INTELLIGENT MICROSCOPY FOR THE STUDY OF FRACTURE AND FATIGUE

摘要

A fundamental goal in the investigation of fracture and fatigue mechanisms is the development of methods for recording critical event inception. The development of in-situ test equipment, and high-resolution microscopy techniques (such as high-resolution orientation imaging microscopy, HROIM) have placed invaluable tools into the hands of researchers. Nevertheless, practical considerations limit the volume of material that can be carefully monitored during a given testing regime. Machine learning techniques offer a promising framework for enhancing efficiency in the search for critical events. This paper presents initial efforts to develop an intelligent microscopy environment for the study of fracture and fatigue, based upon machine learning methods. Classification of local structure as being conducive (or not) to defect nucleation, and Q-learning for improved searches for such local features, are both tested in simple settings as a starting point for more complex developments.