Data-Driven Virtual Flow Rate Sensor Development for Leakage Monitoring at the Cradle Bearing in an Axial Piston Pump

Abstract

The leakage of the tribological contact in axial piston pumps significantly impacts the pump efficiency. Leakage observations can be used to optimize the pump design and monitor the behavior of the tribological contact. However, due to assembly limitations, it is not always feasible to observe the leakage of each tribological contact individually with a flow rate sensor. This work developed a data-driven virtual flow rate sensor for monitoring the leakage of cradle bearings in axial piston pumps under different operating conditions and recess pressures. The performance of neural network, support vector regression, and Gaussian regression methods for developing the virtual flow rate sensor was systematically investigated. In addition, the effect of the number of datasets and label distribution on the performance of the virtual flow sensor were systematically studied. The findings are verified using a data-driven virtual flow rate sensor to observe the leakage. In addition, they show that the distribution of labels significantly impacts the model’s performance when using support vector regression and Gaussian regression. Neural network is relatively robust to the distribution of labeled data. Moreover, the datasets also influence model performance but are not as significant as the label distribution.

A New Approach to Study the Effect of Complexity on an External Gear Pump Model to Generate Data Source for AI-Based Condition Monitoring Application

“…Earlier methods of condition monitoring relied on data sources developed with the aid of sensors. The sensors utilized are accelerometers [8], pressure transducers [9], flow meters [10], and displacement sensors [11] to name a few. The sensors, however, are expensive, prone to failure, difficult to incorporate into existing system design, and lead to overfitting of condition monitoring model [12].…”