Use of a multi-parameter sensor for oil degradation monitoring

Sammanfattning: The continuous monitoring in real time of the degradation of lubricants is of great interest for many different industries to optimize the use of lubricating oil in combustion engines. Traditional oil quality diagnostic methods can measure accurately the different levels of oil degradation, but they cannot be implemented on engines to perform continuous oil quality diagnostics. On the other hand, the different mechanisms that affect the quality of engine oil – e.g. contamination and thermal degradation – affect simultaneously the physical and chemical properties of the oil: this makes it difficult to identify and quantify each degradation form separately. This thesis aims at designing and comparing four diagnostic algorithms to predict continuously and separately the different levels of oil degradation in an engine. These algorithms, called Virtual Sensors, predict simultaneously the different levels of oil degradation based on the measurement provided by a multi-sensor called a Tuning Fork Sensor. To design these Virtual Sensors, regression analyses are performed based on sensor data collected in a lab on oil samples with known levels of degradation. Then, the Tuning Fork sensor is implemented in two locations of the oil circuit of a test-engine to determine the influence of the engine operating conditions and of the engine oil properties on the accuracy of the degradation predicted by each Virtual Sensor. It has been observed that the four Virtual Sensors can predict simultaneously the different levels of oil degradation with a satisfying accuracy when the prediction is based on lab data, but the prediction of these algorithms is highly sensitive to measurement errors. During engine operations, these measurement errors are mostly caused by large variations in the oil pressure in the circuit: this shows that the location in which pressure variations are minimal would optimize the accuracy of the prediction made by each Virtual Sensor. Furthermore, the accuracy of two of the Virtual Sensors depends greatly on the temperature range over which the sensor measures the properties of the oil: the location in which the oil temperature reaches its highest value during engine operations may optimize the accuracy of these algorithms.