Simulation-Based Performance Evaluation of a Three-Input Mamdani Fuzzy Inference System for Adaptive Floodgate Control

Abstract

Flooding remains one of the most destructive natural hazards worldwide, causing substantial damage to infrastructure, economic activities, and public safety. Developing an adaptive floodgate control system is therefore essential to improve flood mitigation under dynamic hydrological conditions. This study proposes and evaluates a three-input Mamdani Fuzzy Inference System (FIS) for intelligent floodgate operation using water level, rainfall intensity, and water flow rate as input variables. The proposeds controller was implemented in the MATLAB R2024a Fuzzy Logic Toolbox using triangular membership functions, a rule base consisting of 27 fuzzy IF–THEN rules, and the Centroid of Area (CoA) defuzzification method. The controller performance was evaluated through twenty representative simulation scenarios using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Output Category Agreement as quantitative evaluation metrics. The reference output values were established from expert-defined floodgate operating categories represented by the designed fuzzy rule base. The simulation results produced an RMSE of 2.1284 and an MAE of 1.2820, while all evaluated scenarios achieved 100% Output Category Agreement, indicating consistent classification of floodgate opening levels. Furthermore, surface response analysis confirmed smooth and continuous controller behavior under different combinations of hydrological input variables. Although the proposed controller was evaluated exclusively through MATLAB-based simulations, the obtained results demonstrate the feasibility of the proposed Mamdani FIS for adaptive floodgate control and provide a foundation for future validation using real-world hydrological data.