基于BP神经网络的柴油机燃烧特征参数前馈预测模型

CA50是柴油机缸压反馈控制技术中的反馈变量,对柴油机的性能有重要的影响。在一台六缸高压共轨柴油机上研究了喷油正时与CA50关系,以及CA50对柴油机经济性和排放的影响。为探究基于神经网络的前馈控制在缸压反馈控制中运用的可行性,建立了通过不同的燃烧边界条件预测CA50的BP神经网络预测模型,进行原机试验得到CA50对发动机性能影响的系列试验点数据。选取190个不同边界条件的试验点作为模型的总样本,其中用于前期神经网络训练的样本125个、用于检测神经网络泛化能力的测试样本65个。结果表明基于BP神经网络的预测模型在误差允许范围内,能较为准确的通过边界条件预测CA50,可以满足柴油机缸压反馈技术中前馈控制的要求。

A Kind of Prediction Model of Characteristic Parameters of Combustion on Diesel Engine Based on Back-Propagation the Neural Network Technology

CA50 is a vital feedback variable in close-loop control strategy of diesel engine based on in-cylinder pressure, whose value effects the performance of diesel engine. The relationship between main injection timing and CA50 is studied in a high pressureScommon railSdiesel engine, and the influence of CA50 on the performance of engine is also studied in this paper. This study select 190 testing points as the total sample , of which including 125 testing points as training sample and 65 testing points as the testing sample , using BP neural network built up an prediction model of CA50 . Result shows that the model is accurate within theSerror-allowedSrange, and it meets the needs of feed-forward control basically.