FAQ and troubleshooting

Installation

'py' is not recognized as an internal or external command

Python, or its py launcher, is not installed. See Install PyFlange.

'pyflange' is not recognized as an internal or external command

The folder where pip installs commands is not in the PATH. Use py -m pyflange instead of pyflange: it always works.

ModuleNotFoundError: No module named 'pyflange'

PyFlange is not installed for the Python interpreter that runs your script. If you have several Python versions, install PyFlange with the same command you use to run the script, e.g. py -3.12 -m pip install pyflange for py -3.12 script.py. In VS Code, check the interpreter selected at the bottom-right of the window.

pip fails with a connection or SSL error

Your company network probably uses a proxy. Ask your IT department for its address and pass it to pip with --proxy, see Install PyFlange.

How do I know which version I have?

py -m pip show pyflange, or py -m pyflange --version. The changes between versions are listed in the Changelog.

Modelling

ValueError: Failure mode D detected, while only failure mode B is supported

validate() found that the flange fails at ULS in a mode for which the polynomial model is not valid (see Validity of the model). Typically the flange is too thin compared with the bolt (mode D or E) or the shell is too weak. Change the geometry, or use FEA.

Which shell stiffness should I use?

By default (k_shell=None) the simplified formula of ref. [2], eq. G.43 is used, which is conservative. A better estimate is the FEA database of ref. [5], 9.3.1: pass k_shell=shell_stiffness(R, s, gap_angle), or "k_shell": "interp" in a JSON case. The database covers diameters of 3 to 10 m, shell thicknesses of 10 to 150 mm and gap angles of 10° to 180°; outside this range shell_stiffness returns nan.

Is Fv the preload before or after losses?

Before losses. The model subtracts the settlement and plastic losses defined by DFT, settlement_factor and PPL; the result is fseg.bolt_preload. If your preload already includes all the losses, leave PPL and DFT at zero. Note that the coefficient of variation term of ref. [2], eq. (23) is not applied by PyFlange: include it in Fv.

Why does the bolt force differ from my FEA results?

The polynomial model is an approximation calibrated on FEA results (refs. [1] and [5]), so some difference is expected; the stress multiplication factor of ref. [5], 6.5 accounts for this model uncertainty. First check that the gap, the shell stiffness, the inclination and the preload are the same in both models.

Can I model T-flanges in a JSON case file?

Not yet: the JSON format and the command line support L-flanges only. Use the Python API, see the T-flange tutorial.

Should I use the AMD1 or the ED2 classes?

IEC 61400-6:2020+AMD1:2025 is the published standard: use the AMD1 classes (PolynomialLFlangeSegment, PolynomialTFlangeSegment) for design. The ED2 classes implement a draft and are useful to anticipate the effect of the next edition.

Results

The damage is zero or extremely small

The bolt force increases significantly only when the shell force approaches the gap opening region. Check that the Markov matrix contains tower bending moments in N·m (not kN·m or stresses), and that Zg is the dead weight per bolt (per bolt pair for T-flanges), in N.

The damage is nan

Usually an input is nan, for example k_shell from shell_stiffness outside the range of the FEA database. Check the intermediate quantities with pyflange determ, or print fseg.shell_force_at_tensile_ULS and the other points.

The Monte Carlo results change at every run

Use --seed on the command line, or numpy.random.seed(...) in Python.

Getting help

If your question is not answered here, search the GitHub issues and open a new one if needed. Include the PyFlange version, a minimal script or case file, and the complete error message.