Prediction of ground vibration caused by road traffic using neural network

点击翻译按钮获取中文摘要

This research consists of two subjects.In the first subject, an estimation method of soil properties is proposed to predict simply the attenuation with distance on ground vibration caused by road traffic. This method is based on neural network in which a real ground is idealized by a half space. In the first year, the fundamental study is performed to build the neural network and it is shown that its system is useful. In the second year, viscous damping of ground is introduced to improve the precision of the prediction and it is cleared that the proposed neural network is fitted for practical use. In the third year, the proposed neural network is improved again to increase the usefulness. Moreover, a falling weight impactor is examined as excitation source, because an easy exciting method is necessary to apply the proposed method to real ground.In the second subject, the introduction of an analytical technique is intended to build a neural network for the prediction of ground vibration at the shoulder of road. In the first year, the neural network is built only by measured data in the field. However, it is seen on the accumulation process of measured data that there is a limit to the neural network based on only measured data. In the second and third years, the introduction of a unit pattern method is considered by referring the case of noise. A simulation method is developed to calculate ground vibration caused by a running vehicle under the consideration of road roughness before the unit pattern is introduced. The characteristics of the unit pattern are made clear by comparison of numerical unit patterns and measured ones. Moreover, a computer simulation method considered traffic flow is developed for the prediction of traffic induced vibration. A neural network for estimation of road roughness is also proposed.