基于混合混沌分类的水下航行器异常压力检测

Based on the Classification of Hybrid Chaotic Abnormal Pressure Detecting Underwater Vehicle

随着水下航行器随下潜深度的逐渐增加,所受的压力也逐渐增加,当下潜至一定深度时,压力出现异常将影响水下航行器的稳定运行。传统的水下航行器异常压力检测方法存在无法克服水下航行器压力多样性和不可预知性问题的弊端,提出一种采用混合混沌思维的水下航行器异常压力检测方法,将一维时序维数扩展至更高维的空间中,挖掘出充分的水下航行器压力信息,恢复原吸引子的动力学特征,依据此刻状态获取下一时刻的状态,对相空间进行重构,合理选择时间延迟与嵌入维数,为时间序列的预测提供可靠基础。对水下航行器样本集进行多次不同k值的k-均值聚类处理,利用评价函数对聚类结果的聚集性及分离性进行评价,获取最优聚类结果,得到小规模类样本,即为水下航行器异常压力样本。仿真结果表明,所提方法为提高水下航行器检测精度提供了依据。

An abnormal pressure detection method based on hybrid chaos for underwater vehicles is proposed. The dimension of one-dimensional time series is extended to the higher dimensional space,the sufficient pressure information of the underwater vehicle is mined,and the dynamics characteristics of original attractor are restored. According to the current state,the next state is obtained,and the phase space is reconstructed. The reasonable choice of time delay and embedding dimension are performed to provide the reliable basis for the time series prediction. The sample sets of underwater vehicle are made repeatedly K-means clustering processing with different K values. The clustering and separation of the results of clustering are evaluated by using the evaluation function. The optimal clustering results and small scale sample are obtained,which are the abnormal pressure samples of underwater vehicle. The simulation results show that the proposed method can improve the detection accuracy of underwater vehicle.