TY - GEN
T1 - Enabling predictive maintenance using machine learning in industrial machines with sensor data
AU - Andriani, Adelina Zian
AU - Kurniati, Nani
AU - Santosa, Budi
N1 - Publisher Copyright:
© IEOM Society International.
PY - 2021
Y1 - 2021
N2 - In line with the advancement of Industry 4.0 which provides opportunities for the utilization of sensors and Machine Learning (ML) technology, make Predictive Maintenance (PdM) practices much easier. Regarding implementing PdM with ML, manufacturers need to provide data that supports the machine learning process. However, the majority of data is unlabeled and still requires manual labeling to support the learning process, which is risky, costly, and laborintensive. Therefore, the current research uses the integration of Active Learning (AL) and Semi-Supervised Learning (SSL) to solve labeling problems and support PdM models with a better level of generalization. First, unlabeled multisensory data stored on the main server database and slight labeled data becomes the research sample. Second, the AL scheme selects the most valuable unlabelled samples, to label and add to the training data set. Third, the SSL scheme to optimize the data usage, using the remaining samples to be labeled. Finally, based on the augmented training data set, the fault diagnostic model is trained to support the failure class prediction. Regarding the selection of the ML algorithm, the result of trained Random Forest Classification (RFC) could predict a fault diagnostic model of approximately 99,85%.
AB - In line with the advancement of Industry 4.0 which provides opportunities for the utilization of sensors and Machine Learning (ML) technology, make Predictive Maintenance (PdM) practices much easier. Regarding implementing PdM with ML, manufacturers need to provide data that supports the machine learning process. However, the majority of data is unlabeled and still requires manual labeling to support the learning process, which is risky, costly, and laborintensive. Therefore, the current research uses the integration of Active Learning (AL) and Semi-Supervised Learning (SSL) to solve labeling problems and support PdM models with a better level of generalization. First, unlabeled multisensory data stored on the main server database and slight labeled data becomes the research sample. Second, the AL scheme selects the most valuable unlabelled samples, to label and add to the training data set. Third, the SSL scheme to optimize the data usage, using the remaining samples to be labeled. Finally, based on the augmented training data set, the fault diagnostic model is trained to support the failure class prediction. Regarding the selection of the ML algorithm, the result of trained Random Forest Classification (RFC) could predict a fault diagnostic model of approximately 99,85%.
KW - Active Learning
KW - And Fault Diagnostic
KW - Predictive Maintenance (PdM)
KW - Semi-Supervised Learning
UR - http://www.scopus.com/inward/record.url?scp=85121132549&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85121132549
SN - 9781792361258
T3 - Proceedings of the International Conference on Industrial Engineering and Operations Management
SP - 2366
BT - Proceedings of the International Conference on Industrial Engineering and Operations Management, 2021
PB - IEOM Society
T2 - 2nd South American Conference on Industrial Engineering and Operations Management, IEOM 2021
Y2 - 5 April 2021 through 8 April 2021
ER -