Abstract: By integrating the quadratic regression analysis of orthogonal composite design and the statistical significance testing, a discriminant method to describe the effect degree of various factors on structural eigenvalue is established. By introducing such a method into the reliability analysis process of response surface for the stability of the surrounding rock of tunnels, an optimum approach for identifying the degree of importance with respect to basic ingredients for state function is developed, which can eliminate the drawback of being incapable of assessing the importance degree of basic ingredients on the structural state when establishing a series of response surfaces for the state function. A MATLAB-based search technology is utilized to solve the reliability index of the improved response surface by means of the optimum approach developed for the state function, which figures out the problem of result distortion. On the basis of the integration between both the above methods, a response surface analysis technique with relatively high computational efficiency is then proposed for the reliability evaluation of the surrounding rock of tunnels in complicated stratum. The proposed technique is illustrated by a certain practical example, in which the running process and operating procedure are demonstrated, and then several aspects related to this technique such as the computational efficiency and accuracy of the results are also verified, showing its validity and applicability.