Dynamic Indicated Torque Estimation for Turbocharged Diesel Engines Based on Back Propagation Neural Network

Abstract

An indicated torque estimation model is presented for turbocharged diesel engines considering both steady-state and transient operating conditions. The proposed model consists of two submodels: a steady-state indicated torque model and a transient torque coefficient model. By combining the steady-state torque with the transient torque coefficient from the two proposed submodels, dynamic indicated torque is obtained. The transient torque coefficient is calculated by training a designed back-propagation neural network (BPNN) using transient test data obtained from the designed experiments based on a DEUTZ BF6M1015 turbocharged diesel engine bench. Only the engine speed, the cycle fuel quantity and the intake air pressure are needed for dynamic torque estimation. The generalization capacity and dynamic torque estimation accuracy of the torque estimation model are validated. The maximum error of the estimated torque is within 8% while the average error is within 2% in both fuel step change and slow change conditions.