Bionic Algorithm And Its Application To Slope And Excavation Engineering

This paper presents the concept of bionic algorithm (BA), and discusses the foundational principle of BA. The important modifications of the usually used BAs, i e. neural network (NN) model, genetic algorithms (GA), ant colony algorithm (ACA) and fuzzy system, are given A lot of uncertainties, for example random uncertainty and fuzzy uncertainty, are existed in the slope and excavation engineering. Therefore, the new design idea and methodology are suggested in terms of BA's principle, and several novel and effective analysis methods are developed The main achievements are as followings 1. For enhancing the speed of search and improving the local convergence, several new models or algorithms are developed A trust-region-based N7N are obtained by leading the trust- region algorithm into back propagation NN The T-S fuzzy system driven by NN is discussed, and an efficient procedure is developed. Three new methods of hybrid genetic algorithm (HGA), i.e. modified complex GA, feasible direction GA and trust-region GA, are proposed. In these HGAs, the grid search technique is used to produce initial population, the elite memory cell is set up to store excellent chromosome, the modified complex method is led into genetic algorithms for improving the local convergence, and the competition mechanism is led into genetic operations. A heuristic ant colony algorithm is presented to solve sequential multi-stage decision problem. 2. Based on the assumption of general slip surface, a general formula of safety factor of slope is presented in which the moment and forces exerted on the slices exactly satisfy the equilibrium condition, and a new generalized function of inter-slice force is given. The heuristic ACA and new HGAs (i e modified complex GA and feasible direction GA) are used to seek the critical slip surface, and the results of several case studies demonstrate that the present algorithms are practical and reliable. Based on the T-S fuzzy system driven by neural network, a fuzzy neural network model used to estimate the slope stability is proposed depending on 80 slope cases, and the predicted results show that the accuracy is obviously higher than other methods 3. A unified optimization model of slope reliability analysis is obtained, and the model抯 structure and the form of limit state equation do not vary with the analysis methods of slope stability. The decision variables involved in this model are composed of two different variables (i e uncertainty variable and certainty variable), and a class of stage HGAs is proposed The results obtained from slope case studies using the stage HGAs show higher accuracy than the traditional methods The compared computation results indicate that the critical slip surfaces sought respectively according to the minimum reliability index /3 and the minimum safety factor F, are considerably different, and more complex slope shows stronger differences. The paper also discussed the influence of statistical features of random parameters c and q on slope reliability index /3. 4. A general slice method is presented for the interior stability analysis of complex nailed wall, in which the shape of slip surface has no restriction and all forces exerted on slices rigorously satisfy the equilibrium condition of force and moment. The interior critical slip surface of nailed wall is sought using the new HGAs. Meanwhile, a new analysis method of the interior stability is developed for simple nailed wall. The two-stage optimization m...