Reduction of noise and vibration in drum type washing machine using Q-learning

Vibration and noise during the spin-dry process in a washing machine are the “pain points” of greatest concern. These are mainly caused by an unbalanced drum that occurs from the uneven distribution of clothes in a drum of a washing machine. Until now, an unbalance in a washing machine has been considered to be a mechanical-design issue focusing on means of suppressing that vibration. Furthermore, while such an uneven distribution of clothes is generally modeled using mass points, simulating the motion of a complicated combination of clothes is impossible. To the best of our knowledge, no control methods have been developed for minimizing this unbalance by eliminating the root cause of vibration by getting clothes to stick evenly to the drum of a drum-type washing machine. In this paper, we propose a model-free control method using Q-learning that handles this unbalance as a stochastic phenomenon. To use Q-learning in a washing machine, the difficult-to-model state of the clothes is properly defined, and the negative effects of a partially observable Markov decision process are avoided even with a small amount of data by actively acquiring and averaging a wide range of unobservable state data. As a result, in the experiment, the unbalance due to the uneven distribution of the clothes during the spin-dry process was reduced by Q-learning, and the vibration was halved and the noise was reduced by 3 dB regardless of the load size. •The clothes unbalancing can be reduced using reinforcement learning.•The state of clothes is represented by the derivative value of unbalanced mass.•The drum speed suitable for unbalance control is analyzed experimentally.•The tub vibration was reduced by half within the same control time.