Details

Autor(en) / Beteiligte

• Nguyen, Hoang
• Bui, Xuan-Nam
• Tran, Quang-Hieu
• Mai, Ngoc-Luan

Titel

A new soft computing model for estimating and controlling blast-produced ground vibration based on Hierarchical K-means clustering and Cubist algorithms

Ist Teil von

• Applied soft computing, 2019-04, Vol.77, p.376-386

Ort / Verlag

Elsevier B.V

Erscheinungsjahr

2019

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Blasting is an essential task in open-pit mines for rock fragmentation. However, its dangerous side effects need to be accurately estimated and controlled, especially ground vibration as measured in the form of peak particle velocity (PPV). The accuracy for estimating blast-induced PPV can be improved by hybrid artificial intelligence approach. In this study, a new hybrid model was developed based on Hierarchical K-means clustering (HKM) and Cubist algorithm (CA), code name HKM-CA model. The HKM clustering hybrid technique was used to separate data according to their characteristics. Subsequently, the Cubist model was trained and developed on the clusters generated by HKM. Empirical technique, the benchmark algorithms [random forest (RF), support vector machine (SVM), classification and regression tree (CART)], and single CA model were also established for benchmarking the HKM-CA model. Root-mean-square error (RMSE), determination coefficient (R2), and mean absolute error (MAE) were the key indicators used for evaluating the model performance. The results revealed that the proposed HKM-CA model was a powerful tool for improving the accuracy of the CA model. Specifically, the HKM-CA model yielded a superior result with an RMSE of 0.475, R2 of 0.995, and MAE of 0.373 in comparison to other models. The proposed HKM-CA model has the potential to be used for predicting blast-induced PPV on-site to control undesirable effects on the surrounding environment. •PPV is one of the dangerous effects and needed to be accurately predicted.•A novel hybrid model (HKM-CA) was developed for predicting PPV with high accuracy.•Performance comparison with the CA, RF, CART, and SVM models.•Performance comparison with empirical techniques for estimating PPV.•W, R, and H are the most influential parameters on PPV.