基于空间声学阵列的采煤机煤岩截割模式识别方法研究

It is of crucial importance for coal-rock cutting pattern recognition of shearer to realize unmanned or less-humanized coal mining. In this project, accurate recognition of coal-rock cutting pattern is regarded as the target and an online identification method based on spatial acoustic array is proposed. The generation and spread mechanism of coal-rock cutting acoustic signal is explored and the distribution characteristic of the signal in time, frequency and space domain is research deeply. Then the sound signal is separated blindly from multiple source signal in the coal mining working face through effective time-frequency points. An adaptive enhancement algorithm and an intelligent denoising algorithm are designed according to the feature of the cutting sound signal. Moreover, the feature vector of the signal is extracted under strong background noise by using the average correlation coefficient, energy coefficient and information entropy. A quantification description model for coal-rock cutting pattern is constructed according to the internal relation between the coal-rock cutting pattern and geological condition of coal seam. Finally, an online recognition method based on the Follow the regularized leader algorithm is established. The potential research result of this project can form a basis for accurate coal-rock cutting pattern identification of shearer. And it can also provide new theory and method to improve intelligent level of the shearer.

采煤机煤岩截割模式识别对于实现“无人化”或“少人化”煤炭开采至关重要，本项目以煤岩截割模式准确识别为目标，提出一种基于空间声学阵列的在线识别方法。探寻采煤机煤岩截割声音信号的产生与传播机理，研究信号在时域、频域及空间域的分布特征，分析工作面的主要声源组成，利用有效时频点进行多源信号中煤岩截割声音信号的盲源分离；针对煤岩截割声音信号特点，设计自适应增强与智能去噪算法，并利用平均相关系数、能量系数及信息熵，实现强背景噪声下煤岩截割声音信号特征向量的有效提取；探求采煤机煤岩截割模式与煤层地质条件的内在关系，构建煤岩截割模式的定量化表示模型，在此基础上，提出基于跟随正则引导算法的煤岩截割模式在线识别算法。本项目的研究能为实现采煤机煤岩截割模式的准确识别奠定基础，并为提高采煤机智能控制水平提供新的理论和方法。

采煤机煤岩截割模式识别是实现高效智能煤炭开采的关键之一，本项目以煤岩截割模式准确识别为目标，提出一种基于空间声学阵列的的煤岩截割模式识别方法。主要研究成果如下：.（1）分析了采煤机煤岩截割声音信号的产生与传播机理，估算出频率分布范围以及幅值大小；根据声场特点选择声学阵列空间构型，采用球形阵列结合双迭代广义逆波束方法对煤岩截割信号进行声源定位分离；研究了综采工作面不同声源之间的耦合关系，设计了基于自适应获取邻域参数的DBSCAN聚类算法对煤岩截割声音信号有效频点，对煤岩截割声音信号进行深度筛选分离。结果表明，自适应获取邻域参数的DBSCN算法聚类算法计算相对比ICA分离算法，得到的SLR提升50.67%，SDR值提升42.11%。.（2）以参数自整定双稳随机共振模型对煤岩截割声音信号自适应增强，提高煤岩截割声音信号在混合信号中的信噪比；提出基于自适应噪声的完全集成经验模态分解（CEEMDAN）和改进果蝇算法的煤岩截割声音信号去噪，目标频率区间能量占比提高至32.15%。构建了评价指标体系提取出煤岩截割声音的特征矩阵，采用主成分分析法缩减参数，消除不同评价指标之间的相互影响。最终可由维度为21维的特征向量所表达煤岩截割特征向量，贡献率达85.3436%。.（3）引入采煤机滚筒侵入岩石系数、煤岩的普式硬度系数、煤岩脆性指数、岩石厚度系数，构建煤岩截割模式多尺度量化模型，将煤岩截割过程中常见工况详细划分为14种模式。提出了一种无监督条件下采煤机煤岩截割模式在线识别方法，实现采煤机煤岩截割模式的自适应识别。.（4）搭建了单滚筒煤岩截割试验台和全尺寸煤岩截割试验台，构建了功能完备的采煤机煤岩截割模式识别实验系统，用于验证煤岩截割模式在线识别算法。实验结果表明，对14种煤岩截割模式的识别率为94.3%，带有煤岩截割模式识别功能的采煤机在面对复杂多变地质条件时，可以在保证采煤机各项运行参数正常的前提下，提高采煤机可靠性和截割效率。