and Big Data
Predicting the likelihood of subsidence of houses and other buildings
In Zaanstad, the municipality, housing corporation Parteon and data scientists from Berenschot Intellerts joined forces to record the likelihood of subsidence in houses and other buildings.
A problem to solve
The subsidence of foundations of houses and other buildings is a serious problem in Zaanstad, a municipality of 152,000 residents. A considerable number of the approximately 70,000 buildings there are built on wooden piles, which are susceptible to bacterial damage. The final prognosis is that a third of the total housing stock has problems with foundations.
Repairing foundations is expensive: on average between 30,000 and 50,000 euros per house. Investigating the quality of the foundations also requires lots of time and money every year. The inspection of the piles is done physically, and the subsidence is also recorded manually, using measuring poles.
Yet bad piles do not necessarily lead to catastrophic subsidence immediately. Many other factors play a part too, such as the type of soil the water level, the presence of trees with long roots, the quality of the drainage system, and vibrations from building work nearby.
There is a great amount of data about houses, other buildings and the ground beneath them available from both the Municipality of Zaanstad and the housing association Parteon, which owns most of the property in the town. Linking all this data could deliver new insights with which a better approach can be formulated for the expensive problem of foundation repair.
Can a predictive model for the quality of the foundations of houses and other buildings be made on the basis of the year of construction, the location, the groundwater situation, the state of the maintenance and the annual rate of subsidence?
The data scientists from Berenschot Intellerts in Utrecht started working on this question.
The inventorying of available details delivered a so-called ‘data lake’ of 140 gigabytes and around 136 million records. The data did not only come from the municipality and Parteon, but also from external parties such as the KNMI and the Kadaster.
“Linking all these files was a time-consuming task,” explains Martin Haagoort from Berenschot Intellerts. “There were a considerable number of transformations required to get all the data at the same level, so that it could be analysed as a whole.”
Satellite data was very important in the analysis. This data, from the SkyGeo company, contains a detailed record of information from the period 1992-2016. By mathematically analysing the radar images from the satellites, Sky Geo can determine the subsidence per building to the nearest millimetre.
In the next step, Intellerts’ data scientists filtered the data for strange outliers. “In some places, the satellite data gave illogical values,” says Haagoort. “For example, as a result of trees that grow precisely above the measuring points in the summer.” For these outliers the data was corrected.
Then the data was ready for the ultimate goal: making a predictive model. For this, Intellerts developed various so-called ‘machine-learning algorithms’. Using these, a first predictive model was created.
The results from this model are comparable with the information that Parteon has on a hundred buildings for which a complete foundation report is available. “The outcome was very encouraging,” says Jurgen de Ruiter, Parteon CFO. “Not only is the model of high value, but it gives extensive information that allows separation into five risk categories, from low (monitoring is enough) to very high (action required).”
135 million data records
For the foundation repair model, data records from the following institutions were selected and processed:
Parteon, expert building data
Zaanstad, foundation data
KNMI, weather data
Kadaster, property data
Sky Geo, subsidence measurements
ESRI, sea level data
Het Waterschap, water level measurements
NLextract, various geographic data
Parteon and the Municipality of Zaanstad benefit
Ultimately, the model can be used to calculate which foundation category all buildings in Zaanstad belong to, even though measurements or expert estimates are not available for most buildings. “This has unprecedented advantages in terms of planning and costs in the management of foundation repair,” according to Levinus Jongmans from the municipality.
“Now we have on record that 11,000 homes are at risk. That is between 15 and 20 per cent of the total housing stock in Zaanstad.” Jurgen de Ruiter from Panteon is also very happy: “We now have a better idea of which buildings we have to work on, and when. This means we can make crucial savings on investigation costs.”
Download the full story (PDF)