论文部分内容阅读
本文利用人工神经网络建立发动机点火提前角预测模型,提出显式优化神经网络模型复杂度的方法,并提出一种在将实验数据划分为训练、验证和测试数据集前的数据预处理方法,能有效提高模型的精度。得到的最优神经网络模型,对训练、验证、测试数据集预测的相关系数均为1,模型误差几乎为零。在划分数据集前通过复制原始数据集得到大量的数据样本,并通过多次迭代训练(999次),优化的发动机点火提前角预测模型具有很高的精度。
In this paper, the prediction model of ignition timing of engine ignition is established by using artificial neural network. A method of explicit optimization of the complexity of neural network model is proposed, and a method of data preprocessing before the data is divided into training, verification and test data sets is proposed. Effectively improve the accuracy of the model. The optimal neural network model obtained has a correlation coefficient of 1 for training, verification and test dataset prediction, and the model error is almost zero. A large number of data samples are obtained by copying the original data set before the data set is divided. After several iterations of training (999 times), the optimized prediction model of ignition timing of the engine has high precision.