Research On Fault Detection And Diagnosis Method For Measurment System Of Dynamic Positioning

Along with the increase of economic and development of science and technology, there is a growing problem of that energy and resources is becoming fewer and fewer. The ocean resources which can get attention rarely in the past is becoming the object which is bitterly contested among countries. Changing environment of ocean is a big challenge in the resource exploitation. And we can't do anything without advanced technology and equipment.In recent years, there is need an efficient equipment to exploit ocean resource, and the fast development of ship dynamic positioning system (DPS) makes it possible. The rapid development of technology has make DPS become more and more complex. Therefore, how to ensure the stability and security of the DPS has always been the focus of this technology.The technology of fault detection and diagnosis (FDD) is a powerful means to solve the security problem. This paper aims at reach the application of FDD in the measurement system of DPS.The traditional method of FDD is based on the analytical model. However, the measurement system of DPS is complex nonlinear, so there is a big crror in the system model.The traditional method of FDD has always show poor reliability?faulty and missing alarm and poor Robustness. The method of FDD which is based on data driven can avoid the problem of inaccurate ship model. And it is designed by process variable data.In order to facilitate the study of FDD, it is necessary to establish ship motion model and get the ship's position of North East coordinate. According to the position and combine with the sensor's model to calculate data of' sensors in time. In order to solve the problem of inaccuracy statistical characteristics of noise, this paper uses wavelet to deal with it.The positioning state of ship can be regarded as a large state system, and all measurements are change in a small range with noise interaction, so principal component analysis of multivariate statistics can be used to study it. Taking the inconvenient of use two indexes to monitor the DPS into account, this paper uses a mixture of statistics as the basis for fault detection. Because the ship positioning state will chang under the perturbation in the interference by external environmental capacity, the system state changes with time, so the use of Recursive Principal component analysis (RPCA) to update the system model, in order to adapt to the changes of normal state. Recursive algorithms are characteristic matrix of cyclic complex eigenvalue decomposition problems, so this paper recursives update matrix eigenvalue and eigenvector efficiently based on first-order perturbation theory, and at the same time, this paper uses the recursive partial decomposition and recursive diagonalization contribution to diagnose the faults, and uses the mixed contribution rate which is based on the two indexes to isolate fault sensor. Through study several common faults and reconstruct multi-sensor fault, this paper proves that the method can detect and diagnose fault timely.Considering the fault-tolerance, the paper designs a fault tolerant structure. In this paper, the method of fault detection and diagnosis has certain theoretical significance and engineering application value.