In this era of Internet and Big Data，the application of machine learning，which is the core of artificial intelligence，in drilling engineering represents a general future development orientation. Drilling engineering is an essential part of hydrocarbon exploration and development technology. Machine learning-based drilling parameter optimization and accident prediction and warning are vital for economy，safety，efficiency，and environmental-friendliness. Given that it is hard to effectively guide the drilling operation for deep and ultra-deep wells using the previous experience and data，this research summarizes the present algorithms used in machine learning first，and gives a preliminary analysis for the benefits of machine learning in drilling engineering consequently. A review of global studies on applications of machine learning in drilling engineering is presented and the challenges are identified. Some suggestions are proposed，in an attempt to promote contributions of machine learning in the development of drilling engineering of China.

Targeting the technical difficulties of well leakage in drilling Shizigou - Yingzhong structural belt of Qinghai Oilfield，geological analysis，operation difficulty summarization and study on plugging measurements are carried out. Taking well SX58 in Shizigou- Yingzhong structural belt as an example，it is found that the low pressure fracture leakage in deep E 3 2 formation is the main factor of well leakage. Based on E 3 2 formation logging data and drilling feedback，combined with 3D printing technology，a simulated fracture model has been designed and developed，and its plugging experiment evaluation method is established. Through laboratory experiments，the plugging material and plugging slurry formula are optimized，which can provide reference and guidance for the follow-up optimal and fast drilling in Qinghai Oilfield.