基于改进K＿means的发动机状态评估方法

A Method of Engine State Evaluation Based on Improved K-means Algorithm

针对目前在缺少先验知识和小样本条件下,进行发动机状态评估难度大的问题,本文中提出了一种基于改进K_means聚类算法的发动机状态评估方法。该方法利用K_means算法的基本原理,避免了评估过程中主观因素的影响;提出相关性指标,对算法进行改进,根据特征参数性质赋予其相应权重;提出最小方差启发式初始聚类中心优选方法,避免小样本条件下初始聚类中心选择中孤点和噪声点的干扰;并充分利用Bootstrap小子样统计方法削弱了试验样本的随机性对评估模型的影响。最后通过实例评估,验证了该方法的可行性和有效性,与传统方法相比,该方法具有更强的客观性与稳定性。

In view of the present difficulty in engine state evaluation under the condition of prior knowledge absence,an evaluation method of engine state based on improved K-means clustering algorithm is proposed in this paper. This method uses K-means algorithm to avoid the influence of subjective factors during evaluation. Correlation indicators are put forward,the algorithm is modified,and the corresponding weights are given according to the properties of feature parameters. A selection method of initial clustering centers with minimum variance heuristic algorithm is used to avoid the interference of acnodes and noise points during initial clustering center selection in the condition of small sample size,and Bootstrap small sample statistical method is adopted to weaken the effects of the randomness of test samples on evaluation model. Finally,the feasibility and validity of the method are verified by real example evaluation,indicating that compared with the traditional method,this method is more objective and stable.