Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Instrumentation of a Parallel Manipulator with Flexible Links: a Neural Network application
Submission Author:
Fabio Felix , MG , Brazil
Co-Authors:
Fabio Felix, Maíra Martins da Silva, Guilherme Serpa Sestito
Presenter: Fabio Felix
doi://10.26678/ABCM.COBEM2021.COB2021-0299
Abstract
Parallel kinematic manipulators (PKMs) present higher speed/acceleration ratios, load capacity, rigidity and compaction when compared with serial manipulators. Some industrial applications exploit some of these advantages, such as high speed/high-precision milling machines, flight and automobile simulators, pick-and-place machine, and surgical robots. The reduction of the inertia of their components can improve their dynamic performance and energy efficiency. This design alternative might yield vibrations requiring the implementation of model-based joint space or task space control strategies. While the former's implementation involves the derivation of precise models, the latter strategy requires adequate computation vision schemes. These requirements impose critical challenges such as the use of model updating and image processing techniques. For this reason, studies regarding the use of redundant sensors have been carried out to obtain relevant data for PKM with flexible links. Experimental data from a planar 3RRR is extracted using encoders, strain gauges and a camera during the execution of pre-determined tasks. An Artificial Neural Network is used for estimating the pose of the end-effector's manipulator. In this work, the implementation of this technique is done using a Multi-Layer Perceptrons (MLP) topology for estimating the manipulator's pose. This estimation can be used for improving dynamic models or for implementing task space control strategies.
Keywords
Instrumentation, Parallel Manipulators with Flexible Links, Artificial neural networks (ANN), Position Estimation
DOWNLOAD PDF VIEW PRESENTATION
