Assessing multi-output Gaussian process regression for modeling of non-monotonic degradation trends of light emitting diodes in storage

Lim, Sze Li Harry; Duong, Pham Luu Trung; Park, H.; Singh, Preetpal; Tan, Cher-Ming; Raghavan, Nagarajan

doi:10.1016/j.microrel.2020.113794

Cited by 5 publications

(4 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Based on our earlier works (Ref. [24] & [25]), we observed a strong negative correlation between the detrending bias and the prediction MAPE. To understand the effects of this bias, various approaches to detrend the test data will be considered in our study here, as summarized in Table 1.…”

Section: B Detrending Mean and Biassupporting

confidence: 61%

“…To minimize this bias, there is a need to explore approaches to detrend the test data, possibly with the mean of the incremental test data observed up to current time of test. A better prediction with lower mean absolute percentage error (MAPE) was obtained by trying this approach in our work [25]. However, this approach was unstable when the number of test data points is low.…”

Section: Introductionmentioning

confidence: 83%

“…The "OFF" condition is akin to LEDs placed under long-term storage, hence, is applicable for our study. In our previous work [25], we have explored some basic flavors of MOGPR to predict remaining storage life for these LEDs and the same data sets have been employed here for this study. Luminous Flux (lumen)…”

Section: Led Storage Data Setmentioning

confidence: 99%

“…Subsequently, we also explored possibility of having bounded training sets with an equally weighted mean of both training data sets to detrend the test data. While this further lowered the bias and MAPE, resulting in improved RSL [25], it greatly constrains the selection of the training sets to be the upper and lower bound of all collected data (which most often may not be the case in practical circumstances) and enforces stringency on the test data sets which have to strictly stay within the bound of the training data sets in order to achieve a lower MAPE. This paper investigates into the effects of bias in the detrending mean on the MAPE (or RSL prediction) while performing MOGPR and proposes more versatile approaches to reduce MAPE in the degradation profile.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Bias Suppression Framework for Detrending Mean of Multi-Output Gaussian Process Regression in LED Remaining Storage Life Prognosis

et al. 2021

View full text Add to dashboard Cite

Advancements in storage prognosis tend to be limited by the inherent challenge to collect sufficient significant degradation data over an extensive period. Using only sparse fleet data, multi-output Gaussian process regression (MOGPR) is one of the few techniques that offers a practical data driven approach to model non-monotonic degradation profiles with low mean absolute percentage error (MAPE). This accuracy in the storage prognosis context is, however, sensitive to the choice of the detrending mean. Working with light emitting diodes (LED) sparse lumen degradation data under storage conditions in this study, the MAPE is observed to be highly correlated to the detrending bias -the difference between the detrending mean and the test mean. We explore various approaches to suppress this bias and advocate a generic framework for fleet storage prognosis. The approaches include detrending using (A) static training data mean, (B) dynamic observed test data mean, (C) static bounded training data set pairs, (D) dynamic weighted mean of unbounded training data set pairs and (E) moving average of weighted mean of unbounded training data set pairs. Our analysis shows that the moving average approach (Method E) of computing weighted mean of unbounded training data set pairs results in the most stable detrending mean to suppress detrending bias and helps achieve an MAPE lower than 1%.INDEX TERMS Storage prognosis, Light emitting diodes (LED), Multi-output Gaussian process regression (MOGPR), Detrending Mean, Detrending Bias, Moving Average (MA).

show abstract

Section: B Detrending Mean and Biassupporting

confidence: 61%

Section: Introductionmentioning

confidence: 83%

Section: Led Storage Data Setmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations