<p>Smart campus infrastructure (SCI) is highly dependent on Internet of Things sensing devices, building automation system and campus private network. Once there is a fault in the operation process, it will often lead to teaching interruption, energy waste and even safety accidents. Traditional operation and maintenance methods that rely on manual experience and single point monitoring are difficult to find complex related faults in time. In recent years, the fault diagnosis method based on graph neural network can describe the topological relationship and time sequence characteristics between devices at the same time in graph structure, and has shown better performance than the traditional deep learning model in industrial systems, energy systems and other scenarios, providing a new idea for intelligent operation and maintenance of complex infrastructure. This study is oriented to the smart campus scene of colleges and universities, and around the multi-source collection and sample construction of infrastructure operation data, a fault diagnosis and early warning model based on graph neural network is designed, and the system architecture and implementation path are proposed. By constructing the equipment association diagram driven by multi-source heterogeneous data, the typical faults are modeled and identified, and the advantages of the model compared with the baseline method in diagnosis accuracy and early warning are verified in the experimental environment. The research provides technical solutions for the active operation and maintenance of SCI and provides method support for the safe and reliable operation in the informatization and intelligent construction of colleges and universities.</p>