Accelerated life tests (ALTs) are widely used in industry to assist in product development. Acceleration models are often obtained from physical principles or past experience. But in some applications, particularly for investigating a new material or a new product, their acceleration models cannot be precisely specified. The uncertainty in model specification may cause serious problems in failure time prediction, and reduce the statistical efficiency of an optimal test plan. In this paper, the D and Ds optimal criteria are proposed for designing ALT plans that are good at selecting the best acceleration model among rival models. A generalized linear model (GLM) is developed for modeling ALT data with censoring. This approach simplifies the derivation of the information matrix of a test plan, and allows the experimenter to develop optimal ALT plans under the GLM framework. The proposed optimal design approach is compared with other conventional approaches through examples. The high design efficiency and design flexibility of the proposed approach are demonstrated in the paper.

Step-Stress Accelerated Life Testing (SSALT) is a special type of experiment that tests a product′s lifetime with time-varying stress levels. Typical testing protocols deployed in SSALTs cannot implement complete randomization of experiments; instead, they often result in grouped structures of experimental units and, thus, correlated observations. In this article, we propose a Generalized Linear Mixed Model (GLMM) approach to take into account the random group effect in SSALT. Failure times are assumed to be exponentially distributed under any stress level. Two parameter estimation methods, Adaptive Gaussian Quadrature (AGQ) and Integrated Nested Laplace Approximation (INLA), are introduced. A simulation study is conducted to compare the proposed random effect model with the traditional model, which pools data groups together, and with the fixed effect model. We also compare AGQ and INLA with different priors for parameter estimation. Results show that the proposed model can validate the existence of group-to-group variation. Lastly, the GLMM model is applied to a real data and it shows that disregarding experimental protocols in SSALT may result in large bias in the estimation of the effect of stress variable.

The lithium-ion (Li-ion) battery is a promising energy storage technology for electronics, automobiles, and smart grids. Extensive research was conducted in the past to improve the prediction of the remaining capacity of the Li-ion battery. A robust prediction model would improve the battery performance and reliability for forthcoming usage. In the development of a data-driven capacity prediction model of Li-ion batteries, most past studies employed capacity degradation data; however, very few tried using other performance monitoring variables, such as temperature, voltage, and current data, to estimate and predict the battery capacity. In this study, we aimed to develop a data-driven model for predicting the capacity of Li-ion batteries adopting functional principal component analysis (fPCA) applied to functional monitoring data of temperature, voltage, and current observations. The proposed method is demonstrated using the battery monitoring data available in the NASA Ames Prognostics Center of Excellence repository. The main contribution of the study the development of an empirical data-driven model to diagnose the state-of-health (SOH) of Li-ion batteries based on the health monitoring data utilizing fPCA and LASSO regression. The study obtained encouraging battery capacity prediction performance by explaining overall variation through eigenfunctions of available monitored discharge parameters of Li-ion batteries. The result of capacity prediction obtained a root mean square error (RMSE) of 0.009. The proposed data-driven approach performs well for predicting the capacity by employing functional performance measures over the life span of a Li-ion battery.

In a paper manufacturing system, it is substantially important to detect machine failure before it occurs and take necessary maintenance actions to prevent an unexpected breakdown of the system. Multiple sensor data collected from a machine provides useful information on the system's health condition. However, it is hard to predict the system condition ahead of time due to the lack of clear ominous signs for future failures, a rare occurrence of failure events, and a wide range of sensor signals which might be correlated with each other. We present two versions of feature extraction techniques based on the nearest neighbor combined with machine learning algorithms to detect a failure of the paper manufacturing machinery earlier than its occurrence from the multistream system monitoring data. First, for each sensor stream, the time series data is transformed into the binary form by extracting the class label of the nearest neighbor. We feed these transformed features into the decision tree classifier for the failure classification. Second, expanding the idea, the relative distance to the local nearest neighbor has been measured, results in the real‐valued feature, and the support vector machine is used as a classifier. Our proposed algorithms are applied to the dataset provided by Institute of Industrial and Systems Engineers 2019 data competition, and the results show better performance than other state‐of‐the‐art machine learning techniques.