Abstract： A Bayesian classification method based on degradation data of highly reliable products was proposed. The products were classified according to the maximum posterior probability. Then, the nonlinear Wiener process model was used to describe the degradation paths of the products. Next,  an algorithm which combined the expectation maximization (EM) method with Kmeans clustering was developed to estimate the unknown parameters of the model. After that, an average cost model was established to make the optimal classification policies. Finally, an example and a simulation were presented to illustrate the effectiveness of the proposed method.