Intelligent damage and fault detection in automatic power operated lift (elevator) doors

Most faults in lift installations occur in the door (entrance) systems. Wear and tear of the door operator mechanism and the door system components/ subsystems will result in defects that lead to damage which in turn leads to faults, understood as a change in the door system that produces an unacceptable reduction in the quality of its performance. The research presented in this report involved the development of an experimental lift door stand to collect vibration signature datasets corresponding to a range of typical damage classes that occur in lift door systems. The installation comprises single speed doors (single panel side opening and two panel centre opening) as well as two speed doors (two panel side opening and four panel centre opening). Once the data are collected the vibration features are extracted and used in supervised learning to train the artificial neural networks designed to recognize patterns and to classify damage. The accuracy / performance varies but the results obtained demonstrate performance of the network with high percentage of correctly classified damage classes involved. The work completed so far forms the basis for the development of decision stage algorithms to analyze the results from the pattern recognition and to decide about appropriate maintenance actions required.