Performance prediction of impact hammer using ensemble machine learning techniques

Abstract

Various mechanical excavation systems such as roadheaders, tunnel boring machines (TBM), and impact hammers, are commonly used in mining, tunneling, and civil engineering projects. However, in recent years, the application of impact hammers in hard rock excavation has been on the rise, especially in fractured geological formations. Impact hammers offer some advantages over drill and blast systems like selective mining, mobility, less over excavation, minimal ground disturbance, elimination of blast vibration, reduced ventilation requirements, and compared to TBMs, low specific energy and small initial investment cost. Prediction of the net machine production rate in terms of net (instant) breaking rate (NBR) plays an important role in estimation of completion time, schedule and cost of the projects. Performance prediction models has been developed based on field data where Impact hammers were used in tunneling operations. While some models are based on statistical analysis of field data, a fewer subset have been developed using artificial neural network (ANN). In this study, 121 data sets, including machine production rate, uniaxial compressive strength (UCS), rock quality designation (RQD), excavator power (P), and weight of excavator (W) have been compiled and using a CRISP-DM data mining technique along with principal component analysis (PCA), a new model for prediction of the impact hammer performance has been introduced with R2 of over 85%.