ABSTRACT
Oil and chemical tankers use the FRAMO system to transport cargo from ship to shore. FRAMO cargo pumps, which are installed inside cargo tanks, are an essential component of the system. The results of the purging operation are used to determine whether or not these cargo pumps are operationally ready. The purging procedures confirm the integrity of the sealing arrangements on both the cargo and hydraulic systems. The cofferdam that separates the cargo from the hydraulic fluid collects any leakage from the FRAMO cargo pump’s seals. The AUDRINO control board and electronic control of various solenoid-signaled hydraulic actuated valves are recommended for this automated purging procedure. Control signals from the shore control center or the inbuilt timer circuit can be delivered via IoT. This control board also oversees the purging sequence. The leak-off liquid is lifted from the cofferdam space to the sample container by automated purging.
The identification of the liquid is essential for obtaining the purging result. The method proposed in this paper employs three distinct sensors to identify the liquid in the cofferdam space in the autonomous ship environment. The three parameters are density meter, pH meter, and a color sensor. These three characteristics distinguish the cargo liquid from the hydraulic oil used in the system. This test result is useful for cargo operation planning. The content received in the cofferdam is revealed by comparing the measured data set to the preloaded database. The major goal of this research is to use physicochemical data to predict oil content. Two distinct data sets were obtained in this investigation. These data sets include three major cargo and hydraulic oil physicochemical properties. Using the random forests algorithm, the instances were effectively identified as cargo oil or hydraulic oil with an accuracy of 98.6229%. The detection of both cargo oil and hydraulic oil was then classified using three distinct data mining techniques: k-near- est-neighbor, support vector machines, and random forests. The random forests algorithm provided the best accurate classification.
