Blog | Tools | Glossary | Search

Share:   | feedback

ECLIPSE Convergence manual

From petrofaq
Jump to navigation Jump to search

This is a summary of ECLIPSE convergence issues. Most of the advice and suggestions apply equally to all the simulators.

Non-linear iteration

ECLIPSE uses an iterative process based on Newton's method to solve the nonlinear equations.

The number of non-linear iterations is a guide to model convergence:

Non-linears per Timestep Guide
1 Very easy to converge
2 to 3 Easy to converge
4 to 9 Increasingly difficult flow situation
> 10 Problem with model?


Check the simulation reports. (see Background to timesteps and iterations). If the problem with:

  • linear or non-linear iterations - check all warning messages for data problems
  • timesteps - reduce when difficult modeling situations arise
  • reports - request reports only when you need them

The greatest improvements in performance are obtained by identifying and correcting the cause of any nonlinear problems.


Controls available from the Schedule section:

  • TUNING sets timestepping, iteration and convergence criteria

– TUNINGL is used for the LGRs in the model

Guidelines:

  • Timestepping controls need alteration fairly frequently
  • Iteration controls seldom need adjustment
  • Convergence controls need adjustment only in highly unusual circumstance



Convergence problem Checklist

All sections

  • Check all problem and warning messages.
  • Does the model run better with no tuning keywords?
  • Remove TUNING or TSCRIT or CVCRIT keywords and try to run.
  • Is the model in equilibrium at the beginning of the simulation?
  • Try running the model without any wells or any aquifers.


Grid and EDIT sections

  • Use MINPV, check LIMITS, check LGRs.
  • Check whether you need to set a minimum pore volume
  • Radial runs: Is the inner radius too small?
  • Dual porosity runs: Is sigma too large
  • LGRs: Are the cells too small? In ECLIPSE 100 try LGRLOCK


PROPS section

  • PVT data: Use EXTRAPMS in ECLIPSE 100. In all simulators, avoid PVT extrapolations
  • SCAL data: Check the slopes of all relative permeability tables.


SOLUTION section

  • Is initial fluid distribution stable?


SUMMARY section

  • Remove RPTONLY
  • Plot PERFORMANCE information


SCHEDULE section

  • Avoid VFP table extrapolations, use EXTRAPMS in ECLIPSE 100.
  • Take care when using MSWs



For more information and details, see the articles from the list below:

List of articles with details

Background to timesteps and iterations‎
How to speed up an ECLIPSE run‎
Eclipse Run Problems And Errors
Eclipse Severity Levels