【作者】 张成良；

【导师】 梁开水；

【作者基本信息】 武汉理工大学 ， 采矿工程， 2004， 硕士

【摘要】 在水电站地下厂房开挖中，岩锚梁开挖难度大、质量要求高，该部位成型质量的好坏将直接影响今后厂房内吊车的运行安全。岩锚梁开挖要求其具有良好的开挖轮廓面和较小的破坏范围。目前，在爆破工程中，技术人员通常凭借个人经验、工程类比或者在相似地段进行爆破试验，来决定爆破开挖参数。这样不仅费工费时，有时爆破效果也不理想。 采用神经网络进行爆破参数设计不需要知识工程师进行整理、总结以及消化领域专家的知识，只需要用领域专家解决问题的实例或范例来训练网络；在知识表示方面，采取隐式表示；在知识获取的同时，自动产生的知识由网络的结构和权值表示，通用性强，便于实现知识的自动获取和并行联想推理。神经网络的非线性和识别功能在预测爆破效果及爆破块度方面的应用取得了良好的效果，促进了爆破技术的发展。 本文以水电站地下厂房岩锚梁爆破开挖为研究对象，重点研究了影响爆破效果的参数及其选择。在首先分析爆破机理的基础上，探讨并确定了影响地下厂房岩锚梁爆破开挖成型质量和稳定性的各种因素。确定以炸药类型、最小抵抗线、孔深、炮眼间距、不耦合系数、线装药量、岩石强度和岩体完整度为主要影响因素；把改进的BP算法应用在爆破参数优化设计中，建立以炸药类型、最小抵抗线、炮孔深度、炮眼间距、线装药量、不耦合系数、岩石强度和岩体完整度为主要影响指标，选择45例国内外成功的爆破参数设计实例为样本，利用BP神经网络进行爆破参数优化设计，并通过现场保护层和岩台修面爆破试验以及对其松动圈范围进行声波测试，其测试结果最优的爆破方案其松动圈最小，这与神经网络进行参数优化设计的爆破方案相吻合。说明采用神经网络进行爆破参数设计在理论和实践上是可行的。更多还原

【Abstract】 The rock anchor beam is the most difficult and demanded in excavating underground factory, its quality will affect its safety of performing later, as a result, a better quality and smaller crushedregion is demanded. During the course of engineer design, those who master at his field make decision depend on his experience, project contrast and test in similar condition, then to decide its parameter. Which waste much time and expenditure. Moreover, the result is not as expect as what we have.Neural network is applied in design of blast parameter, which is not needed the trim and sum up of engineer and knowledge of expertise. What are needed is some succeed examples and stylebooks to train the system. To some knowledge, which is expressed by implication, At the same time to acquire knowledge, much knowledge in the same question were showed in the same network. Those be used in many fields and easy to perform. As a kind of Artificial Intelligence, neural network has been widely used in many scientific fields. In blasting Engineering, it is useful to the development of blast technology, by virtue of its neural networks’ non-linearity and identification ability.Rock anchor beam is selected as the studied object, the parameter of blasting and its parameter selection are studied in the paper. At first, the author has analyzed the destroy style and mechanism of blasting, studied various factors which influence on the effect and supporting design of rock anchor beam of underground hydroelectric power station. The main factors have been determined to be the depth and width of hole, the strength of surrounding rock, the types of detonator, the decoupling coefficient, burden and unit consumptionForm what has been discussed, the author make use of those major factors as standard of affection and studied BP neural networks’ work principle, structure and defectiveness. A model of modified BP neural networks has been used to built model in order to identify selection and optimize of blasting in rock anchor beam. Triumphant parameter design of blasting were selected as a example, in order to perform network of parameter design.At last, those designs of parameter by BP network tested by blasting trials and sound wave. As a result, the smallest range of destory is the best project, which is accorded with design of BP network. It is show that applying the way of BP network is not reasonable but feasible.更多还原