Optimization is increasingly used in science, engineering, economics, management, industry, and other areas and also is central to any problem involving decision making. It deals with selecting the best of many possible decisions. One of the major areas of engineering application of optimization is found in the tuning of existing operations and production plants.
The chemical industry has undergone significant changes during the past decades due to the increased cost of energy, increasingly stringent environmental regulations, and global competition in product pricing and quality and one of the most important engineering tools for addressing these issues is optimization; but formal application of optimization is really not warranted because of the uncertainty that exists in the mathematical representation of the process or the data used in the model of the process, so optimizing the processes considering uncertainties has some merit.
Uncertainty arises in many situations. For example, experts may be uncertain about their own knowledge, there may be uncertainty inherent in the situation being modeled, or uncertainty about the accuracy and availability of information.
Bayesian belief networks offer consistent semantics for representing uncertainty and an intuitive graphical representation of the interactions between various causes and effects, they are really useful where some information is already known and incoming data is uncertain or partially unavailable (unlike rule-based or “expert” systems, where uncertain or unavailable data results in ineffective or inaccurate reasoning). They often produce very convincing results when the historical information in the conditional probability tables or the evidence known is inexact. In simpler terms, a Bayesian belief network is a model. It is especially useful when the information about the past and/or the current situation is vague, incomplete, conflicting, and uncertain. With the historical information stored in the conditional probability tables, they can be used to help make decisions, or as a way of automating a decision-making process.
Since the field of optimization is still a very active research area and in recent years, various new approaches to optimization have been proposed; this article aims using a new probabilistic approach allowing the uncertain parameter to be a random variable and lets us make probability statements about it, posterior to the data sets.
In this study, a Bayesian network, as graphical structure for representing the probabilistic relationship among a number of variables, is used as a model of the process operation. According to uncertainty involved in the data, probabilities will be considered for each parameter. The Bayesian network uses operational data of an operating MTBE production plant located in Bandar Imam Petrochemical Complex, Iran. Thus, by building a probabilistic model, i.e. Bayesian network, the quantity of the plant product is maximized regarding to uncertainties like flow rates and temperatures.