论文部分内容阅读
针对成像测井资料上的裂缝、溶蚀孔洞和燧石结核地质现象在成像测井资料上的形态和分布差异,提出一种用梯度方向直方图统计量(HOG)和熵相结合计算成像测井资料上3种地质现象的特征量、并采用非线性的BP神经网络对这3种地质现象自动识别的方法。首先计算成像测井资料上3种地质现象梯度方向统计量的U_1~U_6和熵U_7,并将其作为特征量,然后分析特征量对这些地质现象的区分性和敏感性,最后采用BP神经网络方法对塔河油田11口井的435个样本分为学习样本和测试样本进行学习回判和测试识别。试验结果表明,对裂缝有区分性和敏感的特征量有5个,分别为特征量U_1、U_3、U_4、U_5和U_6;对溶蚀孔洞有区分性和敏感的特征量是U_2;对燧石结核有区分性和敏感的特征量是U_7。BP神经网络对221个学习样本中裂缝、溶蚀孔洞及燧石的回判正确率均为100%;对214个测试样本,BP神经网络燧石结核识别正确率为85.5%,裂缝识别正确率为88.5%,溶蚀孔洞识别正确率为84.0%。
Aiming at the differences of morphology and distribution of fractures, dissolved pores and chert nodules in imaging log data from imaging log data, a new method is proposed based on the combination of gradient direction histogram statistics (HOG) and entropy to calculate imaging logging data On the three kinds of geological phenomena characteristic quantity, and using the non-linear BP neural network to automatically identify these three kinds of geological phenomena. Firstly, U_1 ~ U_6 and entropy U_7 of gradient direction statistics of three kinds of geological phenomena in imaging well logging data are calculated and used as the feature amount, and then the distinction and sensitivity of the feature amount to these geological phenomena are analyzed. Finally, BP neural network Methods 435 samples of 11 wells in Tahe Oilfield were divided into learning samples and test samples for learning re-judgment and test identification. The experimental results show that there are 5 distinguishing and sensitive features for the fractures, which are the characteristic variables U_1, U_3, U_4, U_5 and U_6, respectively. The distinguishing and sensitive characteristics of the dissolved pores are U_2. The distinguishing and sensitive feature is U_7. The correctness of BP neural network for the judgment of cracks, dissolved pores and flint in 221 learning samples were all 100%. For 214 test samples, the BP neural network flintstone recognition rate was 85.5%, fracture recognition rate was 88.5% , The correct rate of dissolution hole is 84.0%.