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

The primary aim of this project is to develop and refine further the approach based on the application of supervised learning by using the ANN for the detection, localization and classification of damage in elevator door systems. ObjectivesTo achieve the aims the following objectives are defined as below:

1. To develop the experimental setup to generate a comprehensive set of test data. The data should be generated over relevant range of operational parameters and under controlled conditions (tkE will provide relevant support and data to facilitate this development).

2. To design the sensor system for signal acquisition (e.g. vibration) and feature extraction; specification of the type and location of the sensors. The following additional issues will be considered:- integration with / attachment of the sensors to the structure;- monitoring and the operational reliability of the sensors.

3. To develop a pre-processing system (signal conditioning) for feature extraction. This would involve such actions as data cleansing (filtering to de-noise the vibration signal), the dimensional reduction (and further de-noising).

4. To develop the feature extraction procedures. The issue of establishing of a set of possible damage/ fault classes will be re-visited and a more comprehensive set will be determined. The procedure will lead to extraction of low-dimensional data sets (to generate feature vectors)

5. The development of the pattern recognition (PR) algorithm(s) for the pattern processing. This is the key objective which involves ANN training to return the damage type (and possibly the damage severity). In the first instance supervised machine learning (ANN) will be applied but the techniques of unsupervised learning will also be considered.

6. To consider the development of the decision stage algorithm to analyze the outputs from the PR and to decide about appropriate maintenance actions required.