(34e) Integration of Modular Processing in Supply Chain Optimization

Recent studies on modular and distributed manufacturing have introduced a new angle to the economy of scale [1] by taking into consideration the fact that the use of small and standardized modules of fixed size provide several new strategic as well as economic advantages. Small scale and modular plants could shorten planning and construction times, ensure the safety of operation, provide flexibility in production, lower the initial investment, and reduce distribution costs. Moreover, standardized designs are expected to have a lower capital cost per unit of equipment due to the economy of mass production [2]. In our previous work [3], it has been shown that modular designs have an added advantage for addressing the tradeoffs between flexibility and cost of process design. This is achieved by developing machine learning-based approximations for the feasible region of individual modules. These approximations can then be incorporated as constraints to represent the feasibility of the overall process in the design optimization problem. As has been shown in recent publications [4] [5], the potential of modular designs can be utilized at the supply chain level.

In this work, we consider the optimization of a generic supply chain network to determine the optimal process design, and the facility location to minimize the total cost of the supply chain over the desired planning horizon. The economies of numbers are modeled by scaling the capital cost of all installed modules with a coefficient of mass production that reduces the cost per-module as more modules are installed. Support vector machines (SVM) trained using historical production data are used to represent the feasible region of the different size modules. The resulting problem is a mixed integer nonlinear programming (MINLP) problem where binary variables represent facility locations and integer variables represent the number of modular equipment installed. Since the module options vary in cost and the processing capacity, the tradeoff between centralized and distributed manufacturing is implicitly addressed by solving the problem to optimality. It is noted that the coefficient of mass production may vary based on the specific process and the modules under consideration. We present the effect of the coefficient of mass production on the optimal choice and number of the modules, as well as on the optimal location of the production facility. We further investigate the performance of the proposed formulation on a supply chain network of biomass conversion units where distributed manufacturing has an enormous potential to achieve cost-savings and improve the economic competitiveness of new technologies.

1. Chen, Q., Grossmann, I.E.: Recent Developments and Challenges in Optimization-Based Process Synthesis. (2017)
2. Arora, A., Li, J., Zantye, M.S., Hasan, M.M.F.: Design standardization of unit operations for reducing the capital intensity and cost of small‐scale chemical processes. AIChE J. 1–14 (2019). doi:10.1002/aic.16802
3. Bhosekar A., Ierapetritou M.: Modular Design Optimization using Machine Learning-based Flexibility Analysis. Journal of Process Control. Accepted (2020)
4. Allman, A., Zhang, Q.: Dynamic location and relocation of modular manufacturing facilities. Eur. J. Oper. Res. (2020). doi:10.1016/j.ejor.2020.03.045
5. Palys, M.J., Allman, A., Daoutidis, P.: Exploring the Benefits of Modular Renewable-Powered Ammonia Production: A Supply Chain Optimization Study. Ind. Eng. Chem. Res. 58, 5898–5908 (2018). doi:10.1021/acs.iecr.8b04189