基于振动/声发射多源特性分析与深度融合的截齿状态识别

It is a basic work to identify the vibration and acoustic information in the process of cutting coal and rock, and it is necessary to carry out the research on the state of the cutting teeth and the intelligent exploitation of the robot. From the point of view of efficient coal cutting and environmental perception, combined with the theory of rock mechanics, mechanical test technology, signal processing technology, decision fusion optimization theory, considering the influence of the change of cutting parameters and coal rock medium on the cutting vibration and acoustic emission signals, characteristics of time and frequency domain analysis of vibration and acoustic emission. To determine the vibration and acoustic emission energy spectrum of the boundary conditions, vibration and acoustic emission energy spectrum reconstruction picks; considering the vibration signal, the acoustic emission signal of the singularity effect of intercept tooth wear, analysis of coal and rock medium and different cutting parameters on the influence of cutting tooth wear rate, determine the different identification signal of cutting tooth the wear rate of weight, a cutting tooth wear rate sensitivity model identification signal; considering the vibration signal and acoustic emission signal of the influence of measurement errors, analysis of cutting tooth grinding Loss of weight and bias, wear rate on the cutting tooth state feature recognition, mathematical model of cutting tooth state signal recognition - Information coupling, determined to elucidate the relationship between identification information and bit state of the construction of environmental perception cutting and decision fusion method based on information.

探寻截割煤岩过程中蕴含的振动与声信息，是识别截齿状态和机器人智能化开采急需开展的基础性研究工作。从煤炭高效截割和环境感知角度出发，结合岩石力学理论、机械测试技术、信号处理技术、融合决策优化理论，考虑截割参数与煤岩介质变化对截割振动、声发射信号的影响规律，分析振动和声发射的时域与频域变化特性，确定振动和声发射能量谱的边界条件，重构截齿的振动和声发射能量频谱图；考虑振动信号、声发射信号对截齿磨损的奇异性影响，分析不同煤岩介质和截割参数对截齿磨损速率的影响规律，确定不同识别信号对截齿磨损速率权重的影响，建立截齿磨损速率识别信号的敏感性模型；考虑振动信号、声发射信号对测量偏差影响，分析截齿磨损量、磨损速率对截齿状态特征识别的权重与偏置，建立截齿状态-信号识辨-多信息耦合的数学模型，阐明识别信息与截齿状态的判定关系，构建基于截齿截割信息的环境感知与融合决策方法。

本项目以多源信息融合的截齿磨损状态深度学习模型为研究中心，通过结合岩石力学理论、机械测试技术、信号处理技术、融合决策优化理论，项目取得了如下创新性研究成果：. （1）定义截齿4种磨损状态，并依据定义针对不同状态提取截割数据，采用小波包分解及小波降噪方法完成信号的处理，最终得到振动信号频谱图50~100kHz、声发射信号频谱图12.5~50kHz信号的加速度能量和值，及电流信号及磨损量变化值作为特征样本，构建数据样本库。.（2）考虑灰色预测应用情况适合截齿退化数据的变化规律，建立基于退化数据变化的灰色预测模型，检验模型预测精度良好，能够实现对截齿退化的预测。应用样本数据实例应用，结果表明，在振动信号信号下预测值的相对误差仅为0.45%，声发射信号下相对误差为3.92%，预测精度较高，预测效果极好。基于截齿退化数据增量非负且去燥后平稳变化的特点，建立Gamma退化预测模型，并引用贝叶斯更新参数方法更新，提高预测精度，实现截齿退化预测，精度较高，且在贝叶斯参数更新下的预测效果更好。.（3）研究了基于D-S证据理论信息融合截齿磨损状态深度识别，构建了截齿磨损状态识别融合模型。结合实验分析，利用Matlab软件计算得到各个证据体隶属度函数以及不确定度值，在单个证据体下不能准确识别出截齿磨损状态。当四个证据体共同联合作用下能够准确识别出截齿磨损状态，通过Adam算法改进的LSTM深度学习模型实验室识别精确度达到了93%。