인공신경망을 이용한 교량의 주행중 차량하중분석시스템 개발

Abstract

도로교의 설계 및 해석을 위한 설계 활하중 모델의 개발이나 교량의 잔손 수명 예측을 위한 피로하중모델 등의 개발을 위해서는 국내 도로 실정을 반영한 차량하중의 분석이 필요하다. 이에 본 연구에서는 교량을 주행하는 차량의 하중 및 주행 정보를 얻기 위하여 교량 상부구조의 하부 변형률 신호와 인공신경망기법을 이용하여 이러한 정보를 분석하는 시스템을 개발하였다. 신경망 학습 및 테스트를 위하여 수치 시뮬레이션을 통한 학습 데이터 생성이 아닌 실제 주행하고 있는 다양한 임의차량 시험데이터를 이용하였다. 또한 계량소에서 정적 중량을 계측한 시험차량들을 반복 주행시켜 얻은 데이터로 개발된 시스템의 정확도를 평가하였다.; The analysis of vehicular loads reflecting the domestic traffic circumstances is necessary for the development of design live load models in the analysis and design of highway bridges or the development of fatigue load models to predict the remaining lifespan of the bridges. This study intends to develop a ANN (artificial neural network)-based method for the analysis of data to obtain information concerning the loads and running conditions of vehicles crossing bridge structures exploiting the signals measured by strain gauges installed at the bottom of the bridge superstructure. This study relies on experimental data corresponding to the crossing of hundreds of random vehicles rather than on theoretical data obtained through numerical simulations to secure training data for the training and test of the ANN. In addition, data acquired from 3 types of vehicles weighted statically at measurement station and crossing the bridge repeatedly were also exploited to examine the accuracy of the trained ANN. Results obtained by using the ANN-based analysis method, the girder influence line analysis method and the lateral cross beam influence analysis method considering the local behavior of the bridge were compared by means of two examples that are a cable-stayed bridge and a PSC I girder bridge. The values resulting from the developed method were compared to the static values measured at the measurement station together with the values obtained from Low-Speed WIM system (cable-stayed bridge) and High-Speed WIM system (PSC I girder bridge). Moreover, dynamic analysis method was also applied through dynamic modeling of the vehicles in order to perform more accurate analysis of the axle loads. In view of the results related to the cable-stayed bridge, the cross beam ANN analysis method appears to provide more remarkable load analysis results than the cross beam influence line method, and to offer the possibility to conduct ANN-based load analysis even for bridge types of which complexity impeded previous influence line methods to carry out accurate analysis. Consequently, for bridges in which the cross beams are fastened to the deck slab, the ANN-based load analysis using the strain gauge signals of the cross beams is verified to be more accurate than the influence line method using the strain gauge signals of the girder. In addition, the application of the dynamic analysis through dynamic modeling of the vehicles being theoretically considering the effects of the dynamic vibration known as the factor inducing error during load analysis made it possible to also achieve accurate axle loads analysis.