In this paper, adaptive stead-state models are presented for axial flow-pumps used as heart assist devices. These models predict flow rate and power consumption of the pump based on the pressure differential (head) and impeller speed. The developed models are identified using system identification techniques on data obtained from a mock circulatory loop. The mock circulatory experiments include physiologic conditions ranging from a healthy heart function to heart failure. The online adaptive nature of these models is used to estimate effective blood viscosity in real-time. Additional mock circulatory loop experiments are performed to emulate flow obstruction at the pump outlet. Results show that the coefficients from the adapted models can be used to detect, identify and estimate the fault.