Intelligent model for endpoint accelerations of two link flexible manipulator using a deep learning neural network

• Main Authors: Abdulghani, Zidan, Mat Darus, Intan Z., Jamali, Annisa
• Format: Conference or Workshop Item
• Language: English
• Published: 2022
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://eprints.utm.my/id/eprint/99345/1/ZidanAbdulghani2022_IntelligentModelforEndpointAccelerations.pdf; http://eprints.utm.my/id/eprint/99345/; http://dx.doi.org/10.1109/ICSIMA55652.2022.9928949

This article investigates a two-link flexible manipulator (TLFM) that can be modelled utilizing a deep learning neural network. The system was classified under a multiple-input multiple-output (MIMO) system. In the modelling stage of this study, the TLFM dynamic models were divided into single-input single-output (SISO) models. Since coupling impact was assumed to be minimised, the characterizations of TLFM were defined independently in each model. Two discrete SISO models of a flexible two link manipulator were developed using the torque input and the endpoint accelerations of each link. The input-output data pairs were collected from experimental work and utilised to establish the system model. The Long Short-Term Memory (LSTM) algorithm optimised using Particle Swarm Optimization (PSO) was selected as the model structure due to the system's high degree of nonlinearity. The identification of the TLFM system utilizing LSTM optimised by PSO was successful, according to the high-performance result of PSO. Using LSTM-PSO, it is demonstrated that both link 1 and 2 models are accurately identified and that their performance in terms of MSE for links endpoint acceleration 1 and 2 is within a 95% confidence interval.