Application of Levenberg-Marquardt Algorithm in Neural Network Based Actual Engine Torque Estimation and Monitoring

This paper introduces a novel approach for predicting gasoline engine torque using a neural network model based on the Levenberg-Marquardt algorithm (LMA). The model generates an engine torque map by considering engine speed and throttle angle as input variables. To assess its reliability, the technique is validated using a dataset comprising 23,450 real-world gasoline engine data points for training, validation, and testing. The results demonstrate the model's ability to accurately predict and map engine torque at short time intervals. This method significantly enhances the efficiency of monitoring systems in automotive vehicles' functional safety domain by providing precise torque map values through appropriate interpolations.