Computer vision-based lubrication management of ship main engine cylinders

Problems in cylinder lubrication process, such as untimely adjustment, unavailability of direct measurement and inaccuracy in measurement, are always difficult points in marine engineering management. This paper analyzes the influence of cylinder oil alkali value and oil injection rate on cylinder lubrication, and proposes the image processing method of cylinder lubrication based on computer vision in view of the deficiencies in the detection of residual oil alkali value and iron content in the swept airway. The images of piston and cylinder liner are subjected to normalization, clustering, segmentation and morphological feature extraction, and the cylinder lubrication quality analysis software is developed based on Hu invariant moments and the map searching technology of convolutional neural network migration learning, which effectively solves the difficult problem of the uncertainty of the manual adjustment of the cylinder lubrication rate and the high probability of error, and applies to the optimal management of cylinder lubrication for the main engine of very large crude carriers. Application results show that: with the cylinder lubrication image, the computer vision deep learning can optimize the cylinder oiling rate, realizing the optimal control of the cylinder lubrication: iron content of the residual oil in the airway is not higher than 25 mg/kg, residual oil alkali value is not less than 30 mgKOH/g, the cylinder liner wear rate is not higher than 0.1 mm/kh, reducing the risk of excessive wear of the cylinder liner. The method has good popularization value and economic benefits.