「臺灣地區橋梁管理資訊系統」目前系統管理問題，主要為使用者資料填寫不全及資料填寫錯誤，若系統中擁有一套資料偵錯並防止使用者填寫錯誤之機制，不但可提升系統資料之可信度，也能提升後續進行橋梁維修經費之管理效能。因此，本研究提出一套偵錯機制，將系統中之基本資料以及檢測資料中明顯錯誤挑除，並進而防止未來可能之填寫錯誤，以提升系統資料之正確性。 本研究首先將基本資料的錯誤態樣，分為基本資料誤填錯誤以及欄位資料相衝突兩類；透過專家訪談以及資料蒐集後，統整各資料表欄位之合理區間或給予限制條件，篩選出錯誤之基本資料並進而避免未來填寫資料時發生誤填之情形。在檢測資料偵錯方面中，本研究先將橋梁做初步分類，利用集群分析法找出有相似條件之橋梁，並分析同一群組中之橋梁可能的劣化情形，並藉由劣化構件間之關聯性，分析各構件可能之D值，以建立一套判斷填寫錯誤之參考基準。 經實際測試後，本研究所建立之資料偵錯機制確實可篩選出系統中錯誤之基本資料以及檢測資料，可驗證本研究成果對於提升系統資料之正確性具有相當之成效。 ;The major management problem of the Taiwan Bridge Management System (TBMS) is existing of incomplete or error data input by the users. A set of error data detection mechanisms help both in increasing reliability of system data and in effectiveness of bridge maintenance. Therefore, this research aims at establishing such error data detection mechanisms to screen out for correction obvious errors exist in the inventory and in the inspection databases of the TBMS. Furthermore, error proof data input mechanisms are also generated in this research to prevent inputting of incorrect data in the future. For the inventory database, this study categorizes error data patterns of basic data into two types; wrong input of data and data conflicts between fields. After expert interviews were conducted and typical values for bridge attributes were collected, reasonable intervals of various fields and valid relationships among data fields were summarized to become rules for finding error data in the inventory database. In the aspect of detecting possible inspection data, this study classifies bridges into several types first and then uses a cluster analysis method to group bridges with similar characters. In a group, possible D values for various components are analyzed from which components with possible wrong D inspection values can be found. Several tests were performed in the TBMS using the mechanism generated by this research; typos and unreasonable data both in the bridge inventory and in the bridge inspection database were actually found and corrected accordingly. These tests prove that results of this research are meaningful and useful in maintaining the TBMS.
