2019 Spring Meeting and 15th Global Congress on Process Safety

(169b) A Decision Making Model for Chemical Accident Responses Based on Machine Learning

Authors

Phark, C. - Presenter, Ajou University
Jung, S., Ajou University
Chemical accidents are one of the biggest obstacles to the development of the chemical industry. Prevention and mitigation should be considered to reduce the damage of chemical accident ahead. When they are failed emergency response should play an important role. An emergency evacuation order is one of the emergency response method to mitigate the damage after the accident occurred. In order for a chemical emergency response method such as emergency evacuation order to be effective, its accuracy and quickness must be secured.

In the previous study, a model of predicting issuing emergency evacuation order using machine learning was created. The previous prediction model studied 61,563 chemical accident data in NTSIP(National Toxic Substance Incidents Program) from 1996 to 2009 in the United States, and the learning objective was the issuance of emergency evacuation orders under certain accident situation. However, incorrect evacuation order could have been issued in its history and the methodology needed to be improved. In this study, the necessity of emergency evacuation orders was determined by using logical judgment and machine learning in order to solve the problems using the same database. For logical judgment, information on issuing emergency evacuation orders and civilian victim information in the chemical accident database were used. In cases where it is not possible to judge the necessity of an emergency evacuation order based on logical judgment, the unsupervised learning was to evaluate its necessity. Based on the results, the decision making model for emergency response was improved.