基于K-means算法识别瓦斯传感器软故障研究

针对井下瓦斯传感器设备出现的软故障如数据漂移、数据长期低于或高于正常值、数据周期性变动和数据出现大值等问题,提出了一种基于轮廓系数自适应最佳聚类点的K-means算法识别瓦斯传感器出现软故障种类的方法。该方法是利用监控系统采集的瓦斯传感器软故障信号进行小包分解处理后,结合RBF神经网络进行轮廓系数K-means自适应算法的软故障识别训练。K-means自适应算法能够自适应优化聚类中心点,利用聚类中心点的迭代循环计算出最优中心点,选择最佳聚类点进行K-means聚类,从而识别软故障信号的故障类型。实验证明,自适应轮廓系数K-means算法能够有效地识别瓦斯传感器软故障类型,提高了煤矿安全监控系统数据的准确性。

Research on Identifying Soft Faults of Gas Sensor Based onK-means Algorithm

For the problem of soft faults in underground gas sensor equipment, such as data drift, the data is lower than or higher than the normal value for a long time, the data changes periodically and the data has a large value,the K-means algorithm based on adaptive clustering points with contour coefficients is proposed to identify the types of soft faults in gas sensors. The method uses the gas sensor soft fault signal collected by the monitoring system to perform packet decomposition processing, and combines the RBF neural network to perform soft fault recognition training of the contour coefficient K-means adaptive algorithm.The K-means adaptive algorithm can adaptively optimize the cluster center point, calculate the optimal center point using the iterative loop of the cluster center point, and select the best cluster point for K-means clustering, so thereby identifying the type of fault of the soft fault signal. Experiments show that the adaptive contour coefficient K-means algorithm can effectively identify the type of gas sensor soft fault and improve the accuracy of the coal mine safety monitoring system data.