The performance degradation of airborne fuel pump is multi-stage with stationary-accelerated-stationary degradation pattern. Aiming at performance index having characteristic of non-linear and non-stationary

a combined predicting method based on singular value decomposition（ SVD） is presented. SVD method is used to decompose and extract trend and de-trend items of original data. Least square support vector machine（ LSSVM）model of trend and de-trend is build. Particle swarm optimization（ PSO） is applied into adaptive selection of the best parameter of LSSVR. Each prediction result is summed to obtain result of performance data prediction. The proposed method is applied to the experimental data. The analysis results show that the proposed approach can effectively extract the trend information and achieve prediction of performance data.