12CrMo9-10
2011年5月2日 by admin
【Abstract】 Engine is the most important system which determines the driving performance of the vehicle. In this thesis, we mainly study the performance improvement strategies and optimization methods for a 2.0L SI engine. By adopting the devices of variable intake runner length and intake variable valve timing (VVT), and optimizing the intake chamber volume and exhaust valve timing etc., the engine performance has been enhanced significantly.Engine is a complex system characterized by mutiple coupling and non-linearity. Taking the mentioned engine as an example, with the use of new variable actuating devices, such as variable intake runner length and intake VVT, the engine design, optimization and calibration become more complicated, that requires largely increased the resources, time and cost. Therefore, it is necessary to develop multi-variable optimization tools to make the realization of engine design and optimization more efficiently.In this thesis, the engine performance optimization technologies based on Artificial Neural Networks and Genetic Algorithm are studied respectively. The optimization tools based on the coupling of MATLAB/Simulink and GT-Power software are developed.The mainly content in this research are as follows:1. The simulation model of this engine was established by using GT-Power software, and its effectiveness was validated with the engine performance test data. The possible technical approaches to improve engine performance have been studied using simulation, including the variable intake runner length and the intake VVT technologies.2. The engine optimization method based on ANN has been investigated and implemented. First, 3000 GT-Power simulation cases were generated using the Latin Hypercube Sampling (LHS) Algorithm, and input into the GT-Power engine model through Simulink-GT-Power interface. After trained ANN models with engine performance data, Torque, Power, BSFC, Knock Index, Maximum Pressure, and Exhaust Temperature ANN model were obtained respectively. Then, another 200 GT-Power simulation cases were generated with LHS Algorithm and the corresponding performance simulation data were used to test the trained ANN models. The models with best generalization ability were selected. During the optimization process, the torque/power ANN model is the objective function, and the rest ANN models serve as the nonlinear constraints. Considering the practical use and limitation of the engine design and operation, some parameters have to be fixed for the physical engine construction and various operation conditions. The fixed values of those parameters should be selected based on the optimization results, and be used in the following optimization analysis. Then executing optimization with the ANN models again, and the optimal operation parameters and corresponding engine performance were found.3. Another engine performance optimization method was also established by coupling Genetic Algorithm and GT-Power simulation. The Genetic Algorithm optimization was implemented in the MATLAB environment. The GT-Power engine simulation was called repeatively by the Simulink program to iterate the optimal design and operation parameters. After fixed the parameters limited by the engine structure constraints and operation constraints, the optimization was performed once again. Finally, the optimization results from this method were compared with the optimization results obtained from the ANN models. Both of the two optimization methods can achieve satisfactory optimization results, and the optimized engine performance can meet its expected optimization target. However, there still exist some different aspects between the two optimization methods. Users should select the appropriate optimization method according to their practical needs with complete understanding of their merits and limitations.
Tags: 12CrMo9-10
0 Comments, 0 Trackbacks