As the foundation of equipment health management, the degradation process modeling and prediction is an effective way to reduce running risk and maintenance cost. In order to solve the randomness, nonlinearity and multiphase complexity of the degradation process in practice, an adaptive modeling and prediction method for multiphase degradation process based on functional principal component analysis is proposed. This method treats the degradation measurement values as discrete sample values of continuous function, and thus converts the degradation modeling problem into functional data analysis problem. On this basis, this method uses the function principal component analysis method to reduce the dimensionality of the degraded data, and extracts the common information and individual difference information of the equipment degradation. With Bayesian reasoning, the online monitoring data is used to update the degradation model parameters to realize online realtime prediction of equipment health status. Finally, the proposed method is applied to the accelerated life test data of a cooling fan, and the effectiveness of the method is verified. The results show that the proposed method can well model the multiphase complex random degradation process and thus has potential engineering application value.