A Prediction Model for Specific Energy Required by Point Attack Picks Based on a Hybrid Evolutionary Machine Learning Technique

Full-scale linear rock cutting experiment is the most effective method for determining specific energy. Collecting samples for full-scale experiment is usually time consuming, costly, and sometimes impossible. In order to deal with this disadvantage, many researchers have developed models for specific energy estimation. However, those models fail to include all of the cutting parameters that significantly contribute to the value of specific energy (i.e., depth of penetration, cut spacing, ratio of cut spacing to depth of cut, tip angle, attack angle, and skew angle). Investigating the relationships between intact rock properties, cutting parameters, and specific energy required by point attack picks in relieved cutting mode is the objective of this study. A relatively large database including records of uniaxial compressive strength, Brazilian tensile strength, the aforementioned cutting parameters, and specific energy was established. An evolutionary machine learning technique, which is a combination of gene expression programming and particle swarm optimization, was used to fit a non-linear function over the data. The values of coefficient of determination and mean squared error associated with the non-linear model were improved by almost 30% and 60% compared to those associated with the linear model, respectively. The non-linear model included all of the significant cutting parameters.