Abstract:Firstly, the forecast method of unascertained clustering is optimized and then is applied to the thickness prediction of Excavation Damaged Zone (EDZ). Afterwards, combined with the characteristics of underground pressure and the support theory of EDZ, five major factors of roadway, that is, depth, span, the intensity of surrounding rock, rock joint development degree, and excavation basal area are regarded as the discriminant factors for predicting the thickness of EDZ. Based on 17 groups of measured data, a classification model of EDZ thickness and the uncertainty measurement function of each factor as well as its weights are obtained. Meanwhile, the computing formula for the forecasting value is also given. Then, the classification grade for waiting forecast sample is estimated by the unascertained measurement distance, the forecasting value of EDZ thickness is also able to be calculated by combining the average value of EDZ thickness with each classification patterns. With the inspection, the computation results show that the average relative errors of the method are 5.13% and they are 13.61% and 10.17%, respectively by the methods of neutral network and support vector machine, respectively. In order to further test its reliability, the method is used to make prediction on the EDZ in Maluping mine and the predictive value is compared with the measured value as well. The results indicate that predicted value fits measured value quite well; unascertained clustering method is reliable and practical has been proved and it could be applied to the actual engineering.