Abstract Accurate detection of faults in a dynamic system is very beneficial as this information can be used in a wide variety of ways by the machine operators or designers. This advantage becomes many folds when regarding the future condition i.e. time to failure, named remaining useful life, is available in addition to that of the present condition. Thus, prognosis is one of the most useful tools to improve the working of a machine as many critical decisions can be made. Prognosis can be critical for applications that risk loss of life and property. In this paper, a hybrid method, utilizing bond graph and artificial intelligence, is proposed for system health estimation (SHE) and prognosis. The Bond Graph model is used to calculate Energy Activity, which is used as a common metric for both SHE and prognosis. The proposed method is checked by simulation on a spring mass damper system undergoing a fault.