(620b) Distillation Column Conceptual Design In the Presence of Uncertain Parameter Using Bayesian Network

Nowadays, the effect of uncertain parameters in engineering design and operation is so apparent. In different fields of science and engineering, uncertainty effect has been defined and number of methods has been introduced for minimizing error of uncertainty in these fields. Error in calculation will be reduced using uncertainty consideration. Two points of views are applied to uncertain parameters consideration, deterministically and stochastically. In this work, a comparison between deterministic and stochastic methods has been considered. In deterministic state, uncertain parameter has been considered with the most probable value. In the other state, different mathematical approaches can be used to solve uncertainty. Gaining more information about uncertain parameters, for instance by getting more data, they can be exhibited as probability functions. Bayesian Network (BN) is one of the most popular methods for uncertainty consideration. It is used to calculate conditional probability table (CPT).  In this work, a selection of parents and children nodes will be chosen in BN according to their effects. And, the best CPTs are selected to achieve minimum amount of error function. To examine the proposed method, industrial data from a chemical plant were used for BN structure identification and CPT calculation. Here, the main idea is a conceptual design of a debutanizer column in an Olefin plant. It is well known that distillation column is one of the key equipments and is usually present in most chemical plants, therefore, it’s design and operation of is an important issue. Uncertain parameters in this design are:  External reflux flow, Pressure deference between top and bottom of column, the Temperature of the top tray, the Feed temperature and Steam flow rate. FUG method is applied for modeling column, while the specified parameters are replaced by probability functions. Reboiler steam flow rate is the objective function that is minimized; therefore energy consumption will be optimized. As the result of this optimization, investment and operation costs have decreased. Also, this research illustrates results illustrate that considering uncertain parameters in calculations will improve the design results. And, BN structure can be used in operation stage to help the operator in choosing optimum value for every parameter in consideration.