Modelling the relationship between rock properties and the performance of a rock cutting machine

摘要

Quarrying and open-pit-mining techniques have recently been based on continuous surface mining machines which operate in a wide range of hardness conditions, bringing benefits not only of a reduction in the overall mine machinery but also of higher production rates. In order to take full advantage of these new operational methods, a mathematical model is required that is able to describe and predict the behavior of a rock cutting machine in terms of the heterogeneous properties of the rock material to be extracted. The rock properties used to discriminate the different materials are described by geomechanical test results, petrographical study attributes and macroscopic analysis results. The performance of the machine is described by the torque of the diesel engine, pressure of the hydraulic system and speed of the surface miner, measured during cutting work in a pegmatite mine in the north of Portugal. Two model structures are considered. The first model is a first-order Takagi-Sugeno Fuzzy Inference System (ISFIS), where the input variables (rock properties) are fuzzified by using Gaussian membership functions. The consequent part of each fuzzy rule describes each of the measured machine parameters (out put variable) as a linear combination of the rock properties. A clustering technique is applied to the data set in order to extract the fuzzy if then rules as the centers of the rock properties groups (clusters) found. The second model structure is a zero-order Takagi-Sugeno Fuzzy Inference System, which can also be viewed as a Radial Basis Function Network (RBFN). The centers of the radial basis functions are found, in a recursive way, as the input-output training data pairs such that the approximation error is maximally reduced in each selection. Both models give good results, although the best are obtained with the TSFIS model with only two fuzzy if-then rules. These relatively new approaches show that it is possible, with reasonable effort, to model the nonlinear multivariable relationship between the machine's performance parameters and the rock material type.