Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/83248
Title: Model-based failure diagnostics and reliability prognostics for high power white light-emitting diodes lighting
Authors: Fan, Jiajie
Degree: Ph.D.
Issue Date: 2014
Abstract: Nowadays, with increasing concern on environmental protection, energy crisis, and quality of life, in the context of the lighting industry, Solid-state Lighting (SSL) is considered as the next generation green lighting source, following the conventional lighting sources (like incandescent bulbs and fluorescent lamps). As a type of SSL, High Power White Light-emitting diodes (HPWLEDs) lighting, which exhibits higher luminous efficacy, longer lifetime and environmental friendliness, has begun to be used in a broad range of applications, such as general lighting, TV backlighting, traffic signals and so on. However, the mass application of LED lighting in our daily lives still faces several difficulties in areas such as cost control, failure prediction and maintenance. This can be attributed to the limitations of traditional reliability assessment and prediction methods in the new electronics-rich systems. Hence, developing a fast, accurate and effective reliability testing and prediction method to determine the service life for LED lighting is becoming a key issue in popularizing this novel electronic device in the lighting market.
This research developed failure diagnosis and reliability prediction methods for HPWLEDs lighting within the Prognostics and Health Management (PHM) methodology framework. The originality of this work mainly relates to the establishment of failure criteria and data analysis methods for HPWLEDs lighting. Firstly, the Physics-of-Failure (PoF) based PHM approach is used to diagnose the failure modes and failure mechanisms for HPWLEDs lighting, from the chip level to the system level, and then to build damage models to estimate the reliability. Three failure modes: (i) Catastrophic failure; (ii) Lumen degradation; (iii) Chromaticity state shift are firstly categorized for HPWLEDs lighting systems. The data-driven prognostic methods with both statistical and learning approaches are then applied to predict the lumen lifetime, lumen maintenance and the chromaticity state of HPWLEDs lighting. The results show that the proposed Degradation Data Driven Method (DDDM) can predict more reliability messages (e.g. Mean Time To Failure, Confidence Interval and Reliability Function) compared to the IES TM-21-11 standard projecting method, which only can estimate L70. The recursive Unscented Kalman Filter (UKF) method, replacing the ordinary least squares (OLS) implementation recommended in the IES TM-21-11 standard, can improve the prediction accuracy for both lumen maintenance and chromaticity state applications. Finally, a new reliability estimation method is developed to predict the reliability of HPWLEDs lighting by integrating the proposed statistical and filter based data-driven PHM methods. With the proposed new reliability estimation method, the traditional reliability testing procedures can be optimized with a Six Sigma DMAIC step-wise framework by considering both time-and cost-reductions.
Subjects: Light emitting diodes.
Light emitting diodes -- Reliability.
Hong Kong Polytechnic University -- Dissertations
Pages: xvi, 222, [1] leaves : ill. (some col.) ; 30 cm.
Appears in Collections:Thesis

Show full item record

Page views

36
Last Week
0
Last month
Citations as of Mar 24, 2024

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.