The classification of impact signal of 6 DOF cobot by means of machine learning model

Collaborative robot (Cobot) has seen a rise in adoption rate in the industry as the Industry 4.0 era marches in. Cobot were introduced to replace human operators in harsh environments or repetitive work processes. The health condition monitoring of these cobot have not been standardized due to lack...

Full description

Bibliographic Details
Main Authors: Kai, Gavin Lim Jiann, Ahmad Fakhri, Ab. Nasir, P.P. Abdul Majeed, Anwar, Mohd Azraai, Mohd Razman, Ismail, Mohd Khairuddin, Li, Lim Thai
Format: Conference or Workshop Item
Language:English
English
Published: Springer Science and Business Media Deutschland GmbH 2022
Subjects:
Online Access:http://umpir.ump.edu.my/id/eprint/42303/1/The%20classification%20of%20impact%20signal%20of%206%20DOF%20cobot.pdf
http://umpir.ump.edu.my/id/eprint/42303/2/The%20classification%20of%20impact%20signal%20of%206%20DOF%20cobot%20by%20means%20of%20machine%20learning%20model_ABS.pdf
Description
Summary:Collaborative robot (Cobot) has seen a rise in adoption rate in the industry as the Industry 4.0 era marches in. Cobot were introduced to replace human operators in harsh environments or repetitive work processes. The health condition monitoring of these cobot have not been standardized due to lack of widely available standardized fault dataset and the high complexity of diagnostic. This study aims to use machine learning algorithms as a mean to identify the cobot pick and place process offset error using vibrational signals. The vibrational sensor was attached to the end effector of the cobot where the vibration signal of 3 axis were collected. The features were then extracted, standardized, and 544 features were selected from 2337 features based on a hypothesis testing method. The dataset was then spilt into training and testing by a ratio of 80:20. Three machine learning models namely, the k-Nearest Neighbors (k-NN), Neural Network (NN), and Support Vector Machine (SVM) classifier were tested, and the classification accuracy of the models was analyzed. A grid search approach was used to identify the best hyperparameter for each model. The model with the highest classification accuracy of 95.2% was the MLP model compared to SVM (92.4%) and kNN (79%). Therefore, it could be established from the study that a comparable classification efficacy is attainable through the identification of significant features. The findings are non-trivial, particularly with respect to the implementation of the developed classifier in real-time.