Blasting is an important mining operation that usually produce several damaging consequences. Adverse rock fragmentation due to bench blasting is one of them. Hence, analysis of risk level and accurate estimation of particle size distribution of fragment size are of interest. This research developed a new model to simultaneously predict and risk assessment of rock fragmentation using 64 collected data from blasting operated in the Zarshouran gold mine in Iran. In this regard, a newly rock engineering system (RES) is developed based on the reliability information of Z-number theory and causal-effect relationship of the fuzzy cognitive map (FCM). This approach is named the reliability rock engineering causality system (RRECS). To do this, 15 principal effective parameters on rock fragment size were considered in the RRECS modeling process. The uncertainty of the interaction matrix was reduced using Z-number concept. Besides, the weight of effective parameters updated based on the combination of the nonlinear Hebbian algorithm (NLH) and differential evolution algorithm (DE) in FCM. The RRECS performance was validated by statistical linear and non-linear models. The results show R², RSME, BIAS, and Accuracy using the proposed RRECS model calculated to be 0.957, 1.956, 0.001, and 96.741 for training and 0.931, 0.996, 0.016, and 99.053 for testing parts, respectively. Therefore, RRECS has performed better than the exponential, power, logarithmic, polynomial, and linear models. Furthermore, the sensitivity analysis results revealed that the hole diameter and powder factor parameters have the highest and lowest sensitivity on fragmentation, respectively.

爆破是一项重要的采矿作业，通常会产生多种破坏性后果。台架爆破造成的不利岩石破碎就是其中之一。因此，风险水平的分析和碎片大小的粒度分布的准确估计是有意义的。这项研究开发了一个新模型，使用从伊朗 Zarshouran 金矿中收集的 64 个爆破数据同时预测和评估岩石碎裂的风险。对此，基于Z数理论的可靠性信息和模糊认知图（FCM）的因果关系，开发了一种新的岩石工程系统（RES）。这种方法被命名为可靠性岩石工程因果关系系统（RRECS）。为此，在 RRECS 建模过程中考虑了 15 个关于岩石碎片大小的主要有效参数。使用 Z 数概念降低了交互矩阵的不确定性。此外，基于非线性Hebbian算法（NLH）和FCM中的差分进化算法（DE）的组合更新了有效参数的权重。RRECS 性能通过统计线性和非线性模型进行验证。结果显示 R^如图2所示，使用所提出的 RRECS 模型计算得到的 RSME、BIAS 和准确度分别计算为 0.957、1.956、0.001 和 96.741 用于训练和 0.931、0.996、0.016 和 99.053 用于测试部件。因此，RRECS 的表现优于指数、幂、对数、多项式和线性模型。此外，敏感性分析结果表明，孔径和粉末因子参数对碎裂的敏感性分别最高和最低。