Unlike traditional methods, which require additional sensor technologies and complex model architectures, the proposed lean quadrant model uses only hydraulic parameters. The differential pressure, volume flow and pump rotational speed serve as the basis for the monitoring approach, as they represent fundamental control variables within the dosing system. By capturing the behavior of assets within a hydraulic resistance, the system detects anomalies such as clogging, pump wear or leaks with high sensitivity. The model's computational efficiency enables a lean implementation, providing a cost-effective solution that allows for a differentiation between monitoring the pump's health and the asset's status within the quadrant modeling approach.
This paper introduces a novel, lean approach to predictive maintenance for dosing systems, validated within the use case of exhaust gas after-treatment for diesel engines.
Experimental validation demonstrates a model accuracy within a standard deviation of 2.31 ml/min for volume flow monitoring, at a nominal flow of 3,000 ml/min, and a standard deviation of 0.04 bar for differential pressure, at a nominal value of 4 bar, demonstrating its high sensitivity.
This paper presents a computationally efficient predictive maintenance approach based solely on hydraulic parameters, eliminating the need for time-resolved data. A key novelty is the ability to distinguish between wear in the driving machine and the dosing system itself using simple physical relationships. The method is largely independent of system design and fluid type and is validated on external geared pumps, demonstrating a cost-effective and broadly applicable solution for asset health monitoring. The model is not complex and requires only minimal computational effort.
