회귀분석을 이용한 초기 설계 단계의 추진 축계 비틀림 진동 추정

Abstract

Propulsion shaft system in a large vessel propelled by two stroke main engine consists of long shaft from engine crankshaft to propeller. In the propulsion mechanism, vibratory torque from the engine gives repetitive twist which is referred as torsional vibration in such shaft system that results in fatigue damage of the shaft. In order to prevent lethal shaft damage, torsional vibration of propulsion shaft system has been analyzed by solving differential equations. Traditionally, the process of selecting anti-vibration countermeasures is determined and optimized by iterative analysis process with human judgment. In the study a sample set of data was utilized to find regression equations for the torsional vibration aspect – natural frequency and peak torsional response. Multivariate Polynomial Regression (MPR) as one of the machine learning methods is considered the most probable as it enables to provide regression equation with high order and interactions between variables. This approach with the machine learning is able to provide an assistance for an optimized countermeasure in the early shaft design stage.