An official website of the United States government
Remaining Useful Life (RUL) estimation is critical in many engineering systems where proper predictive maintenance is needed to increase a unit's effectiveness and reduce time and cost of repairing. Typically for such systems, multiple sensors are normally used to monitor performance, which create difficulties for system state identification. In this paper, we develop a semi-supervised left-to-right constrained Hidden Markov Model (HMM) model, which is effective in estimating the RUL, while capturing the jumps among states in condition dynamics. In addition, based on the HMM model learned from multiple sensors, we build a Partial Observable Markov Decision Process (POMDP) to demonstrate how such RUL estimation can be effectively used for optimal preventative maintenance decision making. We apply this technique to the NASA Engine degradation data and demonstrate the effectiveness of the proposed method.
more »
« less
Applying Hidden Markov Processes to Optimizing Power Systems Maintenance
Failure identification and prediction in a power system are essential components that are prerequisites for optimizing the maintenance of the system. The incidences of power system failures have increased dramatically in recent times due to the uncertainties inherent in the advent of both man-made and natural disasters. This problem is further exacerbated due to the increasing demand for higher operational efficiency in power systems. Currently, there is a paucity of studies that predict and identify failure in a distribution power system. In this paper, we propose an integrated methodology for selecting the optimal maintenance plan based on predicting and identifying failure modes with the aid of Hidden Markov Models (HMM) and a probabilistic decision-making tool. While the model parameters of previous studies were determined utilizing observable prior knowledge, the use of HMM offers a different approach especially in the absence of such observable prior distributions. Thus, we determine the status of health of a power system by using an HMM to capture the relationship between unobservable degradation state and observed parameters. The preliminary outcome is instructive for the management of power systems especially in response to fortifying the system against aging and degradation.
more »
« less
- Award ID(s):
- 1847077
- PAR ID:
- 10398912
- Editor(s):
- Ellis, K; Ferrell, W.; Knapp, J.
- Date Published:
- Journal Name:
- Proceedings of the IISE Annual Conference & Expo 2022
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Volcanic eruptions pose a significant and sometimes unpredictable hazard, especially at systems that display little to no precursory signals. For example, the 2008 eruption of Okmok volcano in Alaska notably lacked observable short‐term precursors despite years of low‐level unrest. This unpredictability highlights that direct monitoring alone is not always enough to reliably forecast eruptions. In this study, we use the Ensemble Kalman Filter (EnKF) to produce a successful hindcast of the Okmok magma system in the lead up to its 2008 eruption. By assimilating geodetic observations of ground deformation, finite element models track the evolving stress state of the magma system and evaluate its stability using mechanical failure criteria. The hindcast successfully indicates an increased eruption likelihood due to tensile failure weeks in advance of the 2008 eruption. The effectiveness of this hindcast illustrates that EnKF‐based forecasting methods may provide critical information on eruption probability in systems lacking obvious precursors.more » « less
-
Abstract This paper proposes a novel adaptive maintenance policy for degrading systems subject to hard failure. Compared with traditional condition‐based maintenance policies, the proposed predictive maintenance policy makes maintenance decisions adaptively based on model prognostic results. The prognostic model is continuously updated based on newly inspected data. The inspection times and preventive maintenance activities are scheduled online in a sequential manner based on the most current prediction of system reliability. A computationally efficient optimization scheme is proposed for obtaining optimal maintenance parameters. The proposed policy is demonstrated and its performance is evaluated through extensive simulations.more » « less
-
Abstract Forecasting the onset of a volcanic eruption from a closed system requires understanding its stress state and failure potential, which can be investigated through numerical modeling. However, the lack of constraints on model parameters, especially rheology, may substantially impair the accuracy of failure forecasts. Therefore, it is essential to know whether large variations and uncertainties in rock properties will preclude the ability of models to predict reservoir failure. A series of two‐dimensional, axisymmetric models are used to investigate sensitivities of brittle failure initiation to assumed rock properties. The numerical experiments indicate that the deformation and overpressure at failure onset simulated by elastic models will be much lower than the viscoelastic models, when the timescale of pressurization exceeds the viscoelastic relaxation time of the host rock. Poisson's ratio and internal friction angle have much less effect on failure forecasts than Young's modulus. Variations in Young's modulus significantly affect the prediction of surface deformation before failure onset when Young's modulus is < 40 GPa. Longer precursory volcano‐tectonic events may occur in weak host rock (E< 40 GPa) due to well‐developed Coulomb failure prior to dike propagation. Thus, combining surface deformation with seismicity may enhance the accuracy of eruption forecast in these situations. Compared to large and oblate magma systems, small and prolate systems create far less surface uplift prior to failure initiation, suggesting that more frequent measurements are necessary.more » « less
-
Achieving accurate remaining useful life (RUL) prediction for prognostic and health management (PHM) depends upon sufficient prior degradation apprehension of critical components within the system. However, such prior knowledge is not always readily available in practice. We alleviate this shortcoming by proposing a novel data-driven framework that is capable of providing accurate RUL prediction without the need for any prior failure threshold knowledge. Correlative and monotonic metrics are utilized to identify critical features throughout the degradation progress. Subsequently, we append one-hot health state indicators to extracted degrading features, which are utilized together as adversarial training data for a Long Short-Term Memory (LSTM) network-based model. Finally, we utilize a fully connected layer to project the LSTM outputs into the parameters of a Gaussian mixture model (GMM) in conjunction with a categorical distribution, from which the long-term degradation progress is sampled. We verify the performance of the proposed framework using aeroengine health data simulated by Modular Aero-Propulsion System Simulation (MAPSS), and the results demonstrate that significant performance improvement can be achieved for long-term degradation progress and RUL prediction tasks.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
