If you are experiencing convergence problems with a PIPE-FLO Compressible model, there are several troubleshooting techniques you can employ.

PIPE-FLO Compressible's network balancing calculations incorporate a number of advanced numerical solution techniques, including three penalty minimization techniques (steepest descent, Hardy Cross, and a genetic algorithm) as well as a flow averaging algorithm, a pressure averaging algorithm, and the Newton-Raphson method. The program performs the network balancing calculations using a combination of these methods, automatically selecting the best method to drive the solution.  This insures that the best method is used for each type of system during all phases of the calculation process.

If you are experiencing convergence problems when calculating a PIPE-FLO Compressible system, it is sometimes useful to try turning off one or more of the convergence methods.  For example, the Steepest Descent method can depend greatly on the initial estimate, and so it is often the cause of convergence problems.  Also, you can use the program's Pause/Resume feature to view the solution progress at various points in the solution process.

PIPE-FLO compressible uses the average value of the pressure sources in the system to generate the initial system pressure estimate. Convergence problems can arise if the pressure sources do not reflect the final average system pressure. For example, if we have compressors fed from atmospheric pressure (0 psi) to a set flow demand, the initial estimate will be based on the 0 psi pressure rather than the much higher average pressure (100 psi) that appears in the system after calculation. In these cases, adding a pressure demand anywhere on the flow sheet (disconnected from the system is fine) and set to the expected system pressure (~100 psi in the example) will help the system converge faster.

An actual example from a PIPE-FLO Compressible user illustrates these techniques:

A user sent in a PIPE-FLO Compressible system that was a simple two pipeline network with pressure sources at both ends.  He could not get the system to converge to a solution, even after many iterations.

The technical support staff initially ran the system for one iteration, with all of the convergence methods turned on.  When the results were viewed, the flow rate through the system was 5.55 e+023 scfm - something was obviously throwing off the calculation right from the start.

After clearing the results, they started the calculations over again, this time with the Steepest Descent method turned off.  After one iteration they viewed the results - the flow rate through the system was 0.0060 scfm.   They then resumed the calculations (without the Steepest Descent method) and the system converged after 12 iterations, with the final flow rate through the system equal to 0.0037 scfm.   They also tried running the calculations with the Steepest Descent method turned on for all but the first iteration.  The system converged after 25 iterations, with the same flow rate (0.0037 scfm).

This example illustrates some of the techniques you can use to troubleshoot a system that won't converge:

• Try running the calculation for one iteration and then viewing the results.  This will often make it possible to spot what is causing the problem (for example, you may see a very high flow rate or pressure).
• Try turning different convergence methods on or off.
• Use PIPE-FLO Compressible's Pause/Resume feature to view the solution as it progresses.