适用于矿山爆破的“悟空”大模型构建及其平台研发  

Construction and Platform Development of"Wukong"Large Model Suitable for Mine Blasting

作　　者：李萍丰 谢守冬 张宣 张万忠 LI Pingfeng;XIE Shoudong;ZHANG Xuan;ZHANG Wanzhong(Key Laboratory of Safety Intelligent Mining in Non-coal Open-pit Mines,National Mine Safety Administration,Zhaoqing 526000,China;Hongda Blasting Engineering Group Co.,Ltd.,Changsha 410011,China;Beijing Changtu Technology Co.,Ltd.,Beijing 100083,China)

机构地区：[1]非煤露天矿山安全智能开采国家矿山安全监察局重点实验室,广东肇庆526000 [2]宏大爆破工程集团有限责任公司,湖南长沙410011 [3]北京畅图科技有限公司,北京100083

出　　处：《金属矿山》2025年第1期153-159,共7页Metal Mine

基　　金：国家自然科学基金(青年科学基金)项目(编号:52304201,52104074);非煤露天矿山安全智能开采国家矿山安全监察局重点实验室开放基金项目(编号:2024-K03,2024-K05);安徽省爆破器材与技术工程实验室开放基金项目(编号:AHBP2022B-06);安徽省博士后研究人员科研活动经费资助项目(编号:2023B743)。

摘　　要：针对矿山爆破面临的复杂地质条件、爆破效果准确性差及多源数据驱动模型构建困难等问题,为提高矿山爆破作业的智能化和自动化水平,提升作业效率和安全性,构建了一种基于GPT大模型的矿山爆破大模型——“悟空”大模型,该模型集成了GPT大模型的强大自然语言处理能力与深度学习算法,旨在实现智能布孔设计、智能知识问答及智能问数统计等特色功能。进一步研发了配套的智能平台,该平台引入了自动化的爆破布孔方案生成机制,能够根据地质条件、炸药类型及爆破目标等参数,实现快速优化爆破布孔方案,提升布孔设计的合理性和准确性;平台开发了一个知识问答模块,能够实现即时解答爆破作业中的疑难问题,进一步提升作业效率;此外,平台智能问数统计模块能够以对话形式实现数据自动统计分析、自动图表展示、智能数据洞察等功能,赋能矿山数据多场景分析研判。研究表明:①智能布孔设计能够显著提高布孔设计效率和准确性,减少人工干预;②智能知识问答能够有效解决矿山爆破领域中的专业知识获取难题,提升行业人员专业素养;③智能问数统计能够有效收集并分析爆破作业数据,可为矿山爆破的精准管理与决策提供有力支持。所建模型及其配套的智能平台为实现矿山爆破作业智能化提供了新思路,具有重要的理论意义和实用价值。In view of the complex geological conditions,poor accuracy of blasting effects,and difficulties in building multi-source data-driven models faced by mine blasting,in order to improve the intelligence and automation level of mine blasting operations and improve operation efficiency and safety,a mine blasting model based on the GPT big model,the"Wukong"big model,was constructed.This model integrates the powerful natural language processing capabilities and deep learning algorithms of the GPT big model,aiming to achieve special functions such as intelligent hole layout design,intelligent knowledge question and answer,and intelligent question statistics.A supporting intelligent platform was further developed,which introduced an automated blasting hole layout plan generation mechanism,which can quickly optimize the blasting hole layout plan according to parameters such as geological conditions,explosive types,and blasting targets,and improve the rationality and accuracy of the hole layout design;the platform has developed a knowledge question and answer module that can provide instant answers to difficult questions in blasting operations and further improve operation efficiency.In addition,the platform has also developed an intelligent data statistics module that can realize functions such as automatic data statistical analysis,automatic chart display,and intelligent data insight in the form of dialogue,enabling multi-scenario analysis and judgment of mining data.Research shows that:①Intelligent hole layout design can significantly improve the efficiency and accuracy of hole layout and reduce manual intervention;②Intelligent knowledge Q&A can effectively solve the problem of acquiring professional knowledge in the field of mine blasting and improve the professionalism of industry personnel;③Intelligent question answering digital statistics can effectively collect and analyze blasting operation data,and provide strong support for accurate management and decision-making of mine blasting.The established model and i

关 键 词：矿山爆破 GPT模型 爆破设计 智能问答 智能问数统计 

分 类 号：TD235[矿业工程—矿井建设] TD67