Marine engine oils are used for years without an oil change. During this long period of time the oil gets contaminated,
not only by water and fuel but also by solid contaminants due to oxidation of the base oil, overreacted additives soot and
other products of Heavy Fuel Oil combustion.
This paper shows the design, development and assembly of a visible-near infrared (400-1100 nm) sensor that monitors
several characteristics corresponding to in-use marine engine oil condition. Also, chemometric techniques (PLS) are
applied for determining TBN, %insoluble in pentane, soot and water from visible-near infrared spectra, having in mind
the low resolution capability of the extracted on-line sensor signal. Different prediction models for each oil parameter
were obtained. These prediction models were developed by partial least squares regression from the VIS/NIR spectra.
Finally, the sensor has been tested at low-speed crosshead engine (two stroke engine). So that, reference values for TBN,
%insoluble in pentane, soot and water were obtained in the laboratory for every sample. During the validation test, the
models showed: a) a correlation higher than or equal to 0.85; b) the slope for the regression model tends to one; c) low
bias; and d) the root mean square error of prediction (RMSEP) and the standard error of performance (SEP) were similar
and close to the laboratory's estimated error.