A new hybrid method for predicting ripping production in different weathering zones through in situ tests

• Main Authors: Mohamad, E. T., Koopialipoor, M., Murlidhar, B. R., Rashiddel, A., Hedayat, A., Jahed Armaghani, D.
• Format: Article
• Published: Elsevier B.V. 2019
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://eprints.utm.my/id/eprint/91257/; http://www.dx.doi.org/10.1016/j.measurement.2019.07.054

Due to blasting's limitations, ripping as a breaking technique of rock mass is one of the most popular methods in mining and civil engineering applications. The typical practice is that ripping is used for loosening the soils and weak rocks while blasting is used for breaking stronger rocks. With the regulatory restrictions on blasting, there is a growing interest in ripping rocks that traditionally have been blasted. The ripping is typically cheaper than blasting but predicting whether ripping can be done on a particular rock and the estimation of the excavation cost are challenging and a function of rock properties. This study aims at predicting the ripping production based on an extensive database obtained from three sites in Malaysia. The site observations for production rate and the relations with the sandstone and shale rocks were presented. In situ observations/tests (sonic velocity, joint spacing, Schmitdt hammer, weathering zone) were conducted by the site engineers and the results were used as input data for training and proposing a new model for estimating the ripping production. Many hybrid particle swarm optimization-artificial neural network (PSO-ANN) models were created and the best model was identified based on a ranking system. Then, the best PSO-ANN model with coefficient of determination values of 0.982 and 0.978 and root mean square error values of 0.038 and 0.045 for training and testing datasets, respectively, was selected and introduced to predict ripping production. This study documented that the new PSO-ANN achieved higher performance than the ANN method.