Asset availability is one of the key drivers of asset value. Without availability, defined by the system's ability to be in a state to perform as required, the expected return on asset cannot be achieved. Availability strongly depends on the reliability of the system and its components. Asset failures have to be minimized for a sufficient reliability. One major asset management method for achieving high reliability is the implementation of condition-based maintenance. Herein, the operating points are typically discretely or continuously observed by sensors like heat, vibration, ultrasound, volume flow, or pressure measures. The so-called potential failure curve is then put into practice to match the sensor values with the level of asset degradation. Whereas the potential failure curve is the current status quo when implementing an asset health monitoring system, machine learning and classical regression modelling offer new options in data-based condition assessment. This paper outlines the feasibility of recognizing asset operation curves for physical health monitoring and prognostics. Within the selected use case of a gear pump operation, changes in hydraulic resistance throughout its life cycle are used to indicate asset degradation, identify major causes, and predict the remaining useful life (RUL). Instead of using cost-intensive sensors, the data-based method using solely operating data allows health monitoring based on simple analytical models of the operating points without the implementation of an explicit condition monitoring system.