(269b) Systematic Approaches for Discovering Innovations to Enable a Sustainable Circular Economy

The recent push towards Sustainable Circular Economy (SCE) needs to be substantiated with modelling and systems design of value-chains. In our previous work, we have built a general framework to assess and design life-cycle value chains within a superstructure network for establishing SCE [1]. This SCE framework was applied to a carrier bags case study with various plastic and paper alternatives and multiple End-of-Life (EoL) options in the superstructure [2]. Due to the possibility of multiple stakeholders, multi-objective optimization was performed within the framework for objectives from the three domains: Environmental, Economic and Circularity. The trade-offs between these three objective types were quantified using pareto surfaces, and compromise solutions were proposed from the pareto optimal set. This study led to insights about currently available alternatives of the value-chain, e.g., quantifying the paper-plastic dilemma, probing effect of re-use, selecting best EoL strategy etc.

However, this approach is limited by the expanse of the constructed superstructure network. The proposed designs can only be derived from the set of currently available technologies and product systems in the life cycle of products. Evidently, win-win solutions are likely to be obtained by substituting conventional technologies with innovations and emerging technologies. This work is aimed at (i) guiding innovation in value chains and (ii) discovering novel products and pathways. For the first goal, we use hotspot and sensitivity analysis, while the second goal relies on superstructure-free design methods.

First, we identify activities in the carrier bags value-chain which act as hotspots of emissions, cost and ‘loss of circularity’ using life-cycle allocation and displacement methods. For instance, the highest amount of circularity loss in the optimal carrier bags value-chain occurs at the segregation process. These hotspots are obtained for various scenarios of technological advancements and recycling consciousness in society. Through these hotspots, we gain knowledge about where the highest scope of improvement in all the three objective domains lie. This is followed by sensitivity analysis using Monte-Carlo simulations to find effect of perturbations in the parameter space on the pareto surface. We also explore prospects of performing more comprehensive yet tractable sensitivity analysis using (i) analytical approaches: contingent derivatives of perturbation maps [3]; and (ii) stochastic approaches: multi-objective evolutionary algorithms (MOEA)[4]. Sensitivity analysis allows us to rank value-chain activities based on pareto dominance that can be achieved by enhancing their respective efficiencies through innovation . In turn, this helps in guiding innovations towards most promising sectors for SCE. For example, such an approach can point toward the need for innovation in chemical recycling technologies of plastic carrier bags.

For discovering innovations, we explore the use of superstructure-free methods. As stated earlier, the SCE framework only evaluates technologies and connections specified within the superstructure. We expect that novel products and pathways can be found if a more flexible design space is used. In order to do so, we evaluate various PSE methods such as State Task Network (STN), General Modular Framework (GMF) and Unit, Port, Conditioning Stream (UPCS) representation [5] to best describe the nodes and connections within the cradle-to-cradle value chain. We scan the method pool for prevalent analogies between units and life-cycle activities, possibility of directed multigraph required for circular flows and flexibility in designing node connections. However , we do not define or generate the superstructures beforehand. Instead, we let the connections and amounts transferred be decision variables of the design problem. UPCS defines the ‘minimal’ and ‘feasible’ rule s for nodes which ensure that no redundant flows enter a unit, and the necessary flows for transformations do enter, respectively. In addition to these, we add the following rules: (i) each product flow is associated with both input and output ports of UPCS, and (ii) there must exist a unit between every output and input port. The streams linking the ports are captured through an adjacency matrix with binary decision variables in each cell. Each cell of this matrix corresponds to a transformation and refers to a sub-matrix of tuples containing (i) scaling factor of transformation (decision variable), (ii) enthalpy and exergy changes (parameters), (iii) yield per unit input (parameter). Within the manufacturing sector, these transformations can be modelled as reaction-separation networks (e.g., from Reaxys), whereas value-chain activities can be modelled using exergy changes. Physico-chemical constraints such as flow conservation are added as a function of adjacency matrix and scaling factors. We demonstrate this method using an illustrative example and find optimal nodes and connections for a small set of products and elementary flows (to and from nature). We also probe the scalability of this method for larger number of products, and to this effect, consider using techniques like Bender’s decomposition and evolutionary algorithms (like NSGA-II) for superstructure-free design.

Through this presentation we aim to convey further developments in application of our SCE framework [1,2] for guiding innovation and emerging technologies towards activities with most scope and highest impact, using hotspot and sensitivity analyses, respectively. Additionally, we create a highly flexible design space for value chains using superstructure and superstructure-free methods. With a large enough sub-matrix for underlying transformations, we expect that this new methodology can yield optimal solutions which haven’t yet been considered in academia or industry. We also probe the scalability of these methods for large and practically relevant problems using advanced mathematical modeling techniques.

References

1. Thakker, Vyom, and Bhavik R. Bakshi. "Toward sustainable circular economies: A computational framework for assessment and design." Journal of Cleaner Production295 (2021): 126353.
2. Thakker, Vyom, and Bhavik R. Bakshi. “Towards Sustainable Circular Economy: Design Framework and Application to Grocery Sacks”. 2020 AIChE Annual Meeting. Video available at: https://youtu.be/dWj6R0UXMjM
3. Tanino, Tetsuzo. "Sensitivity analysis in multiobjective optimization." Journal of Optimization Theory and Applications3 (1988): 479-499.
4. Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. A. M. T. “A fast and elitist multiobjective genetic algorithm: NSGA-II”. IEEE transactions on evolutionary computation, 6(2) (2002): 182-197.
5. Wu, WenZhao, Carlos A. Henao, and Christos T. Maravelias. "A superstructure representation, generation, and modeling framework for chemical process synthesis." AIChE Journal9 (2016): 3199-3214.