Foreword to the Special Issue on “Data Science in Process Industries”

Process Industries have been an important part of Industrial Statistics for many years. industry data include real-time, high-dimensional measurements as well related to the quality finished products. Machine learning, artificial intelligence, and predictive modeling are increasingly will enrich statistical toolbox in industry. Future IoT Industry 4.0 need these methods develop further be successful. How take advantage them, without relegating a secondary role or even obliviation classical that were workforce addressing problems over years, is challenge concern most us on daily basis. This special issue presents contributions application variety science methodologies effectively address relevant faced by modern Industries. It was launched scope ENBIS 2021 Online Spring Meeting dedicated same theme. ENBIS, European Network Business Statistics, established more than 20 years ago since then has international platform exchange knowledge, scientific results, experiences field applied statistics analytics. The Covid-19 pandemic forced meeting held online. composed nine papers, covering different corners spectra methodologies, processes, problems, can appreciated following short descriptions each contribution. Bobbia et al. addressed problem fusing collected from sensors quality, construct air map. Predictions corrected through kriging residuals instead using pollutant concentration levels. Gaussian processes also used Fleming efficiently capture nonlinearities chemical Bayesian adaptive sequential design. Additionally, Lambardi investigated separable spatio-temporal environmental monitoring. Their approach involves spatial temporal model, which handled separately under composite likelihood approach. model prediction rule temperature office room. natural sequence engineering, inspection/statistical process monitoring/process capability analysis present this issue. Facchinetti designed acceptance sampling plans use generalized beta prior distribution. optimal plan obtained minimizing expected total cost quality. Shu presented quick-switch inspection scheme consideration multiple characteristics based overall capability. Furthermore, they provided web support automation lot deposition correspondingly ultra-mini chip resistors. Moving monitoring, Lepore analyzed multi-stream developed neural network combine outcomes streams into single quantity reflects status process. Borgoni capability, proposing methodology additive quantile regression with inference adopted framework order estimate local planar surfaces. They exemplified dry etching phase semiconductor fabrication. In diagnosis, Ersöz compared mining approaches identify root causes defective networks, classification trees furniture production. On methodological level, Kenett found gap current “automated” cross-validation proposed new one considers structure goals analysis. called befitting cross-validation. contrast conventional automated approaches, it requires assessment domain experts' inputs secure We hope rich mix spikes readership's curiosity read papers bring their thoughts, concerns solutions forthcoming events. looking forward hearing you!