蚁群初始化小波网络的汽油机油膜参数辨识研究

建立瞬态工况小波网络发动机油膜模型,利用蚁群算法对小波网络参数进行初始化寻优,将其作为小波网络参数初始值,以提高小波网络的训练速度和误差精度,并基于该网络模型测试了发动机空燃比瞬态过程,然后利用瞬态工况试验数据进行了仿真,并与台架试验实际数据进行对比。结果表明,基于蚁群算法初始化小波网络模型能有效地辨识发动机瞬态工况油膜参数,高精度地逼近空燃比瞬态过程,不仅具有较强的泛化能力,而且大大缩短了训练时间。蚁群初始化小波网络适用于油膜参数辨识,本研究为发动机瞬态工况空燃比的精确控制奠定了基础。

Dynamic Identification of Transient Fuel Film Parameters of Gasoline Engines Based on ACO Initialized Wavelet Networks

A transient fuel film identification model for gasoline engines was built based on wavelet networks. The ant colony optimization (ACO) was used to initialize wavelet networks parameters as the initial values for the model so as to improve its convergence and precision. Transient air-fuel ratio was tested with this model, simulated with experimental data and compared with the actual air-fuel ratio. Results show that the model initialized with the ACO can effectively and precisely identify the transient fuel film with faster convergence and better generalization. Therefore, such model establishes a foundation for accurate control of transient air-fuel ratio of gasoline engines.