A prognostic driven predictive maintenance framework based on Bayesian deep learning

Article Engineering, Mechanical

RUL prediction of machinery using convolutional-vector fusion network through multi-feature dynamic weighting

Xiaofei Liu et al.

Summary: A novel convolutional vector fusion network (C-VFN) is proposed in this paper to improve the prediction accuracy of remaining useful life (RUL) of machinery by dynamically evaluating the degradation sensitivity of each feature.

MECHANICAL SYSTEMS AND SIGNAL PROCESSING (2023)

Add to Collection

Article Engineering, Industrial

Prognosis-driven reliability analysis and replacement policy optimization for two-phase continuous degradation

Jiantai Wang et al.

Summary: This study proposes a prognosis-driven, multi-threshold inspection and replacement model for two-phase continuous degradation. Inspections are conducted equally spaced to reveal the health state and real-time degradation level of the system. Different control limits are scheduled to manipulate replacement frequencies under separate system states. The cost model seeks to minimize the long-run cost rate through the joint optimization of inspection interval, control limits, and safety coefficients. A case study on high-speed rail bearings justifies the applicability of the proposed model.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2023)

Add to Collection

Article Engineering, Industrial

Deep reinforcement learning for predictive aircraft maintenance using probabilistic Remaining-Useful-Life prognostics

Juseong Lee et al.

Summary: This study proposes a framework that integrates data-driven probabilistic Remaining-Useful-Life (RUL) prognostics with predictive maintenance planning, using aircraft turbofan engines as an example. By employing this framework, the total maintenance cost can be reduced, unscheduled maintenance can be prevented, and the wasted life of engines can be limited.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

A hierarchical scheme for remaining useful life prediction with long short-term memory networks

Tao Song et al.

Summary: Remaining useful life (RUL) prediction is crucial in prognostics and health management (PHM) applications, and data-driven approaches have gained increasing attention for capturing the degradation process of complex systems using operating data. This study formulates RUL prediction as a bi-level optimization problem and proposes a bi-level deep learning scheme to predict machine RUL. The proposed framework outperforms state-of-the-art approaches in case studies.

NEUROCOMPUTING (2022)

Add to Collection

Review Engineering, Industrial

Sustainable manufacturing, maintenance policies, prognostics and health management: A literature review

Pascal Vrignat et al.

Summary: The increasing complexity of industrial processes and the continuing search for higher profits call for a proactive and sustainable maintenance policy. In an Industry 4.0 context, being able to anticipate system breakdowns and propose maintenance interventions in advance is essential.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Industrial

Maintenance optimisation of multicomponent systems using hierarchical coordinated reinforcement learning

Yifan Zhou et al.

Summary: This paper introduces a hierarchical coordinated reinforcement learning (HCRL) algorithm to optimize maintenance of large-scale multicomponent systems, with agent parameters and coordination relationships designed based on system characteristics, and a hierarchical structure established according to components' structural importance measures. The effectiveness of the algorithm is confirmed through validation on different systems, outperforming other methods including deep reinforcement learning.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Industrial

A novel dual-stream self-attention neural network for remaining useful life estimation of mechanical systems

Danyang Xu et al.

Summary: This paper proposes a novel data-driven method called Dual-Stream Self-Attention Neural Network (DS-SANN) for RUL estimation. By employing the multi-head self-attention mechanism and dual-stream structure network, DS-SANN can better capture the correlations and internal differences of monitoring data, leading to improved estimation performance for RUL.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Industrial

Artificial-intelligence-based maintenance decision-making and optimization for multi-state component systems

Van-Thai Nguyen et al.

Summary: In this paper, an artificial intelligence-based maintenance approach is proposed, which uses artificial neural network for predicting maintenance cost at system level and employs deep reinforcement learning algorithm for optimizing maintenance decisions. The experimental results demonstrate the effectiveness of this approach in maintenance cost forecasting and decision optimization.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Industrial

Bayesian deep-learning for RUL prediction: An active learning perspective

Rong Zhu et al.

Summary: This study proposes a Bayesian deep-active-learning framework for RUL prediction, which aims to achieve high accuracy with limited run-to-failure data. The method utilizes Bayesian neural networks for prediction and actively selects target samples for acquiring run-to-failure labels based on prediction uncertainty. A recursive model training and active data selection mechanism are developed to maintain accuracy.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Industrial

Probabilistic forecasting informed failure prognostics framework for improved RUL prediction under uncertainty: A transformer case study

J. Aizpurua et al.

Summary: This paper presents a probabilistic prognostics method for lifetime planning under uncertainty, integrating data-driven probabilistic forecasting models with expert-knowledge based Bayesian filtering methods. The results confirm that the proposed probabilistic transformer lifetime estimate aids in the decision-making process with informative lifetime distributions and associated confidence intervals.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2022)

Add to Collection

Article Engineering, Electrical & Electronic

A Novel Transfer Learning Approach in Remaining Useful Life Prediction for Incomplete Dataset

Shahin Siahpour et al.

