With the recent movement to attack global environmental problems, regulations on everything from vehicle emissions to mileage are becoming stricter year by year to reduce noxious emissions. In order to meet these regulations, it is necessary to achieve higher reliability in those automotive parts which may affect emissions.
Since there is a limit to how much the reliability of parts can be increased against aging and deterioration, software such as learning control utilizing air-fuel ratio lambda control, has been used. However, it has proven difficult to compensate for accuracy, by subdividing areas to memorize the results of learning, and have a higher learning speed.
This paper describes the nest-structuring of learning areas which, unlike conventional air-fuel ratio learning, realize both faster learning and improvements in the compensating accuracy by subdividing the learning area. These achievements reduce the noxious emissions from automobiles and establish a system for compensating for deterioration with long-term durability.