Aplicação de aprendizagem de máquina no alinhamento de eixo propulsor de navios

Abstract

Ship’ shaft alignment is crucial for obtaining a better performance of its propulsion system. This alignment is ensured, yet in the drydock, by adjusting the bearings’ offsets. However, to define an optimum set of vertical displacements it can cost a valuable time for the shipyards. Therefore, this work proposes the use of machine learning (ML) algorithms to predict the suitable set of offsets, following the mandatory rules in force. Thus, a methodology was developed according to which a finite elements model of shaft line of an Anchor Handling Tug Supply (AHTS) ship was assembled to obtain the reactions in the bearings, shaft line deflections, shear forces, bending moments and, at last, the influence coefficients matrix (ICM), and so allowing the dataset creation and the ML models implementation. Results showed that Gradient Boosting, Random Forest and Deep Neural Networks algorithms presented the best results, achieving up to 98.14% of precision in one of the several scenarios tested, and an average precision of around 96.00%, contributing to make quicker the nowadays very time-consuming process of shaft alignment. In this way, the models created were judged to be reliable and capable of being used in the main shaft alignment’ procedures.