A gear shift model for tricycle emissions and fuel use modelling: An application of voting based neural network integration

• Main Authors: Biona, Jose Bienvenido Manuel M., Culaba, Alvin B., Tan, Raymond Girard R.
• Format: text
• Published: Animo Repository 2006
• Subjects: Two-stroke cycle engines—Exhaust gas; Neural networks (Computer science); Mechanical Engineering
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/6621
• Institution: De La Salle University

Importance of vehicle fuel use and emissions models in transport policy development are starting to be recognized in developing countries. One of the pressing issues confronting major cities in the Philippines is the significant amount of two stroke tricycle emissions. A fuel use and emissions model for tricycles is currently being developed to provide a scientific tool to guide policy and technology development in this sector. It consists of three sub models namely: driving and load pattern generator; gear shift model; and emissions and fuel use reference model. Current gear shift models in emissions and fuel use modeling assumes an ideal gear shift logic which may not be representative of the real world behavior of drivers. Gear shift models are sequential in nature and thus are subject to error propagation issues reducing the accuracy of feed forward neural network systems. A self-checking multi-model neural network system utilizing a voting based integration approach is introduced to address this issue. The model was trained utilizing actual tricycle gear shift data gathered from various routes in Metro Manila. Adaptation of the model showed a significant increase in simulation accuracy compared to the individual feed forward neural network gear shift models