Development of optimal fuzzy models for predicting the strength of intact rocks

Fuzzy models have been used in a wide variety of applications particularly problems associated with the strength of rocks and rock masses. However, a systematic approach of modeling has not been presented thus far and developing appropriate fuzzy models is usually carried out by trial and error. In this paper a new soft-computing approach is introduced which benefits from searching capabilities of Multi-Objective Genetic Algorithm (MOGA) to develop fuzzy models optimized in terms of complexity and accuracy. The proposed method is then used to find optimal fuzzy models to predict the strength of intact rock specimens under conventional triaxial stresses. In addition, laboratory tests are conducted on specimens of three rock types to evaluate the models. It is shown that a relatively simple model, with few manageable rules, is able to estimate the strength of intact rocks properly and hence may be selected as the best fuzzy model.

► A new approach was proposed to simply find an optimum ANFIS model using the GA. ► In order to consider both accuracy and complexity we applied a MOGA. ► We predicted the strength of intact rocks with the proposed method. ► It is discussed how to select the best fuzzy models from those found by MOGA. ► It is revealed that complex models are not always the best choice.