Abstract: This study addresses the issues of frequent cutter wear and replacement in composite strata, as well as the high requirements for engineering experience and professional knowledge in shield cutter maintenance decision-making. Geological physical and mechanical parameters, shield cutter design parameters, and shield tunneling parameters such as rotation speed, torque, and thrust were extracted from geological investigation reports, shield cutter design documents, and shield data acquisition systems, respectively. A cutter wear prediction model was established based on the Gradient Boosting algorithm to accurately predict cutter wear conditions. The superiority of this method was verified by comparing it with several other machine learning algorithms. Additionally, a smart maintenance decision-making system for shield cutters was developed to visually display prediction results and assist construction personnel in making scientific maintenance decisions. Case studies show that the system outperforms traditional methods in terms of accuracy and response speed. This research provides scientific support for the management of shield cutter wear during construction, facilitates project decision-making, and enables construction personnel to obtain real-time and comprehensive information about the construction site. It also offers significant support for optimizing cutter maintenance strategies and has important application value.