Prediction and optimization of product quality is the key to product design. A model-based quality prediction and optimization method of refrigeration products was proposed to take into account uncertainties and parameter variations in refrigeration product manufacturing. Product quality (performance and its variation) can be therefore predicted in the early design phase. Moreover

a new method was proposed to get the quality-loss function

specifically a response surface model of design capability index varying with the component tolerances was developed by the Center Composite DOE method and Monte Carlo simulation. Taking the quality-loss function as constraints and minimizing cost as the objective

the product quality and cost were optimized in the end. It turns out that better product quality and lower cost can be achieved with the proposed approach.