基于模糊相关向量机的UUV中缓变故障检测方法  

An Approach Based on Fuzzy Relevance Vector Machine to Deal with Ramp Fault Detection in UUV

作　　者：赵博[1] 王天宇[1] 张润峰 ZHAO Bo;WANG Tianyu;ZHANG Runfeng(Anshan Industrial Technology Research Institute of Harbin Institute of Technology,Anshan 114000)

机构地区：[1]哈尔滨工业大学(鞍山)工业技术研究院,鞍山114000

出　　处：《计算机与数字工程》2023年第4期838-844,共7页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:51909048,52071121,41627801)资助。

摘　　要：作为海洋探索的重要运载平台之一,水下无人航行器(unmanned undersea vehicle,UUV)中导航系统多数时间以惯性/多普勒计程仪组合方式工作,但仍需要接收GNSS信号以修正累积的导航误差。由于受到无线电信号及卫星信号质量的影响,GNSS信息会存在缓变误差,从而降低UUV的导航精度。针对此问题,论文设计一种基于模糊相关向量机(fuzzy relevance vector machine,FRVM)的状态估计器对新息进行预测,克服了传统方法存在较大检测延时的问题。该方法利用模糊隶属度来模糊化样本,避免过度拟合影响模型的泛化能力。与基于自主完好性监测外推法及最小二乘支持向量机(least squares support vector machine,LS-SVM)的故障检测方法相比,该方法能够更精确地描述新息与量测信息之间的相关性。试验结果表明,针对GNSS的缓变故障,该方法能够显著降低了故障检测的延时,从而提高UUV的导航性能。As one of the important platforms for ocean exploration,the navigation system of unmanned undersea vehicle(UUV)works in inertial navigation system/doppler velocity log integration mode at most time,but it still needs to receive GNSS signal to correct the accumulated navigation error.Due to the influence of radio signal and GNSS signal quality,GNSS may have a slow change error,thereby reducing the navigation accuracy of UUV.In view of this problem,this paper designs a fuzzy relevance vector machine(FRVM)based state estimator to predict innovation,overcoming the problem of large detection delay using the traditional methods.Fuzzy membership is employed to fuzzify the samples to avoid overfitting and affecting the generalization ability of model.Compared with the methods based on autonomous integrity monitored extrapolation and least squares support vector machine(LS-SVM),this method can accurately describe the relationship between innovation and measurement information.The test results show that the proposed method can significantly decrease the time delay of fault detection and improve the navigation performance of UUV.

关 键 词：UUV 组合导航 模糊隶属度 相关向量机 缓变故障 

分 类 号：TB567[交通运输工程—水声工程]