風力發電機傳動齒輪之動態分析與疲勞可靠度評估

Abstract

風力發電機齒輪箱為承載變動負載並傳遞動力之關鍵傳動組件，其維修成本高昂，故於設計階段進行可靠度評估應屬重要。為此，本研究建立一套結合動態模擬與疲勞壽命預測之可靠度分析流程，以提供風力發電機齒輪箱設計階段之壽命預測與可靠度評估。本研究首先導入彰化線西地區實測逐時風速資料，據以模擬齒輪系統於隨機風場變化下之動態響應，進而計算各齒輪赫茲接觸應力與齒根彎曲應力；所得應力歷程再透過雨流計數法 (Rainflow Counting Method)、應力壽命法 (Stress-Life Method) 與 Miner 累積損傷法則 (Miner's Rule) 換算成疲勞累積損傷指標，據以評估風力發電機齒輪箱之壽命分布與可靠度。 鑑於風速具高度隨機性，齒輪損傷與壽命亦呈現顯著不確定性，本研究應用蒙地卡羅法 (Monte Carlo Method) 產生大量失效樣本，先依經驗分析法(Empirical Method) 建立可靠度函數，而後結合最大概似估計法 (Maximum Likelihood Estimation, MLE) 與 Akaike 資訊量準則 (Akaike Information Criterion, AIC)，比較多種機率分布模型之適配性。模擬分析結果顯示，風力發電機齒輪箱壽命最符合對數常態分布，平均失效年限為 29.26 年，且於第 29 年後可靠度顯著下降。本研究所建構之分析流程兼具可操作性與擴展潛力，除可應用於風力發電機齒輪箱在特定隨機風場下之壽命分布與可靠度評估外，亦兼可作為齒輪箱之預防性維護與壽命管理決策之輔助工具。

The wind turbine gearbox is a critical transmission component that bears variable loads and transmits mechanical power. Due to its high maintenance cost, conducting reliability assessment during the design phase is essential. This study establishes a reliability analysis framework that integrates dynamic simulation and fatigue life prediction to support lifetime estimation and reliability evaluation in early design stages. Real hourly wind speed data from Xianxi, Changhua, are used to simulate the dynamic response of the gear system under stochastic wind field variations. Time-varying Hertzian contact stress and tooth root bending stress are calculated and converted into cumulative fatigue damage indicators using the rainflow counting method, stress-life method, and Miner’s rule, enabling the estimation of lifetime distribution and system reliability. Given the high stochasticity of wind speeds, gear damage and lifetime show significant uncertainty. The Monte Carlo Method is employed to generate a large number of failure samples. An empirical reliability function is constructed, followed by statistical fitting using the maximum likelihood estimation (MLE) method and comparison via the Akaike information criterion (AIC). Simulation results indicate that the gearbox lifetime best fits a lognormal distribution, with an average failure time of approximately 29.26 years and a notable drop in reliability after year 29. The proposed framework demonstrates strong operability and extensibility. It can be applied not only to reliability evaluation under specific stochastic wind conditions but also as a decision-support tool for preventive maintenance and lifetime management of wind turbine gear systems.