Representation of Engine Data by Multi-Variate Least-Squares Regression

Multivariate least-squares regression techniques have been developed to screen, smooth and interpolate pollutant flows, fuel flow and other parameters of engine performance required for engine control strategy optimization. These methods have been applied to mapping data accumulated at constant speed and load, that is, regression has been performed as a function of air/fuel ratio, EGR and spark timing. In addition, the method has been extended to include engine RPM and torque as additional independent variables, allowing interpolation in the speed/load plane. Concurrently, procedures related to occupancy check for valid interpolation, determination of appropriate order of fit, data set overdetermination and graphical display have been devised. Examples of practical application of regression to engine mapping data are presented. The relative simplicity and accuracy of the regression method make it a valuable tool in the engine control optimization sequence.