Prediction of the Discharge Coefficient of a Labyrinth Weir Type D by an Artificial Neural Network Method

This study presents the use, and its advantages, of artificial intelligence methods to predict the discharge coefficient (C_w), considering the approach conditions of the labyrinth weir type D. The study suggests modifying the training and validation rates in AI tools, which are often fixed without proper justification in previous studies. Unlike most studies that use geometric dimensions as inputs, this work focuses on the approach conditions (the emplacement of the labyrinth weir and filling the alveoli upstream and downstream) of the labyrinth weir type D. The results, based on laboratory experiments, show that these modified inputs significantly impact the e ciency and cost of constructing the weir. Moreover, the C_w predictions based on these inputs are highly satisfactory compared to laboratory test results. In terms of training and validation ratios, the study confirms that the optimal ratio is 70/30 for accurate and highly satisfactory predictions.