Repository logo

The design and optimization of a condition monitoring device using data reduction techniques to estimate the leakage of a load sensing axial piston pump



Journal Title

Journal ISSN

Volume Title





Degree Level



Hydraulic systems are commonly used as solutions to industry challenges. Their excellent power-to-weight ratio can achieve specific design criteria that other power methods may not. In many hydraulic components, precision machining is present. This is to provide hydrodynamic lubrication between contacting components. By design, component life is greatly increased due to limited physical part interaction. Subsequently, any changes to the machined surfaces can result in accelerated and even catastrophic damage. Pressure compensated load sensing (PCLS) axial piston pumps are common in heavy duty hydraulic applications and provide flow in hydraulic systems. Typically, when a pump is exposed to common environmental contamination, internal machined surfaces can become damaged in the form of scoring. Depending on the degree of damage, this can result in increased leakage across lubricating boundaries or catastrophic failure due to adhesion. Component failure can then manifest in several ways. On a pump, slight wear can result in increased case drain leakage and the operator may not notice any performance issues, however, catastrophic failure may result in immediate system changes. A current method of evaluating the condition of an axial piston pump is by measuring the case drain leakage flow. This procedure involves installing a test flowmeter between the case drain leakage line and the reservoir and recording the flow at certain pressures. This can be an involved procedure and any time a closed hydraulic circuit is disassembled, the risk of introducing contamination is high. Additionally, robust, heavily used flowmeters can be inaccurate and unreliable due to wear and calibration errors. There is an obvious need to further develop the method of evaluating the health of a load sensing axial piston pump. The research contained in this thesis provides a potential cost effective alternative to case drain flow monitoring of PCLS axial piston pumps through the analysis of dynamic pump data. A nonlinear dynamic model of a load sensing axial piston pump and circuit is developed and validated with experimental dynamic pressure and swash angle position signals. The dynamic response of the pump outlet pressure, control piston pressure, and swashplate angle of a load sensing pump is shown to change with case drain leakage, both with the model and experimentally. iii A statistical procedure, Principal Component Analysis, (PCA), is applied to a large training dataset developed by the dynamic model. PCA is a fundamental piece of the leakage prediction algorithm developed in this research. In a simulation study, the designed leakage prediction algorithm is able to predict leakage using clean training and test data with a root mean square (RMS) error of less than 1%. Further algorithm development includes determining the best dynamic measurements to obtain, the amount of training data, a filter design for the raw experimental data, and training data manipulation. A simulation study shows that the signal combination that gives the best prediction performance is a combination of the pump pressure, control piston pressure, and the swashplate angle. This was confirmed by evaluating the leakage prediction performance with experimental pump response data. Having determined the optimal sensor data, the amount of training data is investigated. This was shown to improve from 100 samples and peak at 1000 samples. An optimization using experimental data was performed to determine the best filter to apply to the experimental response data. It was determined that a low pass filter with a cutoff frequency 10% below the piston pumping frequency gave the best leakage prediction results. This research includes a thorough investigation into the manipulation of the training data. The detailed optimal noise addition parameters give a predictive error of less than 20% using a signal combination of pump pressure, control piston pressure, and swashplate angle for experimental pump response data. Using just the pump and control piston pressure transients results in approximately 40% prediction error. Swashplate response data give conflicting results as the predictive error for the minimally worn pump is much different than the high wear pump (<10% for minimally worn pump and >20% for severely worn). This research is an investigation into the feasibility of a load sensing axial piston pump condition monitoring device that measures case drain leakage via dynamic measurements. A comprehensive analysis is performed to optimize a leakage predictive algorithm and the design is tested in simulation as well as with experimental data and shows good potential.



Condition monitoring axial pump hydraulics fluid power



Master of Science (M.Sc.)


Mechanical Engineering


Mechanical Engineering



Part Of