Citation

BibTex format

@article{Caputo:2022:10.1115/1.4052299,
author = {Caputo, C and Cardin, M-A},
doi = {10.1115/1.4052299},
journal = {Journal of Mechanical Design - Transactions of the ASME},
title = {Analyzing real options and flexibility in engineering systems design using decision rules and deep reinforcement learning},
url = {http://dx.doi.org/10.1115/1.4052299},
volume = {144},
year = {2022}
}

RIS format (EndNote, RefMan)

TY  - JOUR
AB - Engineering systems provide essential services to society e.g., power generation,transportation. Their performance, however, is directly affected by their ability to cope withuncertainty, especially given the realities of climate change and pandemics. Standard designmethods often fail to recognize uncertainty in early conceptual activities, leading to rigidsystems that are vulnerable to change. Real Options and Flexibility in Design are importantparadigms to improve a system’s ability to adapt and respond to unforeseen conditions.Existing approaches to analyze flexibility, however, do not leverage sufficiently recentdevelopments in machine learning enabling deeper exploration of the computational designspace. There is untapped potential for new solutions that are not readily accessible usingexisting methods. Here, a novel approach to analyze flexibility is proposed based on DeepReinforcement Learning (DRL). It explores available datasets systematically and considers awider range of adaptability strategies. The methodology is evaluated on an example waste-toenergy system. Low and high flexibility DRL models are compared against stochasticallyoptimal inflexible and flexible solutions using decision rules. The results show highly dynamicsolutions, with action space parametrized via artificial neural network. They show improvedexpected economic value up to 69% compared to previous solutions. Combining informationfrom action space probability distributions along expert insights and risk tolerance helps makebetter decisions in real-world design and system operations. Out of sample testing shows thatthe policies are generalizable, but subject to tradeoffs between flexibility and inherentlimitations of the learning process.
AU - Caputo,C
AU - Cardin,M-A
DO - 10.1115/1.4052299
PY - 2022///
SN - 1050-0472
TI - Analyzing real options and flexibility in engineering systems design using decision rules and deep reinforcement learning
T2 - Journal of Mechanical Design - Transactions of the ASME
UR - http://dx.doi.org/10.1115/1.4052299
UR - http://hdl.handle.net/10044/1/91265
VL - 144
ER -