基于车用柴油机不平衡数据集的故障识别组合模型

基于车用柴油机的不平衡数据集，根据对应故障发生频次高与低，将模型建立对象分为样本丰富的大数据量故障与样本集不完备的小数据量故障两种。面向前者，基于XGBoost (Extreme Gradient Boosting)分类算法构建故障识别模型，面向后者，基于模糊神经网络构建故障识别模型，然后针对两类模型进行参数调节以获得最优效果，并分别建立评估机制。模型评估结果表明，该故障识别组合模型能够较为精确、全面地识别大多数故障种类，是一种对数据量要求不高且总识别率超过80%的多适应性识别模型算法，可作为汽车维保工作中的重要工具使用。

Fault Identification Combination Model Based on Imbalanced Dataset of Automotive Diesel Engines

Due to lack of engine fault identification models in the automotive industry, a combination model for fault identification is proposed based on operating data of automotive diesel engines. Because of different occurrence frequency of the faults, based on the imbalanced dataset of automotive diesel engines, the faults are divided into two types: the faults with a large number of samples and the ones with incomplete samples.For the former, the fault identification model is constructed based on XGBoost (Extreme Gradient Boosting)classification algorithm, and for the latter, the model is built based on fuzzy neural network. Then the parameters in the two models are adjusted to achieve the optimal effect, and the evaluation mechanisms are established respectively. The model evaluation results show that identification accuracy of the model is more than 80%. The proposed multi-adaptive model with low requirement of training data can be used for vehicle maintenance.