33. The apparatus of claim 29 wherein said pattern recognition means performs a back propagation analysis.

34. The method for identifying selected conditions in a combustion engine comprising the steps of:

conducting a plurality of engine operating trials using a test engine, each trial including at least one respective one of a plurality of known selected conditions;

collecting engine operating data during said operating trials corresponding to a plurality of engine parameters, said parameters including individual firing events, said individual firing events including deviation of acceleration of a single cylinder from median filtered acceleration over one engine-cycle;

inputting said engine parameters and respective condition codes are training vectors to a pattern recognition system to correlate patterns in said engine parameters with respective selected conditions;

conducting an engine operating trial using an engine to be diagnosed;

collecting engine diagnostic data corresponding to said plurality of engine parameters; and

inputting said engine parameters based on said diagnostic data as input vectors to said pattern recognition system to generate at least one condition code corresponding to one of said selected conditions present in said engine to be diagnosed.

35. Apparatus for identifying selected conditions in a combustion engine comprising:

trial means for conducting a plurality of engine operating trials using a test engine, each trial including at least one respective one of a plurality of known selected conditions;

data means coupled to said trial means for collecting engine operating data during said operating trials corresponding to a plurality of engine parameters, said parameters including individual firing events;

pattern recognition means for receiving said engine parameters and condition codes as training vectors and for recursively adjusting to a configuration of said pattern recognition means for matching input vectors with said condition codes, each condition code identifying one of said selected conditions; and

operating means coupled to said pattern recognition means for collecting engine diagnostic data corresponding to said plurality of engine parameters in an engine to be diagnosed;

said engine parameters based on said diagnostic data being input as input vectors to said adjusted pattern recognition means to generate at least one condition code corresponding to one of said selected conditions present in said engine to be diagnosed.