A fractal-theory-based multi-agent model of the cyber physical production system for customized products

Abstract

The manufacturing of customized products relies on precise control and close collaboration between distributed production processes. However, discrete manufacturing units cannot achieve consistently efficient collaboration owing to varying grades of smart manufacturing capabilities. Traditional cyber physical production systems have already achieved service- and data-driven resource allocation optimization, but the difficulty of discrete resource collaboration management remains. To address such an issue, this paper proposes a low-complexity and flexible factory operation model, the smart fractal factory (SFF). First, we propose the model definition of the cyber–physical fractal (CPF) based on fractal theory, whose information composition and evolution mechanism are used to formally describe the multi-scale self-similarity of structural characteristics and the operation mode of the smart manufacturing system. The core of the CPF is essential information element (EIEs) and typical information flow (TIFs), which are critical to upgrading manufacturing capabilities intelligently. Second, considering the realizability of the CPF pattern, the attribute configuration of EIEs and the operation mechanism of TIFs are established. Among them, self-organizing, self-adaptation, and self-learning TIFs are utilized for the construction of resource collaboration relationships, the evaluation and optimization of dynamic collaboration, and the enhancement of autonomous analysis or decision-making capabilities. Then, by leveraging agent modeling technology, we present a CPF implementation method to encapsulate distributed resources intelligently and homogeneously into smart manufacturing units that can autonomously optimize and collaborate. Thus, the multilayered architecture of traditional factories in the industrial Internet environment is transformed into an intelligent collaboration architecture based on CPF clusters in the SFF system. Finally, the usability and effectiveness of SFF were verified in a real manufacturing case of a customized product.