Determining the amount of water and the total acid number index in the hydraulic oil of the Austoft 7000 sugarcane harvester with visible/near-infrared spectroscopy technology

The current era witnesses significant challenges in the realm of sugarcane harvesting, primarily stemming from the assessment and high maintenance costs associated with sugarcane harvesting machines. Consequently, there arises a crucial need for a rapid and non-destructive method of monitoring the hydraulic oil of sugarcane harvesters, facilitating the determination of pollution levels and the total acid number (TAN) index. This study delves into the efficacy of visible/ near-infrared spectroscopy as a non-destructive means of measuring and predicting the water content and TAN index in hydraulic oil samples extracted from Austoft 7000 sugarcane harvesters with varying numbers of operational hours. To achieve this objective, hydraulic oil samples are subjected to analysis within the spectral range of 400-1800 nm. Multivariate regression models are developed based on reference measurements and preprocessed spectral data. Various preprocessing methods including moving average (MA), Savitzky-Golay (SG) smoothing, standard normal distribution and first derivative are employed to enhance the accuracy of water content and TAN index predictions. The findings underscore the viability of visible/near-infrared spectrometry for rapid and non-destructive assessment of the water content and TAN index in the hydraulic oil across different operational phases of Austoft 7000 sugarcane harvesters. Notably, the partial least squares (PLS) model utilising a moving average preprocessing method exhibits the most promising results in predicting the water content in the hydraulic oil (cross-validation correlation coefficient (r cv ) = 0.98, root mean square error of cross-validation (RMSECV) = 1.32, predictive correlation coefficient (r p ) = 0.93 and root mean square error of prediction (RMSEP) = 2.49), boasting excellent accuracy (standard deviation ratio (SDR) = 3.99). Similarly, the PLS model incorporating a combination of moving average preprocessing and standard normal distribution demonstrates exceptional accuracy in predicting the TAN index (SDR = 3.87) (r cv = 0.98, RMSECV = 0.004, r p = 0.93 and RMSEP = 0.008). Consequently, the utilisation of visible/near-infrared spectroscopy technology is advocated within the agricultural and industrial domains to expedite the monitoring of hydraulic oil quality and pollution control processes.