AI-based condition monitoring of a variable displacement axial piston pump

Conventional condition monitoring involves integration of additional sensors for fault detection and diagnosis. They are costly and sensitive to faults themselves. To overcome these issues and data scarcity, simulation model data is used as a source of training data for Artificial Intelligence based condition monitoring of the axial piston pump. The sensitivity of the simulation model is improved by performing data augmentation. The classification of faults for condition monitoring in the model is performed by developing a classifier utilizing machine learning algorithm. This was tested for experimental, simulation, and augmented simulation data with respective accuracy scores of 84.8%, 70.1%, and 75.7%. Hence, augmented simulation data is a suitable option for online condition monitoring.