Robust Identification Algorithms for Adaptive Engine Controls

The current engine systems to achieve extremely low emission has a wide range lambda sensor positioned upstream and a switching lambda sensor downstream a catalyst. The system is required to maintain the output of the switching lambda sensor to optimal target value under all engine load and catalyst aging conditions in order to optimize the conversion rate of catalyst. Therefore, a standard STC (self-tuning controller) and the robust adaptive controller composed of an identifier, a predictor and a sliding-mode controller are applied to the system at the beginning of this research. However, the standard STC caused the drift phenomena of adaptive parameters and could not provide sufficient control performance. The identification algorithm of the robust adaptive controller cannot correctly identify the gain characteristic of the system by the influence of frequency weight characteristic of RLS (recursive least square) algorithm. Consequently, the identification algorithms of two adaptive controllers are modified to avoid these issues. As a result, the control performance of the output of the switching lambda was improved and the emissions from the engine were dramatically reduced to the level meeting LEV-II emission standard in California. These adaptive controls were applied to all mass production vehicles of Honda