GROUND VIBRATION ISOLATION USING ARTIFICIAL NEURAL NETWORK

Abstract

Ground vibration is a growing concern in urban areas due to increased construction activities and infrastructural development. Common sources include pile driving, heavy equipments, road and train traffics. Due to land shortage in urban areas, buildings and other structures are often situated near source of vibrations. This proximity has potential of causing structural damage and discomfort to nearby residents. These vibrations can be mitigated by controlling the source, inserting wave barrier in the transmission medium (soil), or isolating the base of the target building or structure. However, using trenches as wave barrier are most practical because of their low-cost, rapidity and simplicity. Several studies were conducted on the use of trenches for vibration screening. It was demonstrated that trenches are very efficient for screening vibrations. This thesis presents the use of Artificial Neural Network (ANN) for ground vibration isolation. An open trench was employed as a wave barrier for mitigating ground vibration. A 3D finite Element (FE) model was developed using COMSOL Multiphysics. The model was first studied in the absence of trench. It was later studied with an open trench in the propagating path. These studies were done for nine soils with distinct soil parameters such as young modulus, Poisson ratio and density. A database was generated composing soil parameters and trench parameters. A Neural network was created using MATLAB. Levenberg-Marquardt Algorithm was used to train the neural network. Input includes soil young modulus, Poisson ratio, soil density, trench width, trench depth and trench length. Output was the amplitude reduction ratio (Ar) i.e., ratio of peak acceleration with trench to peak acceleration without trench. Datasets of six soils was used for training while three datasets were used for testing the neural network. A different dataset was used for making predictions. According to predicted output, the highest isolation level achieved is 68.28%and this requires a trench which is 1.5m wide, 10m deep and 17.5m long.