Optimizing De-Icing Operations

Challenge Accepted: Optimizing the De-Icing Operations of Delft

Quantillion is first and foremost focused on changing the primary metal industry so they can become leaders instead of followers in the era of Industry 4.0. To accomplish this, Quantillion brings innovative technologies from various fields together. Two such technologies, in which Quantillion is specialized, are Data Science (DS) and Artificial Intelligence (AI). The applications of DS & AI however are not only restricted to heavy industry. They can also be used to, for example, optimize the de-icing operations of a Dutch city. And that is exactly what we did after we found out that there was a close parallel between the operations of primary aluminium and de-icing due to their vehicle routing component. Together with the municipality of Delft, we accepted the challenge to optimize the de-icing activities of Delft by making use of the process data.

While many Dutch citizens see the de-icing of highways, roads, bike lanes and walkways during the winter season as given, the underlying decisions that have to be made can have serious consequences for humans and nature. For example, imagine that the roads are de-iced by way of salt sprinkling while it is not necessary at all. This results of course in a financial loss due to the underlying labor and material costs that are wasted. Moreover, unnecessary salt sprinkling has a negative impact on the environment which is very difficult to quantify. Even worse, imagine that the municipality makes the faulty decision not to de-ice the roads while it must have been done. In the worst case, this can have life-threatening consequences…

The fair question that immediately comes up is: “How hard can it be to always de-ice the roads when they become icy?” In a sense, the answer is: “Not hard at all”. However, by diving into the operational processes of de-icing, it becomes clear that large efficiencies can be obtained by optimally deciding when the actual de-icing takes place and how it takes place.

In short, the de-icing process can be divided into two general approaches. The first approach is called curative de-icing. In this case, the decision to de-ice the roads is based on whether the roads are already becoming icy. The advantage of curative de-icing is that you do not have to predict the near future icy roads. However, by applying the curative de-icing method, the opportunity of careful planning is lost. For example, if the roads are becoming slippery during off working hours, the de-icing workers are forced to work. Hence, this is accompanied with higher than normal labor costs. The second approach is known as preventive de-icing. To apply this approach, accurate predictions must be made about the slipperiness of the roads. Preventive de-icing however makes it possible to accurately plan when (for example the day before it becomes icy) and how the roads are to be de-iced. Due to this up-front planning, significant material and labor costs can be saved while still serving all the citizens as expected.

Quantillion and the municipality of Delft investigated and developed a quantitative-based solution to make it possible for the Delft de-icing operations to switch from curative to preventive de-icing. Quantillion has combined advanced DS/AI prediction models together with local weather forecasts and surface ground data of Delft to accurately predict when and where in Delft the roads will become icy. In 91 out of 100 cases, on average, we are able to predict whether to de-ice preventively or not.

Accurate predictions however mean nothing if no significant improvements to the operations can be realized. To see what the real added business value of switching from curative to preventive de-icing is, Quantillion successfully developed a business case. By switching from curative de-icing to preventive de-icing (based on our AI models), we calculated that the overall de-icing costs would be reduced by at least 8%, which is comparable to average savings that AI/DS bring to aluminium industry.

While the current beginnings look promising, there is still a lot to improve on. It is for example crucial to keep gathering high quality weather prediction data. Only then, it is possible to precisely predict day in day out where in Delft it becomes icy. If this step is mastered, then the next step is to optimally plan how the roads will be de-iced, so that the overall costs are minimized. An important component in this optimal planning is the smart routing of the de-icing assets. Due to our deep knowledge of the primary aluminium industry, this is where we as Quantillion come into the familiar territory. 


  • Through the use of data science it is possible to accurately plan when and how the roads are to be de-iced.
  • Switching from curative to preventive de-icing allows to reduce the overall de-icing costs.
  • The next step in this development process is implementation of smart routing of the de-icing assets.