Although the behavior of chisel picks in soft rock can be understood within classical Evan's model, recent studies have brought its accuracy into question. Previously, researchers had recognized several deficiencies in the model: symmetrical geometry cutting assumptions, inverse proportionality between cutting forces and uniaxial compressive strength, and the cutting force not reducing to zero when the pick angle fell to zero. In this paper we use an artificial neural network framework to develop a model for predicting peak cutting forces in shale. Artificial neural networks are a powerful computational tool which has gained popularity in recent years. Our methodology utilizes physical rock tests and full-scale linear cutting tests conducted at Colorado School of Mines. The linear cutting machine measures full scale cutting forces while cutting actual rock, featuring a large stiff reaction frame on which the cutter is mounted. The metrics from the linear cutting machine include cutter forces, penetration, specific energy, cutting sequence, cutter speed, angle of attack and cutter tilt. The benefits of this work are more efficient designs of excavation machines such as shearers and road headers, and improved prediction of machine performance in weak to medium strength rock such as shale.
Introduction and Background
Pick cutting tools have been found to be superior to disc cutters when applied to soft to medium strength rock (Bilgin et al., 2012). The behavior of chisel picks and point attack picks (conical picks) can be understood within the framework of Evan's theory (Evans, 84), which is often represented as shown in Equation 1.
Evan's theoretical model predicts maximum cutting forces assuming tensile failure mode based on symmetrical pick geometry, rock properties and cutting depth. The underlying assumptions are that that the penetration of the pick is smaller than the cutting width and that the resulting tensile crack is circular towards the direction of movement. However the reliability and accuracy of Evan's model for predicting peak cutting forces for conical picks have recently been brought into question (Kuidong et al., 2014). Some of the problems with Evan's model include symmetrical geometry cutting assumptions, the counterintuitive inverse relationship in the model between the cutting forces and the uniaxial compressive strength, and that the cutting force does not reduce to zero when the pick angle (cone angle) falls to zero (Figure 1). Although Roxborough (1978) suggests that Evan's model produces good results where tensile failure occurs; Nishimatsu (1972) has demonstrated instances of shear rock failure for low to medium strength rock.