Research On The Multi-sensor Information Fusion Methods Of Vessel Dynamic Positioning System

With the development of dynamic positioning (DP) technology and measurement andcontrol theory, the modern DP system equips with the position, attitude and environmentalsensors which form a multi-sensor measurement system. Information fusion technologyprovides all effective way to multi-source information processing problem. The accuracy andreliability of the measurement system of DP can be effectively improved by using themulti-sensor information fusion technology. This paper studies the application of informationfusion theory in the domain of dynamic positioning system.For the multi-sensor fusion system of DP, ship motion model, sensor measurement modeland filtering algorithm are the basis of the optimal filtering of ship motion state. They directlyaffect the stability and accuracy of filter. The sensor fault detection and fault tolerance are theimportant measures to ensure the reliability and validity of multi-sensor fusion. The fusionmethod determines the accuracy of the fusion system. Therefore, this paper focuses onresearching deeply into the theory of the above five aspects of multi-sensor data fusion of DP.The main works are as follows:The maneuvering characteristics of DP vessels are researched, and a continuous whitenoise acceleration motion model of DP ship is built. The measurement mechanisms ofposition reference systems (PRS) of DP system are studied, and the linear measurementmodels of position reference systems in the North-East coordinate frame are proposed.The sensor mutant fault detection algorithm is studied. According to the different faultconditions some thresholds are set. This makes the sensor mutant fault algorithm to detect thesensor mutant failures, freezing signals and wild points. To overcome the limitation of sensordrift fault detection by voting or median method, based on multi-sensor information fusion asensor drift fault detection algorithm is proposed. By the reconstruction of filter residual withthe fusion information of multi-sensor, the sensor fault detection algorithm based on residualis improved effectively to detect sensor drift fault.To overcome the effects of inaccuracies of measurement noise statistical characteristics,based on the innovation covariance matching principle an adaptive square-root cubatureKalman filter (SRCKF) algorithm is built. The method calculates the adaptive coefficients toreal-time adjustments the measurement noise covariance matrices. For the uncertainty ofsystem model and the theory limitations of the strong tracking filter (STF), the equivalentcalculating method of fading factor of STF is built, and the strong tracking SRCKF and strong tracking adaptive SRCKF algorithms are proposed by combining the SRCKF and adaptiveSRCKF with STF separately.For the problem of processing asynchronous information by time registration method,two asynchronous multi-sensor information fusion algorithms are proposed based on themultiscale estimation theory. Using wavelet decomposition, the measurement models andmeasurements on different scales are transformed to one scale, and then asynchronousmulti-sensor information multiscale optimal fusion estimation is achieved by the synchronousmulti-sensor fusion on the specified scale. The state estimation information on the finest scaleis feedback to other scales by wavelet decomposing. With wavelet reconstruction, the stateestimations on other scales are transformed to the finest scale, and the asynchronousmulti-sensor information optimal distributed fusion estimation is realized by the synchronousinformation fusion on the finest scale.According to the actual needs of the redundant dynamic positioning system, thestructures of synchronous and asynchronous position reference systems and synchronousattitude measurement system are designed. The proposed fusion structures form themulti-sensor information fusion system of DP. With the vessel semi-physical simulationsystem, the validities of the motion model of DP ship, sensor measurement models, the fusionconfigurations of synchronous PRS and synchronous heading measurement system areverified by the simulations.