One of the early indicators of malfunctions in machines is the condition of the oil circulating inside. With sophisticated measurement techniques, the condition of the oil and therefore the health of the machine can be determined. One such measurement is the total acid number (TAN), which is used to measure acid compounds in oils and is known to be a reliable indicator of the degree of oxidation. However, reliable TAN measurements can only be made in the laboratory with expensive equipment, resulting in lost time and high costs while also producing non-environmentally friendly waste.
Quantag, a nanotechnology spinoff of the well-known Opet-Fuchs Petroleum company, is developing new sensor solutions aimed at replacing cumbersome TAN measurements with real-time sensor data. A crucial piece in this solution is an AI model which can predict TAN values from Fourier transform infrared spectroscopy (FTIR) spectral measurements made by the nanotechnology sensors. Here, Quantag relied on neural network technology from Kaizen Intelligence. Without the need for laboratory tests, our model is able to estimate the TAN value with 89% accuracy and in real time, relying only on sensor measurements. Laboratory tests are now only required for calibration at rare intervals. This not only saves precious time and money, but also opens up new opportunities in real time tracking and predictive maintenance.