Demand‐based optimization of a chlorobenzene process with high‐fidelity and surrogate reactor models under trust region strategies

Abstract

This work demonstrates the optimization of the industrial scale chlorobenzene process, which continuously produces multiple products and includes a multiphase reaction with bubble column reactors (BCRs). The trust region filter (TRF) method is applied to carry out the demand‐based optimization of large chlorobenzene process with high‐fidelity BCR models. The TRF method uses surrogate models that substitute the high‐fidelity BCR models in the process model, and avoids the direct implementation of high‐fidelity models, which leads to a large and intractable optimization problem. The surrogate models are constructed based on basis functions that apply first order corrections from the gradients of high‐fidelity models. Different basis functions, CSTR and linear models, are studied in this work. As a result, the usage of CSTR models for the basis function leads to fewer function evaluations of the high‐fidelity model because CSTR model is a reasonable approximation of the high‐fidelity models and an initial guess of the optimization problem. Also, the TRF with surrogate models successfully provides an optimal solution of the high‐fidelity process model with few iterations and function evaluations of the high‐fidelity model itself. From the comparison with a low‐fidelity CSTR model, the solution with the TRF presents more accurate results. The surrogate approaches also make a smooth transition from low‐ to high‐fidelity models in process development. We apply this approach to a demand‐based optimization that integrates nontrivial business options, including optimal shortage of customer demands for profitable operation.

Synthesis of Combined Heat and Mass Exchange Networks Via a Trust Region Filter Optimisation Algorithm Including Detailed Unit Designs

“…These ROMs are updated at each iteration of the TRF algorithm to ensure that they are accurate within a small trust region around the previous solution, obtained from the running the detailed discretised DAE model of Kazi et al (2020a). For more details on the TRF strategy, see the work of Yoshio and Biegler (2020). The NLP is solved at each iteration of the TRF, including all NLP subproblems, with the HEN NLP linked to the MEN and REN NLPs via the temperatures and flowrates of the lean streams in the MEN, which are also the flowrates of the rich streams in the REN.…”

“…(Kazi et al, 2020b). More recently, Kazi et al (2020c) developed a novel algorithm for including these models into HENS as surrogates via a TRF algorithm from Yoshio and Biegler (2020).…”

Challenges and Opportunities to Enhance Flexibility in Design and Operation of Chemical Processes

“…Recent work uses, e.g., artificial neural networks to set up a surrogate model enabling multi-scale optimization exemplified with a membrane process [91] or thin film growth processes [92]. Other solutions proposed to deal with multiscale problems are model reduction methods or surrogate modeling approaches to reduce computational effort for complex nonlinear systems to allow for efficient control and scheduling [93][94][95][96]. Some aspects concerning the abovementioned challenges are expected to be treated using machine learning methods in chemical engineering in the future [97].…”

Energy Optimization through Heat and Power Integration on a Chlorobenzenes Production Plant

“…Nevertheless, chlorobenzene remains widely applicable in many fields such as rubber, plastics, pharmaceuticals, dyes, and pigments. On the other hand, dichlorobenzene is used as a precursor for the production of paints and coatings and has applications in electronics and agrochemicals industries [1]. Dorota et al found that the solubility of N-alkylated naphthalene diimides (NDI) in dichlorobenzene or chloroform was highest for NDIC5-NDIC8, i.e., medium-length alkyl chains NDI which are used in electronics, photovoltaics, and sensors [2].…”