In order to effectively control the safety risk of helicopter flight training, a safety risk assessment index system for helicopter flight training based on the "man-machine-environment-management" (MMEM) theory was established and a method based on order relation analysis method-complete ensemble empirical mode decomposition with adaptive noise (G1-CEEMDAN) was proposed to determine index weight. G1 was used to determine the subjective weights, and CEEMDAN was used to determine the objective weights. Afterwards, the two methods were combined by the principle of minimum relative information entropy to determine indexes combination weights for safety risk assessment of helicopter flight training. A typical helicopter flight training institution was taken as an example for analysis. The results show that the adverse weather has the greatest impact on the system safety. The proposed method is simple and practical, and it can effective in assisting the institution to implement risk control and risk management purposefully.