岩石灾变关键破裂声发射特征信息实时识别研究

Currently, little effective quantitative study on the relationship between acoustic emission information and rock failure, and lack of effective means of acoustic emission recognition. In this project, we carried out mechanical experiments on rock foundation rupture, composite fracture and crack propagation, took the mechanical characteristics and acoustic emission signals corresponding to the main rupture events in the experiments as the main research objects. The principle and method of speech recognition are used to extract the acoustic emission characteristic information which reflects the main rupture events, and then constructed the learning training model and standard characteristics library. On the basis of research results of the learning training model and the standard characteristics library, building acoustic emission signals recognition method based on Hidden Markov theory to realize the real-time matching of acoustic emission information during rock failure and the rupture status, and to inquire into the quantitative relationship between the mechanical characteristics and acoustic emission characteristics corresponding to rock failure events in detail. These research results provides a theoretical basis for the acoustic emission characteristic signals recognition in rock failure process and thus provides new ideas for real-time warning technology.

基于当前缺乏声发射信息与岩石破裂之间关系的有效量化及关键时段声发射有效识别手段研究。本课题开展岩石基础破裂、复合破裂和裂纹扩展等力学实验，以实验中关键破裂事件对应的力学特征和声发射信号为主要研究对象，采用语音识别的原理与方法提取反映关键破裂事件的声发射特征信息，构建用于关键特征信息识别的学习训练模型和标准特征库；在标准特征库和训练模型研究成果基础上，构建基于隐马尔科夫理论的信号识别方法，实现岩石破裂灾变过程中声发射特征信息与破裂状态的实时匹配识别，并深入探究岩石破裂事件对应的力学特征和声发射特征的定量关系；为岩石破裂灾变前兆的声发射特征信号识别提供理论依据，为实时预警技术提供新思路。

本项目以岩石破裂灾变的关键声发射信号为研究对象，紧密围绕声发射关键信号自动化优选、声发射声学模型、学习训练算法模型、声发射信号识别方法等开展了系统的理论研究和实验研究，达到预期目标。取得了系列重要的创新成果：（1）优化改进了岩石破裂声发射关键信号提取方法，实现了关键信号全自动优选，为后续岩石破裂关键声发射特征分析、实时识别奠定了坚实基础。（2）提出了基于理想模拟信号特征分析，优选评价实际信号的新思路，并构建了标准声发射信号模拟仿真方法，有效解决了岩石破裂物理实验难以获得理想的初始信号样本的难题。（3）研究了基于声纹识别技术原理的声发射声谱分析方法，并基于该方法构建了岩石破裂模式分类方法和判别准则，解决了现行所有分类方法无法识别一个声发射信号内复合叠加的多破裂模式事件科学难题，为信号细节特征的精准分析和后续的声发射信号识别标准解译库的研究奠定了基础。（4）研究了关键声发射信号破裂模式与岩石破裂演化过程的对应关系，提出了“裂纹渐进扩展型声发射信号”、“失稳解离型声发射信号”、“解离启动信号”、“解离尾声信号”、“主裂纹渐进扩展型声发射信号”、“主裂纹扩展前的局部前兆破裂信号”和“主裂纹扩展后的局部微扩展破裂信号”等概念，进一步完善和明确了关键声发射信号与岩石破裂演化过程的对应关系。（5）首次提出了基元信号、岩石破裂过程信号、事件簇信号组的声发射信号层级体系，完成了标准解译库的总体设计与构建。在此基础上，构建了基于隐马尔科夫链的声发射声学模型、学习训练算法模型、声发射信号识别方法，成功地将语音识别的相关理论交叉应用到岩石破裂声发射信号识别领域，为相关研究开辟了新思路，奠定了理论基础，也为岩石破裂灾变声发射前兆信息实时识别奠定坚实基础。