Estimation of Engine Torque from a First Law Based Regression Model

A first law based regression model for estimating mean value engine torque on-board a diesel engine is presented. The model uses first law terms across the engine control volume in a regression built from least squares to predict engine torque. Torque information is often required by the engine ECM for torque based control and torque broadcast purposes. In the absence of real-time torque measurement torque estimation is usually achieved through look-up tables or empirical models. Given the increase in engine operating parameters as well as engine operating regimes as a result of emission control and exhaust aftertreatment technologies, accurate torque estimation has become more challenging as well as necessary. The present work suggests that the ‘gray-box’ modeling approach described might generalize better and be more robust than other commonly used empirical approaches such as regression or neural networks using raw engine operating parameters as inputs instead of energy balance terms across the engine control volume.