A Hybrid Deep Learning Framework for Real-Time Fault Diagnosis and Prediction of Elevator Systems

Abstract

With the rapid rise of the number of elevators in China, the connection between elevators and daily production and life is more and more close. How to ensure the normal operation of elevators and timely diagnosis and early warning of faults has become the focus of the government and relevant scholars. In this context, this paper monitors and warns the corresponding faults in the main components of traction elevator (control system, mechanical system), and visually displays the monitoring and early warning results. The main research contents are as follows: analyzes the typical faults of the elevator, mainly analyzes the main characteristics of broken wire defects based on the basic type of wire rope defects; expounds the basic principle of magnetic leakage detection and the basic design idea of the excitation device; designs an elevator fault detection and early warning system. The system collects the vibration acceleration signal of the elevator car through the sensor and uses Ethernet to transmit the data flow processing engine using Fume, Kafka and Flink, aggregate and distribute the data to MySOL and database; builds the visual interface based on InteliJ IDEA development tool to realize the real-time monitoring and fault warning of elevator operation, providing the basis for fault diagnosis and preventive maintenance of the elevator.