Catalyst monitor with direct prediction of hydrocarbon conversion efficiency by dynamic neural networks

• US 5,625,750 A
• Filed: 06/29/1994
• Issued: 04/29/1997
• Est. Priority Date: 06/29/1994
• Status: Expired due to Term

• Current Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Original Assignee: Ford Motor Company
• Inventors: Puskorius, Gintaras V., Feldkamp, Lee A.
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/267,736
• Time in Patent Office: 1,035 Days
• Field of Search: 395/22, 395/23, 395/24, 395/903, 395/20-25, 395/27, 364/424.03, 364/150, 364/151, 364/577, 60/274, 123/672, 123/571, 382/155-159
• US Class Current: 706/21
• CPC Class Codes: F01N 11/00 Monitoring or diagnostic devices for exhaust-gas treatment apparatus , e.g. for catalytic activity safety, indicating or supervising devices for internal combustion engines F02B77/08; testing of machines G01M13/00
• CPC Class Codes: F01N 2550/02 Catalytic activity of catalytic converters
• CPC Class Codes: F02D 41/1405 Neural network control
• CPC Class Codes: G05B 13/026 using a predictor
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: Y02T 10/40 Engine management systems

Abstract

A process and apparatus for monitoring catalyst conversion activity includes a predictor of feedgas emissions and a predictor of tailpipe emissions, each predictor providing an output for generating a ratio of conversion activity. Each predictor comprises a trained neural network receiving at least one of, and preferably a plurality of, the engine operating condition signals available from an electronic engine control. Preferably, each neural network is trained by inputting accumulated data acquired from performance evaluation of a plurality of vehicles having consistent powertrains but with different degrees of deterioration.

Claims

1. A process for monitoring catalytic converter activity comprising:

training a first neural network to predict feedgas emissions from a plurality of signals derived from an electronic engine control module by sensing emission constituents under dynamic conditions;

training a second neural network to predict tailpipe emissions from a plurality of signals derived from an electronic engine control module by sensing emission constituents under dynamic conditions;

predicting feedgas emissions at said first trained neural network by inputting a first set of a plurality of engine operating condition signals to said first neural network;

predicting tailpipe emissions at said second trained neural network by inputting a second set of said plurality of engine operating condition signals to said second neural network;

predicting catalyst conversion activity by determining the ratio of feedgas emission predictions to tailpipe emission predictions; and

indicating when said predicted catalytic converter activity is below a predetermined level.

2. The invention as defined in claim 1 wherein said training steps comprise:

accumulating a plurality of feedgas emissions measurements and accumulating a plurality of tailpipe emissions measurements; and

determining average catalyst conversion activity by determining the ratio of accumulated feedgas emissions measurements to accumulated tailpipe emissions measurements.

3. The invention as defined in claim 1 wherein said training a first neural network step comprises modeling a prediction of feedgas emissions by accumulating a plurality of measured feedgas emission outputs from a plurality of vehicle operating cycles.

4. The invention as defined in claim 3 wherein said training step comprises accumulating a plurality of measured emission outputs from a plurality of vehicles.

5. The invention as defined in claim 1 wherein said first plurality of engine operating condition signals comprises at least two signals taken from the group consisting of mass air flow signal, engine speed signal, front EGO signal, vehicle speed signal, fuel requested signal, spark advance signal, and EGR signal.

6. The invention as defined in claim 1 wherein training said second neural network step comprises modeling a prediction of tailpipe emissions by accumulating a plurality of measured tailpipe emission outputs from a plurality of vehicle operating cycles.

7. The invention as defined in claim 6 wherein said training step comprises accumulating a plurality of measured tailpipe emission outputs from a plurality of vehicles.

8. The invention as defined in claim 1 wherein said second plurality of engine operating condition signals comprises at least two signals taken from the group consisting of mass air flow signal, engine speed signal, front EGO signal, vehicle speed signal, fuel requested signal, spark advance signal, EGR signal, rear EGO sensor and predicted feedgas emissions.

9. The invention as defined in claim 1 wherein at least one neural network is a recursive neural network.

10. The invention as defined in claim 1 wherein each neural network is a recursive neural network.

11. A monitor for monitoring compliance with hydrocarbon conversion activity levels of an internal combustion engine having an electronic engine control with a plurality of engine operating condition signals, comprising:

a first plurality of said engine operating condition signals;

a second plurality of said engine operating condition signals;

a first trained neural network predictor generating a feedgas emission constituent prediction signal in sensed response to said first plurality of engine operating condition signals;

a second trained neural network predictor generating a tailpipe emission constituent prediction signal in sensed response to said second plurality of engine operating condition signals;

a comparator for determining compliance of a ratio of said feedgas emission prediction to said tailpipe emission prediction with a predetermined ratio standard for said predictions; and

an indicator for advising of a failure to detect compliance in response to said ratio comparator.

12. The invention as defined in claim 11 wherein said second plurality of engine operating condition signals includes said feedgas emission constituent prediction.

13. A process for monitoring catalytic converter activity comprising:

inputting a plurality of engine operating condition signals;

predicting a first level of emissions at a feedgas pipe entering the converter in response to inputting a first set of said plurality of engine operating condition signals through a dynamically trained neural network;

predicting a second level of emissions at a tailpipe exiting the converter in response to inputting a second set of said plurality of engine operating condition signals through a dynamically trained neural network;

comparing the predicted first and second levels of emission predictions to determine a ratio of the tailpipe emission prediction to the feedgas pipe emission prediction; and

indicating the activity of said catalyst by signaling in response to a determined ratio of predictions greater than a predetermined ratio.

14. The invention as defined in claim 13 wherein said predicting a first level comprises inputting said first set of engine operating condition signals to a trained neural network.

15. The invention as defined in claim 13 wherein said predicting a second level comprises inputting said second set of engine operating condition signals to a trained neural network.