An investigation of flight vehicle performance parameters via the modern Hilbert-Huang transform and the improved sequence alignment methods

淡江大學機構典藏 > 工學院 > 航空太空工程學系暨研究所 > 學位論文 > Item 987654321/88054

請使用永久網址來引用或連結此文件: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/88054

題名:	An investigation of flight vehicle performance parameters via the modern Hilbert-Huang transform and the improved sequence alignment methods
其他題名:	利用現代希爾伯特-黃轉換及改進的序列比對法之飛行性能參數研究
作者:	鄭宇軒;Cheng, Norman
貢獻者:	淡江大學航空太空工程學系碩士班宛同;Wan, Tung
關鍵詞:	飛航操作品保系統;希爾伯特-黃轉換;改進的基因序列比對法;二維模擬拍撲翼空氣動力學;本質模態函數錯位現象;飛行資料分析;信號分析;ATR-72;生物資訊學;Flight Operational Quality Assurance (FOQA);Hilbert-Huang Transform (HHT);Improved Sequence Alignment (ISA);2-D Simulated Flapping Wing Aerodynamics;IMF Staggering Effect;Flight Data Analysis;Signal Processing;ATR-72;Bioinformatics
日期:	2012
上傳時間:	2013-04-13 11:58:41 (UTC+8)
摘要:	本研究的主旨是利用希爾伯特-黃轉換（Hilbert-Huang Transform, HHT, 包含Empirical Mode Decomposition, EMD和Hilbert Spectral Analysis, HSA）及改進的基因序列比對法（Improved Sequence Alignment, ISA）分析飛行操作品質系統（Flight Operational Quality Assurance, FOQA）之飛行性能參數資料及二維模擬拍撲翼之空氣動力係數。本研究最初目的是利用HHT分析原始訊號，從隱藏且以頻率為基底的本質模態函數群（Intrinsic Mode Functions, IMF）找出規律性的模式。盡管EMD作為一種濾波器，透過不同的頻率層級分解出具有數學意義的IMF，但不一定具有強健的物理意義。EMD的特性是將原始訊號分解出8~12個IMF，這增加分析問題的困難度，尤其是物理意義不顯明的函數群。於是本篇論文研發之改進的基因序列比對法（Improved Sequence Alignment, ISA）用於量化IMF函數群，以統計的方式記錄函數群的五個本質字母（5-Elemental Alphabets），並且存放置基因譜（Genetic Profile），用於實現兩個序列的序列比對。本研究裡，使用二維模擬拍撲翼在固定風場下的升阻力訊號作為基準，以證明HHT和ISA整合的實用性，因為其結果是可預測的。實際的案例則以ATR-72班機的三把正常航班及一把異常航班為例，異常航班的引擎資料是一把No.1引擎在巡航階段空中熄火的狀況，以系統和方法分析之。IMF錯位現象（IMF Staggering Effect）被發現在拍撲翼的HHT分析結果，並且從ISA的排列方式得到證實。飛機扭力參數的序列排列資料說明潛在因子可被偵測的可能性，但是因子仍然不明顯。這個原因主要因為引擎參數變化通常發生在為秒之間，原始資料則是1Hz紀錄，造成資料先天不精準性。拍撲翼的分析為基準，證實了IMF錯位現象可被ISA排列出來，譬如Lift和Lift_2ms_h有著sequence A = (c0, c1, c2, c3, -, c4, c5, c6, c7, c8, c9, c10, c11, c12)和sequence B = (c0, -, c1, c2, c3, c4, c5, c6, c7, c8, c9, c10, c11, c12)的排列。另外，頻率相似度F是支配排列結果的重要因子，以拍撲翼為例，在升力排列的出現次數為37.10%，阻力為47.89%。飛行資料中找出相似模式的機率為55.13%。本研究的結論是證實HHT和ISA的可行性，未來展望期望加入更進階的經驗搜尋法則，從排列的結果中更快速找到相似模式，從排列的模式裡還另外找出資料同源的可能性（Homology）。 A set of Flight Operational Quality Assurance (FOQA) flight data and a set of flapping wings lift and drag computational fluid dynamics (CFD) results were examined by the aid of Hilbert-Huang Transform (HHT) and Improved Sequence Alignment (ISA) method. The purpose of applying HHT to data is to find the hidden frequency-based (Intrinsic Mode Functions) IMF patterns from the origin signal. Despite HHT works as a filter to extract the variability of signals with different frequency levels and has the ability to decompose original signals into several mathematical subcomponents, the results often being very hard to explain the physical meaning. It is often difficult to analyze a population of data when these data were recorded. Hence, the ISA method of quantizing all the signals to statistical measurements by statistical measurements was invented here, and was applied for data dimensionality reduction. The approach of this thesis study is to first apply HHT and ISA on flapping wing CFD data as the benchmark since the expected aligned results are explicit and then apply the system and methods on FOQA data. The flight data of three normal flights and one abnormal flight, a case of engine No.1 failure during cruise flight, was extracted from a specific ATR-72 twin-engine turboprop airliner. The IMF staggering effect was first found in flapping wings HHT results and was fully verified via ISA. Several important alignment results of different flight phases were found in the parameter Torque of ATR-72 flight data that indicates the existence of potential risk factors before the event flight. The alignment results of flapping wings are well as a very good benchmark case, but the alignment results of FOQA engine failure during operation in flight are vague. This reason is majorly because of the engine failure event usually appears to be a sudden change within milliseconds. The data time interval used in current work is 1 Hz. This is considered as the congenital missing of data that leads to a fundamental problem lacking of precise measures. Facts are found using HHT and ISA method. The benchmark case of aligning Lift and Lift_2ms_h as sequence A = (c0, c1, c2, c3, -, c4, c5, c6, c7, c8, c9, c10, c11, c12) and sequence B = (c0, -, c1, c2, c3, c4, c5, c6, c7, c8, c9, c10, c11, c12). The case shows perfect alignment with staggering effect included. Another important fact is the F (degree of frequency measurement) of the 5-elemental alphabets is a dominating factor since HHT is frequency-based. The probability of formats containing alphabet F in Lift is 37.10% and Drag is 47.89%, indicating alphabet F is very common in alignments and in IMF as well. For FOQA data alignments, the pattern finding probability is approximately 55.13%. An empirical system and procedure for pattern searching is needed to handle a population of data. The hidden patterns were found in both cases and the concept of data homology was illustrated.
顯示於類別:	[航空太空工程學系暨研究所] 學位論文

文件中的檔案:

檔案	大小	格式	瀏覽次數
index.html	0Kb	HTML	291	檢視/開啟

在機構典藏中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....