Bangalore’s undercarriage bared

Bangalore’s undercarriage bared
Deccan Herald

This project is about studying what lies beneath the burgeoning surface of Bangalore city. The data for this project was collated by IISc in association with TRFI.

A 3-D map of the very foundations of Bangalore, up to a depth of 30 to 40 metres, is nearing completion. The project uses ‘neural networks’, a computational analysis, which mimics the working of the human brain to map what lies beneath the city.

A paper on this project was presented by three researchers from the Department of Civil Engineering, Indian Institute of Science (IISc). T G Sitharam, Pijush Samui and P Shyam Sunder presented the paper at Geopractice - 2005, a national conference on geotechnical engineering held in the City recently.

Groundwork

Geotechnical engineers are concerned about what lies beneath the soil. Ultimately the weight of a building, bridge, flyover or metrorail, which is being built, rests on the soil and thus it becomes very important to know what the soil underneath is like.

An engineer has to know its density at various depths, the stresses that exist in it and a host of other data, which all add upto what is known as the ‘N-values’ of soil.

A geotechie preparing the groundwork for a building, has to get his ‘N-values’ based on limited results because he simply doesn’t have access to enough data. The data cannot be generalised and also the resulting ‘site characterisation’ may not be as good as it could have been.

The project is based on the data gathered from 766 bore holes, dug over a 220 sq km area. The bore holes were about 30 to 40 feet deep and soil properties were studied at half-metre intervals.The data was collated by Torsteel Research Foundation India (TRFI) and the Indian Institute of Science (IISc), for the geotechnical investigations carried out in various parts of Bangalore, for major projects undertaken between 1995 and 2003.

Nervous breakdown?

The Geographic Information System (GIS) of the subsurface model of Bangalore City uses these penetration tests, which are ‘trained’ with artificial neural networks to recognise the pattern, based on the information available.

Neural networks simulate the way neurons (the nerve cells, which help us sense, remember and think) are put together in the brain. (A single neuron can be connected to many other neurons, so that the overall structure of the network can be very complex.)

The network keeps ‘learning’ the rules of calculation and applies it to the next set of values.

Upon receiving the new results, it adjusts the rules accordingly and keeps doing so until the picture is complete.