Adaptive PID controller design for wing rock suppression using self-recurrent wavelet neural network identifier

详细信息    查看全文

• 作者：Milad Malekzadeh ; Jalil Sadati ; Mojtaba Alizadeh
• 关键词：Wavelet neural network ; Wing rock motion ; PID controller ; Identification
• 刊名：Evolving Systems
• 出版年：2016
• 出版时间：December 2016
• 年：2016
• 卷：7
• 期：4
• 页码：267-275
• 全文大小：947 KB
• 刊物类别：Engineering
• 刊物主题：Artificial Intelligence and Robotics
  Complexity
  Statistical Physics, Dynamical Systems and Complexity
  
• 出版者：Springer Berlin / Heidelberg
• ISSN：1868-6486
• 卷排序：7

文摘

This paper presents a novel control scheme based on auto-tuning PID controller to suppress wing rock phenomena. Due to having a complex dynamic, wing rock motion identification is not a simple task, and this complexity can adversely affect the performance of PID controller. Employing a wavelet neural network based identifier, this paper develops an auto tuning adaptive PID controller to tackle the problem. Since having an acceptable control performance inevitably involves having a meticulously trained identifier, the training performance is of utmost importance. Aiming at boosting the training efficacy, a two-phase algorithm encompassing Bees algorithm and Back-Propagation (BP) is proposed by this paper to train the proposed identifier respectively in off-line and on-line modes. Due to its inherent capability in sifting the global minima, Bees algorithm is employed to find initial values of weights around which it is then possible to conduct a local search by means of BP based online training. Therefore, the identifier can precisely furnish the proposed PID controller with the system sensitivity in on-line mode. The adaption of PID controller can thus be performed in each time step. The performance of this method has been presented in simulation results and the comparison section confirms the effectiveness of proposed scheme.