| Titre : | Deep Learning for predictive maintenance |
| Auteurs : | Ikram Remadna, Auteur ; Sadek Labib Terrissa, Directeur de thèse |
| Type de document : | Thése doctorat |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2023 |
| Format : | 1 vol. (115 p.) / couv. ill. en coul / 30 cm |
| Langues: | Anglais |
| Mots-clés: | Prognostics and Health Management (PHM), Remaining useful life (RUL), Predictive Maintenance (PdM), C-MAPSS dataset, Ensemble learning, Deep learning |
| Résumé : |
Recently, with the appearance of Industry 4.0 (I4.0), machine learning (ML) within artificial intelligence (AI), industrial Internet of things (IIoT) and cyber-physical system (CPS) have accelerated the development of a data-orientated applications such as predictive maintenance (PdM). PdM applied to asset-dependent industries has led to operational cost savings, productivity improvements and enhanced safety management capabilities. In addition, predictive maintenance strategies provide useful information concerning the source of the failure or malfunction, reducing unnecessary maintenance operations. The concept of prognostics and health management (PHM) has appeared as a predictive maintenance process. PHM has become an unavoidable tendency in smart manufacturing to offer a reliable solution for handling industrial equipment’s health status. This later requires efficient and effective system health monitoring methods, including processing and analysing massive machinery data to detect anomalies and perform diagnosis and prognosis. Prognostics is considered a key PHM process with capabilities for predicting future states, mainly based on predicting the residual lifetime during which a machine can perform its intended function, i.e., estimating the remaining useful life (RUL) of a system. The prognostic research domain is far from being mature, which is still new and explains the various challenges that must be addressed. Therefore, the work presented in this thesis will mainly focus on the prognostic of monitored machinery from an RUL estimation point of view using Deep Learning (DL) algorithms. Capitalising on the recent success of the DL, this dissertation introduces methods and algorithms dedicated to predictive maintenance. We focused on improving the performance of aero-engine prognostic, particularly in estimating an accurate RUL using ensemble learning and deep learning. To this end, two contributions have been proposed, and the results obtained were validated by an extensive comparative analysis using public C-MAPSS turbofan engine benchmark datasets. The first contribution, for RUL predictions, we proposed two-hybrid methods based on the promising DL architectures to leverage the power of the multimodal and hybrid deep neural network in order to capture various information at different time intervals and ultimately achieve more accurate RUL predictions. The proposed end-to-end deep architectures jointly optimise the feature reduction and RUL prediction steps in a hierarchical manner, intending to achieve data representation in low dimensionality and minimal variable redundancy while preserving critical asset degradation information with minimal preprocessing effort. The second contribution, in a practical situation, RUL is usually affected by uncertainty. Therefore, we proposed an innovative RUL estimation strategy that assesses degrading machinery’s health status (provides the probabilities of system failure in different time windows) and provides the prediction of RUL window. |
| Sommaire : |
Contents Abstract i Résumé ii List of Publications iv Acknowledgements vi I INTRODUCTION 1 I.1 Context and Motivation . . . . . . . . . . 2 I.2 Purpose and Research Questions . . . . . . . . . 4 I.3 Assumptions . . . . . . . . . . . .. . . . . . . 5 I.4 Contributions . . . . . . . . . . . . . . 6 I.5 Manuscript organisation . . . . . . . . . . . . 7 II Predictive Maintenance and PHM 9 II.1 Introduction . . . . . . . . . . . . . . . . . . . . . . 10 II.2 Towards the 4th Industrial Revolution . . . .. . . 10 II.3 Evolution of maintenance . . . . . . . . . . 11 II.4 PHM paradigm . . . . . . . . .. . . . . . . 15 II.4.1 Data Acquisition . . . . . . . . . . . 15 II.4.2 Data Processing . . . . . . . . . . . . . . 15 II.4.3 Condition Monitoring . . . . . . . . . 17 II.4.4 Diagnostics . . . . . . . . . . . . . . . . 17 II.4.5 Prognostics . . . . . . . . . . . . . . 17 II.4.6 Decision support . . . . . .. . . . . . . . 19 II.4.7 Human-Machine Interface . .. 