A Strategy for Early Detection of Hazardous States in Batch Reactors

• Main Authors: Kui-Yi Chen, 陳奎伊
• Other Authors: Jyh-Shyong Chang
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/24845582921495755397

碩士 === 大同大學 === 化學工程研究所 === 89 === In order to establish a strategy for early detection of hazardous states in batch reactors, the off-line parameter sensitivity analysis was first applied to evaluate the possible dynamic behavior. If the reactor system is operated in parameter sensitive region, a least variation of system parameter will induce an increased reacting temperature to an allowable maximal temperature. Above such a maximal temperature, sub-reaction, cracking-reaction, or even runaway reaction may occur. An on-line parameter sensitivity analysis under closed-loop operation was developed for the operation of batch reactor. It is found that if the reactor system is operated in the parameter sensitive region, the maximal sensitivities of all state variables appear at the same time. To avoid any possible measurement error of the initial concentration, a state-observer was applied to obtain correct states for parameter sensitivity analysis. For on-line detection of hazardous states, a neural network was built and applied to evaluate whether the operation conditions were located in the parameter sensitive region. Applicability of the developed method was demonstrated in the acetic anhydride reaction (with or without sulfuric acid as catalyst) carried out in a batch reactor. On-line parameter sensitivity analysis can be proceeded under closed-loop operation for early detection of hazardous states. Based on the prediction, in-time adjustments of operation conditions are required to prevent any hazardous condition that will be induced.