Service Needs

Reference No.

N-0149

Title

Intelligent Lift / Escalator Inspection Checklist with the Use of Image Analytic and Machine Learning Technology

Business Needs / Challenges

There are currently about 70,000 lifts and 10,000 escalators across the territory, of which about 54,000 installations (i.e. 67.5%) are in excess of 20 years of age.  Continuous aging of components would undoubtedly impose potential safety hazards on these installations.  Currently, about 100 staff are conducting about 28,900 inspections on lifts and escalators every year, and we have planned to inspect each installation at least once every five years.  In order to further enhance the efficiency and effectiveness of the daily management and monitoring of lifts/escalators, we plan to develop a platform, which is assisted by artificial intelligence, so as to facilitate joint participation of Responsible Persons of lifts/escalators, Registered Lift/Escalator Contractors (RCs) and the enforcement body.  This platform would be able to ensure the lifts/escalators components are in suitable conditions, as well as the safe and normal operations of these installations.

Application Areas

City Management
Transport

Expected Outcomes

Our preliminary idea is to develop an intelligent inspection checklist for recording inspection findings during inspections.  The duly completed checklists would be uploaded to a cloud storage instantaneously for Engineers’ easy retrieval, and advisory letters to responsible persons and/or registered contractors would also be automatically generated once anomalies are identified for Engineer’s further review before issuance.

To this end, we hope artificial intelligence (AI) model can assist in some of the inspection items.  Reference conditions will be established for newly installed or existing lifts/escalators using a set of photos (e.g. 360 degree photo inside the lift machine room, or photos of key components like suspension ropes of lifts etc.).  We need to use numerous photos to train up an AI model by deep machine learning.  The trained AI model would compare and analyze the photos taken during inspections, or even the photos of the same / similar kind submitted by RCs periodically, with the reference photos set.  In case anomalies (e.g. rouging of suspension ropes), or any other critical condition of components are detected by the AI model, apart from informing Engineers and issuing advisory letters, the parties involved, especially RCs, would be immediately alerted, triggering them to carry out suitable rectification works as soon as possible.  At the same time, the responsible persons can also acknowledge the latest conditions of the installations under their purview.

The platform and system may be developed in stages (i.e. implementation of module from time to time, or we may first start with critical component like suspension ropes of lifts as trial, and later other components like overspeed governor rope, door locks, brake etc.).

Technologies to be Used

Artificial Intelligence (AI)
Cloud Computing
Data Analytics
Deep Learning
Machine Learning
Predictive Analytics