Intelligent modeling of double link flexible robotic manipulator using artificial neural network

• Main Authors: Jamali, A., Mat Darus, I. Z., Mohd. Samin, P. P., Tokhi, M. O.
• Format: Article
• Language: English
• Published: JVE International 2018
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://eprints.utm.my/id/eprint/85575/1/IntanZaurahMatDarus2018_IntelligentModelingofDoubleLink.pdf; http://eprints.utm.my/id/eprint/85575/; http://dx.doi.org/10.21595/jve.2017.18575
• Institution: Universiti Teknologi Malaysia

The paper investigates the application of the Artificial Neural Network (ANN) in modeling of double-link flexible robotic manipulator (DLFRM). The system was categorized under multi-input multi-output. In this research, the dynamic models of DLFRM were separated into single-input single-output in the modeling stage. Thus, the characteristics of DLFRM were defined separately in each model and the coupling effect was assumed to be minimized. There are four discrete SISO model of double link flexible manipulator were developed from torque input to the hub angle and from torque input to the end point accelerations of each link. An experimental work was established to collect the input-output data pairs and used in developing the system model. Since the system is highly nonlinear, NARX model was chosen as the model structure because of its simplicity. The nonlinear characteristic of the system was estimated using the ANN whereby multi-layer perceptron (MLP) and ELMAN neural network (ENN) structure were utilized. The implementation of the ANN and its’ effectiveness in developing the model of DLFRM was emphasized. The performance of the MLP was compared to ENN based on the validation of the mean-squared error (MSE) and correlation tests of the developed models. The results indicated that the identification of the DLFRM system using the MLP outperformed the ENN with lower mean squared prediction error and unbiased results for all the models. Thus, the MLP provides a good approximation of the DLFRM dynamic model compared to the ENN.