橋梁在其生命週期中，無可避免地會產生結構性或經常性的損壞，而為瞭解橋梁之狀況並預防損壞問題之擴大，橋梁檢測已成為維護工作重要之一環。橋梁定期檢測極為重要且有期程之限制，為防止檢測期程延遲並提升檢測效率，在考量人力及成本的限制下，如何於有限之檢測期程內妥善規劃檢測之行程，已成為橋梁維護之重要課題。 本研究首先利用專家問卷篩選影響橋梁目視檢測時間之因子，而影響目視檢測時間之因子為21項『臺灣地區橋梁管理資訊系統』模組資料欄位。問卷調查結果顯示主要影響因子為(1)總橋孔數、(2)是否為跨河橋、(3)橋長、(4)橋版面積、(5)橋下淨高和(6)橋梁劣化構件數。接著本研究利用六項時間影響因子與橋梁檢測時間進行多元線性迴歸，以建立評估橋梁目視檢測時間模式。而多元線性迴歸式將(1)總橋孔數、(2)橋版面積和(3)橋梁劣化構件數作為重要時間影響因子。 本研究旨在建構橋梁目視檢測行程最佳化模式。為令模式更有彈性並具實務價值，本研究除考量橋梁檢測時間外，亦將每日可工作之時間納入考量，並採用多路線之設計，以達成目視檢測時間為最短之目標。本研究之行程最佳化模式共分兩階段，第一階段利用四項啟發式規則(Heuristic Rule)建立行程規劃之初始解，而第二階段以基因演算法(Genetic Algorithm，GA)改善初始解。本研究之模式建立完成後，採用桃園縣政府轄區內鄉道橋梁進行案例驗證。 本研究完成之成果包含(1)篩選影響橋梁目視檢測時間之因子，(2)建立評估橋梁目視檢測時間之數學模式，以及(3)建構橋梁目視檢測行程最佳化模式。藉由此橋梁目視檢測行程最佳化模式之建立，橋梁檢測單位可縮短檢測期程，提升目視檢測之效率，並可依本模式建立之檢測行程規劃結果進行檢測進度之控管。 Taiwan is a small island with area of 36,188 square kilometers. Based on a recent study, around 30% of the 28,000 bridges on the island are over 30 years old. Therefore, bridge visual inspection has become a major task of bridge maintenance to identify deteriorations and damages. The current national standard of bridge inspection in Taiwan follows the DER&U methodology, which is an evaluation system based on the degree, extent, relevancy, and urgency of the faults identified for all structure components. There are usually time limitations for bridge management agencies to complete their bridge inspections. Therefore, developing an efficient bridge inspection plan has become a crucial issue for all bridge management agencies. The goal of this research is to establish a heuristic approach for optimizing inspection routes with minimum workday numbers. A series of surveys was conducted to identify the potential factors that may affect the bridge inspection time. A multiple regression model based on the potential factors was then established to estimate the bridge inspection time. Finally, A two-phase approach was established in this research to optimize the bridge inspection routes for any given district. The first phase utilized four heuristic rules obtained from experienced route planners to generate an initial solution; the second phase then improved the initial solution by using Genetic Algorithm. A computer program was designed to perform the two-phase approach. A real world example with around 200 bridges was studied to validate this proposed approach.
