TheJashShah/LeakageDetectionPaper-Repository ? reverse-engineered prompt

Build me a Python project that reproduces the leak detection workflow from the paper on water distribution networks.

I want to be able to download the LeakDB Hanoi data, turn the raw network files and scenarios into clean training data, train a simple baseline neural network and a physics informed version that uses conservation of mass and the Hazen Williams equation, then run inference on unseen data using the best validation model and a 0.5 threshold.

Please also include the rule based leak localization step so it can point to the most likely node or pipe, plus the sensitivity and robustness checks from the paper. Make the whole thing easy to run in order from raw data to final results, with clear scripts and simple outputs I can inspect.

If you need to check anything about the dataset or library behavior, look up the current docs online first.