基于新型RBF神经网络的四缸发动机活塞-轴系仿真研究

Simulation of Piston-Crankshaft System of Engine Based on Improved Radial Base Function Neural Network

将一种新型的RBF神经网络和可视化等技术引入四缸发动机活塞-轴系的动力学建模中,建立了四缸发动机的活塞-轴系的仿真模型。提出的神经网络考虑了发动机运行具有周期性和不同缸存在点火相位差等特点,能重构发动机各缸燃烧气体作用于活塞的压力和其它方法难以再现的由二维雷诺润滑方程计算得到的油膜力,其有效性也被证明。再对神经网络进行训练、模块化并耦合到四缸发动机活塞-轴系动力学模型中,开发了MATLAB/SIMULINK环境下的四缸发动机活塞-轴系动力学仿真模块。这种方法也适合于其它类型发动机建模。

By the use of introducing a new type of Radial Base Function neural networks （RBFNN）and visualization technology into the simulation of the piston-crankshaft coupling dynamic system of a four-cylinder engine, the simulation models of this type of engine were developed. Under the consideration of operation performances of the above four-cylinder engine, the RBFNN proposed could reconstructed the oil film forces calculated by two-dimensional Reynold lubrication equation and pressure applied on the piston by combustion gas in a cylinder, and its validity was demonstrated. Based on coupling the successfully trained and modularized RBFNN into the dynamic equations for the above piston-crankshaft system, the simulation models of this system were developed by SIMULINK in MATLAB. The proposed method can also apply to the modelling of other type engines.