Dual in-filled trenches for vibration mitigation and their predictions using artificial neural network

Machinery at construction sites, pile driving, road and rail traffic and blasting are some of the main sources of ground vibration. Most of these actions are unavoidable and solutions should be found to mitigate their detrimental effects on the buildings in the vicinity and the activities conducted in them. Isolation of this vibration has become an important issue, especially in highly populated urban areas and around buildings with sensitive equipment and/or sensitive activities. Even though base isolation of the target structure is well accepted, its implementation cost is high due to the technology, materials and equipment required for effective isolation. Base isolation techniques are therefore constrained to buildings in areas with high seismic activities. On the contrary, the use of other isolation methods such as the use of ground barriers and ground improvements are cost effective and some experimental and numerical studies have been conducted on using trenches as ground barriers. It has been proven that trenches can be effectively used for ground vibration screening. However, since the problem involves a large number of parameters and the studies are case specific, conclusions made become difficult to correlate. The use of a single trench has been studied previously and at times it requires unrealistic depths to achieve the required vibration mitigation. To address this issue, this paper develops and presents a method for evaluating the effectiveness of mitigating ground vibration with dual or two trenches using artificial neural network (ANN). Towards this end, a finite element (FE) model was first developed and the modelling techniques were validated using results in the literature. It was then used to carry out a parametric study to provide information on the effects of controlling parameters on vibration mitigation. This information was used to develop an ANN to predict vibration mitigation with dual in-filled trenches under different conditions and constraints. The present method can be further extended to a wider range with more numerical simulations which will allow the development of a generic prediction model.