Using Degradation Measures to Estimate a Time-to-Failure Distribution

Lu, C. Joseph; Meeker, William O.

doi:10.1080/00401706.1993.10485038

Cited by 731 publications

(331 citation statements)

References 37 publications

Supporting

Mentioning

294

Contrasting

Order By: Relevance

“…To overcome the shortcomings of these two methods, researchers have proposed some new methods that combine probability statistics with performance degradation, which mainly include an approximate method, an analysis method or a two-stage method. [10] Lu et al [11] established a degradation model as a function of time with multidimensional random variables. Fan et al [12] used a degradation data-driven method (DDDM), which was based on the general degradation path model, to predict the reliability of high-power white light LED (HPWLED) by analysing lumen maintenance data.…”

Section: Introductionmentioning

confidence: 99%

Life prediction for a vacuum fluorescent display based on two improved models using the three‐parameter Weibull right approximation method

Zhang

Zong

et al. 2017

Luminescence

View full text Add to dashboard Cite

To obtain precise life information for vacuum fluorescent displays (VFDs), luminance degradation data for VFDs were collected from a group of normal life tests. Instead of exponential function, the three-parameter Weibull right approximation method (TPWRAM) was applied to describe the luminance degradation path of optoelectronic products, and two improved models were established. One of these models calculated the average life by fitting average luminance degradation data, and the other model obtained VFD life by combining the approximation method with luminance degradation test data from each individual sample. The results indicated that the test design under normal working stress was appropriate, and the selection of censored test data was simple. The two models improved by TPWRAM both revealed the luminance decaying law for VFD, and the pseudo failure time was accurately extrapolated. It was further confirmed by comparing relative error that using the second model gave a more accurate prediction of VFD life.The improved models in this study can provide technical references for researchers and manufacturers in aspects of life prediction methodology for its development. Due to the very long life of a VFD, the reduction in life test time and the improvement in test efficiency have become some of the important and urgent issues faced by business. As a result, accelerated life tests for VFD have been proposed. [3][4][5] Accelerated life tests of these products are conducted under abnormal operating conditions, so their rationality and their accuracy need to be verified by a normal life test. The large amounts of information contained in normal life tests not only reveal the characteristics of VFD luminance degradation, but can also give real-life information. What is more, these test results can be compared with those estimated from accelerated life tests to verify the rationality of the accelerated failure mechanism and the accuracy of life prediction models. As a complete normal stress test time is lengthy, it is necessary to establish a life prediction model that can calculate life using partially finished test data under normal working stress.Many researchers have predicted the life of optoelectronic products using a performance degradation method, and most of these investigators chose the exponential decay model. [6,7] Other researchers have used probability statistics methods to evaluate the life of the product. [8,9] Although these two methods have been widely used and studied, they have some deficiencies. The first method can only calculate the average life of the product, but can not obtain other life information, such as median life, reliability function and confidence interval. The reliability of the second method depends on the number of failure times, but test time and cost are a great challenge due to Abbreviations used: AIC, Akaike Information Criterion; DDDM, degradationdata-driven method; LSM, least square method; MLE, maximum likelihood estimation; TPWRAM, three-parameter Weibull right ap...

show abstract

Section: Introductionmentioning

confidence: 99%

Life prediction for a vacuum fluorescent display based on two improved models using the three‐parameter Weibull right approximation method

Zhang

Zong

et al. 2017

Luminescence

View full text Add to dashboard Cite

show abstract

“…We will not use this degradation model to feed into CBM policies later. The original application of random coefficient models is to estimate the time-to-failure distribution based on degradation data; see Lu and Meeker (1993). We provide an example of this below.…”

Section: Random Coefficient Modelsmentioning

confidence: 99%

Maintenance, Modeling and Optimization

Ben‐Daya¹,

Duffuaa²,

Raouf³

2000

View full text Add to dashboard Cite

“…In this case, we model the degradation process X(t) per maintenance cycle by the following two approaches: (i) the Random coefficient model (cf. [19]), because it is relatively flexible and convenient for describing the degradation paths derived from the physics of failures; (ii) the Gamma process (cf. [31]), which is a process that is often used in the literature.…”

Section: Case Studymentioning

confidence: 99%

“…Wang [33] proposed a CBM model based on the general random coefficient model (cf. Lu and Meeker [19]) to determine the optimal control limit and the monitored interval in terms of cost, downtime and reliability. Gebraeel et al [9,10] extended the general degradation model to estimate the RUL distribution from sensor signals by a Wiener process and Bayesian updating.…”

Section: Introductionmentioning

confidence: 99%