Research On Safety Production Early-warning System For Ammonia Plant

In modern chemical production there are many characteristics, including complex process, intensive energy and many kinds of hazardous substances, and most of the production activities are in the condition of high temperature and high pressure that is easy to cause a fire, explosion, toxicants leakage or other major accidents which have a significant negative impact on the lives and property of citizens and social life and seriously damage the manufacture of chemical enterprises. So it has become a core problem about how to predict potential dangers, originally master all kinds of unsafe factors and furthest reduce or eliminate the probability of accident and the severity of the consequences of accident.Combining the actual situation in safety production of ammonia plant, and using the Delphi Method, this paper has established the early warning index system from the perspective of Human, Machine and Environment System. It has been divided into the target layer, element layer and index layer. The element layer is divided into four elements of physical factors, staff quality, working environment and safety management. The index layer is subdivided into18indicators. And it has calculate the index weight using the Analysis Hierarchy Process to obtain the order of affecting ammonia plant’s early-warning indicators. Based on this, it has introduced the Support Vector Machine theory, and established the early-warning model based on the Support Vector Machine. And the warning degree is divided into four grades:serious warning limit, moderate warning limit, light warning limit and non-warning warning limit. It has achieved the safety production early-warning for ammonia plant through the application of MATLAB programming.According to the training tests on historical data, it has been proved that the early-warning model based on SVM can quite accurately predict the safety production conditions of ammonia plant. At this point the safety production early-warning system for ammonia plant, which will early warn the safety production status of ammonia plant, is established to make the unsafe factors can be found timely by security managers and correspond measures to prevent major accidents.