Parameters Identification of Multi-Axis Systems by Haar Wavelet

• Main Authors: Apisit Pinitnanthakorn, 李佳榮
• Other Authors: CHEN, SHYH-LEH
• Format: Others
• Language: en_US
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/72k2fy

碩士 === 國立中正大學 === 前瞻製造系統碩士學位學程 === 106 === Model based control approach is vital to achieve high precision in multi-axis dynamics system. Physical parameters in dynamics model have to be concerned in order to design and develop model-based control algorithm. The study of parameter identification in multi-axis dynamics system is always challenge and draws the attention to many researchers to propose several techniques of parameter identification. There are two crucial industrial applications in this study, which are playing a key role to the advance manufacturing nowadays. The first is five-axis machine tool and the second one is five degree of freedom serial manipulator. The system dynamics of both industrial applications and identification procedure will be described in this thesis. We are going to use approach of orthogonal basis function so-called Haar wavelet basis function. Orthogonal basis of Haar wavelet can be utilized to represent time function into Haar wavelet series. Hence, the parameter identification procedure in each mentioned industrial application has been developed base on Haar wavelet transform. Dynamics model of both industrial applications are formed by Lagrange’s equation. In five-axis machine tool, Haar wavelet basis function will be used to represent a regression model that is reconstructed from dynamics model. Finally, least square technique is applied to estimate unknown parameters. In robotic arm identification, Haar wavelet basis will be used to represent all state vectors in robotic dynamics model, which implied that the size of each of state vectors were reduced since applying Haar wavelet transform. The dynamics model of five degree of freedom serial manipulator robotic arm is tremendously larger than five-axis machine tool, which is very complicate to establish the regression form. Therefore the famous evolutionary algorithm approach so-called particle swarm optimization is use to search the best fit parameters in search spaces. The experiment of both multi-axis systems will be conducted through PC-Based control. The results of identification can be validated via simulation in Simulink/MATLAB. The error of measured response from measurement device and predicted response from simulation can be quantified by statistical tool called root mean square error and relative root mean square error.