Operational Resilience Strategies for Geopolymer Concrete Production under Raw Material Supply Variability

The advent of low-carbon construction has made geopolymer concrete (GPC) a sustainable material for construction. However, the supply uncertainty of the raw materials needed for GPC production makes this a challenge. This research aims to develop and design an integrated digital twin-reinforcement learning framework for optimizing geopolymer concrete production processes. The problem statement concerns the uncertainty involved when producing geopolymer concrete. This paper focuses on building a digital twin structure for optimizing the geopolymer concrete process. The authors also designed a reinforcement learning framework for optimizing the geopolymer concrete production process. The objective is achieved since the digital twin is a computer representation of a production environment. The computer simulation will utilize reinforcement learning. This will ensure that the production is done at a lower cost. Additionally, the digital twin can predict the supply uncertainty. The computer simulation will determine the supply uncertainty level. Performance was evaluated for three supply conditions: stable, with a moderate and severe level of variability, based on a set of indicators: throughput, downtime, energy consumption, CO2 emission, and quality variability. In all cases, it has been shown that the Digital Twin–Reinforcement Learning (DT–RL) approach results in a considerable improvement of production resilience and sustainability performance by as much as 22% relative to downtime performance, as well as saving 13% of energy and a decrease of CO2 emission by as much as 15% relative to static planning. Additionally, a strongly negative correlation between resilience and quality variability of manufactured products was shown to exist. This research shows that applying digital intelligence to green material production leads to an improvement in efficiency and green performance.