(147y) Development of Mathematical Programming Models for the Synthesis of Reactor Networks

The reactor network is typically considered the heart of a chemical process due to its role in converting raw materials to valuable target products. Superstructure-based methods have been extensively applied for the synthesis of reactor networks. In this approach, the superstructure embeds candidate unit operations and interconnections, and can be translated into an optimization model. The optimal solution to the model indicates the best configuration and operating conditions among all the alternatives contained in the superstructure. To find better solutions, the interactions between reactor and other subsystems (e.g., separation and heat exchanger networks) must be captured by the superstructure, i.e., the subsystems should be synthesized simultaneously. However, most superstructure-based models for reactor network synthesis are not amenable for simultaneous process synthesis; thus, models for reactor, separation, and heat exchanger network synthesis must be used sequentially. The goal of my research is to develop methods for reactor network synthesis that are suitable for simultaneous process synthesis.

First, I identified that existing reactor network synthesis models are typically used in isolation from the synthesis of the other subsystems and address a narrow problem statement. Most models consider a set of reactions carried out in every reactor (i.e., no catalyst selection decisions) and a single inlet stream with fixed composition, obstructing the full potential of a simultaneous approach. For example, we expect that the inlet stream composition can change when representing a recycle stream from the separation network. To fill this gap, I proposed a generalized problem statement for reactor network synthesis that considers additional degrees of freedom in terms of the number and composition of inlet and outlet streams, and candidate sets of reactions that can be (de)activated in different reactors. These features add flexibility to the models, allowing rich connectivity with other subsystems and the formulation of new problems for reactor network synthesis during the early stages of chemical process design (e.g., catalyst selection decisions).

Accordingly, I proposed a generalized framework for reactor network synthesis that systematically generates a flexible superstructure through the solution of graph-based optimization models. The candidate reactions are modeled through tasks assigned to reactors; a task is defined by a subset of reactions, a catalyst, and the operating conditions (e.g., temperature, pressure). Graphs are used to represent the reactions and tasks available, and the concept of task competition is introduced to define groups of tasks with similar roles. I proposed integer optimization models based on task competition to determine the number of candidate reactors and interconnecting streams, generating a rich but compact superstructure. The framework is suitable for simultaneous process synthesis, allowing the seamless coupling of reactor network synthesis with approaches for separation and heat exchanger network synthesis. In particular, the seamless integration of the reactor and separation networks enables the representation of the same superstructure using fewer streams, significantly improving solution times.

Furthermore, I studied the cost interactions between the reactor and separation subsystems through the attainable region approach. This approach allows us to determine the bounds of productivity (targets) of a desired component achievable by any reactor–separation network, for a given reaction system and feed stream. However, the separation subsystem is typically very idealized (e.g., no cost considerations), resulting in targets expensive to achieve in practice (e.g., requiring large recycle streams). I proposed modeling methods to account for costs while accurately identifying the attainable region for reactor–separation synthesis. Improved separation system models that better describe the costs associated with separation tasks are used. The proposed approach allows the analyses of different trade-offs within the attainable region and aids in establishing practical targets for novel and existing designs.

Finally, in a collaborative project, I performed technoeconomic analysis of isobutanol production in lignocellulosic biorefineries. Biomass-based isobutanol can be an effective and sustainable alternative to fossil fuel-based isobutanol. I use experimental data to estimate the minimum fuel selling price of isobutanol and identify the major cost drivers of the biorefinery. I establish targets of key technological parameters (e.g., fermentation yields, pretreatment operating conditions) of the biorefinery, providing effective future research directions for cheaper biomass-based isobutanol production.