Development of Washing Machine Dehydration Unbalance Control Specifications Through Bayesian Optimization

Optimizing control specifications to prevent unbalance during the dehydration process in washing machine development is a complex task that consumes significant time and resources. Traditional methods involve expert engineers conducting various experiments and data analyses to develop optimal control specifications. However, these approaches are time-consuming and struggle to optimize diverse performance factors simultaneously. Additionally, the quality of the results heavily depends on the engineers’ experience and capabilities, making it challenging to maintain consistent quality. To overcome these limitations, a new data-driven approach is proposed. This study proposes a methodology that uses Bayesian Optimization to predict the unbalance during the dehydration process and derive optimal control specifications to minimize it. Bayesian Optimization builds a predictive model based on collected data and uses an acquisition function for efficient exploration to find the optimal solution. Through this method, we automated the optimization of unbalance prevention control specifications. Applying the proposed methodology to an actual washing machine model achieved performance equivalent to that derived by expert engineers. Specifically, we succeeded in maintaining the maximum vibration during the dehydration process below the target level and reducing the time to reach high-speed rotation (RPM) ranges. The main contribution of this study is the rapid derivation of machine learning-based optimized control specifications with minimal human intervention and small-scale experiments by building a test automation system within the home appliance development process. This approach shortened the development period and improved quality consistency.