Blind Image Super Resolution Using Deep Unsupervised Learning

The goal of single image super resolution (SISR) is to recover a high-resolution (HR) image from a low-resolution (LR) image. Deep learning based methods have recently made a remarkable performance gain in terms of both the effectiveness and efficiency for SISR. Most existing methods have to be trained based on large-scale synthetic paired data in a fully supervised manner. With the available HR natural images, the corresponding LR images are usually synthesized with a simple fixed degradation operation, such as bicubic down-sampling. Then, the learned deep models with these training data usually face difficulty to be generalized to real scenarios with unknown and complicated degradation operations. This study exploits a novel blind image super-resolution framework using a deep unsupervised learning network. The proposed method can simultaneously predict the underlying HR image and its specific degradation operation from the observed LR image only without any prior knowledge. The experimental results on three benchmark datasets validate that our proposed method achieves a promising performance under the unknown degradation models.

A Predictive Maintenance Model for Flexible Manufacturing in the Context of Industry 4.0

The Industry 4.0 paradigm is the focus of modern manufacturing system design. The integration of cutting-edge technologies such as the Internet of things, cyber–physical systems, big data analytics, and cloud computing requires a flexible platform supporting the effective optimization of manufacturing-related processes, e.g., predictive maintenance. Existing predictive maintenance studies generally focus on either a predictive model without considering the maintenance decisions or maintenance optimizations based on the degradation models of the known system. To address this, we propose PMMI 4.0, a Predictive Maintenance Model for Industry 4.0, which utilizes a newly proposed solution PMS4MMC for supporting an optimized maintenance schedule plan for multiple machine components driven by a data-driven LSTM model for RUL (remaining useful life) estimation. The effectiveness of the proposed solution is demonstrated using a real-world industrial case with related data. The results showed the validity and applicability of this work.

Identifiability Analysis of Degradation Model Parameters from Transient CO2 Release in Low-Temperature PEM Fuel Cell under Various AST Protocols

The detrimental effects of the catalyst degradation on the overall envisaged lifetime of low-temperature proton-exchange membrane fuel cells (LT-PEMFCs) represent a significant challenge towards further lowering platinum loadings and simultaneously achieving a long cycle life. The elaborated physically based modeling of the degradation processes is thus an invaluable step in elucidating causal interaction between fuel cell design, its operating conditions, and degradation phenomena. However, many parameters need to be determined based on experimental data to ensure plausible simulation results of the catalyst degradation models, which proves to be challenging with the in situ measurements. To fill this knowledge gap, this paper demonstrates the application of a mechanistically based PEMFC modeling framework, comprising real-time capable fuel cell performance, and platinum and carbon support degradation models, to model transient CO2 release rates in the LT-PEMFCs with the consistent calibration of reaction rate parameters under multiple different accelerated stress tests at once. The results confirm the credibility of the physical and chemical modeling basis of the proposed modeling framework, as well as its prediction and extrapolation capabilities. This is confirmed by an increase of only 29% of root mean square deviations values when using a model calibrated on all three data sets at once in comparison to a model calibrated on only one data set. Furthermore, the unique identifiability and interconnection of individual model calibration parameters are determined via Fisher information matrix analysis. This analysis enables optimal reduction of the set of calibration parameters, which results in the speed up of both the calibration process and the general simulation time while retaining the full extrapolation capabilities of the framework.

Online Model-Based Remaining-Useful-Life Prognostics for Aircraft Cooling Units Using Time-Warping Degradation Clustering

Remaining-useful-life prognostics for aircraft components are central for efficient and robust aircraft maintenance. In this paper, we propose an end-to-end approach to obtain online, model-based remaining-useful-life prognostics by learning from clusters of components with similar degradation trends. Time-series degradation measurements are first clustered using dynamic time-warping. For each cluster, a degradation model and a corresponding failure threshold are proposed. These cluster-specific degradation models, together with a particle filtering algorithm, are further used to obtain online remaining-useful-life prognostics. As a case study, we consider the operational data of several cooling units originating from a fleet of aircraft. The cooling units are clustered based on their degradation trends and remaining-useful-life prognostics are obtained in an online manner. In general, this approach provides support for intelligent aircraft maintenance where the analysis of cluster-specific component degradation models is integrated into the predictive maintenance process.