ارزیابی ثبات حفره های ورودی زیرزمینی با استفاده از چند رگرسیون سازگاری و رگرسیون لجستیک

The mining industry relies heavily on the use of empirical methods and charts for the design and assessment of entry-type excavations. The commonly adopted empirical design method, commonly referred to as the critical span graph, which was specifically developed for the assessment of rock stability in entry-type excavations, was based on an extensive database of cut and fill mining operations and case histories in Canada. It plots the critical span versus the rock mass rating for the observed case histories and has been widely accepted for an initial span design of cut and fill stopes. Different approaches, either based on classical regression and classification statistical techniques or even the supervised machine learning methods, have been proposed to classify the observed cases into stable, potentially unstable and unstable groups. This paper presents a new assessment approach which combines the use of a multivariate adaptive regression splines (MARS) approach and the logistic regression (LR) method. The proposed MARS_LR model can capture and describe the intrinsic, complex relationship between input descriptors and the dependent response without having to make any assumptions about the underlying relationship. Considering its simplicity in interpretation, predictive accuracy, its data-driven and adaptive nature plus the ability to map the interaction between variables, the use of MARS_LR model in evaluating stability of underground entry-type excavations is promising.