Kristupas Bajarunas

Deep learning has demonstrated significant potential for prognostics in complex systems (Fink et al., 2020). Recent advances in physics-informed machine learning have integrated physics-of-failure principles within data-driven models (AriasChao, Kulkarni, Goebel, & Fink, 2022). Beyond physical laws, fleet-level time-to-failure (TTF) distributions provide valuable prior knowledge for individual asset life predictions.In this paper we derive a probabilistic analytical health index(HI) model based on power-law degradation, enabling a probabilistic description that reconciles individual variability with fleet-wide trends. We show that, under Weibull, Gamma, and Pareto-distributed TTFs, the HI evolution follows an analytical form, allowing explicit characterization of time to reach intermediate degradation levels. Therefore, this work provides a theoretical foundation for integrating reliability principles with deep learning, advancing towards Reliability-Informed Deep Learning (RIDL). The approach is validated on synthetic turbofan engine data and real-world battery degradation datasets. This work establishes a rigorous basis for embedding reliability engineering principles into deep learning, improving predictive maintenance and remaining useful life (RUL) estimation.

Accurately estimating a Health Index (HI) from condition monitoring data (CM) is essential for reliable and interpretable prognostics and health management (PHM) in complex systems. In most scenarios, complex systems operate under varying operating conditions and can exhibit different fault modes, making unsupervised inference of an HI from CM data a significant challenge. Hybrid models combining prior knowledge about degradation with deep learning models have been proposed to overcome this challenge. However, previously suggested hybrid models for HI estimation usually rely heavily on system-specific information, limiting their transferability to other systems. In this work, we propose an unsupervised hybrid method for HI estimation that integrates general knowledge about degradation into the convolutional autoencoder's model architecture and learning algorithm, enhancing its applicability across various systems. The effectiveness of the proposed method is demonstrated in two case studies from different domains: turbofan engines and lithium batteries. The results show that the proposed method outperforms other competitive alternatives, including residual-based methods, in terms of HI quality and their utility for Remaining Useful Life (RUL) predictions. The case studies also highlight the comparable performance of our proposed method with a supervised model trained with HI labels.

Hybrid models combining physical knowledge and machine learning show promise for obtaining accurate and robust prognostic models. However, despite the increased interest in hybrid models in recent years, the proposed solutions tend to be domain-specific. As a result, there is no compelling strategy of what, where, and how physics-derived knowledge can be integrated into deep learning models depending on the available representation of physical knowledge and the quality of data for the development of prognostic models for complex systems. This Ph.D. project aims to develop a general strategy for hybridizing prognostic models by exploring multiple methods to incorporate physical knowledge at various stages of the learning algorithm. The project will prioritize expert knowledge as the primary source of information, while domain-specific knowledge will serve as an additional feature when applicable.

Developing Health Indicators (HI) is a crucial aspect of prognostics and health management of complex systems. Previous research has demonstrated the benefits of accurately determining the HI, which can lead to better performance of prognostic models. However, the existing methodologies for determining HI in complex systems are mostly semi-supervised and rely on assumptions that may not hold in real-world scenarios. The existing methods usually involve using a reference set of healthy sensor readings or run-to-failure data to infer HI. But the unsupervised inference of HI from sensor readings, which is challenging in scenarios where diverse operating conditions can mask the effect of degradation on sensor readings, has not been extensively researched. In this paper, we propose a novel physics-informed unsupervised model for determining HI. Unlike previous methods, constrained by assumptions, the proposed method uses prior general knowledge about degradation to infer HI, thereby eliminating the need for a reference set of healthy sensor readings. The proposed unsupervised model is an Autoencoder that incorporates constraints on its latent space to ensure consistency with knowledge about degradation. We assess the efficacy of the proposed model by analyzing a prevalent prognostic case study, specifically the turbofan engine dataset (N-CMAPSS). Our analysis considers the model's sensitivity to data availability and the resulting Health Index's quality, including trendability and monotonicity. Additionally, we investigate the impact of incorporating the Health Index in predicting Remaining Useful Life (RUL). We demonstrate that our proposed method generates a Health Index that exhibits greater monotonicity and trendability than the current state-of-the-art semi-supervised approach. Moreover, our approach for identifying the Health Index leads to enhanced prognostic performance compared to the existing semi-supervised approach.