摘要 |
PROBLEM TO BE SOLVED: To make control convergency and stability compatible with a higher dimension, and implement stable operation at knocking limit to minimize an output loss. SOLUTION: Knock strength knklvl is acquired (S150). When it is strong knock, the whole learning value is significantly updated on a retard side to clear out a partial learning value (S152), and then the learning is shifted to the whole learning (S160). If it is intermediate knock, the learning is shifted to the whole learning (S160) when the whole leaning is not finished, and the learning is shifted to partial learning to update the partial learning value on the retard side (S161) when the whole learning is finished. If it is weak knock, a knock F/B value is updated on the retard side (S162). If there is no knock, the learning is shifted to the whole learning to update the whole learning value on an advance side (S160) when the whole learning is not finished, the knock F/B value is updated on the advance side (S162) when the whole learning is finished and during a retard control, and the partial learning value is updated on the advance side (S161) when the retard control is finished. COPYRIGHT: (C)2007,JPO&INPIT |