||Current abnormal diagnosis of engineering drawings is performed with heuristics based on engineer experience because of lots of design data to be considered over a limited time. This is based on individual guesses and opinions, not on the scientific method. However, with Artificial Intelligence algorithms, it is possible to automate the tasks of human by considering more design data. Therefore, "drawing digitization" and "rule-based abnormal diagnosis" can be performed automatically by applying the appropriate computer science technology. In this paper, the methodology framework is presented and the feasibility is demonstrated through case study for line sizing abnormal diagnosis.