<![CDATA[Automated Peritoneal Dialysis (APD) requires accurate dialysate flow regulation to ensure stable fluid transfer during fill and drain phases. Flow instability may prolong transfer processes and potentially reduce effective dwell time. This study proposes a performance-target-oriented Proportional-Integral-Derivative (PID) tuning framework based on Extreme Learning Machine (ELM) for APD peristaltic pump flow-rate control. A multi-domain simulation model was developed by integrating a DC motor actuator model and a lumped-parameter fluid dynamic model of a roller-type peristaltic pump in MATLAB/Simulink. The ELM model was trained using simulation-generated data to map desired response-performance parameters, including rise time, settling time, percent overshoot, and Integral Time Absolute Error (ITAE), into PID gains. The proposed method was evaluated under five target-performance scenarios and compared with the conventional Ziegler–Nichols tuning method. The balanced-focus ELM configuration achieved the most suitable overall response, with rise time of 0.1382 s, settling time of 1.8860 s, percent overshoot of 17.6694%, ITAE of 78.1001, and steady-state error of 0.0162. Compared with Ziegler–Nichols, the proposed method reduced overshoot by 74.87%, settling time by 55.46%, ITAE by 83.04%, and steady-state error by 96.11%. These results indicate that ELM-based PID tuning can improve APD flow-control stability and tracking accuracy under ideal simulation conditions for APD]]>
