Neuro-Genetic Optimization of Transmission Ratios for Automotive Performance and Emissions 2004-01-2932
Nowadays In product development, the virtual prototyping has a vital role. It is becoming more important to design and test products by simulator packages before building physical prototypes. The advantages are lower expenses, shorter time to market, more iteration in design cycle, and thus higher quality of products.
Vehicle power train is complex integrated systems, which need to be designed for numerous thermodynamic, economic and environmental factors. The simulation of this system is a difficult task because a proper simulation includes complexity and non linearity that tend to be computationally expensive. In addition there is a great deal of uncertainty and variability in the component data, and in the evaluation of potential objective criteria. To be acceptable to a large public a vehicle must be a tradeoff between fuel economy, performance and emissions.
In this project a computer simulation package of automotive Power train was developed as a tool for design engineers, called VPTA (Vehicle Power Train Analyzer). Because of design study for complicated mechanical system, there is need to solve system in the loop, so the simulation must be very rapid. The neural network modeling is proper tools for such a simulation. A new Neuro-Genetic evolutionary multi-objective optimization algorithm has been developed and applied to optimize gear box ratio, but the OOP nature of developed code made it easy to include various problems including vehicle drive train layout and configurations.