WEKO3
アイテム
Separation of Gravity Anomaly Data considering Statistical Independence among Signals : Application to Severely Contaminated Data Obtained by Prototype Mobile Gravimeter
http://hdl.handle.net/2261/53029
http://hdl.handle.net/2261/530298321be2f-3137-49d1-a8eb-835afe7af1d4
名前 / ファイル | ライセンス | アクション |
---|---|---|
![]() |
|
Item type | 学位論文 / Thesis or Dissertation(1) | |||||
---|---|---|---|---|---|---|
公開日 | 2013-01-22 | |||||
タイトル | ||||||
タイトル | Separation of Gravity Anomaly Data considering Statistical Independence among Signals : Application to Severely Contaminated Data Obtained by Prototype Mobile Gravimeter | |||||
言語 | ||||||
言語 | eng | |||||
資源タイプ | ||||||
資源 | http://purl.org/coar/resource_type/c_46ec | |||||
タイプ | thesis | |||||
その他のタイトル | ||||||
その他のタイトル | 信号データ間の統計的独立性を用いた重力異常データの分離 : 可搬型重力測定機の試作機による観測記録への適用 | |||||
著者 |
Khatri, Prem Prakash
× Khatri, Prem Prakash |
|||||
著者所属 | ||||||
著者所属 | 東京大学大学院工学系研究科社会基盤学専攻 | |||||
Abstract | ||||||
内容記述タイプ | Abstract | |||||
内容記述 | The ground motion (GM) characteristics are affected by local subsurface structure. Gravity method is one of the useful methods to know the information on subsurface structure. The gravity anomaly data obtained by gravity survey can be correlated with the lateral variation of subsurface rock densities. For gravity survey, spring type gravimeter has been used so far. This gravimeter gives accurate resolution but they are very expensive and diffcult to handle. Recently, Team Morikawa have developed a prototype mobile gravimeter that uses Force-Balanced (FB) accelerometer. This prototype is light weight, compact, easy to handle and inexpensive. It also offers the resolution that is good enough for preparing gravity map for subsurface modelling. However, unlike the conventional spring-type gravimeter, this newly developed FB gravimeter is highly sensitive to high frequency noise. The observed data by this gravimeter are easily contaminated by various kinds of disturbances in a small size carrier like engine vibration, carrier acceleration, wind velocity and carrier tilting accompanied by sensor drifts, electrical noise etc. The amplitudes of such noises can be upto 100,000 times larger than the gravity anomaly. In order to extract the gravity anomaly from such observation, data processing is essential. Conventionally, the data was observed in a large carrier (ship) on a more stable environment and the sensor was not sensitive to high frequency noise, so the noise contamination was not severe. The data processing techniques like low pass filtering and Second order statistics method (such as SOBI) were used. However, in case of severely contaminated data, low pass filtering might not be enough. SOBI is an advanced blind source separation (BSS) technique that separates source and noise blindly by exploiting the statistical property of data. It separates the target source by assuming that source and unwanted data are un-correlated at various time-lags. The gravity anomaly and other noises are generated from independent physical sources. It can be safely assumed that gravity anomaly and other data are independent but, it can not be strictly claimed that they have no correlation. So, further improvement than second order statistics method is desired. As a scheme of considering independence of signals to blind source separation, Independent Component Analysis (ICA) has been used in the field of BSS since 1990's. It separates the sources by maximizing the independence of linearly transformed observed signals. Both mixing matrix and source signals are identified when only the mixed data are available. Further, independence between signals has nothing to do with their amplitudes. The huge difference in amplitudes among gravity anomaly and noise does not affect their independence. So ICA is suitable for our purpose. ICA renders ambiguity in amplitude of separated signal but this problem has little significance in our case since an appropriate scalar multiple can be estimated with the help of information of gravity at few known points. Thus it is proposed to use ICA for separating gravity anomaly data from its mixture with several noises. The survey data is observed at Toyama bay, Japan. The National Institute of Advanced Industrial Science and Technology (AIST), Japan has provided the gravity map for the same place. This map is used to calculate the reference data that facilitates us to verify the performance of the proposed scheme. The prototype gravimeter consisted of group of sensors. Since ICA requires at least two sets of data, the major data obtained by Analog servo (VSE) was combined with data by other sensors as supplementary data. Following Team Morikawa's approach, the performance of various sensors are compared. The application of low pass filtering(LPF) as a pre-processing to ICA is realized to be important. The presence of high frequency noise in the data is found to be unfavourable for the separation of gravity anomaly data. Both SOBI and ICA work only after the application of LPF. The choice of an appropriate cut-off filter was also observed to affect the results. The combination of VSE data and vertical component of Accelerometer Titan (Taurus-Z) as an iput to ICA gives good result. When other horizontal components were used with VSE data the results are not satisfactory. Further, ICA is found to perform better at certain conditions of data acquisition environment. At the portions when ship motion is unidirectional the trend of ICA separated data is harmonious with reference data. When the ship velocity was lesser while proceeding towards the sea, the ICA result is matching very well with reference data. When the ship was highly unstable during ship stopping time ICA result are deviating away from the reference data. At other relatively stable sections the ICA separated data follows the trend of reference data well. The separation of input data by ICA into different output components verifies that the source gravity anomaly and other data are independent. Thus it satisfies our assumption. The harmony of ICA separated data with trend of reference data at major sections verifies the applicability of ICA, under certain data acquisition environments. The accuracy of properly separated data by ICA is good enough for preparing gravity map for the purpose of subsurface modelling. However, there is still a room for further improvement. An effort is made to study time-frequency characteristics of data without observing any clear merit so far. The further improvement in methodology is considered to be the part of future works. Based on the results and considering the applicability of ICA so far, it can be concluded that a positive sign is observed for the improvement of mobility of gravity method. | |||||
書誌情報 | 発行日 2012-09-27 | |||||
日本十進分類法 | ||||||
主題 | 501 | |||||
主題Scheme | NDC | |||||
学位名 | ||||||
学位名 | 修士(工学) | |||||
学位 | ||||||
値 | master | |||||
研究科・専攻 | ||||||
工学系研究科社会基盤学専攻 | ||||||
学位授与年月日 | ||||||
学位授与年月日 | 2012-09-27 |