Data Driven Automotive Engine Modelling and Calibration using Artificial Neural Network

Abstract

The evaluation of automobile diesel engine models equipped with BS6 technology and above is still in its early stages, making both physical testing and predictive modeling novel areas of research. This study aims to develop an Artificial Neural Network (ANN) model for predicting the performance parameters of automobile diesel engines, including brake power (BP), brake-specific fuel consumption (BSFC), brake thermal efficiency (BTE), nitrogen oxides (NOX), volumetric efficiency (VE), and exhaust gas temperature (EGT). Currently, engine modeling and calibration rely on expensive tools that require skilled manpower and licensing costs, limiting accessibility for research. Additionally, the accuracy of conventional modeling tools is typically around 60%, which can be improved to 90% through parameter optimization. Furthermore, engine modeling is a time-consuming and repetitive process across different vehicle models and variants, necessitating more efficient alternatives. To address these challenges, this research focuses on two key objectives: 1. Developing a data-driven automotive diesel engine model for Hardware-in-the-Loop (HiL) simulation using ANN. 2. Predicting engine performance and emissions under real-world drive cycles. The Feedforward Backpropagation Network is trained using the Levenberg-Marquardt algorithm to achieve optimal performance. Experimental validation shows that the trained ANN engine model accurately predicts engine emissions and performance within the provided drive cycle, demonstrating its potential as an efficient and cost-effective alternative to traditional modeling techniques.