Computer-Aided Calibration Methodology for Spark Advance Control Using Engine Cycle Simulation and Polynomial Regression Analysis

The increasing number of controllable parameters in modern engine systems has led to increasingly complicated and enlarged engine control software. This in turn has created dramatic increases in software development time and cost. Model-based control design seems to be an effective way to reduce development time and costs and also to enable engineers to understand the complex relationship between the many controllable parameters and engine performance. In the present study, we have developed model-based methodologies for the engine calibration process, employing engine cycle simulation and regression analysis. The reliability of the proposed method was investigated by validating the regression model predictions with measured data. From the results it was clear that the engine cycle simulation, which was tuned using both measured and predicted data obtained from more detailed models that consider intake and exhaust pipe flow, was useful enough to alternate with calibration bench testing. An F-test was demonstrated to yield the F-ratio and its false probability at each of all candidate terms in a polynomial equation. Using the false probability, statistically non-significant terms were deleted from the original polynomial equation. It was evident that the developed method could optimize the polynomial regression model with regard to the tradeoff relationship between goodness of fit and the CPU-state. The analysis demonstrated that this method could be applicable to the engine calibration process.