The ENBIS‐23 Quality and Reliability Engineering International Special Issue

This year's Quality and Reliability Engineering International (QREI) special issue is related to the 23rd Annual Conference of the European Network for Business and Industrial Statistics (ENBIS, www.enbis.org). The conference took place in Valencia from September 10 to 14, 2023, and was hosted by the Department of Applied Statistics, Operation Research and Quality at the Technical University of Valencia (UPV). Our call for papers for the ENBIS-23 special issue attracted a lot of interest in the statistics community, ranging from researchers in academia up to practitioners in industry, which nicely reflects the mission of ENBIS: “connecting people in the field of business and industrial statistics”. Altogether, it was possible to collect 14 articles from a wide range of topics in the area of business and industrial statistics, covering design of experiments (DoEs), degradation and failure modeling, signals and time series, as well as sensitivity analysis (SA) of model outputs and machine learning interpretability. Concerning the DoE-area, Hamre and Tyssedal [1] introduce a new approach, called the decoupling method, for analyzing foldover designs. This method is shown to be less prone to error than others, making also possible to assess whether effects of an order greater than two should be included in the fitted model. Otava and Mylona [2] demonstrate how a split-plot design based on a Bayesian D $D$ -optimality criterion, built to ensure more precise pure-error estimation of the variance components, can be easily adapted to fit additional needs, such as a speedy implementation and good prediction properties of the fitted model. This technique is applied within a case study inspired from pharmaceutical synthesis. In the field of space filling DoEs, Grömping [3] studies the stratification patterns and dimension by weight tables to complement the established criteria for space filling and the General Stratum Orthogonal Arrays. This paper deep-dives the ideas behind the stratification pattern and the related ranking, which arise as a byproduct of the implementation in the R package SOAs. Regarding degradation and failure modeling, Akkus and Testik [4] implement a failure mode and effect analysis (FMEA) in an intuitionistic fuzzy environment, in order to reduce the subjectivity found in traditional FMEA. The authors propose a unique weighting methodology for the decision makers (accounting for their expertise and insights), and apply their approach to the production department of a defense company. The article of Boros et al. [5] is about deep-learning methods for estimating the Hurst parameter in fractional stochastic processes, where it reflects the memory decay, self-similarity, or fractal dimension of these models. The authors explore the reliability of Long Short-Term Memory (LSTM) networks and apply their methodology to Li-Ion battery degradation data. Lamalle et al. [6] propose a novel clustering technique tailored to randomly censored data in survival analysis. It is based on an underlying mixture model of Weibull distributions and consists in estimating its parameters by means of a variant of the Expectation-Maximization method in the presence of random censorship. In Han et al. [7], the reliability of repairable subsystems within a system is modeled under the competing and complementary risks framework, assuming that the lifetime of each subsystem type by each failure type follows non-identical exponential distribution. The developed inferential procedures are applied to the daily maintenance activity data of aircraft systems. Among the contributions to signals and time series, Obst et al. [8] suggest to include textual sources of information, such as daily weather reports, into the forecasting of electricity-demand and meteorological time series. For this purpose, they consider two different approaches of numerically encoding the (preprocessed) texts, one being based on the frequency of individual words, the other relying on a neural word embedding. If receiving an out-of-control (OC) signal in monitoring a multiple stream process, an important task is the post-hoc identification of the (group of) streams being responsible for this OC signal. Lepore et al. [9] propose to specifically train an artificial neural network for this purpose and illustrate their approach by a case study on heating, ventilation, and air conditioning (HVAC) systems in train coaches. Torres et al. [10] use the wavelet transform (WT) to decompose the vibration signal of spur gearboxes, in order to detect potential failures and to judge their severity. Their main objectives are to establish the appropriate level of decomposition, to identify the most suitable mother wavelets and filters, and to study the influence of the sensor's positioning and inclination on the feature extraction process. Lai et al. [11] investigate the impact of time series cross-validation on traditional time-series models and machine learning methods. This benchmark involves a case study predicting the incidence of COVID-19 based on wastewater data. Let us finally turn to the field of SA of model outputs and machine learning interpretability. The contribution of Fellmann et al. [12] is concerned with SA, which studies the impact of uncertain inputs onto the output of a model. More precisely, inspired by an application to two-dimensional pollutant concentration maps with various meteorological and traffic-related inputs, the authors consider the scenario of a model with set-valued output, and they propose and investigate new set-based SA indices. Otero et al. [13] provide a comprehensive and detailed examination of the influential variables, by using standard SA methods, on a model of panel block assembly in shipbuilding. Babaei and Guidici [14] present a framework and metrics that allow to identify fairness in machine learning models, using group-based fairness measures. The authors also propose a propensity matching approach that can improve fairness and apply it to a real use case that concerns credit lending. We would like to thank QREI for supporting ENBIS with the publication of this special issue. We are very grateful to the authors for their contributions to this special issue, to the anonymous peer-reviewers for carefully reading the submissions as well as for their constructive feedback, and to the Wiley editorial staff for their support during this work.