基于多源数据融合的掘进机截割岩壁硬度识别方法

Hardness recognition method of roadheader’s cutting rock wall based on multi-source data fusion

针对煤矿井下掘进机截割岩壁硬度识别难度大的问题,利用其悬臂振动信号、升降油缸和回转油缸压力信号、截割电机电流信号,提出了一种基于多源数据融合的截割岩壁硬度识别方法。该方法首先对各类信号进行小波包分解,单支重构各频带信号并组建时频矩阵,通过奇异值分解得到包含时频信息的若干特征奇异值,以构造特征向量;再利用LDA算法实现数据特征级融合,得到类可分性更好的低维特征。为解决概率神经网络(PNN)平滑参数无法确定和网络结构复杂的问题,提出了基于差分进化算法(DE)和QR分解的PNN优化方法,并通过优化PNN对低维特征进行硬度识别。实验结果表明:所提出的特征量提取和模式识别方法是有效的,与目前常用的其它模式识别算法相比,优化PNN在掘进机三种工况下均有更高的硬度识别准确率。

Aiming at the large difficulty problem of hardness recognition of roadheader’s cutting rock wall in coal mine,a method for hardness recognition of cut rock wall based on the multi-source data fusion algorithm was proposed by using the roadheader’s cantilever vibration signal,pressure signal of its lift cylinder and rotary cylinder and current signal of its cutting motor.Firstly,the various signals were decomposed with the wavelet packet transformation method,then single branches were used to reconstruct various signals within different frequency bands and form a time-frequency matrix.The SVD was used to obtain several feature singular values containing time-frequency information,and construct feature vectors.Furthermore,the LDA algorithm was used to realize data feature fusion,and low-dimensional features with better class separability were obtained.In order to solve problems of a probabilistic neural network(PNN)’s smoothing parameters being not able to be determined and complicated network structure,a PNN optimization method based on differential evolution(DE)algorithm and QR decomposition was proposed.The low-dimensional features were used to recognize hardness with the optimized PNN.The test results showed that the feature extraction method and pattern recognition one proposed here are effective;compared with other pattern recognition algorithms currently used,the optimized PNN one has a higher hardness recognition accuracy rate under three working conditions of roadheaders.