The New Calculation Engine In SimaPro 8
SimaPro 8 has been equipped with a completely new calculation engine. The reason for this is simple. The old calculation engine worked well with life cycle assessment (LCA) models of a few thousand processes. But with the introduction of ecoinvent 3 - resulting in much larger models - SimaPro needed a software update to continue supporting the available data sources.
Why A New Engine?
In 2002, the “matrix inverse” principle was widely adopted for calculation in LCA software tools. Recently, the matrix inverse principle started presenting a practical problem in LCA databases and models that are growing in size: its memory use. The matrix inverse principle memory use increases quadratically with the number of processes in an LCA model. For example, when your LCA model becomes four times as big, memory use increases 16 times.
Solving The Memory Size Problem
Fortunately, there is an alternative called the “Leontief inverse” principle. Unlike the matrix inverse approach, Leontief inverse is an iterative process, which means that the calculation engine refines the numbers with each iteration until the desired precision is reached.
Before it was possible to implement Leontief inverse in commercial LCA software, several practical obstacles had to be overcome. After intense investigations together with Nanne van der Zijp of ModelIt and rigorous programming work by Vincent Cleij of PRé, SimaPro’s calculation engine is now ready for the future.
What The New SimaPro Can Do
The results are convincing. SimaPro’s new calculation engine can be described as follows:
- Low memory use. Memory use is linearly related to the number of links between processes in an LCA model instead of quadratically related to the number of processes.
- High precision. All results are calculated with a precision of 8 digits or better, except the cumulative indicators shown in the network diagram and the intermediate results of the uncertainty analysis (Monte Carlo) function.
- Better error detection. Incorrect modelling loops are detected. For example, SimaPro will give an error message when 100% or more of the assessed product output is recycled.
In practice, this means that SimaPro can now support LCA models with a large number of processes. For example, you can now make links to unit processes in the new ecoinvent 3 libraries without having to worry about computer memory limitations. Some users will experience that loading of large LCA models into the computer’s memory takes time. But once loaded, the data stays in the memory until the SimaPro database is closed, making recalculations much faster.
Ready For The Future
With this new approach, SimaPro’s calculation engine is ready for large LCA models. In the upcoming years, we expect to incorporate several new data libraries with medium to large size LCA models in SimaPro, for example for the food sector. When linking your model to several LCA models in different libraries, you may want a computer with a faster hard disk and processor for faster loading of the data into the computer’s memory, but SimaPro’s calculation engine will no longer be the bottleneck.
Soon, we will publish a white paper that will describe the new calculation engine and its practical implications in more detail.
Tommie joined the Consultancy Team in 2012, working with databases and methods. He collaborated in projects such as Prosuite and improving the ReCiPe method. He worked at PRé from 2012 until 2015.