Internal Leakage Fault Detection for Variable Displacement Axial Piston Pump

Abstract

In mobile machines, reliable condition monitoring (CM) of hydraulic system would be very beneficial because it could decrease the maintenance costs in case of a failure. It should also foresee potentially harmful, slight defects in the system, before they lead to system downtime. Very often the focus of the CM is in short term performance. For example, a neural network-based CM solutions require an extensive data collection and teaching phase. A model-based CM on the other hand an accurate model of the process which is tuned into a known operation points. However, it is typical that the operation conditions change. As a result, the process models become outdated and cause-action definitions do not apply to the changed situation. Therefore, a new model structure, Multi-Variable Histogram (MVH), for CM purposes is introduced. The MVH model is statistical nonlinear model of variable relations. The model is based on schematics where the system operation point changes are taken into account. The system input variables, as explanatory variables, define the operation point for the variable being observed. When the effects of the system input variables excluding faults are taken into account, faults can be seen as operation point changes. MVH model based CM solution includes the key elements of a long-term manageable CM solution. Furthermore, the experimental tests are carried-out with a variable displacement axial piston pump to verify the performance of the solution.