The accuracy of aero-engine model-based fault diagnosis is directly influenced by the matching accuracy between the self tuning model and real engine.The parameter selection of support vector regression by adaptive genetic algorithms was proposed and applied to the real-time on-board self tuning model

thus effectively improving the matching accuracy.After analyzing the influences of parameter selection of support vector regression(SVR) on the performances of the SVR model

optimal parameters in the given region were selected in the search area by genetic algorithms.Different methods such as back propagation neural network(BP-NN)

SVR

adaptive genetic algorithms-least square support veitor regression(AGA-LSSVR) were compared in the compensation of aero-engine model error;the results indicate that the proposed AGA-LSSVR is most feasible.