【摘要】：海洋平臺作為海上油田開發(fā)的重要設備,其兼有結(jié)構(gòu)復雜和重量大的雙重特點,這種大型工程結(jié)構(gòu)物的健康狀況是海上油田開發(fā)的關(guān)鍵所在。近幾年,發(fā)生不少海洋平臺由于疲勞破壞而引起垮塌的案例。對于海洋平臺而言,其損傷在海洋平臺結(jié)構(gòu)的服役期間是不可避免的。海洋平臺結(jié)構(gòu)的健康監(jiān)測與損傷診斷已成為刻不容緩的重要課題。采用振動測量和分析技術(shù)對海洋平臺結(jié)構(gòu)進行識別,既能對平臺結(jié)構(gòu)的完整性適時地做出總的評價,又能為系統(tǒng)識別提供必要的數(shù)據(jù)�，F(xiàn)有的海洋平臺健康分析方法主要是針對海洋平臺結(jié)構(gòu)或者力學角度分析,此類方法對于海洋平臺大型機械作業(yè)長期累積作業(yè)狀態(tài)分析很有成效,但實時的監(jiān)控海洋平臺健康狀態(tài)分析,以及建立一個系統(tǒng)的評價方法來評價其實時作業(yè)環(huán)境的優(yōu)劣問題,還沒有解決。本文運用馬田系統(tǒng)的理論方法,將一種新的多元數(shù)據(jù)定量決策的模式識別方法應用到海洋平臺數(shù)據(jù)分析中。以“102號”海洋平臺數(shù)據(jù)為例,建立了海洋平臺健康狀態(tài)分析系統(tǒng),通過運用MATLAB、SPSS、MINITAB等數(shù)據(jù)處理分析軟件對各個監(jiān)測傳感器獲得監(jiān)測量數(shù)據(jù)進行處理,將提取到的正常與異常狀態(tài)分析樣本進行分類,使用正交表篩選所建系統(tǒng)的特征變量,分析信噪比數(shù)據(jù)結(jié)果得到優(yōu)化后的特征數(shù)據(jù)集合。最后,采用ROC曲線分析和費歇爾判別分析兩種閾值確定方法進行馬氏距離臨界值的計算。通過對比兩種方法的計算結(jié)果,確定出更準確的閾值范圍。作為海洋平臺健康狀態(tài)分析的新方法,為后期未知樣本提供一種模式識別與預測的新方法。
[Abstract]:As an important equipment for offshore oil field development, offshore platform has the dual characteristics of complex structure and heavy weight. The health condition of this large engineering structure is the key to offshore oil field development. In recent years, there have been many cases of offshore platform collapse due to fatigue damage. For the offshore platform, the damage is inevitable during the service of the offshore platform structure. The health monitoring and damage diagnosis of offshore platform structure has become an important task without delay. Using vibration measurement and analysis technology to identify offshore platform structure can not only evaluate the integrity of platform structure in good time but also provide necessary data for system identification. The existing health analysis methods of offshore platforms are mainly aimed at the structural or mechanical analysis of offshore platforms. This kind of method is very effective for long-term cumulative operation state analysis of large-scale mechanical operations of offshore platforms. However, the problems of real-time monitoring and monitoring of the health status of offshore platforms and the establishment of a systematic evaluation method to evaluate the merits and demerits of real-time operating environment have not been solved. In this paper, a new pattern recognition method of multivariate data quantitative decision is applied to the data analysis of offshore platform by using the theory method of Martian system. Taking the data of "102" offshore platform as an example, the health status analysis system of offshore platform is established. By using MATLAB,SPSS,MINITAB and other data processing and analysis software, the monitoring and measuring data obtained from each monitoring sensor are processed. The extracted normal and abnormal state analysis samples are classified, and the orthogonal table is used to filter the feature variables of the system, and the optimized feature data set is obtained by analyzing the results of SNR data. Finally, two threshold determination methods, ROC curve analysis and Fischer discriminant analysis, are used to calculate the critical value of Markov distance. By comparing the calculation results of the two methods, a more accurate threshold range is determined. As a new method of health state analysis for offshore platform, it provides a new method for pattern recognition and prediction for later unknown samples.
【學位授予單位】：大連海事大學
【學位級別】：碩士
【學位授予年份】：2015
【分類號】：TE95

【引證文獻】

相關(guān)會議論文 前1條

1 曾江輝;曾鳳章;陳嵩輝;;基于支持向量機的馬田系統(tǒng)閾值確定方法研究[A];第三屆中國質(zhì)量學術(shù)論壇論文集[C];2008年

，

本文編號：2278022