24 October 2018

A unique solution gives greater certainty about uncertainty underground

GeoBIM is on the way to becoming the world's largest geotechnical database. This creates new possibilities. By extracting data from GeoBIM and having AI robots analyse it, we can do more than just create better models of how things look underground. We can also gain control of the degree of uncertainty in our analyses, regarding the level of a bedrock surface for instance.

Digitalization is currently one of the big buzzwords in the urban construction industry – everything must be measured with sensors and cameras. But all this information needs new tools to analyse it.

"Our only chance of managing this is via robots with artificial intelligence, sometimes called 'machine learning'. Tyréns now has a lot of expertise in this area. And thanks to GeoBIM, we also have a unique platform for retrieving geotechnical information," says Mats Svensson, one of the originators behind the new database.

GeoBIM already acts as the central communication hub for many of Sweden's major infrastructure projects. Instead of having to print out maps and other geotechnical information as hard copy and pinning these up on a wall, the database allows all saved geotechnical data to be coordinated and visualised in 3D.

"It means that everyone can access the same updated and quality-assured information simultaneously, and we can get an overall picture of all aspects." 

There are currently around 57,000 survey points in the database. Mats Svensson believes that, once all the data has been uploaded, GeoBIM will be the world's largest geodatabase with half a million points in Sweden.  

It gives us a unique edge and fantastic opportunities. One of the things Tyréns has done is perform tests to calculate rock levels underground using AI technology. By training the mathematical algorithms, they have also managed to find a method for analysing uncertainty in their conclusions.

"We have simply become a lot more certain about the degree of uncertainty in our calculations. Suppose we have determined that the bedrock level where a railway tunnel is to be constructed is 10 metres below the ground surface. We can then also establish the degree to which that result may vary, e.g. by plus/minus 0.5 metres."

"Previously these analyses were founded on what was often a limited statistical basis and a fair amount of gut feeling, resulting in significantly poorer control of the degree of uncertainty."

With more accurate information about how things look underground, contractors can now calculate the cost of an infrastructure project and plan their workforce levels and what machinery will be needed. This will make things easier as well as saving money. 

The new tools have appeared at just the right time – the Swedish Transport Administration has recently started requiring that an uncertainty model should accompany the construction model for underground spaces.

"There is no tradition in the industry of working with uncertainty models. Some success was achieved in attempts to find a suitable method in individual major projects, such as the Stockholm Bypass. But often it has been a matter of informed guesswork. So there are a lot of engineers at the moment who are tentatively looking for ways to gain control of the uncertainty factor. It's good to know that we are a step ahead here!"

"Our main priority right now is to distribute the tools so that more people in the industry can use them. This development will progress that much quicker because the demand is so great."

Tyréns is also well ahead on improving the way in which uncertainties relating to soil strength and deformation characteristics are calculated. This will bring further improvements for underground models.

"At the end of the day we will have a complete concept that will allow us greater certainty about the uncertainties involved in the geometry and material characteristics of underground spaces."