We propose a prognostics and health management (PHM) model for integrated circuits (ICs) based on back propagation (BP) neural network. The model is not only independent of physical mechanism of ICs aging, but can effectively fit the nonlinear function of IC failures as well. We conduct a large number of experiments on a programmed FPGA, and take the extracted experimental parameter samples as training samples to train the PHM model. Experimental results verify the trained model. The results show that the proposed PHM model is in agreement with the experiment and can meet the requirements of PHM for ICs.