An FSPCA-BP algorithm and its application to aircraft overhaul makespan prediction problem

Abstract

The implementation strategy for meticulous production management of aircraft overhaul is facing barriers, such as the lack of an effective method that can achieve accurate prediction of makespan in the whole aircraft overhaul process. In this paper, a novel prediction method for makespan of aircraft overhaul by extending feature selection principal component analysis (FSPCA) and backpropagation (BP) neural network (FSPCA-BP) is proposed. Firstly, the FSPCA algorithm is developed to achieve feature selection and dimension reduction simultaneously. Then, the principal components factors obtained from the FSPCA algorithm are used as the input vectors for the BP neural network to predict the makespan of aircraft overhaul. Finally, the proposed method is demonstrated and tested in an application scenario of a partner company and is further compared with PCA-BP and BP approaches, respectively. The result shows that the FSPCA-BP approach has higher accuracy than PCA-BP and BP approaches in aircraft overhaul makespan prediction. Finally, the managerial implications of the proposed method for shop-floor departments are discussed in detail.