基于信息增益率的WNB水下机器人故障分类

提高故障诊断能力对于确保水下机器人系统的稳定运行具有重要意义,故障分类是目前水下机器人故障诊断所面临的一个重要问题。针对水下机器人推进器系统数据特征,提出一种基于信息增益率的加权朴素贝叶斯故障分类算法。首先,计算故障训练样本的先验概率,将各属性的信息增益率作为权值；其次,构建基于增益率加权的朴素贝叶斯分类模型；然后,对检测的故障数据利用分类模型获取具有最大后验概率的故障模式,实现故障分类。与朴素贝叶斯算法和决策树算法相比,仿真实验结果表明基于信息增益率加权的朴素贝叶斯算法的分类成功率更高,能够有效地实现水下机器人的故障分类。

WNB algorithm Based on Information gain Ratio for AUV Fault ClassificationZhou Yue1,2,3Xing Yanyan1,3 Guo Wei3

It is very important to improve the fault diagnosis ability to ensure the stable operation of the autonomous underwater vehicle (AUV) system. Fault classification has recently been the focus of fault diagnosis for AUV. A weighted naive Bayesian (WNB) algorithm based on information gain ratio is proposed to classify the fault patterns according to the data feature of AUV propeller system. Firstly, the prior probability and each attribute information gain ratio of AUV fault training samples are calculated. Secondly, the WNB model is built based on the information gain ratio. Then, the classification model is used to realize the fault classification by obtaining the maximum posteriori probability of the fault pattern. The Simulation results demonstrate the feasibility and effectiveness of the proposed algorithm, which has higher classification success rate, compared with naive Bayesian algorithm and the decision tree algorithm.