基于免疫危险感知的水下接驳盒故障检测方法

针对现有水下接驳盒故障检测中存在仅通过特定的观测指标来判断系统故障，且观测指标的选取以及故障阈值的设定主要依赖人工经验的问题，本文借鉴机体免疫防御机制，提出了一种基于危险感知的水下接驳盒故障检测方法。借鉴机体免疫中的危险理论，在原DCA算法（dendritic cell algorithm）的基础之上，保留DCA算法的信号转换机制，对它的输入信号定义和异常评价方法进行改进以适合水下接驳盒工作环境，实现故障的在线检测。首先，依据变化是系统危险发生的征兆和外在表现的思想，提出了一种基于系统特征变化的危险信号提取方法，以提高DCA输入信号分类的自适应性。其次，针对原DCA算法只对数据项进行异常评价，无法检测系统级故障，提出采用成熟DC的浓度作为系统故障的评价指标。最后，采用水下接驳盒的真实数据进行模拟实验，并与PCA主元分析方法进行性能对比。实验结果显示本文方法不仅能有效检测出渐变故障，且比PCA方法具有更高的准确率，并能更早的发现故障。因此，本文提出的方法在水下接驳盒的故障检测中具有可行性。

Fault Detection Method of Underwater Junction Box Based on the Danger Theory of Immune System

Currently,system-level fault of underwater junction box can be detected only through specific indicators.Furthermore,the selection of indicators and the fault threshold value setting depends on experience.To address these problems,a fault detection method of underwater junction box based on the body''s immune defense mechanism was proposed.By reference to the danger theory of body immunity,the signal conversion mechanism of the DCA (dendritic cell algorithm) was adopted,and the definition of input signals and anomaly appraisal method of the DCA were improved for online fault detection of underwater junction box.Firstly,according to the observation that change is usually the symptom of system in danger,the method of extracting danger signal based on the change of system features was proposed to improve the self-adaptability of input signals classification of DCA.Secondly,to solve the problem that DCA can only detect anomaly data items and cannot detect system-level failures,an appraisal method of system-level anomaly based on cell concentration was proposed.Finally,our method was tested with the gradual failure data of underwater junction box,and compared performance with PCA (principal component analysis).The results showed our method can effectively detect the gradual failures with higher accuracy.Moreover,it can detect fault earlier than PCA.In conclusion,our method is effective in fault detection of underwater junction box.