AI Based Predictive Maintenance as a Key Enabler for Circular Economy: The KYKLOS 4.0 Approach

Circular economy (CE) is a recent model of production and consumption. According to the European Parliament, this model simply extends the life cycles of products through sharing, leasing, reusing, repairing, refurbishing, and recycling existing materials as much as possible. Digitalization plays a crucial role in the transformation towards a sustainable circular economy. By providing accurate information about appliances and machines conditions, minimizing waste and promoting a longer life for them can be achieved. Predictive maintenance (PdM) is a service using data analytics and aiming at detecting machine failures, degraded performance, or a downtrend in product quality before one of these occur. Due to the advantages that artificial intelligence (AI) techniques currently offer, more and more predictive maintenance solutions start incorporating them in order to better analyse the gathered data. This paper gives an overview of the Deep Learning toolkit that has been developed within the European project KYKLOS 4.0, and which provides a bunch of functionalities including data collection and preprocessing, models definition, and models validation. This toolkit is also endowed with a graphical user interface facilitating its use. It has also been tested with publicly available datasets as well as datasets collected in manufacturing environments. In the current paper, the toolkit will be described in the context of a pilot where the data were collected from a shipyard located in the Astander city, in Spain.