To meet the current demands to reduce pollution, electric vehicles are becoming more powerful and are utilising faster charging. This puts the battery cells into heavy and demanding usage causing heat generation. Also, as electric vehicles become more popular and are used in more extreme conditions, batteries require pre-heating in some cold environments. To meet these demands, electrically non-conductive liquids can be used in direct contact with immersed power electronics to provide cooling, heating, and health maintenance capabilities. These liquids are commonly known as dielectric fluids.
Just like any other oil, dielectric fluids deteriorate throughout their service life due to chemical, electrical, and thermal stresses. Oil degradation is a very complex and unique phenomenon to each environment and application. The main cause of oil deterioration is oxidation, and the rate at which the oil oxidises can be increased with particle and moisture contamination. Each individual degradation feature such as oxidation, particle contamination and water contamination can decrease the electrical insulation of a dielectric fluid. Due to the complex nature of oxidation, this project investigates whether particle and water contamination can be monitored in order to preserve the flow, cooling and electrical insulation properties of the dielectric oil.
Viscosity is a property that changes with oil degradation, and is measured ex-situ using conventional viscometry methods. An available non-invasive technology that can be used to measure the viscosity is ultrasound. This measurement technique has many advantages including low cost, intrinsically high shear rates, small installation area, capable of working in harsh environments, and requires no moving parts. The use of ultrasonic shear waves in a three layered system (a transducer propagating sound through three dissimilar layers such as aluminium, polyimide and oil) has been previously utilised to measure the viscosity of oils in-situ. In order to illustrate ultrasound's potential to monitor a dielectric oil's degradation, water and copper were used to simulate contamination in static conditions without the presence of air bubbles. Hence by using the three layered ultrasonic viscometer to measure their impact on viscosity, a potential in-situ condition sensor could be established.
In this thesis a viscosity investigation was conducted using frequencies of 1.79, 2.98, 10.1 and 13.86 MHz. Water contaminated mixtures consisted of 3, 10 and 20 \% volumes. This high and relatively large percentage range was chosen as in many electronic environments, cellulose is also used to increase insulation. It can hold as much as 20 \% water. Copper contamination consisted of 0.1, 0.5 and 1 \% volumes, which were low in comparison to water. This was because most electronic cooling environments would not typically produce more than 1 \% copper from erosion and chemical reactions during their service life.
Analytical and numerical models were developed in order to interpret the measured ultrasonic data. Modelling the physical aspects of acoustic shear wave motion throughout three layers provided a link between theory and experiment. To validate the models, calibration oils were measured and variations between theory and experimental data were justified and accounted for.
The measured ultrasonic viscosity of the pure coolant fluid showed Newtonian behaviour at low frequencies that aligned well with conventional viscometer results. However at high frequencies the viscosity was much lower, which appeared as shear thinning behaviour between two Newtonian plateaus. A similar shear thinning behaviour between high and low ultrasonic measurements was also seen in contaminated fluids, which further supported the proposed shear thinning behaviour. With the addition of copper there was little effect on the viscosity, which was due to such low fractions of contamination. Water contamination showed distinguishable viscosity data with increasing dispersed phase volume across all frequencies. The results did not match the conventional rheometer data but they did illustrate shear thinning, which appeared more drastic with an increasing shear rate.
Based on the ultrasonic viscosity results, shear transducers show the potential to be used to measure the condition of battery coolants similar to the fluid tested in this thesis. A combination of low and high frequencies can provide viscosity results similar to conventional methods and at shear rates potentially within a transition zone. Finally, it was a promising investigation as to whether water and copper could be detected in a dielectric oil based on their impact on viscosity.
