Most mining and tunneling projects usually require blasting operations to remove rock mass. Previous studies have mentioned that if the blasting operation is not properly designed, it may lead to several environmental issues, such as ground vibration. This study presents various machine learning (ML) techniques, i.e., hybrid extreme learning machines (ELMs) with the grasshopper optimization algorithm (GOA) and Harris hawks optimization (HHO) for controlling and predicting ground vibrations resulting from mine blasting. Actually, the GOA–ELM and HHO–ELM models are improved versions of a previously developed ELM model, and they are able to provide higher performance capacity. For the proposed ML modeling, a database was established consisting of 166 datasets collected from Malaysian quarries. The efficacy of the proposed ML techniques was observed in the training stage as well as in the testing stage, and the results were evaluated against five parameters constituting the fitness criteria. The results showed that the GOA–ELM model delivered more accurate ground vibration values compared to the HHO–ELM model. The system error values of the GOA–ELM model for the training and testing datasets were 2.0239 and 2.8551, respectively. The coefficients of determination of the GOA-ELM model for the training and testing datasets were 0.9410 and 0.9105, respectively. It was concluded that the new hybrid model is able to forecast ground vibration resulting from mine blasting with high level of accuracy. The capabilities of this hybrid model can be extended further to mitigate other environmental issues caused by mine blasting.

大多数采矿和隧道工程通常需要进行爆破作业以清除岩体。先前的研究已经提到，如果爆破操作设计不当，可能会导致一些环境问题，例如地面振动。这项研究提出了各种机器学习（ML）技术，即具有蚱hopper优化算法（GOA）和哈里斯霍克斯霍克斯优化（HHO）的混合极限学习机（ELM），用于控制和预测爆破产生的地面振动。实际上，GOA–ELM和HHO–ELM模型是先前开发的ELM模型的改进版本，它们能够提供更高的性能。对于拟议的机器学习建模，建立了一个数据库，其中包含从马来西亚采石场收集的166个数据集。在训练阶段和测试阶段都观察到了所提出的机器学习技术的有效性，并根据构成适合度标准的五个参数对结果进行了评估。结果表明，与HHO-ELM模型相比，GOA-ELM模型可提供更准确的地面振动值。用于训练和测试数据集的GOA–ELM模型的系统误差值分别为2.0239和2.8551。训练和测试数据集的GOA-ELM模型的确定系数分别为0.9410和0.9105。结论是，新的混合模型能够高度准确地预测爆破引起的地面振动。该混合模型的功能可以进一步扩展，以减轻由爆破引起的其他环境问题。