Data processing: vehicles – navigation – and relative location – Vehicle control – guidance – operation – or indication – With indicator or control of power plant
Reexamination Certificate
2000-01-19
2001-09-18
Wolfe, Willis R. (Department: 3747)
Data processing: vehicles, navigation, and relative location
Vehicle control, guidance, operation, or indication
With indicator or control of power plant
C701S111000
Reexamination Certificate
active
06292738
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to a method for adaptive detection of engine misfire and more particularly, to a method which accurately detects misfires within an internal combustion engine by the use of multiple neural networks which are automatically and adaptively trained to compensate for the effects of vehicle variability and aging.
BACKGROUND OF THE INVENTION
Engine misfire detection systems are employed within internal combustion engines in order to reduce the likelihood of harmful emissions being discharged by the engine. Conventional engine misfire detection systems, assemblies, and methodologies typically receive and/or measure the angular velocity of the engine's crankshaft to derive angular acceleration values that are used to determine the occurrence of misfires within the cylinders of the engine. Particularly, engine misfire detection systems analyze the derived acceleration values to determine whether any “acceleration deficits” (i.e., less than normal or desirable acceleration values) are present, and compare these “acceleration deficit” values to predetermined deficit values, which are expected in the event of a misfire, to determine if a misfire has occurred. The predetermined deficit values typically vary based upon the specific cylinder that is firing, and upon the operating condition of the vehicle's engine (e.g., the engine speed and load).
For example and without limitation, because the crankshaft is not “perfectly stiff”, it gives rise to torsional oscillations, which are manifested as additions or subtractions to/from the acceleration of the crankshaft. These torsional oscillations vary based upon the cylinder that is firing and the operating conditions of the engine. Under certain operating conditions, accelerations resulting from these torsional oscillations may exceed or equal the acceleration deficits caused by a misfire, thereby significantly and adversely effecting the accuracy of misfire detection.
Various attempts have been made to compensate for the effects of torsional oscillations. For example and without limitation, engine misfire detection systems and methods have used adaptive schemes and static neural networks to compensate for the effects of torsional oscillations, such as the systems described within U.S. Pat. No. 5,531,108 of Feldkamp et al., and U.S. Pat. No. 5,774,823 of James et al., which are each assigned to the assignee of the present invention and which are each fully and completely incorporated herein by reference.
Other prior engine misfire detection systems have implemented dynamic neural networks to compensate for the effects of torsional oscillations. In one type of misfire detection system, described in U.S. Pat. No. 5,699,253 of Puskorius et al., which is assigned to the present assignee, and which is fully and completely incorporated herein by reference, a dynamic or “recurrent” neural network is trained to convert observed acceleration values to values which are nearly representative of values sensed by an ideal sensor operating in the absence of torsional oscillations. In this type of system, the neural network acts as a “nonlinear filter” or as an “inverse model”.
In another type of misfire detection system, described in U.S. Pat. No. 5,732,382 of Puskorius et al., which is assigned to the present assignee and which is fully and completely incorporated herein by reference, a dynamic or “recurrent” neural network is trained to “detect” or to directly determine whether a misfire has occurred. In this type of system, the neural network effectively and operatively combines the functions of an inverse model and a classifier. While these neural network-type systems have been proven to be more effective in compensating for the effects of torsional oscillations than other prior methods and/or systems, they suffer from several drawbacks.
For example and without limitation, because these prior neural network type systems are typically “trained” as part of a development process, they have a limited ability to handle variations that arise from vehicle-to-vehicle variability and the effects of vehicle aging. Particularly, because the “training” process is relatively complicated, time consuming and computationally intensive, it is not suited to be carried out “on-line” or during the normal use of a vehicle or an engine.
Additionally, known or predetermined output target values, which are required to train the neural networks, are typically not available during the normal use of the vehicle. During the development of the neural networks, output target values are made to be available and/or are artificially and precisely “synthesized” by the use of, for example and without limitation, ideal laboratory grade sensors and equipment, complex filtering techniques, and artificially induced engine misfires. During the normal operation of an engine or vehicle, these types of artificially “synthesized” target output values are not available. Hence, these systems cannot be adaptively trained “on-line”.
Another drawback associated with these prior systems is that they do not take into account the effect that misfires have on future firing events. By failing to consider the accelerations and torsional oscillations that are generated by a misfire, classifications or determinations of firing events, which occur after a misfire, may be incorrect or inaccurate.
Applicant's invention addresses these drawbacks and provides a method for detecting misfire within an engine which automatically adapts to the effects of vehicle variability and aging, and which takes into account the effect that misfires have on future firing events.
SUMMARY OF THE INVENTION
It is a first object of the invention to provide a method for detecting misfire within an engine which overcomes at least some of the previously delineated drawbacks of the prior systems, devices, and/or methods.
It is a second object of the invention to provide a method and for detecting misfire within an engine which utilizes a plurality of neural networks to determine whether a misfire within an engine has occurred.
It is a third object of the invention to provide a method for detecting misfire within a vehicle engine which utilizes a plurality of trainable neural networks to adaptively compensate for the effects of vehicle variability and aging.
It is a fourth object of the invention to provide a method for detecting misfire within an engine which takes into account the effects that misfires have on future firing events.
According to one aspect of the present invention a system for determining whether a first firing event within an engine is a normal event or a misfire is provided. The system includes at least one sensor which measures first engine operating data which is associated with the first firing event, and second engine operating data which is associated with at least one second firing event which occurs before the first firing event. The system further includes a controller which is communicatively coupled to the at least one sensor and which receives the first and the second engine operating data. The controller is effective to determine whether the first firing event is a misfire or a normal event based upon the first engine operating data and the second engine operating data.
According to a second aspect of the present invention, a method for determining if a firing event within an engine is a misfire is provided. The method includes the steps of: identifying a plurality of potential misfire states for the engine; providing a plurality of neural networks, each of the neural networks corresponding to a unique one of the potential misfire states; determining a first probability that each of the plurality of neural networks corresponds to a current misfire state of said engine; determining a second probability that each of the plurality of neural networks corresponds to a misfire state of the engine which occurs before the event; determining a third probability that each of the potential neural networks corresponds to a misfire state of the engine which occur
Feldkamp Lee Albert
Feldkamp Timothy Mark
Marko Kenneth Andrew
Prokhorov Danll Valentinovich
Ford Global Tech, Inc.
Ford Global Tech., Inc.
Wolfe Willis R.
LandOfFree
Method for adaptive detection of engine misfire does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for adaptive detection of engine misfire, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for adaptive detection of engine misfire will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2524378