During the process of quality assessment, aiming at the dynamics of measurement error and the two kinds of decision mistakes, namely accepting a defect and rejecting a qualified component, an online assessment strategy of component quality was proposed. Dynamic measurement process analysis and Bayesian minimal cost decision rule were combined to evaluate the quality of components. A Kalman filtering model for dynamic estimation of systematic measurement error was established and the adaptive windowing approximation method was adopted to online adjust the variance of dynamic random measurement error. In the decision process, Bayesian minimal decision cost method was used to evaluate component quality and the decision confidence was provided as well. Finally, by taking the rotor dynamic balancing process as an example, the proposed method was validated. The results indicated that this method could trace the change of measurement error, decrease the risk of the two wrong decisions, and improve the reliability of quality decisions.