An analytical engine to gather data and predict maintenance schedules
-Qantas Airlines, Australia
Predictive maintenance is an AI base system which gathers a huge amount of data from multiple system to analyze and produce meaningful output based on which maintenance schedules can be predicted. Existing system at the airport doesn’t use any sort of analytics and works on electronically controlled relays and control units.
Since this project required a good amount of knowledge on AI based systems and IoT implementation, the primary target among our team members was to level up our knowledge base within a week with relevant information. We were quite good at IoT based systems but cloud based analytics and AI was a different game altogether. Nevertheless, the information required to kick start the project was gathered and hardware design started along with first stage of software. A dual core processor based on Intel Quark SoC and integrated WiFi BLE interface was selected. Other required hardware interfaces and circuit designs were planned and implemented as the project moved forward. The system runs a basic version of Linux distro with no GUI. Coding platform selected for this project was Node.js due to its Async feature and highly effective parallel processing techniques.
About 20 different parameters are read and analyzed each second in a cloud platform that uses Microsoft Azure and IBM Bluemix systems. A completely independent web GUI takes care of all the additional features required or the system including remote firmware upgradation. Our previous firmware upgradation system was revisited and redesigned to be much more effective as this system deals with critical machines operating in an airport. Starting from 3 phase 415V machines to smaller temperature, humidity sensors and power factor measurements, the entire hardware was built to track the movement and conditions of each machines and report it back to the main processing engine for further analytics.