Abstract This study presents a Time to Failure (TTF) aligned methodological framework for analyzing sensor degradation and predicting Remaining Useful Life (RUL) in Industrial Internet of Things (IIoT) systems. Addressing critical challenges of sensor drift and measurement uncertainty,... This study presents a Time to Failure (TTF) aligned methodological framework for analyzing sensor degradation and predicting Remaining Useful Life (RUL) in Industrial Internet of Things (IIoT) systems. Addressing critical challenges of sensor drift and measurement uncertainty, the proposed architecture employs a seven layer pipeline to transform noisy, high frequency telemetry into actionable health indicators. A key innovation is the shift from absolute cycle alignment to TTF aligned trajectory construction, which mitigates survivorship bias and reveals consistent degradation signatures that are otherwise obscured during early operational life. Empirical validation on smart manufacturing datasets (FD001 subset) demonstrates the framework's efficacy: statistical metrics including effect size and Pearson correlation identified sensor_11 and sensor_4 as primary degradation indicators, exhibiting monotonic trends and low inter engine variability. Predictive modeling via Random Forest regression confirmed that TTF aligned features significantly enhance failure prediction accuracy compared to raw sensor inputs. Furthermore, phase aware degradation analysis using change point detection enables robust health state segmentation by distinguishing healthy and transition phases. Static sensors were systematically excluded to reduce computational overhead while preserving prognostic relevance. The framework bridges raw sensor data and reliable decision support, enabling proactive maintenance scheduling that minimizes unplanned downtime and optimizes operational safety in Industry 4.0 environments. Scalability is ensured through containerized inference services, while encryption protocols protect sensitive industrial data. Future work should explore generative models for data recovery and expand validation across heterogeneous fleets to enhance resilience in complex industrial networks, ultimately supporting sustained production quality through data driven prognostics. Read More Read Less

Abstract This study demonstrates how unsupervised machine learning techniques can audit and optimize technical knowledge repositories through systematic semantic analysis of a curated Python FAQ corpus comprising 163 question answer pairs. Addressing the challenge that over 80% of... This study demonstrates how unsupervised machine learning techniques can audit and optimize technical knowledge repositories through systematic semantic analysis of a curated Python FAQ corpus comprising 163 question answer pairs. Addressing the challenge that over 80% of organizational data exists in unstructured text form, we implement a transparent, eight stage natural language processing pipeline encompassing preprocessing, TF-IDF feature extraction, cosine similarity analysis, Latent Dirichlet Allocation (LDA) topic modeling, Principal Component Analysis (PCA) dimensionality reduction, and network-based co-occurrence analysis. Our methodology emphasizes contextualized preprocessing decisions, responding to contemporary gaps in methodological transparency within organizational text mining research. Results reveal five coherent thematic clusters corresponding to core Python programming concepts: language fundamentals, data structures, syntax and control flow, function operations, and advanced mechanisms. Cosine similarity analysis identifies non trivial semantic overlap among FAQ entries, highlighting actionable opportunities for content consolidation to reduce redundancy and improve retrieval efficiency. Network analysis establishes "python," "function," "list," and "dictionary" as high centrality conceptual anchors within the corpus topology. These findings translate into practical recommendations for technical documentation management, including evidence based FAQ merging, hierarchically organized navigation schemas, and gap identification for underrepresented subject areas. By bridging unstructured text data with structured organizational intelligence, this reproducible framework supports educational chatbot development, curriculum design, and knowledge base optimization. The study underscores that rigorous, contextually justified preprocessing combined with multi perspective unsupervised analytics enables researchers and practitioners to unlock significant value from complex text corpora while ensuring methodological transparency and analytical validity in organizational research applications. Read More Read Less

Abstract Anomaly detection (AD) in textual data remains challenging due to semantic complexity and the scarcity of labeled anomalous examples. This study proposes a comparative unsupervised framework to identify irregularities within mixed code text corpora, specifically addressing the... Anomaly detection (AD) in textual data remains challenging due to semantic complexity and the scarcity of labeled anomalous examples. This study proposes a comparative unsupervised framework to identify irregularities within mixed code text corpora, specifically addressing the heterogeneity of technical data from developer forums. We evaluate two distinct paradigms Isolation Forest (density-based isolation) and PCA Autoencoder (reconstruction based error) on a stratified subset of 10,000 Stack Overflow samples, derived from 52,270 total entries and represented via 500 dimensional TF-IDF features. Results indicate that while both methods identified 500 anomalies (5% contamination rate), they exhibited markedly different detection patterns. Isolation Forest showed a strong bias toward natural language, flagging 9.54% of text samples versus 0.08% of code samples. Conversely, the PCA Autoencoder detected a more balanced distribution, identifying 3.15% of code and 6.71% of text samples as anomalous. Agreement between methods was low (~3%), with only 15 consensus anomalies, suggesting complementary notions of irregularity rather than redundant signals. Isolation Forest captured global outliers, such as verbose questions, while the autoencoder detected local irregularities, such as unusual code constructs. These findings underscore the value of multiperspective anomaly detection strategies for comprehensive coverage of irregular patterns. Ultimately, systematic anomaly analysis supports dataset cleaning, model robustness evaluation, and improved preprocessing for downstream tasks, including code understanding and automated question answering systems in software engineering applications Read More Read Less