INDUSTRY WATCH
AI and robotics driven applications should be considered to boost capacity on existing infrastructure.
Fault detection
Airfield Ground Lighting is crucial to airport safety, so it is critical to maintain their uptime. By combining historic data, light uptime, telemetry, weather data and static data, which is light position with connected light sensors, real time temperature, vibrations or humidity levels, a training set of labelled examples could define the correlation between those parameters and the actual airfield light performance.
Benedicte Hennebo, CMO, VP Growth and
Transformation, High Tech, Transport and Logistics
The identified pattern would then be validated against a test set to confirm findings. In a live environment this unsupervised machine learning application could be used as a tool to prioritise airfield segments that need urgent replacement compared to standard maintenance.
Automated workflows
Applying Natural Language Processing to maintenance notes and large databases of common failures, a Fault Search Engine would provide technicians most probable root causes and repair suggestions. Through automated voice commands and, or VR support, technicians could ask the machine questions related to light life expectancy, actual usage or abnormal behaviour.
The tool could provide factual answers by correlating large amounts of ERP and other data sources, weather records, traffic peaks, suggest what to do next, circuits check, CMS adjustment, etc and even place work orders such as POs for asset replacement or special tools request.
In front of too elaborated questions that would bring the machine answer accuracy below a probability threshold, an automated connection to the rightly skilled engineer would be established with prompted documentation to guide field technicians.
The additional resolution data points provided by the expert would then be recorded together with its context as a reinforced learning point enabling the machine to improve its answers, hence further automating maintenance and repair prescription.
Warehouse improvement
Tapping into historic purchase orders and light controller data, a Machine Learning system could correlate spikes in spare parts orders based on type, age of the installed base, time of the year, traffic movements and O & M delivery lead time leading to automate purchase orders when stocks are low.
In addition, interacting with the airfield maintenance machine learning system and other robots or humans, wheeled delivery robots could select and dispatch parts and tools for a faster and zero error field task execution.
54 INTELLIGENTCIO MIDDLE EAST www. intelligentcio. com