Temperature Prediction of Solar Array Vacuum Heat Test Based on Deep Belief Network

Abstract

The vacuum heat test is an indispensable test item in the development of spacecraft. In the vacuum heat test, the correct interpretation and prediction of temperature data is very difficult and very important. This paper proposes a thermal test temperature prediction method based on deep belief network. The method includes the construction of deep belief network, layer-by-layer pre-training of model parameters, and adopts the Adam algorithm to replace the traditional steepest gradient descent method to complete the supervised fine tuning process, so that it converges to the method described in this paper can effectively completoptimal solution more quickly. Finally, the case analysis is carried out with real test data. The results show that the e the thermal test temperature prediction.