軸旋轉振動之實驗設計及學習向量量化神經網路研究

EXPERIMENTAL DESIGN ANALYSIS OF SHAFT ROTATION VIBRATION AND MODELING BY LEARNING VECTOR QUANTIZATION NEURAL NETWORK

船的動力系統係透過引擎軸、中間軸及艉軸傳至螺槳，螺槳將推進力傳於船體。此動力旋轉機構長期運作下因振動及扭轉等因素造成無法預期之損壞，不僅需花費許多時間與經費進行修復。一般是以有限元素分析作動力旋轉機構的異常診斷，難以用於即時分析。故本研究係針對船用軸旋轉振動進行實驗設計分析，針對軸的材質種類、軸的直徑及轉速共三個因子，進行量測時域之振幅變化，並計算其軸向振動及徑向振動之觀測值的變異量大小，完成實驗的組合表。由柏拉圖了解各因子的貢獻度及各因子間的交互作用，並利用學習向量量化（LVQ）神經網路進行軸旋轉振動狀態分類與識別。本研究係針對艦艇動力系統長期運作下因振動等因素所造成之損壞問題，以建立一套可適用於艦艇動力系統的振動分析模式。利用人工智慧系統取代傳統人力判斷，以達到降低人為失誤以及人力成本之需求。

The ship's power system is transmitted to the propeller employing an engine shaft, an intermediate shaft, and a tail shaft, and the propeller transmits the thrust to the hull. The long-term operation of this power rotating mechanism due to vibration, torsion, and other factors causes unpredictable damage. It needs a lot of time and costs to repair and reduces war readiness status. Generally, using finite element analysis in the abnormal diagnosis of power-rotating mechanisms is difficult for real-time analysis. Therefore, this study is in light of experimental design analysis for the rotational vibration of the ship's shaft. Through the parameters measured by the experiment, measure the amplitude change in the time domain with the three factors of material type, diameter, and speed of the shaft. Calculate the variation of the observed values through the axial and radial vibration, and complete the factor level combinations of the experiment. Understand the contribution degree of each factor and the interaction by the Pareto diagram and use the learning vector quantitative (LVQ) neural network to classify and identify the vibration state of shaft rotation. This study aims to establish a set of vibration analysis modes applied to the ship's power system for the damage caused by vibration and other factors during the long-term operation of the ship's power system. Use artificial intelligence systems to replace traditional human judgment to achieve the need to reduce human error and labor costs.