NEURAL NETWORKS FOR THE PREDICTION OF MINE-INDUCED VIBRATIONS TRANSMISSION FROM THE GROUND TO BUILDING FOUNDATION

Abstract eng:
This paper deals with the application of neural networks for the prediction of the transmission of mine-induced accelerations of ground vibrations to the foundations of apartment buildings. Results from measurements in situ at the same time on the ground level and on the building foundation were used as the neural networks training and testing patterns. Rockbursts in Legnica-Glogow Copperfield (LGC), the most seismically active mining region in Poland with surface horizontal vibrations reaching 0,2 acceleration of gravity (g) and vertical components reaching 0,3g, were the sources of vibrations. The mining tremors are not subject to human control and they are random events in view of their time, place and magnitude, similarly as earthquakes. Full-scale tests were carried out many times in a period of a few years. Comparison of a huge number of records of vibrations measured at the same time on the ground and on the building foundation level leads to conclusion that they differ generally significantly. However the more precise estimation of harmfulness of the mine-induced vibrations to actual buildings can be performed on the basis of the foundation vibrations. Therefore the prediction of foundation vibrations is necessary if the measured ground vibrations are accessible only. Reduction of maximal values of accelerations of horizontal vibrations as well as differences between non-dimensional ground and foundation acceleration response spectra were taken into account in this paper. Various pre-processing of the experimental data (compression, linguistic variables introduction, scaling of input and output data) is proposed. The influence of some mining tremors as well as ground vibrations parameters (e.g. energy, epicentral distance, direction of wave propagation, direction of vibrations parallel to the transverse or longitudinal axis of the building, values of accelerations of the ground vibrations) on the transmission is considered in the input vectors. The main advantage of the neural approach is that the prediction of the parameters of vibrations of building foundation can be performed on the basis of ground vibrations taken from experimental data. The obtained results show that application of relatively simple neural networks enables us providing for building foundation vibrations based on ground vibrations with satisfactory accuracy.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket