DRIVING CYCLE DEVELOPMENT USING K-SHAPE AND CONVOLUTIONAL NEURAL NETWORK FOR ENERGY CONSUMPTION AND EMISSIONS ANALYSIS

Abstract

One of the key challenges in the automotive industry is enhancing fuel efficiency and reducing emissions while meeting regulatory requirements. Common issues include the lack of context-specific driving cycles, limitations of conventional clustering methods in capturing non-linear driving behaviours, inefficiencies in real-time system integration with Siemens Totally Integrated Automation (TIA) Portal, and delays from traditional data exchange methods, leading to inaccuracies in fuel consumption and emission analysis. This research focuses on integrating MATLAB scripts for fuel consumption and emission analysis with real-time execution in the Siemens TIA Portal. A major focus is on seamless integration strategies between MATLAB and Siemens environments during execution to enhance analysis efficiency. The study involves collecting driving cycle data for Ipoh City using MATLAB Mobile and DC-TRAD and constructing the Ipoh City driving cycle using the K-shape clustering technique, which identifies complex patterns more accurately than conventional clustering methods. Additionally, a convolutional neural network (CNN) algorithm is applied for effective and precise driving cycle development. The research includes a detailed analysis of execution cycle time, fuel consumption, and emissions across both MATLAB and Siemens environments. A significant improvement in execution performance is achieved, with model cycle times in the Siemens environment reduced by over 90%, reaching a maximum of 100 milliseconds compared to 45 seconds in MATLAB. This substantial reduction in cycle time is accomplished without compromising accuracy, as the results from MATLAB are successfully replicated in the Siemens environment, leading to the selection of route 6 as the optimized route for the Ipoh City driving cycle. Addressing challenges related to computational power and system integration for real-time processing, this research outlines strategies to optimize MATLAB scripts for real-time deployment within Siemens systems. Ultimately, this integration aims to provide efficient and accurate solutions for analysing energy consumption and emissions in automotive applications, contributing valuable advancements to the field.