Purwadi Joko Widodo Widodo, Heru Sukanto, Budi Santoso, Lullus Lambang Govinda Hidayat, Joko Triyono, Iwan Istanto, Rahman Wijaya, Fitrian Imaduddin

A Comparative Study of Traditional PID Tuning Techniques and AI-Based Algorithmic Approaches Utilizing the Python Control Library

Introduction

A comparative study of traditional pid tuning techniques and ai-based algorithmic approaches utilizing the python control library. Compare PID tuning: traditional Ziegler-Nichols vs. AI algorithms (GA, PSO) in Python. AI yields superior control performance (lower ITAE) despite longer execution.

Abstract

This study aims to compare PID parameter settings with conventional tuning methods and tune methods using AI (artificial intelligence) algorithms. This study was conducted by means of simulation using a computer program created in Python and utilizing AI libraries to solve the problem of determining PID (proportional-integral-derivative) parameters. Two AI algorithms used in this study, namely the Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) methods, were compared with the conventional Ziegler-Nichols method. The study was conducted by applying the PID parameters obtained to a certain transfer function and then comparing them on several related aspects. The results of the study showed that the solution obtained using the AI method requires a longer execution time, more than 2 seconds for PSO and more than 3 seconds for GA, while ZN requires less than 1 second. However, the AI method can provide better solutions, as can be seen from the magnitude of the ITAE that occurs, where GA and PSO provide ITAE less than 1 while ZN is more than 22.

Review

This study presents a timely and relevant comparative analysis of PID controller tuning methods, juxtaposing the conventional Ziegler-Nichols (ZN) technique with two prominent artificial intelligence (AI) algorithms: Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). Utilizing a simulation-based approach implemented in Python, and presumably leveraging the Python Control Library, the research systematically investigates the efficacy of these diverse methodologies in determining optimal PID parameters for a specific transfer function. The premise of comparing established conventional techniques against modern computational intelligence approaches for a fundamental control problem like PID tuning is well-conceived and contributes to understanding the practical benefits and drawbacks of each. The core findings of the study reveal a critical trade-off between computational efficiency and control performance. As expected, the ZN method demonstrated significantly faster execution times, completing the tuning process in less than 1 second. In contrast, the AI-based methods, PSO and GA, demanded substantially longer processing times, exceeding 2 and 3 seconds respectively. However, this increased computational burden was demonstrably justified by a marked improvement in the quality of the derived PID parameters. Quantified by the Integral of Time-weighted Absolute Error (ITAE), the GA and PSO algorithms achieved significantly lower values (less than 1), indicating superior control performance, especially when compared to ZN, which yielded an ITAE greater than 22. While the study effectively highlights the potential of AI algorithms for achieving superior PID tuning, several areas could benefit from further elaboration and investigation. The abstract mentions applying the parameters to "a certain transfer function"; expanding this analysis to a broader range of system dynamics (e.g., higher-order systems, non-minimum phase systems, systems with significant dead-time) would substantially strengthen the generalizability of the conclusions. Future work could also explore the robustness of these AI-tuned controllers to system uncertainties and disturbances. Additionally, for practical implementation, a more detailed discussion on the implications of the observed execution times in real-world, real-time control applications, where quick re-tuning might be necessary, would add valuable context to the comparative analysis. Nonetheless, this research provides a clear and valuable initial comparison, underscoring the potential for AI-based tuning to significantly enhance PID controller performance, albeit with an acknowledged computational overhead.

Full Text

Comments

You need to be logged in to post a comment.