Early detection of imbalance in load and machine in front load washing machines by monitoring drum movement

Official URL: https://risc01.sabanciuniv.edu/record=b2553780 _(Table of contents)

Balance issues in washing machines manifest themselves in the form of vibrations. These unwanted vibrations become more prominent at high spin speeds. They can be detrimental to the machine’s performance and shorten lifespan by causing permanent physical damage. Detecting these vibrations early in the wash cycle and at spin speeds below the machine’s resonant frequency is critical in devising proper measures to alleviate their effects. In this thesis, we focus on the two common balance issues observed in washing machines. The first one is machine imbalance, which stems from the improper adjustment of leveling legs. The second balance problem is the load imbalance, which is the result of an uneven distribution of the load inside the drum. We specifically investigate the possibility of detecting these imbalances as early as possible using models trained on sensory data collected from the drum. For this aim, we collect vibration data on the two types of imbalance scenarios throughout the wash cycle. Using these data, we build supervised classification models using different feature extraction techniques on the multivariate times series data and different machine learning models. We compare models that are trained with different partial data collected at different time segments early in the wash cycle. Our results show that we can attain a 95% F1-score with input as short as 500 ms of the wash cycle, indicating that early prediction of these two imbalances during the wash cycle is possible. The collected data are shared for the research community.