FLBCS: A Secure Autonomous System for Digital Twins in Concrete Mixing Plants Driven by Federated Learning and Blockchain Collaboration

Abstract

Traditional digital twin systems in concrete mixing plants face challenges such as data privacy leakage, difficult quality traceability, and low collaborative efficiency. In contrast, the privacy-preserving feature of federated learning and the decentralized trust mechanism of blockchain can effectively enable secure data sharing and trustworthy decision-making. However, the production data of mixing plants is characterized by high frequency, multi-source, and privacy sensitivity, and there remains a lack of the federated learning and blockchain collaborative architecture that balances privacy and trustworthiness. To address this challenge, this paper proposes a secure autonomous system for digital twins in concrete mixing plants driven by federated learning and blockchain collaboration. Specifically, a privacy-preserving federated training mechanism based on homomorphic encryption and a lightweight blockchain consensus algorithm are integrated to construct a secure collaborative learning paradigm, where data does not leave its domain, models can be shared, and decisions are trustworthy. Experimental results demonstrate that the proposed scheme achieves favorable model convergence and system stability while ensuring data privacy.