Towards energy and emission reduction in industry with Tiny Machine Learning
Sensor technology, machine learning and artificial intelligence can make a major contribution to reducing emissions and using energy more efficiently in production and maintenance processes
By making smart use of digital technologies and data analytics (Smart Energy), industrial and logistics processes can be managed and optimised more efficiently in various areas. At the same time, the implementation of these technologies is complex for many companies; often the knowledge about sensors, machine learning, artificial intelligence and data -analysis and is not in-house, the sensors offered in the market are expensive and wired, one depends on expensive cloud subscriptions , there are security risks for the network and the company itself is not able to analyse the data.
Tiny Machine Learning (Tiny ML) is a form of machine learning that addresses these concerns. Tiny Machine Learning is a new technology that generates data in a cheap, safe and simple way. Tiny ML focuses on developing algorithms and techniques that can be implemented on very small devices, such as microcontrollers and Internet of Things (IoT) devices. The goal of Tiny ML is to enable these devices to make local decisions based on data collected by their (very low-cost) sensors, without the need to connect to a remote server or cloud. Implementing Tiny ML in the manufacturing industry can be done low cost and at significantly lower cost than traditional AI solutions. This makes it possible to start collecting data with sensors at a lot of additional places in processes and machines to monitor and predict. It is therefore a promising technology that contributes to the industry's digital transformation towards Industry 4.0 and makes it accessible to less-wealthy organisation such as SMEs.
Nevertheless, there is a lot of cold feet, especially among SMEs, to engage in this new promising digital technology on a large scale: there is insufficient knowledge about the approach and the intended results, there is a lack of exemplary projects, the available supply is too expensive for SMEs and does not adequately respond to the specific needs of SMEs, and there is no known or available supply of services or support to realise testing, implementation and scaling up .
Therefore, a living lab is needed, a Fieldlab, where knowledge seekers (asset owners/producers) and knowledge carriers (knowledge institutions, data analysts, sensor suppliers) find each other in setting up an infrastructure for testing Tiny ML applications, getting results on a small scale (executing use cases), where calculations are made for individual business cases and where companies (asset owners, suppliers, maintenance firms , system integrators, sensor suppliers) , education and knowledge institutions work on individual proof and implementation of Tiny ML and collective knowledge building about this technology and its applications. The aim of the project Fieldlab Smart Energy is to make Tiny Machine Learning technology sustainably accessible to industrial SMEs and to support these businesses in the targeted deployment of this technology with the aim of reducing energy consumption and emissions as well as the efficient use of renewable energy sources. Already 25 parties have committed to this project.