Tee/Wye Junction Complexity

Tee/wye junctions contain powerful and sophisticated models for determining the pressure losses at flow splits and converging flow streams. This power comes at the cost of additional model complexity, which can lead to convergence problems in models with numerous tees. Here are a couple of recommendations.

Tee/Wye junctions have two loss options: Simple and Detailed. The Simple loss model ignores losses at the tee. The Detailed model uses the correlations summarized in the Tee/Wye Loss topic. The potential convergence problems occur when using the Detailed loss model.

The potential problem is not with the Detailed loss model itself. Rather, it is due to multiple tees coupling together in such a way that the Solver cannot converge on all of the flow splits. The tee loss coefficients are dependent on the flow splits, but the flow splits are dependent on the other losses in the system. Therefore, if a pipe network contains many tees using the Detailed loss model, when one tee's loss coefficient is updated from an intermediate Solver iteration, the entire network needs to be re-calculated to determine how the flow splits changed. Sometimes this process is not able to terminate within the Tolerances specified by the model.

A common issue with detailed tees is having the pipes connected in the Properties window differently than how connections are set up in the Workspace. Make sure to double check those connections before running the models.

You can quickly and easily check if a convergence problem is being caused by using the Detailed loss model. Create a child scenario, then using the Global Junction Edit window, you can change all tees to the Simple loss model and rerun the model. If it now converges then the problem was likely due to the use of Detailed tee modeling.

Experience has shown that this problem can frequently be resolved by using absolute tolerance criteria on the Tolerance panel of Analysis Setup rather than relative tolerance. Use this option at your own discretion and avoid setting the tolerances too loose such that you get false convergence. Often times, using the "Both Absolute and Relative" option gives the best of both worlds.

You also have the option of just using the Simple (lossless) tee model for all tees. In many systems the loss due to tees/wyes is negligible, especially for systems which see significant loss from other components that dominate the system.