Predictive Analysis of Robotic Manipulators Through Inertial Sensors and Pattern Recognition

Predictive Analysis of Robotic Manipulators Through Inertial Sensors and Pattern Recognition

Jorge Alonso Moro (Universidad Europea de Madrid, Spain), Carlos Quiterio Gómez Muñoz (Universidad Europea de Madrid, Spain) and Fausto Pedro García Márquez (University of Castilla-La Mancha, Spain)
Copyright: © 2020 |Pages: 11
DOI: 10.4018/978-1-7998-0106-1.ch015

Abstract

Industrial robotics is constantly evolving, with installation forecast of about 2 million new robots in 2020. The predictive maintenance focused on industrial robots is beginning to be applied more, but its possibilities have not yet been fully exploited. The present study focuses on the applications offered by inertial sensors in the field of industrial robotics, specifically the possibility of measuring the “real” rotation angle of a robotic arm and comparing it with its own system of measure. The study will focus on the measurement of the backlash existing in the gearbox of the axis of a robot. Data received from the sensor will be analysed using the wavelet transform, and the mechanical state of the system could be determined. The introduction of this sensing system is safe, dynamic, and non-destructive, and it allows one to perform the measurement remotely, in the own installation of the robot and in working conditions. These features allow one to use the device in different predictive functions.
Chapter Preview
Top

Introduction

Nowadays, the diagnosis of the structural state of the robotic manipulator arms is made in a slightly optimized way, which involves the supervision of an operator of the movement of each axis of the arm. Thus, it is not possible to have a clear measurement of the dysfunction of the robot arm by observing the evolution of the problems during the work of a robotic cell. It must be taken into account that there are mechanical breakdowns, due to backlash in their gearboxes, which involve production stops of around 12 hours. In the case that the spare part is not available in the necessary time, this stop would be extended until the arrival of the replacement. Moreover, the possibility of completely replacing an old robot with a new one involves a great investment and a time of intervention in the production line that would force to stop production during the same.

Currently, preventive and corrective rather than predictive maintenance is performed, replacing the mechanics parts when it reaches its useful life, as suggested by the manufacturer and making preventive ranges of change of greases and oils.

Several studies like Sarkar, Ellis, and Moore (1997), detects mechanical backlash through signal modelling of the torque produced by a motor in a series of gear wheels. This type of case is undesired since the mechanics of the robot itself are usually cycloidal gearboxes. It would be also necessary to take measurements with an external equipment at the ends of each axis of the robot. It is possible to find systems that try to reduce these problems during production, such as the TCP tool calibration. TCP tool is a device in a circular hollow shape, in which the robot tool moves every certain time to determine externally the real position in the space of the robot tool, and it allows to correct the trajectories of the same in dynamic (Cox and Freeman, 2005a). This type of device corrects the possible dispersions that may have during the process, but it will not indicate any information about the mechanical state of the manipulator.

There are patented systems which perform a neural network to determine what type of failure has the robot with the help of several sensors that run through the manipulator and the monitoring of these data (Sjöstrand, Blanc, and Tavallaey, 2010). The analysed parameters are: temperature of the gearboxes, friction of the axis and temperature of the motor, among others (Cox and Freeman, 2005b & Huang, 2017 & Stoica and Moses, 1997 & Vallely and Papaelias, n.d. & Shi et. Al, 2018).

Nowadays, it is possible to find integrated predictive maintenance systems that read different variables of the robot, such as the intensity, voltage and temperature of its motors (Amini, 2016 & Shi, Soua, and Papaelias, 2017 & Muñoz, Jiménez, and Márquez, 2017 & Jiménez et. Al, n.d. & Jiménez et. Al, n.d.). But the analysis of all this data is done externally to the robot, from embedded computers from other manufacturers.

Complete Chapter List

Search this Book:
Reset