基于U-Net的断层检测实验教学设计

Design of Experimental Teaching of Fault Detection Based on U-Net

断层反映了地下性质的突变,是控制油气运移和聚集成藏的主要因素,断层检测是地震数据解释中的一项重要工作。为了拓展地球物理专业学生的理论知识视野和深化学生对断层检测方法理论知识的理解,设计了基于深度学习理论方法的断层检测教学实验,将断层检测过程视为一个图像分割任务,建立数据库来训练所搭建的U-Net网络,训练完成后应用于地震数据断层检测。通过教学实验的训练,激发了学生对新理论和新知识的学习兴趣,使学生掌握了深度学习的基本知识和原理,深化了学生对断层检测理论知识的理解,锻炼了学生处理实际问题的能力。结果表明,该实验教学的设计对于提升教学质量效果显著。

Faults reflect the abrupt changes of underground properties, and are the main factor controlling the migration and accumulation of oil and gas. Fault detection is an important task in seismic data interpretation. In order to expand the theoretical knowledge of geophysics students and deepen their understanding of the theoretical knowledge of fault detection methods, the article designs a tomographic detection teaching experiment based on deep learning theoretical methods, regards the tomographic detection process as an image segmentation task, and establishes a large number of databases to train the built U-Net network, and apply it to fault detection in seismic data after the training is complete. Through this teaching experiment, students’ interest in learning new theories and new knowledge has been aroused, so that students can master the relevant knowledge and basic principles of deep learning, deepen their understanding of the theoretical knowledge of fault detection, and students’ ability of dealing with practical problems is trained. The results show that the design of this experimental teaching has a significant effect on improving the quality of teaching.