Steady state solution
Previous Top Next

Steady state solution

Solaria offers two steady state and two transient solvers.  They both solve the same resistor/capacitor network.  They just use different methods to solve it.  You can choose the solver of choice by selecting it through the menus.

Select Solve>Steady State>Finite Diff or hit the graphic toolbar button.

Which solver to use for steady state solutions

For steady state solutions, the choice is easy.  Use the solver that solves the fastest.  If the finite difference solver is taking a long time, switch to the GCG solver.  Use the “Use Current Temperatures” option when restarting a solution.  For steady state solutions, it doesn’t matter how you get to the answer.  It only matters that the final solution has converged.  Types of models that converge slowly using the finite difference solution are those where the conduction part of the problem has a lot of detail and there is poor heat transfer to the heat sink.  This could be a model of a detailed electronic component in natural convection.  In this case, the GCG solver will solve extremely fast.  A pure conduction problem usually solves faster using the finite difference solver, unless there are magnitude differences in resistance between the heat source and heat sink.

Finite Difference Solver

The program will first check that there is at least one boundary node in the model.  If there is not, Solaria will give a message to that effect and then terminate the solution.  If there are no boundary nodes, which includes fluid flow nodes, a steady state model cannot converge to a solution.

Maximum Number of Iterations

Select the Maximum Number of Iterations edit box and enter a value.  The default value is 5000 iterations.  Performing a heat balance around each node in the model once is defined as one iteration in a finite difference solution.

Convergence

Select the Convergence edit box and enter a value.  The default value is 0.001 degrees. Convergence is defined as the maximum temperature difference between successive iterations for the entire model.  A convergence of 0.001 degree means that all nodes in the model have less than a 0.001 degree change in temperature in one iteration.

Heat balance in Percent

Select the Heat Balance in Percent edit box and enter a value.  The default value is one percent. This percentage is given as:

graphic

where graphic is the total heat load applied to the model and graphic is the heat load calculated by Solaria going to the boundary nodes in the model.  If there are no heat loads applied to the model, only boundary nodes at different temperatures, the sum of the heat load calculated by Solaria going to the boundary nodes in the model is given.  For a converged solution, theoretically this value should be zero.

Final Temperature

The value entered in the Final Temperature edit box is assigned to all nodes in the model except boundary nodes just prior to solving the model.  These temperature are used as initial temperatures for the solution.  The value entered does not effect the results though the temperature entered does effect the time to solve the model.  The better the guess, the faster the solution will converge on the answer.  If the model has already been run and some parameters have changed, selecting the Use Current Temperatures checkbox will use the temperature from the last run as the initial "guess" for the current run.  This will usually result in a very quick convergence.

Variable property calculation interval

This option defines the number of iterations between the recalculation of temperature dependent heat loads and properties for steady state solutions.  Regardless of the setting, whenever the convergence is above 1.0, the temperature dependent heat loads and properties are updated every iteration.  After this, the defined interval is used.  If the interval is too large or the properties vary significantly with temperature a lower interval should be used.  When Solaria performs a recalculation, you will noticed that the convergence displayed on the screen will increase for several iterations and then continue to drop.  For transient solution, variable properties are calculated every iteration.

In the Max. Variable Resistance Iterations enter a iteration increment.  A typical value is 50.  If the problem has highly temperature dependent properties then a value of 1 should be used.  A lower value will only slightly slow the solution.

Initial damping

For steady state solutions, a damping value can be defined.  A value lower than 1.0 can add stability to a solution which may have trouble converging.  A value greater than one, with a maximum of 1.95, can significantly speed the solution, by as much as 50 times.  While the solution is in progress this damping value can be interactively changed.  The initial value of this variable can be set by selecting the drop down list in the Initial Damping edit box and selecting a value.

If a model has radiation, fluid flow or temperature dependent properties, the steady state solution will be nonlinear.  In some nonlinear solutions it is necessary for the program to add damping to maintain stability.  Using this feature, damping can be added to the steady state solution prior to solving.  It is not necessary to manually add damping.  The program will detect instability and automatically add damping.  If a model terminates within a few iterations during the solution, this indicates a very unstable solution.  This should happen very rarely.  If it does then some initial damping should be added.

Select the Solve button to solve the model.




graphic

Preconditioned Conjugate Gradient Solver

Select Solve>Steady State>GCG or hit the graphic toolbar button.

Maximum Number of Iterations

Select the Maximum Number of Iterations edit box and enter a value.  The default value is 5000 iterations.  Performing a heat balance around each node in the model once is defined as one iteration in a finite difference solution.

Convergence1 and Convergence2

Select the Convergence edit box and enter a value.  The default value is 0.001 degrees. Convergence is defined as the maximum temperature difference between successive iterations for the entire model.  A convergence of 0.001 degree means that all nodes in the model have less than a 0.001 degree change in temperature in one iteration.

The Termination Criteria area of the dialog is used to define when the solution will be considered solved.  Note the residual terminations are based on absolute heat flows and not relative like the finite difference solver heat flow percentage.  That is, if you specify the termination to be ABS(MAX(R)) less than 0.001 and the bulk heat flow to the model is 1000 Watts, then the GCG will solve to a relative error of 0.001/1000 or  0.000001 which is 0.0001%.  This is more stringent than the finite difference solution and may force more than necessary iterations.  The check boxes allow the user the flexibility to terminate on SUM or MAX of the absolute values of the residual.  By unchecking these, the user can force a fixed or maximum number of iterations in the GCG solver.  The default is SUM and MAX and 1000 iterations.

Final Temperature

The value entered in the Final Temperature edit box is assigned to all nodes in the model except boundary nodes just prior to solving the model.  These temperature are used as initial temperatures for the solution.  The value entered does not effect the results though the temperature entered does effect the time to solve the model.  The better the guess, the faster the solution will converge on the answer.  If the model has already been run and some parameters have changed, selecting the Use Current Temperatures checkbox will use the temperature from the last run as the initial "guess" for the current run.  This will usually result in a very quick convergence.

graphic