摘要
提出了基于时间序列相似性度量的瓦斯报警信号自动识别技术.基于动态时间弯曲(DTW)距离对从山西某高瓦斯煤矿2010年瓦斯监测数据库中提取的150组数据采集周期为9～70s的采掘工作面瓦斯含量超限报警时间序列进行聚类分析,获得了7种典型的瓦斯超限报警时间序列模式;以此为数据源,采取分段形态度量方法,提取并筛选出3个重要指标,建立了瓦斯报警时间序列形态特征库,并提出了基于分段形态度量的瓦斯报警信号快速辨识算法.对另外150组瓦斯报警时间序列进行辨识,实验表明,准确率达92%以上.尤其是通过联合分析瓦斯报警时刻前后的k0和k1值可以快速辨别瓦斯超限原因是炮后瓦斯还是突出警报,统计表明,k0>0.1,k1>0时,100%发生煤与瓦斯突出.
The time series similarity measure technique is presented. In the case of one high gas coal mine from Shanxi province,one hundred and fifty gas warning time series has been got out from the large scale gas time series history database and from which seven representative kinds of time series patterns are listed out by clustering analysis based on DTW distance. With the piecewise morphological measure methods, three key indicators are extracted and filtered out. Then the morphological character table of the gas warning time series can be established. The gas monitoring warning signal identification algorithm is accordingly presented. The experi- ments indicate that the accuracy rate of this identification method reaches more than ninety two percent. With the values of k0 and k1, it is easily to identify if the alarm is caused by the gas e- mission after blasting or the gas outburst. From the statistical data, it could be concluded that the gas outburst will happen when k0〉0. 1 and k1〉0.
出处
《中国矿业大学学报》
EI
CAS
CSCD
北大核心
2012年第3期474-480,共7页
Journal of China University of Mining & Technology
基金
国家自然科学基金项目(40971275,50811120111)
