The effect of model complexity on the performance of a simulation model of an external gear pump for ai-based condition monitoring applications

Abstract

External gear pumps can be found in many stationary but especially mobile applications due to their cost-effective, simple, and robust design. The sudden failure of a pump can lead to substantial maintenance breakdown time and production loss as well as a chain reaction that may lead to failure in other components. It is therefore important to identify failure at an early stage by developing a condition monitoring system. Condition monitoring requires a large amount of data and traditional approaches tend to rely on experimental data. However, experimental data is not widely available, and the sensors used in the experiments may incur faults thereby making the system unreliable. It is also difficult and expensive to damage a component realistically to generate experimental data. Hence, a simulation model that imitates the behavior of the system in the healthy and faulty states can be utilized to generate additional data required for condition monitoring. The effect of model complexity in this course is investigated to build a robust simulation model. The simulation model of an external gear pump at three levels of complexity, namely simple, geometric, and detailed geometric is developed using MATLAB Simulink. Data analysis is performed to validate the simulation model by comparing with measured data. An AI-based classifier is developed to test the reliability of the simulation <br/>model to generate data for condition monitoring application. The model validation was best for the geometric model and the test classifier also demonstrates an accuracy score of 90.19%. Hence, the geometric model is <br/>suitable to be utilized as a tool to generate data for AI-based condition monitoring of external gear pump.