TY - JOUR
T1 - Reliability Analysis of Cyclic Accelerated Life Test Data Using Log-Location-Scale Family of Distributions Under Censoring With Application to Solder Joint Data
AU - Zhang, Wenhan
AU - Zhu, Xiaojun
AU - He, Mu
AU - Balakrishnan, Narayanaswamy
N1 - Publisher Copyright:
© 1963-2012 IEEE.
PY - 2024
Y1 - 2024
N2 - Accelerated life testing is widely employed due to the high cost involved in testing high-quality products under normal operating conditions. For products exposed to continuously fluctuating stress in the working environment, cyclic stress tests become necessary. The Coffin-Manson model is commonly used when product failure is solely attributed to temperature changes (Δ T). However, this assumption does not always hold in many practical situations. The Norris-Landzberg model, which considers both maximum temperature and cyclic change frequency, offers much flexibility in modeling fatigue life due to cyclic temperature fluctuations. Several studies have been conducted based on the Norris-Landzberg model. However, using the multiple linear regression method without any distributional assumption may fail to provide satisfactory inferential results. This article assumes the log-location-scale family of distributions and then shows that the weighted least-squares method based on order statistics of failure times yields the best linear unbiased estimators (BLUEs) of parameters based on complete as well as Type-II censored data. We then study some properties of these BLUEs using both theory and Monte Carlo simulations. Next, we present an illustrative example involving solder joint data to demonstrate the model and the associate inferential results developed here. Finally, the optimal design procedure is discussed.
AB - Accelerated life testing is widely employed due to the high cost involved in testing high-quality products under normal operating conditions. For products exposed to continuously fluctuating stress in the working environment, cyclic stress tests become necessary. The Coffin-Manson model is commonly used when product failure is solely attributed to temperature changes (Δ T). However, this assumption does not always hold in many practical situations. The Norris-Landzberg model, which considers both maximum temperature and cyclic change frequency, offers much flexibility in modeling fatigue life due to cyclic temperature fluctuations. Several studies have been conducted based on the Norris-Landzberg model. However, using the multiple linear regression method without any distributional assumption may fail to provide satisfactory inferential results. This article assumes the log-location-scale family of distributions and then shows that the weighted least-squares method based on order statistics of failure times yields the best linear unbiased estimators (BLUEs) of parameters based on complete as well as Type-II censored data. We then study some properties of these BLUEs using both theory and Monte Carlo simulations. Next, we present an illustrative example involving solder joint data to demonstrate the model and the associate inferential results developed here. Finally, the optimal design procedure is discussed.
KW - Accelerated life-test
KW - best linear unbiased estimator
KW - cyclic tests
KW - log-location-scale family
KW - Norris-Landzberg model
KW - test of solder joints data
KW - type-II censoring
UR - http://www.scopus.com/inward/record.url?scp=85212950709&partnerID=8YFLogxK
U2 - 10.1109/TR.2024.3509446
DO - 10.1109/TR.2024.3509446
M3 - Article
AN - SCOPUS:85212950709
SN - 0018-9529
JO - IEEE Transactions on Reliability
JF - IEEE Transactions on Reliability
ER -