I am working on simulation and created a model made of many equations. The simulation takes very long time to simulate and make results. So my aim is I want to be able to speed up the process using the results.
The simulation created a table of results for each timestep of the simulation. Here is an example of the look of my table.
| Time | HeatIn | Flow|| Conc1 | HeatOut | HEArea | UsefulEn|
|:-----|- -----:|:---:||:--- --|--------:|:------:|---------|
| 0 | 234 | 45 || 0.45 | 455 | 34 | 46 |
| 1 | 678 | 45 || 0.45 | 345 | 34 | 24 |
| 2 | 345 | 34 || 0.45 | 456 | 34 | 56 |
| 3 | 346 | 23 || 0.45 | 123 | 34 | 56 |
| 4 | 278 | 45 || 0.45 | 456 | 34 | 34 |
| 5 | 235 | 45 || 0.45 | 458 | 34 | 43 |
My data is much bigger but this table for example use. My data have 150,000 time observation and >10,000 variable.
I want to use the time output (table data) of the complicated model to create a simple model of the equations used to generate the table. This simple model will then be used to produce a new time series but more fast. It would also be good to change the independent parameter in the simple model to create slightly different results.
But I am not sure what is available and how to proceed. I researched reduced order model using single value decomposition but it seems this methods is only useful for CFD and finite volume analysis. Can these method take time series as input data? If not, what is best approach to use to generate a simplified time series model from the data created by a complex model.