鋪面養護模式對於道路養護單位及業者一直扮演重要的角色，隨著經濟的發達與發展，臺灣地區的民眾與產業更加仰賴公路串連生活圈及產業鏈，公路的服務品質一向是普羅大眾所關心的重點，現行臺灣地區的省道由交通部公路總局進行養護及管理，然而自2004年起公路總局之養護預算逐年減少，相對其逐年增加之養護里程，如何在有限的資源下將養護公路面積盡量達到最大為公路總局近年面對的問題。為實際了解現行臺灣道路養護狀況，本研究訪談目前臺灣道路養護單位及蒐集鋪面狀況資料，並運用統計學之迴歸分析及人工智慧方法建構鋪面成效預測模型，線性迴歸分析結果顯示其關係式可用來進行未來鋪面成效之預測，雖然支持向量迴歸分析之判定係數R2較線性迴歸分析之R2高，然而因氣象資料較難從現地檢測取得，因此本研究使用線性迴歸結果之關係式進行後續研究預測；為計算省道預期總養護成本，本研究採用美國聯邦公路局開發之生命週期成本分析軟體RealCost進行計算，藉由線性迴歸分析之結果及相似之交通量來挑選作為單位成本計算之道路，經計算後發現省道一年養護總共需52億台幣，其已超過公路總局2014年度決算數7億之多，將該數值代入關係式後得到預期之IRI值為2.60，代表公路總局每年花費52億時其預期之道路IRI值為2.60，結合臺灣與美國各州之IRI標準後，若公路總局期望其省道IRI在良好需至少花費39億台幣，而花費75億台幣則可使IRI值達到非常好之狀況，因此，本研究認為利用鋪面成效預測式及RealCost軟體可以提供道路機關一制定鋪面養護策略之參考。;Pavement Maintenance Model has been playing an important role in the pavement industry. As the economy and technology progress, people and industries in Taiwan rely more on highway systems to connect their livings. In this case, highway’s service quality has always been a public concern. The Taiwan Provincial highways currently are managed and maintained by the Directorate General of Highways(DGH). However, the maintenance budget of DGH has been decreasing since 2004, while the maintenance mileage of Provincial Highway has been increasing. Hence, how to maximized the maintenance efficiency under limited source and budget has become a great challenge to DGH. In order to understand the current highway pavement maintenance status, this research visited several Taiwan pavement maintenance agencies and collect pavement condition data. After the maintenance status and cycle are known, statistic method and artificial intelligence model are used to construct pavement performance forecast model. The result of linear regression analysis showed that using correlation equation from the linear regression analysis is capable of forecasting future pavement performance. Although the result of support vector regression has better R2(0.9) than linear regression (0.7), the weather data is difficult to obtain from field inspection. Therefore, this research uses the correlation equation from the linear regression analysis to perform future pavement performance prediction；In order to calculate the Provincial Highway’s total expected maintenance cost, this research use RealCost which is the Life-Cycle Cost Analysis (LCCA) software developed by Federal Highway Administration (FHWA) to calculate the overall maintenance cost. A case study for calculating the unit maintenance cost has been chosen through the result of linear regression and the similarity of traffic volume. After the calculation the agency cost of Provincial Highway is NTD 5.2 billion per one year, which is NTD 7 billion higher than the DGH 2014 final cost. By inputting the total expected cost into the correlation equation the predicted IRI is 2.60, which means that if DGH spends NTD 5.2 billion per one year in pavement maintenance the expected IRI is 2.60. Combine and input the IRI standard from United State and Taiwan Highway agency, if DGH want to maintain their pavement in Fair condition, the agency will have to spend at least NTD 3.9 billion, while spending at least NTD 7.5 billion will keep the pavement in very good condition. This research believes by using the equation and the RealCost software can provide Highway agencies a more viable option in pavement maintenance strategy.
