User Tools

Site Tools


simulation_issues

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
simulation_issues [2015/08/13 11:25]
erik [Checklist]
simulation_issues [2015/08/13 14:45]
Line 1: Line 1:
-====Introduction==== 
-The dynamic nature of the environmental models make them very sensitive to input data. This means that with inconsistent [[parameter]] and [[lookup table|time-series]] input data, the numerical [[solver]] might not be able to solve the equations or might end up with crazy results. 
- 
-With the risk of annoying the reader, it must be re-iterated that the most common source of problems is that the user has started too big. By working in smaller increments many problems can be avoided.  ​ 
- 
- 
- 
- 
- 
- 
-  
-====Checklist==== 
-  - If you can't start a simulation, check the [[error messages#​errors|error message]]  ​ 
-  - If the simulation halts due to an error, consult the [[error messages#​simulation errors|simulation error listing]].  ​ 
-  - If you can start the simulation but crashes, or if the results are crazy, try to lower the absolute and relative error settings of the [[solver]] and run a new simulation. This helps in some situations and is an easy thing to try before digging deeper.  ​ 
-  - Disable all sub-systems except for one: right-click each [[sub-system]] in either the [[graph]] or [[interaction matrix]] and select **Disable** from the menu. Run a simulation and make sure that the results seem reasonable. Then enable the sub-systems one by one and run simulations after each, until you see which sub-system causes the problem. ​ 
-  - Make sure that all connections are correctly set by opening (double-clicking) each [[connector]] going to and from the [[sub-system]].  ​ 
-  - Review the parameter values. If **min** and **max** values have been suggested, see if any of your parameter values are outside of this range. 
-  - If you feel you have the understanding of the model to do it, disable [[transfer|transfers]] in the sub-system until you have found which transfer causes the problem. Use the [[user_interface#​Information]] panel to see which objects are used by the transfer and that might cause the problem. ​ 
- 
- 
- 
- 
-====Probabilistic simulations==== 
-If you experience problems with [[probabilistic simulation|probabilistic simulations]],​ before doing anything else, make sure that you can run a [[deterministic simulation]] and that the results are reasonable. ​ 
- 
-In a [[probabilistic simulation|probabilistic simulations]] it is even harder to avoid inconsistent parameter values; when samples are randomly drawn from the parameter [[PDF|PDF'​s]] it is easy to end up with odd parameter combinations. It is also not uncommon for parameters to receive illegal values, like parameters that end up with negative values or zeroes. Even if a [[simulation]] completes, some [[iteration|iterations]] might take a long time to finish (this is a good indication that there is a problem with parameter values). ​ 
- 
-===Opening the tool box=== 
- 
-By default, the [[simulation]] is halted once an [[iteration]] cannot be run. The reason is, of course, to avoid a situation where the user is not aware that something has gone wrong during the simulation and that the results are not to be trusted. But when trying to find out what is causing the problem, it is sometimes useful to be able to run all of the iterations. 
- 
-To remedy this problem, go to the [[simulation_screen#​simulation_settings|advanced]] simulation settings tab and make sure that **Halt on error** is unselected. ​ 
- 
-Even with sound and healthy [[PDF|PDF'​s]],​ some parameter combinations can cause make iterations take a long time. If one or two iterations take an hour each, then it doesn'​t matter if the remaining iterations take 1 second. Therefore, enter a (reasonable) **time out**, eg. 5 times the time for a normal [[deterministic simulation]]. ​ 
- 
-Finally, the software can help you with some **output statistics**. Make sure that this box is checked. 
- 
-===Raw data table=== 
- 
-Your new best friend is the [[raw data table]]. For a given output, it shows all values that have been generated for a specific time point. ​ 
- 
-===Time chart=== 
- 
-Another useful friend is the [[time chart]], which can help you identify [[wp>​outliers]] for time dependent outputs. Before you proceed, be aware that if you have run many (>500) simulations,​ drawing the chart will take some time.  
- 
-  - Create a [[time chart]] of the troubling outputs. 
-  - Right-click and **Edit** the chart. 
-  - In the **Data** tab, for **Iterations**,​ select **All**. 
-  ​ 
-===Histogram=== 
- 
-Finally, for parameters it is sometimes faster to quickly identify a problem using a [[histogram]]. The histogram will display the distribution for your parameter. ​ 
- 
-====Simulation fails to run==== 
- 
- 
- 
- 
- 
-  
-  
  
simulation_issues.txt ยท Last modified: 2015/08/13 14:45 (external edit)