Biomanufacturing holds promise for long-term crewed space exploration as a low-cost means to transform on-site resources into products that meet mission needs. Bioproduction targets currently include pharmaceuticals, sustenance, fuels, and polymers that can be used in diverse additive manufacturing applications. A distinguishing feature of the resultant space biomanufacturing factory [1] compared to Earth-manufacturing factories is how versatile and robust a space factory must be to ensure crew survival. Efficiently exploiting available space resources to biomanufacture products in an integrated way while subject to unique space-related constraints is an open research problem. Such efficient exploitation goes beyond bioengineering for harsh environments, and includes determining optimized operation protocols for individual biomanufacturing processes as well as for the whole factory. Ascertaining optimal protocols is complicated by interdependencies in constituent system inputs and outputs, time-varying product demand and resource availability, and high process uncertainties that result from remote and extreme operation conditions. In this work, we address the "operation under uncertainty" problem, building upon recent optimized space biomanufacturing scheduling results [2]. We consider processes for carbon fixation into methane and acetic acid, nitrogen fixation, polymer production from methane and acetic acid, and crop cultivation. We integrate new dynamical bioprocess models within our previous system optimization framework, and construct a large-scale Mixed Integer Linear Program (MILP) to minimize the equivalent system mass (ESM) [3] of a mission-specific biomanufacturing factory. ESM refers to a prevalent infrastructure and logistics metric that accounts for mass, volume, power, and thermal energy demand. Our MILP includes bioprocess-specific constraints and equipment dimensions, as well as new restrictions on joint process durations that are imposed by the integration of our new bioprocess models. We resolve bilinear MILP terms with additional variables and bounding constraints [4]. We then study updated factory scheduling strategies --- without any intermediate storage between processes, with intermediate storage, and with debottlenecking of slow steps to change effective batch throughput for a fixed demand value --- to identify the best operational and scheduling strategies for space biomanufacturing. We also solve this MILP problem while accounting for high model parametric uncertainty and variable
in situ Martian resources and product throughputs. Uncertainties in process dynamics arise from unknown changes in biochemical reaction pathways in the space environment, such as the effects of microgravity [5], which can reduce yield. Examples of variabilities that we address include changes in input light due to a Martian dust storm, and changes in output product demand after an update to mission tasks. To efficiently handle such uncertainties and variabilities in the objective function and in the constraints, we adopt a robust multi-stage scheme [6, 7]. Thus, our work simultaneously clarifies optimal operation for extraterrestrial biomanufacturing, satisfies safety-critical constraints, mitigates endogenous system uncertainty, and manages external environment variability.
[1] Berliner A et al. (2020). Towards a Biomanufactory on Mars, Preprints.
[2] Sen Gupta S, Menezes AA. (2021). Optimal design and operation of a Mars biomanufacturing factory.
[3] Levri JA, Drysdale AE, Ewert MK, Fisher JW, Hanford AJ, Hogan JA, Jones HW, Joshi JA, Vaccari DA. (2003). Advanced life support equivalent system mass guidelines document. NASA/TM—2003-212278.
[4] Rodriguez MA, Vecchietti A. (2013). A comparative assessment of linearization methods for bilinear models. Computers and Chemical Engineering, 48:218-33.
[5] Zea L, Prasad N, Levy SE, Stodieck L, Jones A, Shrestha S, Klaus D. (2016). A molecular genetic basis explaining altered bacterial behavior in space. PloS one, 11:e0164359.
[6] Sen S. (2005). Algorithms for stochastic mixed-integer programming models. Handbooks in Operations Research and Management Science, 12:515-58.
[7] Xie F, Huang Y. (2018). A multistage stochastic programming model for a multi-period strategic expansion of biofuel supply chain under evolving uncertainties. Transportation Research Part E: Logistics and Transportation Review, 111:130-48.
