An Internal Environment Anomaly Detection System for Secondary Water Supply Pump Houses Based on the Improved PatchCore Algorithm

With the growing public awareness regarding tap water safety, the secondary water supply pump room, acting as the final checkpoint before water enters homes, plays a crucial role in ensuring a safe internal environment. However, pump rooms face challenges such as disordered spatial distribution and a shortage of maintenance staff, leading to significant difficulties and costs in maintenance and supervision. This paper introduces an anomaly detection system for the internal environment of secondary water supply pump rooms, based on an enhanced PatchCore algorithm. The system is designed to alleviate the burden and costs associated with manual monitoring by providing continuous 24-hour anomaly detection through automated equipment. It incorporates 5G communication and edge computing technologies within edge-end devices, showcasing superior capabilities in data transmission speed, detection efficiency, information security, and intelligent processing. On the software front, the system employs an improved PatchCore algorithm that has been trained and tested on a dataset constructed from real-world scenarios of secondary water supply pump rooms, achieving an AUROC metric of up to 0.996 and an inference speed of 23.6ms per image. Anomaly information from the pump rooms is transmitted to an IOT platform, email, and server for remote access by personnel. This research offers an effective solution for the intelligent monitoring of pump rooms and provides valuable practical insights for safeguarding the security and reliability of secondary water supply systems.