19 II.5 Benchmarking datasets for System-Level Prognostics . . . . . 19 II.5.1 Occupancy data set . . . . . . . . . . . . 19 II.6 Conclusion . . . . . . . . . . . . . . . . . . 24 III Prevailing ML and DL 25 III.1 Introduction . . . . . . . . . . . . . . . . 26 III.2 Machine Learning Basics . . .. . . . . . . 26 III.2.1 Learning scenarios . . . . . . . . 27 III.2.1.1 Supervised learning . . . .. . . 27 III.2.1.2 Unsupervised learning . . . . . . 29 III.2.1.3 Semi-Supervised learning . . .. . 30viii III.2.1.4 Reinforcement learning . . . . . . 30 III.2.1.5 Multi-Task Learning . . . . 30 III.2.1.6 Transfer Learning . . . .. . . . 31 III.3 A brief overview of deep neural network architectures . . . . . . . . . 31 III.3.1 Recurrent Neural Network . . . . . . . . 33 III.3.2 Convolutioanl Neural Network . . . . . . 35 III.3.3 Auto-encoders . . . . . . . . . . . . . . 37 III.3.3.1 Variations of AEs . . . . . . 37 III.3.4 Regularization for Deep Learning . .. . 39 III.4 Deep Learning frameworks . . . . . . . 41 III.5 Data-driven prognostic challenges and issues . . . . . . . . . . . . . . . 41 III.5.1 Data-related challenges . . . . . . . 43 III.5.2 Model-related challenges . . . . . . 43 III.5.3 Synthesis: Toward enhanced data-driven prognostics . . . . . . 44 III.5.3.1 Issues to be addressed . . .. . . 44 III.5.3.2 Assumptions . . . . . . . 45 III.5.3.3 Objective and contributions . . . . 45 III.6 Conclusion . . . . .. . . . . . 46 IV Leveraging the Power of Multimodal and Hybrid Deep Neural Network Techniques for RUL Estimation Enhancement 47 IV.1 Introduction . . . . . . . . 48 IV.2 State-of-the-art Deep Learning Methods for Engines RUL Estimation . 49 IV.2.1 CNN . . . . . . . . . . . . . . . . 49 IV.2.2 RNN and its variants . . . . . . . . 49 IV.2.3 DNN using auto-encoders . . . . . . 50 IV.3 Proposed Hybrid Deep Neural Network architectures . . . . 51 IV.3.1 Convolutional Auto-encoder with BDGRU-BDLSTM Hybrid Model . . . . . . . . . . . . . . . . . . . 52 IV.3.2 CNN-BDGRU Hybrid Model . . . . . 54 IV.4 Experiment study . . . . . . . . . . . . . . . 55 IV.4.1 Data pre-processing . . . . . . . . . . . 56 IV.4.1.1 Data Normalization . . . . . . 56 IV.4.1.2 Masking and padding . . . 56 IV.4.2 Evaluation Metric . . . . . .. . . . . . 56 IV.4.3 Prediction procedure . . . . . . .. . . 57 IV.4.3.1 CNN-BDGRU Training procedure . . . . . . . . . . . 58 IV.4.3.2 CAE with BDGRU-BDLSTM Training procedure . . . 59 IV.4.4 Results and Discussions . . . . . . . . . 60 IV.4.4.1 Prediction performance . . . . . 60 IV.4.4.2 Computational Cost Analysis . . .. . 64 IV.4.4.3 Compared with other approaches . . .. 65 IV.4.4.4 Effect analysis . . . . . . . . . . . . 67 IV.4.4.5 Comparison with the latest works . . . 67ix IV.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 V RUL Prediction using a fusion of Attention-based Convolutional Variational AutoEncoder and Ensemble Learning Classifier 70 V.1 Introduction . . . . . . . 71 V.2 Related work . . . . . . . . . . 72 V.2.1 Data-driven methods for RUL estimation . . . . . . . . . . . . . 72 V.2.2 Dimensionality Reduction and Visual Explanation techniques . 73 V.2.3 Model’s Hyperparameters optimization . . . . . . . . . . . . . . 74 V.2.4 Research Gaps and Contribution . . . 74 V.3 Methodology . . . . . . . . . . . . . . 76 V.3.1 Problem Formulation . . . . . 76 V.3.2 Remaining useful life estimation based on CVAE with attention mechanism . V.4 Results analysis . .. 80 V.4.1 Evaluation metrics .. . . . 81 V.4.2 Data preprocessing . . . . . . 82 V.4.2.1 Feature selection . . . . . . 82 V.4.2.2 Data Normalization . . .. . 82 V.4.2.3 Sliding Window . . . . . . . . 84 V.4.2.4 Data rebalancing . . . . . . 84 V.4.3 Results . . . . . . . .. . . . 85 V.4.3.1 Data Pre-processing Parameters Analysis . . . . . . . 85 V.4.3.2 Visualisation of latent vectors and identification of the conflict zone . . .. . . . . . 89 V.4.3.3 Performance analysis . . . . 92 V.5 Conclusion . . . . . . . . . . . . . . . 97 VI Achievements and Conclusions 98 VI.1 Thesis aims . . . . . . .. . . . 99 VI.2 Thesis Contributions . . . . . . . . . . 100 VI.2.1 First Contribution . . . . . . . . 100 VI.2.2 Second Contribution . . . . . .. 100 VI.3 Perspectives . . . . . . . . .. . . . 101 Bibliography 10 |
| En ligne : | http://thesis.univ-biskra.dz/6039/1/Thesis_Final_Version__Copy_-3.pdf |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| TINF/184 | Théses de doctorat | bibliothèque sciences exactes | Consultable |
