ABSTRACT This study is devoted to the estimation of the process capability index (PCI), , within the framework of the Fréchet distribution in the purview of modified version of a multiple interval type‐I censoring scheme. The estimation of is approached through three distinct methodologies: maximum likelihood estimation, maximum product spacing estimation, and a Bayesian framework. In the Bayesian paradigm, independent gamma priors are specified, and Bayes estimators of are derived under the squared error loss, general entropy loss and the LINEX loss functions. Inference is further augmented by constructing asymptotic confidence intervals and highest posterior density intervals. A comprehensive simulation study is undertaken to evaluate the finite‐sample performance of the proposed estimators in terms of mean squared error. Moreover, an optimal censoring scheme from a collection of potential schemes under various censoring scenarios has been investigated. The practical efficacy of the proposed methodologies is illustrated through the analysis of a real‐world data set.
ABSTRACT The integration of artificial intelligence into industrial quality control (QC) and management encompassing automated defect detection, real‐time process monitoring, and zero‐defect manufacturing has generated a rapidly expanding yet fragmented research landscape. This study addresses the need for comprehensive mapping of quality and reliability engineering practice through a computational systematic review of 9618 publications spanning the 2020–2025 period. A hybrid natural language processing (NLP) pipeline combining FlashText, Sentence‐BERT, KeyBERT, and spaCy performs semantic extraction of AI architectures, learning paradigms, defect typologies, and deployment contexts. The review pursues three objectives: mapping the architectural evolution of deep learning in industrial inspection, characterizing operational challenges of AI‐driven manufacturing, and identifying methodological frontiers for improving production reliability and supporting intelligent quality assurance (QA) systems. Results reveal that AI‐based quality control advances through coevolution of neural architectures, learning paradigms, and deployment requirements. CNNs remain central to discriminative visual inspection, while Transformers extend capabilities for multimodal reasoning. The field exhibits strong bias toward surface defects while overlooking safety‐critical internal defects. Emerging trends show transition from algorithm‐centric research toward deployment‐oriented paradigms including Uncertainty Quantification, Edge AI, Explainable AI, and Digital Twins. This work provides practitioners with an evidence‐based foundation for developing reliable, scalable AI quality systems aligned with Industry 4.0 principles.
ABSTRACT This systematic review examines a progress in Acceptance Sampling Plans (ASPs) research development from 2016 to 2026, highlighting the critical shift from static models toward dynamic, intelligent methodologies that enhance efficiency and decision‐making under uncertainty. Following Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines, we analyzed peer‐reviewed articles from Scopus and Web of Science, employing thematic analysis to categorize literature based on methodological evolution, optimization techniques, statistical advancements, and non‐classical uncertainty management. The review identifies several major developmental trends. Adaptive sampling schemes such as repetitive and group ASPs reduce required sample sizes by optimizing decision rules. Accelerated Life Testing (ALT) integrated with step‐stress modeling, ranked set sampling, and hybrid censoring schemes cuts test times for highly reliable products. Complex failure‐pattern modeling has evolved beyond traditional distributions to incorporate flexible semi‐parametric classes, lifetime percentiles, and process‐capability loss indices. Uncertainty frameworks, including Neutrosophic statistical interval methods and Bayesian‐Neutrosophic models, improve decision robustness under indeterminate data conditions. In general, this synthesis provides practitioners with guidance to reduce inspection time and costs, supporting adaptive quality control in modern manufacturing environments.
ABSTRACT The rapid proliferation of Internet of Things (IoT) devices has created new security risks, particularly botnet infections. This paper proposes an unsupervised monitoring framework that combines a ‐variational autoencoder (VAE) with multivariate statistical process control (SPC) for botnet detection in IoT networks. Categorical fields of network flows are represented with per‐field embeddings, trained jointly with the VAE and concatenated with the scaled numeric features. The VAE is trained only on benign traffic; the resulting latent code and reconstruction residuals are monitored with Hotelling's and squared prediction errror (SPE) control charts, with control limits calibrated by a moving block bootstrap that preserves local temporal dependence of network flows during Phase I calibration. A per‐flow EWMA variant of both charts is also evaluated as an alternative to fixed‐length monitoring windows. On 12 IoT‐23 botnet families, the proposed charts reach pooled AUCs of 0.9906 () and 0.9946 (SPE) against held‐out benign traffic, with 100% flow‐level recall in every family. However, at the bootstrap‐calibrated operating point, the held‐out benign false‐positive rates are 75.3% and 75.1%, respectively, revealing a substantial distribution shift between the Phase I reference and the subsequent benign traffic. The framework is benchmarked against seven unsupervised baselines and four supervised classifiers. A component‐wise ablation shows that the learned embedding representation is a major contributor to AUC, whereas a deterministic autoencoder using the same embeddings achieves comparable AUC; a separate threshold‐sensitivity analysis evaluates the effect of the operating threshold. Quantitative latent‐space separability measures (silhouette and Davies–Bouldin) identify the botnet families that remain hardest to separate from benign traffic.
ABSTRACT Accurate remaining useful life prediction is critical to the reliable operation of operational amplifiers in modern electronic systems. However, existing methods face two key challenges: predictions typically lack statistical guarantees on uncertainty quantification (UQ), and randomly missing monitoring data caused by sensor or transmission failures further degrade model performance. To address these issues, this paper proposes a data augmentation‐driven RUL prediction and conformal UQ method, termed DiffUQ. First, an Electronic Component Degradation Pattern Encoder (ECDPE) is constructed by integrating contrastive learning into a variational autoencoder framework to extract discriminative degradation representations from multivariate monitoring data. Building on these representations, a diffusion‐based data generation framework (DiffGen) equipped with a specially designed ST‐Denoiser is developed to impute and augment data with missing values while capturing the underlying spatio‐temporal correlations. Finally, deep metric learning is combined with enhanced conformal prediction to construct prediction intervals that retain a coverage guarantee under departures from exchangeability, with the loss of coverage bounded explicitly. Experiments on the accelerated degradation dataset of SX8042MF operational amplifiers demonstrate that DiffUQ achieves superior point prediction accuracy and uncertainty quantification compared to several state‐of‐the‐art baselines under various missing‐data ratios. Moreover, DiffGen consistently improves multiple underlying prediction models and outperforms widely used data augmentation methods such as CGAN, WGAN‐GP, VAE, and TimeVAE.
