Study of prognostics for spacecraft based-on particle swarm optimized neural network

As the number of spacecraft growing and the spacecraft playing a significant role in many aspects, it is necessary to research the prognostics in order to find the spacecraft fault in advance. This paper puts forward a method of particle swarm optimized neural network used to realize prognostics for spacecraft considering the importance of prognostics in the field of PHM study. Firstly based on the principle and operation traits of the spacecraft, the fault modes and fault characteristics are analyzed to obtain the feature that the prognostics can be achieved through the telemetry parameters of the spacecraft. Secondly the neural network is invoked to approximate and model the key telemetry parameters of the spacecraft due to the neural networks high nonlinearity. After fixed the input and hidden layers nodes, the neural network is optimized particle swarm optimization algorithm with considering the parameter slow convergence in the neural network and the specialty of particle swarm optimization algorithm. Thirdly times series method is incurred to predict the data of the telemetry parameters while there access the real-time downloading of telemetry data. Furthermore the predict data are contrasted to the outputs of the particle swarm optimized neural network to inspect whether the relations between the output layers and the input layers of the neural network is broken. Then the prognostics for spacecraft are achieved with the inspection information. Finally an experiment using there years telemetry data with fault data in the power system of a spacecraft is set up to verify the feasibility of the method proposed in this paper, and the experimental results prove the effectiveness of the particle swarm optimized neural network in prognostics for spacecraft.