基于高斯混合聚类的综合物探方法及其在岩溶勘探中的应用

A Comprehensive Geophysical Prospection Method Based on Gaussian Mixture Clustering and its Application in Karst Exploration

综合物探是一种有效的岩溶勘探技术,但其预测结果中存在人为影响大、溶洞边界模糊等缺点.首先,基于机器学习技术,采用高斯混合模型,分别对高密度电法和面波法勘探数据做分类处理;然后,提出Category-boundary算法,进一步细分上述分类得到的边界,提高高斯混合模型分类精度;最后,根据专家经验与地勘资料制定分类融合规则,在勘察数据驱动和工程地质知识引导的有机结合下,形成一套综合物探的高精度分类融合新方法.将新方法应用于浙南某岩溶勘探工程,获得了边界清晰的溶洞探测结果,与实际钻孔信息对比高度吻合,验证了新方法的有效性.

Comprehensive geophysical prospection is an effective technique for karst exploration,but its prediction results usually suffer from significant artificial influence and fuzzy boundaries of karst caves.Based on the machine learning technology,a Gaussian mixture model is used to classify the exploration data of high-density electrical method and data of surface wave method respectively.Then,a Category-boundary algorithm is proposed to further subdivide the classification results,which improves the accuracy of Gaussian mixture model classification.Finally,the classification fusion rules are formulated based on expert experience and geological exploration data.Under the organic combination of survey data-driven and engineering geological knowledge guidance,a new set of high-precision classification and fusion methods is proposed for comprehensive geophysical exploration.By applying this new method to the karst exploration project in southern Zhejiang,a karst cave prediction is made with clearer boundaries.Compared with the actual drilling information,cave prediction and drilling information are highly consistent,which verifies the effectiveness of the method proposed.