Convergence Exception in case of negative pressure?

[dlohmeier]

In my PR #234 , I adapted the convergence behavior such that PipeFlowNotConverged exceptions are always raised after a hydraulic or heat transfer calculation, and all component result tables are initialized with NaNs in the very beginning. So, when such an exception is raised, the result tables are all filled with NaNs, hinting the user at an error in the calculation.

There is however still one exception in which the pipeflow results are obviously wrong, but it converges. In some cases, the pressure at some node(s) can become negative and the pipeflow converges to a physically incorrect solution. A warning is printed in the junction component during result extraction, if such a problem occurs. But shouldn't this still lead to a convergence exception?

This issue is also linked to #188 .

[dlohmeier]

Another thought on this: It seems to me that there might be systems in which negative pressure compared to the system pressure (which is currently usually ambient) is allowed. Therefore, I believe that the logged warning could be sufficient and it might be better to keep it up to the user if this behavior is allowed. Yet in order to give more control over what happens, especially in case of time series simulations, we might add an option that switches between the two states. Something like: allow_negative_pressure . If this is set to false, an exception will be raised.