A Machine Learning-based Soft Sensor for Laundry Load Fabric Typology Estimation in Household Washer-Dryers

Abstract Fabric care manufactures are striving to make more energy efficient and more user-friendly products. The aim of this work is to develop a Soft Sensor (SS) for a household Washer-Dryer (WD) that is able to distinguish between different fabrics loaded in the machine; the knowledge of load composition may lead to a more accurate drying, faster processed and lower energy consumption without increasing the production costs. Moreover, automatic classification of load fabric will lead to an enhanced user experience, since user will be required to provide less information to the WD to obtain optimal drying processes. The SS developed in this work exploits sensors already in place in a commercial WD and, on an algorithmic point of view, it exploits regularization methods and Random Forests for classification. The efficacy of the proposed approach has been tested on real data in heterogeneous conditions.