Causal analysis of flight exceedance events, e.g. hard-landing, is a key task for mod-ern airlines performing Flight Operation Quality Assurance (FOQA) programs. The main objective of the program is to learn from experience: detect early signs of major problems and correct them before accidents occur. It has been found that flare operation would greatly influence the landing performance. According to the finding, we proposed a deep learning approach to assist airlines performing causal analysis for hard landing events. Experimental results confirm that compared with the other state-of-the-art techniques, the proposed approach provides a more reliable results. The technique can be the basis of de-veloping advanced models for further revealing the relationships between pilot operations and flight exceedance events.