An intelligent quality prediction and autonomous decision system for natural products manufacturing processes

Abstract

Zero Defect Manufacturing (ZDM) is an advanced production paradigm aimed at eliminating quality defects. Based on ZDM concept, in this article, an intelligent quality prediction and autonomous decision system was proposed to improve the quality management ability of the natural products manufacturing process. Firstly, a foundational framework is introduced, which includes five key elements for implementing product-oriented ZDM and process quality management strategies. Based on this framework, a quality prediction model was developed. The model reveals the quality propagation patterns within the material-process–product chain. Furthermore, to enhance the model's data processing and decision-making capabilities in a multi-stage system, we propose a process correction method originated from multi-agent reinforcement learning. Lastly, the proposed framework underwent validation using a dual-system manufacturing process. Following three validation iterations, production efficiency was increased by 15.33%, 15.25%, and 15.40% individually while meeting product quality requirement at the same time. These results suggest that the proposed framework offers substantial promise for realizing ZDM in multi-stage systems in natural product manufacturing.