Please use this identifier to cite or link to this item:
Development of a Novel Risk Assessment Method Using Fuzzy Modelling and Its Application in Maritime Safety
|Contributors: ||NTOU:Department of Merchant Marine|
fuzzy-rule-base approach;fuzzy modeling;risk assessment
|Issue Date: ||2012-04-13T01:47:40Z
|Abstract: ||摘要:風險評估為運用科學的方法，針對變數資料或資訊進行系統化分析，將風險 量化並進行排序。當資料充足時風險評估得以統計學為基礎並應用如錯誤樹分析 或事件數分析等方法進行分析，然在海事風險評估方面，因為歷史數據的不存在 或不足夠，導致上述傳統方法無法進行有效評估。 有鑒於此，部分以模糊理論為基礎之近似推論研究可將專家針對風險之質化 意見轉為量化數值，進而建立風險排序。然傳統模糊理論仍有不足之處：當語言 詞歸屬函數圖形為一個非完全評估之情況時，會導致模糊資料相近之各情境無法 鑑別出風險值之高低；傳統方法常用之模糊推論法在針對變數權重實施敏感度分 析時，將會鈍化模糊推論之結果，該現象在變數增多的情況下更為嚴重；傳統方 法對於法則之前項語言詞代表值之間均為等差關係之假設以及後項語言詞門檻 範圍之設定方式，將會使語言詞歸屬函數模型中各語言詞間之相對關係無法被正 確表示，而影響模糊結論之鑑別度，造成不合理之排序。本計畫預計建構一套新 模糊風險評估方法論，可改善上述傳統方法之缺點，其步驟為建立完全評估語言 詞歸屬函數模型，使用新語言詞歸納法建立模糊法則庫，以Product-Max 模糊推 論法得到模糊結論，最後運用最大歸屬度平均法解模糊得出風險排序值。本研究 預計以海難報告及學術文獻之案例進行分析，以驗證本研究之方法論較傳統方法 可得到更精確且合理之風險排序。|
Abstract:Risk assessment can be regarded as a systematical study based on scientific methods to evaluate risks based on statistical data or information. A reasonable risk assessment is capable of providing a list of scenario risk priority. A variety of risk assessment techniques has been developed. When the data under consideration is sufficient, traditional methods such as Fault Tree Analysis (FTA) or Event Tree Analysis(ETA) are commonly applied. However, in circumstances where the lack or incompleteness of data exists especially in the domain of the maritime risk assessment, such methods are incapable of generating reasonable risk results. Approximate reasoning based on fuzzy set theory, can be regarded as one of the methods cable of providing appropriate risk outcomes under the aforementioned circumstances. However, such approaches still have drawbacks. First, when the membership functions of linguistic terms do not overlap appropriately causing the summation of the memberships does not equal to 1, the distinguishability of the results generated is questioned in conditions where the fuzzy data of scenarios are close to each other. Secondly, Min-Max approaches are often applied to acquire fuzzy conclusions. However, such a method may cause the loss of important information when conducting sensitivity analysis for the weights of each variable. The phenomenon deteriorates when factors considered are numerous. Thirdly, the representative values for each linguistic term in the fuzzy rules are in equal-difference relationship. The fuzzy modeling developed based on such a practice may not generate reasonable results if the distribution of membership functions is extremely disproportionate. In this study, a methodology capable of resolving such difficulties is proposed. This is achieved by first the development of a new set of linguistic terms and the summation of the memberships of such terms is set to 1. This is followed by the development of a fuzzy rule base using a new algorithm for the linguistic term selection for the fuzzy rules. The fuzzy conclusion obtained will then be defuzzified using the Product-Max method. Finally a risk ranking can be established. The methodology will be demonstrated using the cases studies selected from marine casualty reports and academic literatures.
|Appears in Collections:||[商船學系] 研究計畫|
Files in This Item:
There are no files associated with this item.
All items in NTOUR are protected by copyright, with all rights reserved.