Silo.AI and ÅA (Åbo Akademi) organised a hackathon on Edge computing and Machine learning at the university’s Agora building in Turku. Teams from each organisation set out to discover and learn about the methods and technologies needed to perform Edge computing. The hackathon was organised as part of a research and development project ‘EDGE – Analytics for Smart Diagnostics in Digital Machinery Concept’.
What is Edge computing?
Edge computing is an umbrella term for technologies that aim to make data analysis as close to the source as possible. In other words, the computing would happen near the sensors capturing the data, instead of transferring the data to the cloud. This type of computing is particularly interesting in the manufacturing industry and elsewhere where it can complement cloud computing. In the future, this technology could become crucial in various other areas, such as unmanned cargo ships and self-driving cars.
The need for Edge computing will rise in the era of IoT and 5G. Already today cars may have as many as 100 sensors that produce terabytes of data daily. Edge computing will permit to perform basic analytics of the data close to the sensor, and consequently, to improve the analytics and the overall awareness of the situation.
Edge computing is sometimes necessary in areas with poor connectivity. This is common in remote areas and could be useful in the aforementioned cargo ships. Depending on the case, the amount of data might be so large that transferring the data to be computed in the cloud might take too much time. Edge computing will permit early-stage analysis and thus produce more accurate data to be then analysed in the cloud.
Infrared camera for predictive maintenance
At the hackathon the EDGE project team were preparing for cases where Edge computing might be used in the future. The event focused on learning about a camera that captured both visual light (just like any camera) but also infrared radiation. The team tried to build workflows and tools that would make the initial analytics at the sensor-level possible.
Temperature tracking is great for discovering leaks or sudden changes of temperature, that might be an indication of something failing. Infrared cameras are typically used in industrial environments to track temperature. As almost everything that uses or transmits power gets hot before it fails, infrared cameras offer a great data source for predictive maintenance. With Edge computing applied to predictive maintenance, the initial data could be analysed at the Edge level, before getting sent to the cloud.
The team started by taking data from a near-industrial level camera, a FLIR AX8, which is mostly used in boats used for recreational purposes (see photo below). The FLIR camera captures both visual light, in other words, images, but also temperature data based on infrared radiation. As it was the first time the team used that type of camera, a large part of the work included testing different modes and seeing what type of data came out from the API.
To perform the actual computing, the team looked into using Raspberry Pis. These inexpensive but powerful single board computers are able to track the devices they are connected to, orchestrate all the data and use a device specific system to report the data through a low-range network (bluetooth, IR, wifi) or by direct wiring. Bunch of these can be seen in the header image of this blog post.
Hackathon was the first step for the two year project
The main challenges at the hackathon focused around the data infrastructure, the data itself and the analytics. As the Edge project goes on, the team will start to work with actual industry grade cameras, that are used in energy audits, factories and engine rooms.
In the future the Edge project also aims to look at anomalies in sound, for which the team will continue to explore other technologies to be used on top of the infrared camera data.
The hackathon was part of EDGE – Analytics for Smart Diagnostics in Digital Machinery Concept -project that Silo.AI and ÅA are part of. In addition to these, the project includes 10 other infrastructure and production companies. The project is funded by Business Finland, Åbo Akademi, University of Vaasa and Tampere University of Technology and the companies participating.