Simulate loading plans with ease on Azure – ONDRAF/NIRAS

ONDRAF/NIRAS,  the Belgian  Agency  for  Radioactive Waste and  Enriched  Fissile Materials,  is  responsible  for  the  management of  all  radioactive  waste on  Belgian  territory. In  doing  so,  ONDRAF/NIRAS  is  aiming  at developing  and applying  sustainable  solutions, which can guarantee  the  protection of  humans  and  the  environment,  now  and  in the future.

For the long-term management of the low- and intermediate level short-lived waste (category A waste), a license application is underway to construct and operate a surface disposal facility at Dessel. In this facility the category A waste will be post-conditioned in monoliths (concrete box) and then disposed in concrete modules.

The briefing

As a first step of the disposal operations, a global loading plan should be established. This global loading plan is the process of arranging the monoliths in the different modules taking into account the different constraints (e.g. radiological constraints, operational constraints). In order to define an optimized global loading plan, different simulations need to be realized and compared.

The initial set-up for simulating different loading plans included numerous manual steps in Excel, Pentaho and Power BI. ONDRAF/NIRAS’ engineers had to manually upload data in Excel to perform simulations. After the on-premises ETL tool had run, a manual refresh of Power BI was necessary to see the simulated result.

Main goals for building an Azure Data Platform:

• Reduce manual steps and risk of errors in simulation process
• Reduce lead time per simulation scenario
• Make simulation process accessible to multiple users

Our solution

During the first phase of the project, we rebuilt ONDRAF/NIRAS’ on-premises solution on the Azure Cloud in a highly secure setup.

1. First, we used Azure Data Factory to automatically load data from different sources to Azure SQL DB, where data is stored. Simulations can be made by triggering stored procedures with a specified set of input parameters. This eliminates the use of Excel files in the simulation process.
2. Next, Power BI is used for visualising and comparing different simulated scenarios. After a simulation has run in Azure SQL DB, the result is automatically visualised in a Power BI dashboard.
3. There is also daily partitioning and archiving on Azure Data Lake Storage allowing the engineers to easily compare different snapshots within Power BI.
4. Last, we embedded PowerApps into the Power BI dashboard to automatically start a new simulation based on new data input. Using PowerApps, data can be uploaded via a graphical user interface, and a new simulation can be triggered within Power BI.
5. To make continuous release management and version control possible we produced an automated deployment solution for release management using Azure Devops.

Rebuilding the original on-premises set-up on Azure Cloud makes it possible to eliminate all manual tasks, hence automatically integrating all components and reducing the chance of mistakes during data input.

Later, we added additional functionalities to the initial set-up to optimize data input and made it possible to take more parameters into account for the simulation of loading plan.

The outcome

After implementing a fully integrated and highly secure platform, ONDRAF/NIRAS is able to simulate and compare different loading plans in a less complex and more efficient way using Power BI as a one-stop shop.

Because all different components of the simulation process are integrated into one platform, simulating varying loading plans does not require any manual steps, except for data input. Although data input remains a manual step (by design), manual labour and risk of errors is reduced through the use of a graphical user interface.

The integrated solution with a user-friendly user interface provides an easy way for multiple people with the right business knowledge to use the tool and generate simulations.