SYSTEM AND METHOD OF DATA-DRIVEN DEEP LEARNING MODELS FOR DETECTING ANOMALIES IN A STEEL WIRE ROPE

The present invention teaches a data-driven system (10) and method using deep learning models for detecting anomalies in a steel wire rope (SWR) (20) for elevators comprises: a multi-channel data pre-processing module (22); a warning layer (24) further comprises a binary classifier (33) and an anomaly indicator (37); a distinguishing layer further comprises a multi-class classifier (35), wherein the binary classifier (33) detects anomalies along a targeted SWR at a position and the multi-class classifier (35)identifies a known defect and warns an unknown defect on the targeted SWR; and a feedback module (28) configured to record and feed the anomalies detected back into the warning layer (24) and the distinguishing layer (26) for parameters updates and re-training and its method thereof.