1 January 2022 in Numerical Algebra, Control & Optimization
In this paper, we survey the primary research on the theory and applications of distributionally robust optimization (DRO). We start with reviewing the modeling power and computational attractiveness of DRO approaches, induced by the ambiguity sets structure and tractable robust counterpart reformulations. Next, we summarize the efficient solution methods, out-of-sample performance guarantee, and convergence analysis. Then, we illustrate some applications of DRO in machine learning and operations research, and finally, we discuss the future research directions.
8 March 2022 in IEEE Transactions on Industrial Informatics
Infrared thermography provides abundant spatiotemporal degradation information, facilitating non-contact condition monitoring. Reducing domain shift between simulated and industrial infrared images is significantly desired for leveraging labeled simulated data to tackle practical insufficiency of run-to-failure samples. Recently, adversarial-based domain adaptation (DA) techniques have aroused broad concern in solving domain shifts. However, simultaneously aligning marginal and conditional distributions in cross-domain remaining useful life (RUL) prediction is rarely researched in adversarial-based DA. In this article, an adversarial regressive domain adaptation (ARDA) approach is, thus, put forward to address this challenge. First, a regressive disparity discrepancy is designed to describe the dissimilarity between distributions and derive the generalization bound for cross-domain prognostics. Guided by this bound, the ARDA effectively aligns marginal and conditional distributions by learning indistinguishable features and considering the relationship between samples and prediction tasks. Simulated and experimental infrared degradation image datasets are used to demonstrate the effectiveness and superiority of the proposed approach over existing methods for cross-domain RUL prediction.
2 April 2019 in Journal of Manufacturing Science and Engineering
In mass customization, a manufacturing line is required to be kept in reliable operation to handle product demand volatility and potential machine degradations. Recent advances in data acquisition and processing allow for effective maintenance scheduling. This paper presents a systematic framework that integrates a sensor-driven prognostic method and an opportunistic maintenance policy. The prognostic method uses degradation signals of each individual machine to predict and update its time-to-failure (TTF) distributions in real time. Then, system-level opportunistic maintenance optimizations are dynamically made according to real-time TTF distributions and variable product orders. The online analytics framework is demonstrated through the case study based on the collected reliability information from a production line of engine crankshaft. The results can effectively prove that the real-time degradation updating and the opportunistic maintenance scheduling can efficiently reduce maintenance cost, avoid system breakdown, and ensure product quality. Furthermore, this framework can be applied not only in an automobile line but also for a broader range of manufacturing lines in mass customization.
8 March 2019 in Technometrics
This paper proposes a new methodology to predict and update the residual useful lifetime of a system using a sequence of degradation images. The methodology integrates tensor linear algebra with traditional location-scale regression widely used in reliability and prognostics. To address the high dimensionality challenge, the degradation image streams are first projected to a low-dimensional tensor subspace that is able to preserve their information. Next, the projected image tensors are regressed against time-to-failure via penalized location-scale tensor regression. The coefficient tensor is then decomposed using CANDECOMP/PARAFAC (CP) and Tucker decompositions, which enables parameter estimation in a high-dimensional setting. Two optimization algorithms with a global convergence property are developed for model estimation. The effectiveness of our models is validated using two simulated datasets and infrared degradation image streams from a rotating machinery.
26 March 2021 in Journal of Manufacturing Science and Engineering
With the development of the product-service paradigm, many original equipment manufacturers (OEMs) have begun to provide remanufacturing services besides maintenance services as the lessors. An integrated contract including remanufacturing and maintenance service brings extra leasing profits and complex scheduling challenges as well. The use of machine remanufacturing in conjunction with system-level opportunistic maintenance (OM) has not been previously studied. This paper tries to integrate remanufacturing planning into maintenance optimizations for improved system performance in terms of saved leasing profits. An integrated remanufacturing and opportunistic maintenance (R&OM) policy is developed and applied to a leased manufacturing line in batch production. The system-level saved profit achieved via the R&OM policy is estimated and maximized through a sequential decision-making programming. Performing this integrated policy, it is possible to utilize the sequential R&OM opportunities raised from the random batch orders. In addition, preventive maintenance (PM) and preventive remanufacturing (PR) can be dynamically optimized to form profit-effective schemes. Numerical examples based on the collected information from a leased batch production line of engine crankshaft demonstrate that this proposed R&OM policy could efficiently achieve saved leasing profit maximization, reduce joint decision complexity, and expand OM theory.
13 August 2020 in IISE Transactions
Multi-stream degradation signals have been widely used to predict the residual useful lifetime of partially degraded systems. To achieve this goal, most of the existing prognostics models assume that degradation signals are complete, i.e., they are observed continuously and frequently at regular time grids. In reality, however, degradation signals are often (highly) incomplete, i.e., containing missing and corrupt observations. Such signal incompleteness poses a significant challenge for the parameter estimation of prognostics models. To address this challenge, this paper proposes a prognostics methodology that is capable of using highly incomplete multi-stream degradation signals to predict the residual useful lifetime of partially degraded systems. The method first employs multivariate functional principal components analysis to fuse multi-stream signals. Next, the fused features are regressed against times-to-failure using (log)-location-scale regression. To estimate the fused features using incomplete multi-stream degradation signals, we develop two computationally efficient algorithms: subspace detection and signal recovery. The performance of the proposed prognostics methodology is evaluated using simulated datasets and a degradation dataset of aircraft turbofan engines from NASA repository.