add friction model "Hofer"

[jkisse]

The friction factor "Hofer" is introduced. It is an explicit implementation of Colebrook, used by some established Gas Grid Simulation tools.
Please refer to the added paragraph in the docs for more details.

[dlohmeier]

Looks interesting. I just thought that the transition area is relevant for different models. Maybe we should think about how to correct equations for laminar / turbulent flow. I think there is an issue on this topic and this implementation could be a nice approach.

[jkisse]

Thanks @dlohmeier!
I'd also appreciate your opinion on the usage of np.divide etc. in numba functions, e.g. here. It seems to me as if the advantage of array-processing cannot be used in the for-loops, thus it might be slightly more efficient to use the native "/" directly instead of np.divide? Did you investigate that a bit when you implemented the numba-functions?

(analogously for np.abs and np.power)

[dlohmeier]

As far as I can recall, numba was good at exactly this: identify numpy code within jit-decorated functions and compile them correctly. I always found it more readable and wrt. timings I remember that I found no difference or a slight advantage of the numpy ufuncs over native code. It might have changed with newer numba versions.

[kbensafta]

Is the PC-derivative indeed applicable in this case or is it an approximation, since Hofer is just an explicit form of Prandtl-Colebrook?

[jkisse]

Not sure - I found it acceptable, also because of the quite complicated derivative of the Hofer formula. Apparently it is working ; )
Maybe the convergence behaviour could be strengthened if a separate "Hofer-derivative" is implemented, but I find it sufficient so far.

[jkisse]

We found out that max_iter has been renamed rather silently a year ago, thus I'd like to use this PR to add a warning for those who still use it.

When calculating large transmissions grids, we often need more than 10 iterations. Therefore, I'd appreciate it if we could set it to 30 by default. It shouldn't have any significant effects on the runtime.

[dlohmeier]

We found out that max_iter has been renamed rather silently a year ago, thus I'd like to use this PR to add a warning for those who still use it.

When calculating large transmissions grids, we often need more than 10 iterations. Therefore, I'd appreciate it if we could set it to 30 by default. It shouldn't have any significant effects on the runtime.

Increasing the default number of iterations is indeed necessary in quite some calculation scenarios for us, so I would welcome that change. Wrt. the max_iter kwarg, I am a little surprised, since I thought we really intended to make this change backwards compatible. A warning is indeed appropriate then.

[dad616610]

Some nitpicking

[jkisse]

Increasing the default number of iterations is indeed necessary in quite some calculation scenarios for us, so I would welcome that change. Wrt. the max_iter kwarg, I am a little surprised, since I thought we really intended to make this change backwards compatible. A warning is indeed appropriate then.

There is a logger.info if non-standard options are passed to set_user_pf_options (here), but has no effect for options passed as kwargs to pp.pipeflow.