ANFIS Controller Design Using PSO-Tuned PID Data for pH Regulation in Industrial Cooling Towers

The Adaptive Neuro-Fuzzy Inference System (ANFIS) controller is a modern alternative to the conventional PID controller. This paper presents the design of the ANFIS controller for pH regulation in cooling towers based on the dataset from the PID controller. This paper aims to design the ANFIS controller to achieve a lower RMSE than benchmark models, with improved transient response specifications such as rise time, settling time, and overshoot to optimize pH regulation characteristics. One of the fundamental requirements for designing the ANFIS controller is the availability of a dataset. The challenge in the design process is how to prepare the dataset. This issue was addressed by recording the PID controller dataset, which includes the error (e), the change in error (∆e), and the output. The methodology consists of the following steps: (1) modeling the pH loop of the cooling tower; (2) tuning a PID controller using the Particle Swarm Optimization (PSO) algorithm; (3) recording the PID controller's dataset, including error (e), change of error (∆e), and output; (4) training the ANFIS model using the MATLAB ANFIS Toolbox; and (5) enhancing the transient response by modifying the dataset. Following Instructions for design, the simulated results showed that the ANFIS controller achieved a root mean square error (RMSE) of 0.0081. The transient response characteristics of the best-performing ANFIS controller (Modified ANFIS_4) include: rise time = 0.5863 s, settling time = 1.4867 s, overshoot = 2.7958%, and peak = 7.6548. In comparison, the baseline PSO-tuned PID controller yielded a rise time of 0.6046 seconds, a settling time of 2.3155 seconds, an overshoot of 8.6770%, and a peak value of 8.0916. The results confirm that the ANFIS controller outperforms the PID controller in all key transient response parameters, offering improved accuracy, faster stabilization, and reduced overshoot. These findings demonstrate the effectiveness of the ANFIS design based on real PID controller data, supported by Instructions for design for reliable implementation in nonlinear industrial control systems.