With increasing digitization, models are not only used during the design phase but throughout the life cycle of systems. Especially the use of models as soft sensors during operation offers opportunities in cost saving, easy data acquisition and therefore additional functionality of systems. Soft sensors are models of components that use easily accessible auxiliary quantities to estimate target quantities that are difficult to measure. Networks of soft sensors are the prerequisite for redundant data acquisition in a system and thus encourage the occurrence of data-induced conflicts, i.e., inconsistent values from different soft sensors, which may result from: (i) the breakdown or defect of a measuring sensor, (ii) model uncertainties of the soft sensors, (iii) change of component characteristics, e.g. due to wear. The resolution of these conflicts either leads to greater confidence in the model-based system quantities or allows the detection of changing components characteristics. Hence soft sensor networks can be used to detect wear in system components.

Wear in pumps and valves leads to a change in the flow rate and the inner leakage. Therefore, the detection of wear with soft sensors requires the detection of small changes in the system flow rates. In the full paper an analysis of the influence of small flow rate variations on redundant soft sensor outputs is carried out. For this, small flow rate variations are implemented on a test bench for positive displacement pumps. Furthermore, a systematic analysis of parameter and data uncertainties and their propagation in models for positive displacement pumps is carried out. The resulting flow rates and the measurement uncertainties from the models of the pump and the throttle valve of the test bench are compared and discussed with respect to data induced conflicts and the detection of wear.