車輛反應式平坦儀之平坦度指標演算法

Abstract

國內各縣（市）對於道路平坦度之重視程度逐年增高，亦見採用科學儀器進行量測之概念，目前國際上較為客觀且常見之平坦度評估指標為國際糙度指標（IRI），然其檢測成本較高，難以廣為應用為經常且全面之平坦度量測工具，導致鋪面管理所需之長期性平坦度資料不易建立。有鑑於此，本研究以具有經濟性之簡易型車輛反應式平坦儀作為主要量測設備，建立一量測車輛反應垂直加速度為主之平坦度指標演算法，應用傅立葉轉換（Fourier Transform）分析車輛反應垂直加速度對於影響道路行駛舒適性之敏感波長範圍，同時搭配巴特沃斯濾波器（Butterworth Filter）進行加速度訊號之濾波分析，以提升檢測結果與IRI間之相關性。經道路實測及數據分析後，本研究發現保留波長為7.5至30公尺之加速度值所計算出之均方根能與IRI有相當良好之相關性。經由一系列的演算及兩次速度調整後，研究建立了與IRI具有相同尺度範圍之加速度均方根指標（AARI）。此外更以不同檢測速度行駛於大量之實測道路來驗證此演算法之適用性，結果顯示經本研究所建立之演算法計算後，能夠使AARI與IRI之判定係數R2從原先之0.50大幅提升至0.88，證實本研究所提出之AARI指標能夠相當反應路面之平坦度與舒適性。 本研究藉由多次道路實驗數據之分析結果，探討車型與懸吊系統差異、簡易型平坦儀之擺放位置、道路坡度、GPS資料遺失及儀器擷取頻率等因素之影響。結果顯示利用希爾伯特黃轉換（HHT）之總體經驗模態分解法（EEMD）能夠解決不同車型與懸吊系統對於檢測時之差異影響；且利用量測速度與距離資料之內插計算可大幅降低GPS資料遺失之影響。最後本研究亦將簡易型平坦儀納入鋪面糙度績效分析，進行大規模之200車道公里市區道路實測，並利用邏輯斯迴歸（Logistic Regression）與接收者操作特徵（ROC）曲線進行敏感度分析，訂定一篩選正確率高達94%之AARI市區道路養護門檻值，作為日後道路管養單位之維修參考標準。

Inertial profilers have been the mainstream measuring device on pavement roughness detecting for decades, but many local government agencies in Taiwan still using profilographs and other simple measuring devices due to relatively high cost of inertial profilers. In this research, a response-based measuring device （RBD） was first introduced, followed by a computational algorithm to enhance the analysis procedure of measured acceleration data. Then, the effects of several experimental factors were also considered. The developed algorithm has several steps, including speed normalization on acceleration data, conducting Fourier transform, and applying Butterworth filter which includes low-pass and high-pass filter on the acceleration data of the frequency domain. Four test sections were surveyed by both RBD and inertial profiler. For any given measuring speed, the retention wavelength of acceleration between 7.5-meter and 30-meter will result in the best index calculation. Through a series of calculation procedures and a double measuring speed calibration, adjusted acceleration root-mean-square index （AARI） which has the same scale of measure with IRI was developed. The algorithm was successfully verified by additional city-road field survey. The coefficient of determination R2 between AARI and IRI has been significantly improved from 0.50 to 0.88 after the enhancement procedure was conducted. In order to evaluate the roadway roughness accurately, the effects of several factors such as suspension system, device placement, roadway slope, missing data on GPS and collection frequency of device were studied. According to the experimental results, the influcence of different suspension systems on AARI can be significantly reduced by ensemble empirical mode （EEMD） decomposition method of Hilbert-Huang transform （HHT）. Also, the calculation of interpolation can solve the loss of GPS position coordinates. The study also conducted a large-scale 200-km network city-road survey. The logistic regression and receiver operating characteristic curve methods were performed for the sensitivity analysis to determine appropriate maintenance threshold for screening roadway network. The result shows that the accuracy rate of selecting pavement sections for maintenance is 94%.