Development of IoT-Based Real-Time Monitoring System and Prediction of Blast-Induced Ground Vibrations in Opencast Mines using Soft Computing Techniques

Abstract

Blasting is an economical and viable operation for reliable excavation of hard rock in mining and civil construction. An ambiguous ground vibration generated by blasting is unenviable and causes grievous damage to nearby inhabitants, residential premises, and other sensitive sites. Consequently, proper monitoring and prediction of ambiguous ground vibration is an indispensable prerequisite to pinpoint the safe limits in and around mines to reduce their hazardous effects. Currently, conventional monitoring systems (seismographs) are widely used to measure the ground vibrations purposes. The existing systems have few limitations such as expensive, need an expert to operate, tedious, and time-consuming process. To mitigate the flaws of existing system, in this work, designed and developed a real-time, economical, reliable, continuous monitoring wireless system with Internet of Things (IoT) technology for blast-induced ground vibration (BIGV) measurement. The recent proliferation of wireless sensor networks (WSNs) evolution into the IoT vision enables a variety of low-cost monitoring applications which allows a seamless transfer of information via embedded computing and network devices. As Micro-Electro-Mechanical-System (MEMS) based accelerometer sensors are becoming widely prevalent in vibration and condition monitoring applications.
Additionally, these sensors are integrated within a wireless sensor network (WSN) to allow monitored data to be transmitted wirelessly. The developed system was integrated with threeaxis, low-g, cost-effective ICM-20600 MEMS accelerometer and 32-bit ATSAMV71N21B microcontroller. In addition, a General Pocket Radio Service (GPRS) for SIM800C device was used as a radio frequency module and integrated to design an effective prototype. The experiment has been carried out by installing IoT prototype at Dungri limestone of ACC Limited, Bargarh, India, and twenty-two blast-events PPV was recorded at variable distances from blast source. The experiment results ensure that the Peak Particle Velocity (PPV) ranges from 0.081 to 2.94 mm/s at different monitoring locations. Similarly, in this study, to evaluate and predict the ambiguous PPV, seven conventional predictor models proposed by the United States Bureau of Mines (USBM), Ambraseys–Hendron, Langefors–Kihlstrom, General predictor, Ghosh–Daemen predictor, Central Mining Research Institute (CMRI) predictor, Bureau of Indian Standards, as well as Multiple Linear Regression (MLR), were applied and established a relation between PPV and its influencing parameters. The results were compared based on evaluation performance models such as Coefficient of Determination (R2), Mean Absolute Deviation (MAD), Root Mean Square Error (RMSE), and Normalized Root Mean Vii Square Error (NRMSE) between monitored and predicted values of PPV. The outcomes of empirical predictors exhibit that the MLR model yields significant R2 (0.86420), MAD (0.24122), NRMSE (0.23539), and less RMSE (0.28940) as compared to other conventional predictor equations. Although, empirical predictor models have two major flaws such as lack of generalizability and a limited number of input variables. Therefore, a study on the development of an alternative method of accurate PPV prediction using soft computing techniques was undertaken. An endeavor has been made in this research to apply three soft computing prediction models, namely, Feed-Forward Back Propagation Multilayer Perception (MLP) Neural Network, Radial Basis Function Neural Network (RBFNN), and Support Vector Machine (SVM). In this context, eleven input parameters such as number of holes, average top stemming, average spacing, average burden, average hole depth, hole diameter, maximum charge per delay, powder factor, total explosive, total depth, as well as absolute distance, and one output: PPV, was used and trained. The obtained results reveal that the RBFNN approach provides high R2 (0.99891), Accuracy (99.62956), MAD (0.00370), NRMSE (0.02287), and low RMSE (0.04897) among all other soft computing techniques and empirical predictor approaches for accurate prediction of blast-induced ground vibration. Hence, the RBFNN model yielded better performance as compared to the other prediction models to estimate the PPV.