D.-S. Chen and T.-C. Wang (Taiwan)
Neural network, clustering, MINLP, DOE
Finding more profitable and safer operating conditions for processes are essential factors for a company to remain competitive. The conventional approach for such tasks is primarily based on empirical/semi-empirical model building. In order to have a reliable and precise model, it is not uncommon to require a considerable amount of data. However, the more data required, the more time and money cost. This paper present a systematic way to search optimal operational conditions for processes under minimum data requirement. The basic structure of our method is based the loop modeling, searching, clustering, recommending experimental conditions and remodeling. The proposed algorithm not only can be used to find optimal operating conditions but also do well for mixed-integer nonlinear optimization problems.
Important Links:
Go Back