Summary: Thanks to the successful implementation of intelligent data-driven approaches, predicting the remaining useful life (RUL) problems has gained significant attention. Transfer learning approaches are used to transfer knowledge from source domain data to target domain data. However, there is a discrepancy between the data distribution of source and target domain datasets, which is solved by domain adaptation techniques. This article proposes a transfer learning approach for RUL prediction using consistency-based regularization to handle missing information in the incomplete target domain dataset.

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT (2022)

Add to Collection

Article Engineering, Electrical & Electronic

Dynamic Predictive Maintenance Scheduling Using Deep Learning Ensemble for System Health Prognostics

Chuang Chen et al.

Summary: This paper presents a dynamic predictive maintenance strategy for modern engineering systems using a deep learning ensemble model. The model, consisting of a deep autoencoder and bidirectional long short-term memory, accurately estimates system health state and remaining useful life. The effectiveness of the proposed strategy is demonstrated by comparing it with recent publications using a dataset from NASA.

IEEE SENSORS JOURNAL (2021)

Add to Collection

Review Computer Science, Artificial Intelligence

A review of uncertainty quantification in deep learning: Techniques, applications and challenges

Moloud Abdar et al.

Summary: Uncertainty quantification (UQ) methods are essential in reducing uncertainties in optimization and decision making processes. Bayesian approximation and ensemble learning techniques are widely used types of UQ methods.

INFORMATION FUSION (2021)

Add to Collection

Article Computer Science, Interdisciplinary Applications

Reinforcement learning-driven maintenance strategy: A novel solution for long-term aircraft maintenance decision optimization

Yang Hu et al.

Summary: This paper proposes a novel reinforcement learning-driven aircraft maintenance strategy, which can provide real-time maintenance decisions based on aircraft's future mission requirements, repair costs, spare parts storage, and PHM outputs. Testing in simulated scenarios demonstrates that the strategy can better adapt to the varying requirements of different maintenance situations.

COMPUTERS & INDUSTRIAL ENGINEERING (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

Aleatoric and epistemic uncertainty in machine learning: an introduction to concepts and methods

Eyke Huellermeier et al.

Summary: Uncertainty plays a major role in machine learning, with aleatoric and epistemic uncertainty being two key types. As machine learning becomes increasingly relevant for practical applications, handling uncertainty becomes more important.

MACHINE LEARNING (2021)

Add to Collection

Article Engineering, Industrial

Multicellular LSTM-based deep learning model for aero-engine remaining useful life prediction

Sheng Xiang et al.

Summary: A novel multicellular LSTM model is proposed for predicting the remaining useful life (RUL) of aero-engines, successfully mining different levels of degradation trends. By combining MCLSTM with a deep neural network (DNN), health indicators (HIs) are extracted and RUL is predicted.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2021)

Add to Collection

Article Automation & Control Systems

A Risk-Averse Remaining Useful Life Estimation for Predictive Maintenance

Chuang Chen et al.

Summary: This work develops a risk-averse RUL estimation method by incorporating degradation feature selection, support vector regression, and long short-term memory network, which successfully reduces the over-estimation rate while maintaining a reasonable under-estimation level to enhance prediction robustness and marginal utility. The proposed method shows feasibility and effectiveness in predictive maintenance based on verification using an aero-engine dataset from NASA.

IEEE-CAA JOURNAL OF AUTOMATICA SINICA (2021)

Add to Collection

Article Computer Science, Hardware & Architecture

Failure Prognosis and Applications-A Survey of Recent Literature

Mojtaba Kordestani et al.

Summary: Fault diagnosis and prognosis are crucial functionalities in complex and safety-critical engineering systems, enabling early fault detection, isolation, prediction of fault propagation, preventive maintenance, and serving as a countermeasure to catastrophic incidents. This article provides a survey of recent research on fault prognosis, reports on significant application domains, and outlines potential directions for future research.

IEEE TRANSACTIONS ON RELIABILITY (2021)

Add to Collection

Article Engineering, Industrial

An adaptive predictive maintenance model for repairable deteriorating systems using inverse Gaussian degradation process

K. T. Huynh

Summary: An adaptive predictive maintenance model is developed for deteriorating single-unit systems, considering past dependency of maintenance actions, system remaining useful life, and maintenance duration for dynamic decision-making, evaluated using semi-regenerative theory. It shows flexibility and cost-effectiveness through comparative studies with benchmarks under different maintenance costs and degradation characteristics.

RELIABILITY ENGINEERING & SYSTEM SAFETY (2021)

Add to Collection

Article Computer Science, Information Systems

A Data-Driven Maintenance Framework Under Imperfect Inspections for Deteriorating Systems Using Multitask Learning-Based Status Prognostics

Lei Zhang et al.

Summary: This paper introduces a new data-driven maintenance framework utilizing deep learning for multitask learning to predict equipment performance degradation and failure probability, improving prediction accuracy and the optimization of the maintenance process.

IEEE ACCESS (2021)

Add to Collection

Article Automation & Control Systems