An optimized identification method of coal-bearing stratum lithology
-
摘要: 针对现有煤矿井下含煤地层岩性识别方法存在地层信息参数获取难度大、岩性识别精度低的问题,提出了一种基于主成分分析(PCA)算法和核模糊C均值聚类(KFCM)算法的含煤地层岩性优化识别方法。利用钻进试验台获取机械钻速、回转扭矩、钻压、转速、回转压力和泥浆泵流量6种钻进敏感参数,构造高维钻进参数集,作为识别数据来源,包括训练样本和测试样本;结合PCA算法的特征提取优势和KFCM算法具有较好聚类效果的特点,建立基于PCA-KFCM算法的岩性识别模型;采用PCA算法对训练样本进行特征提取和降维处理,得到训练样本的特征值和特征向量;采用KFCM算法对训练样本主成分数据集进行模糊核聚类分析,将试验岩样分为若干类型;通过马氏距离判别法建立判别准则,利用最小马氏距离完成对测试样本的地层岩性识别。测试结果表明,基于PCA-KFCM算法的含煤地层岩性优化识别方法能够有效识别地层岩性,与常规KFCM算法相比,识别精度提高了23.2%。Abstract: In view of difficulties in obtaining stratum information parameters and low accuracy of lithology identification in existing lithology identification method of coal-bearing stratum in coal mine underground, an optimized identification method of coal-bearing stratum lithology based on principal component analysis (PCA) algorithm and kernel fuzzy C-means clustering (KFCM) algorithm was proposed. A high-dimensional drilling parameters set was constructed by using drilling test rig to obtain six kinds of drilling sensitive parameters, such as penetration rate, rotary torque, drilling pressure, rotational speed, rotary pressure and mud pump flow rate, which was taken as identification data sources, including training samples and test samples. Combining feature extraction advantage of PCA algorithm and good clustering effect of KFCM algorithm, a lithology identification model based on PCA-KFCM algorithm was established. The PCA algorithm was used to extract features of the training samples and reduce the dimension of the data to obtain eigenvalues and eigenvectors of the training samples. KFCM algorithm was used to conduct fuzzy core clustering on principal component data sets of training samples, and the test rock samples were divided into several types. The criterion was established by the Mahalanobis distance method, and the formation lithology of the test samples was identified by the minimum Mahalanobis distance. The test results show that the optimized identification method of coal-bearing stratum lithology based on PCA-KFCM algorithm can effectively identify formation lithology, and the identification accuracy is improved by 23.2% compared with the conventional KFCM algorithm.
点击查看大图
计量
- 文章访问数: 58
- HTML全文浏览量: 13
- PDF下载量: 9
- 被引次数: 0