Development to Emergency Evacuation Decision Making in Hazardous Materials Incidents Using Machine Learning

Abstract

Chemical accidents are the biggest factor that hinders the development of the chemical industry. Issuing an emergency evacuation order is one of effective ways to reduce human casualties that may occur due to chemical accidents. The present study proposes a machine learning-based decision making model for faster and more accurate decision making for the issuance of an emergency evacuation order in the event of a chemical accident. To implement the decision making model, supervised learning by the 1-Dimension Convolutional Neural Network based model was carried out using the HSEES and NTSIP data of ATSDR in the United States. An action—victim matrix was devised to determine the validity of emergency evacuation orders and the decision making model was made to learn the matrix so that the decision making model could recommend whether to execute the emergency evacuation orders or not. To make the decision making model learn the chemical accident situations, the embedding technique used in text mining was applied, and weighted learning was carried out considering the fact that learning data are asymmetric. The AUROC value for the results of the decision making by the model is 0.82, which is at a reliable level. Establishing such an emergency response decision making model using the method proposed in the present study in the mitigation stage will help the process. Among the chemical accident emergency management stages, constructing a database for the model, and using the model as a tool for quick decision making for an emergency evacuation order, is also thought to be helpful in the establishment and implementation of emergency response plans for chemical accidents.

Research on Integrated Scheduling of Multi-Mode Emergency Rescue for Flooding in Chemical Parks

“…The actual rescue completion time of the case (256 h) is taken as the upper limit of the initial rescue time, i.e., T = 256 h. The data in Table 2 are brought into the proactive emergency rescue scheduling model. The baseline rescue plan under flooding is calculated as S b = (0, 0, 134, 0, 138, 16,47,32,154,40,129,64,100,112,100,189,165,165,129,118,134,129,129,236,206,168,236,255). The corresponding execution model is X b = (0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0), and the loss caused to the affected people is 4722.24.…”

“…For example, Xu et al [15] constructed a multi-indicator emergency risk assessment and emergency route-planning model by considering chemical accident scenarios and emergency behavior characteristics of different individuals. Phark et al [16] used machine learning algorithms to study the decision model for the emergency evacuation of chemical accidents. In addition, some scholars have studied the problem of emergency resource scheduling.…”