PID Controller Tuning Optimization with Genetic Algorithm for a Ball-Beam Balancing System

Ow Yong, Hui Yan (2018) PID Controller Tuning Optimization with Genetic Algorithm for a Ball-Beam Balancing System. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Ow Yong Hui Yan - FULL THESIS.pdf Restricted to Registered users only Download (3MB)

Abstract

This project aims to identify the optimized PID parameters using Genetic Algorithm (GA) for the balancing control of a single axis ball-beam system powered with two propellers at each end. A prototype of the dual-propellers-powered ball-beam balancing system controlled by Arduino microcontroller was setup where a gyroscope was placed at the pivot point of the ball-beam. The Arduino would read the position of the beam based on the feedback from gyroscope and compare it with the desired balanced position i.e. 0 degree which was set before test runs. The transfer function of ball-beam system was studied and applied with the Genetic Algorithm optimization method to find the optimum PID parameters for the system in MATLAB. The PID parameters computed by MATLAB was then applied into the prototype to test its feasibility in balancing the beam. After setting up the prototype, the PID parameters optimization for the ball-beam system was obtained in MATLAB with the derived transfer function. The parameters were obtained from five different bound range setting with three sets of result for each bound range. Hence, a total of 15 results were generated. In overall, the results of parameters for the component P and component D were very close with the values of between 0.928 to 1.000 for P and values of between 0.960 to 0.999 for D. Only component I showed some differences in the results which ranging from the smallest value of 0.019 to the largest value of 0.705. The 15 sets of the optimized PID parameters were then applied in the prototype for actual balancing testing purpose. Time to achieve the state of balance (within the allowable angle of ±5°) or settling time was then recorded for each testing. The best result obtained was 12.57 seconds with the worst was 41.12 seconds. All the results however were still far better than manual tuning which using Ziegler-Nichols method. The transfer function was found to be workable in this project to run the GA in obtaining the optimized PID parameters in the MATLAB. GA was able to generate optimum results of the PID parameter in the bound range 0 – 20 (for P), 0 – 0.04 (for I) and 0 – 15 (for D). This is the best settling time for the ball-beam system to achieve balanced state, which was 12.57 seconds. The major limitation is the transfer function used in this project could only approximately represent the actual ball-beam system i.e. the prototype. Hence there is still room for improvement. A better transfer function is suggested to be investigated for better result. The project can be served as a foundation for quadcopter drone balancing in-the-air PID control tuning applications.