Performance prediction of impact hammer using ensemble machine learning techniques

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%.