AI Agent-Based Smart O&M for Centrifugal Pumps

This paper presents an intelligent operation and maintenance (O&M) system based on an AI Agent. The system is composed of three main layers: the lower-computer multi-source sensing data acquisition layer, the upper-computer intelligent analysis platform, and the AI Agent decision-making layer. In the data acquisition layer, sensors for vibration, temperature, pressure, and flow are networked via a gateway, enabling real-time monitoring of high-speed pump operating conditions. The upper-computer intelligent analysis platform is built on a browser/server (B/S) architecture, providing visualized displays of operational data, equipment status, and diagnostic results, while supporting historical data retrieval and diagnostic record queries. At the AI Agent decision-making layer, an intelligent diagnostic module based on the ReAct framework is developed, which integrates a knowledge base and real-time data analysis capabilities to achieve automatic fault identification and early warning. The system has undergone both simulation and field testing. Simulation results show that when abnormal temperature or vibration is detected, the AI Agent is triggered to perform automatic analysis and quickly generate a diagnostic report. Field test results confirm the system’s ability to collect data over long periods and maintain stable operation. This system offers reliable support for maintenance decision-making and provides a novel solution for the intelligent O&M of rotating machinery.