Selection of a best-fit soil profile from a pre-computed database for the prediction of railway induced ground vibration

Reliable prediction of railway induced ground vibration requires detailed rolling stock, track and soil characteristics. Soil databases coupled to Geographical Information Systems offer access to local soil data on a nationwide scale from which a soil profile for a specific site can be estimated. In order to get a quick estimate of the ground vibration levels, a best-fit soil profile is selected from a large pre-computed database of soil impedance and transfer functions, avoiding the expensive real-time solution of a track-soil interaction problem. We present an automated selection procedure yielding the best-fit soil profile in the database. The selection procedure uses the relative error between soil indicators for soil profiles in the database and the user-defined profile, which should be computed efficiently to ensure fast selection. The procedure is split into two steps. In a first off-line step, for each soil profile in the database, soil indictor with respect to the dynamic soil characteristics is computed and stored in the database. In a second step, the distance between the soil indicators of the user-defined profile and all soil profiles in the database is computed in very limited time. The soil profile with the lowest distance is selected as the best-fit soil profile. % The paper illustrates the selection procedure with several examples. The first is concerned with homogeneous half-space, while the second one considers a layered soil.