| Summary: | 碩士 === 國立高雄第一科技大學 === 環境與安全衛生工程研究所 === 99 === Abstract
Numerical solutions relying on computer software are so commonly seen that most come to believe every manufacturing process must have gone through certain stage of so-called optimization.As the computer hardware gets cheaper, applications exploring the system transfer function between input and output parameters based on daily monitoring data from the discrete control system of the plant started to appear.
Petrochemical industry after the Industrial Revolution, It’s became the most important energy , but the last years people began to understand the natural resources was limited, and the trend of global competition, we must continue to promotion our ability to international
Process of operation products, except trying to reduce costs and improve efficiency, another key point would be the precise control of production allocation , the important issue of production ratio, the reaction system would be the critical core, all kind of parameter link with efficiency, it’s can influence on the production allocation.Therefore, It is important to understand and control the parameter.
In this research, the subject was use a number of product manufacturing process in the reaction system , to find out the best research in explore the product configuration ratio of operating conditions, , for the impact response system process operating variables, Study to collect reaction system operating variables of historical data and organize to regression analysis to parse the data, and then simulated using neural network process model associated with different operating variables, identify the most critical parameter; use this method with the theory of learning, to establish a multi-property prediction models to improve the process analysis of variance efficiency.
In this research, We set up a simulation model, although the results isn’t as our expected, but in the current adjustment process, can provide valuable indicators to accurately predict the proportion of product configuration and optimization of operating parameters selected to provide plant operators reference And timely information to accurately monitor and adjust in order to control the proportion of product configuration and stability operations program.
Keywords: back-propagation neural network, multiple regressions
|
|---|
