Aimed at the problems that the fitting method is limited by the system model and the model order in establishing a small deviation state variable model of aircraft engine, a method based on multiple-improved-chaotic fruit fly optimization algorithm (MICFOA) is proposed. Firstly, the method is divided into two sub-processes: first thing is to optimize the system matrix, input matrix, and to find the optimal results, and then is to optimize the output matrix and the transfer matrix. Simultaneously, work is done according to the principle that the SVM's dynamic response is consistent with the nonlinear model's, fitness functions are constructed that are unaffected by the variable value domain. Secondly, synergisitic sub-population strategy and chaos mapping strategy are introduced into FOA to improve the diversity of fruit fly populations by using the adaptive adjustment strategy introduced to balance the relationship between global search and local search to avoid premature convergence. Finally this method is used to establish a turbo-shaft engine's SVM, and to design LQ/H∞ disturbance-rejection controller. The simulation results show that MICFOA can improve the accuracy of 5-10 orders of magnitude compared with FOA, the SVM has good dynamic and static performance, and the newly built model is consistent with the nonlinear model.