Abstract: Projection pursuit is an efficient uncertainty analysis method in solving complex stability problems for surrounding rock. Its result is affected and controlled by the projection direction. To find an optimum projection direction vector and reflect the uncertainty of the process, we adopted an opposition-based learning fruit fly optimization algorithm (OBLFOA) to modify the projection pursuit evaluation model for surrounding rock stability. The classification threshold of projection pursuit is difficult to be divided in some special cases. In view of this, we constructed the normal cloud models for different grades of surrounding rock by using the optimal projection value according to the optimum projection value of measured samples. To determine the reliability of the model evaluation results and the fuzziness of the evaluation process, we introduced the fuzzy entropy E as the auxiliary parameter with the evaluation result L forming a two-dimensional evaluation mode (L, E). Since the evaluation indexes of surrounding rock stability are varied and the evaluation system is not unique, we selected two cases with different evaluation index systems for application and compared with other methods. Results indicated that the proposed model is feasibile and effective and the evaluation results are objective and accurate.

Key words: surrounding rock stability, opposition-based learning fruit fly optimization algorithm, projection pursuit, backward cloud model, fuzzy entropy