چكيده لاتين
The aim of this study is to develop a data-driven method for estimating laundry weight in washing machines using machine learning algorithms, without the need for external sensors. Although conventional sensor-based methods can provide acceptable accuracy, they tend to be costly, vulnerable, and require periodic calibration. In contrast, the proposed approach in this research relies on internal washing machine data and the design of an intelligent soft sensor.
For data collection, over 1,120 real-world tests were conducted on a direct drive washing machine using various fabric and clothing loads. Motor current, instantaneous speed, and motor stop duration were recorded as initial input variables. Then, using feature extraction and selection methods, the most influential variables were identified and redundant data was eliminated.
In the modeling phase, several machine learning algorithms including Random Forest, XGBoost, Gradient Boosting, and Logistic Regression were tested. The results showed that the Logistic Regression model, when optimized, achieved an accuracy of 97% while remaining feasible for implementation on standard washing machine hardware. This presents a key advantage over studies utilizing deep neural networks, which typically require additional hardware and high-performance processors.
Overall, this study, by leveraging real-world data, eliminating the dependency on external sensors, and selecting lightweight models, successfully addresses some of the main gaps in previous research and offers a cost-effective, sustainable, and generalizable approach to estimating laundry weight.
