Mills College

The rapid adoption of renewable energy sources has increased the complexity of infrastructure maintenance, making early fault detection and predictive analytics critical for sustainable operations. This study presents a novel spatiotemporal Graph Neural Network (GNN) framework that captures both topological and temporal dependencies in wind energy systems. Unlike conventional models, it integrates graph attention mechanisms with gated recurrent units for enhanced fault detection, Remaining Useful Life (RUL) prediction, and anomaly identification using multi-task learning. The methodology is validated using the CARE dataset comprising 89 years of SCADA data from 36 turbines, achieving 94.2% fault detection accuracy, a <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{6.4\%}$</tex> MAPE for RUL prediction, and up to 8.4 weeks of early warning. The model demonstrates strong crossdomain generalization and scalability across diverse wind farm conditions. This framework enables more proactive, cost-efficient maintenance and contributes to the reliability and resilience of clean energy infrastructure, aligning with global goals for sustainable and uninterrupted power supply.

This article explores how White, award-winning history teachers have been differently affected by legislative efforts to restrict teaching about race and racism, and teachers’ various resistances in response. Findings describe how teachers’ feelings of constraint (or lack thereof) around race-restrictive policies, rhetoric and legislation were often locally determined, resulting from actions by parents, local policymakers, school administrators and other community-based actors. The article then reports on a subset of teachers’ various responses to restrictions, illuminating how teachers exerted agency and tactical manoeuvring in order to continue to teach about race and racism in contradiction to official policy and unofficial pressure. Findings showed how teachers in a variety of sociopolitical contexts continued teaching about race and racism in history despite the risks of official or unofficial retribution. They did so through wielding explicit denial, through selective silences and by directing student attention to topic suppression in ways that conveyed meanings about race and racism even when remaining officially silent.

Cristian Moyano-Fernández, 'Rewilding and Ecological Justice: The Ethics and Politics of Wildlife Regeneration'. Un article de la revue Philosophy in Review (Volume 45, numéro 4, novembre 2025, p. 1-38) diffusée par la plateforme Érudit.

Laboratory safety is an important issue in academic settings. To address the challenge of effectively teaching lab safety, we developed an educational card game titled Stocked. This game incorporates gamification elements to enhance student engagement and familiarity with common laboratory safety items and equipment. This paper outlines the game’s development and preliminary evaluation. A survey was conducted to gain feedback from the players, and a SWOT analysis was performed. Players found Stocked to be a fun, fast-paced, and engaging game.

Ushbu maqolada mangʻitlar sulolasi hukmronligi davridagi Buxoro amirligining XVIII asr ikkinchi yarmi – XIX asrning oʻrtalaridagi amirlikning soliq tizimi, dehqonlardan olinadigan soliqlar miqdori va ularning taqsimlanishi tadqiq qilingan. Tadqiqot ishida ma’lum davrda yozilgan manba va adabiyotlardan foydalanilganligi koʻrsatilgan. Soliqlarning dehqonlar yashash tarziga salbiy oqibatlari ham koʻrsatib oʻtilgan.

This study investigates the spatial and temporal heterogeneity of soil moisture on slopes of China’s northeastern black soil region, focusing on the effects of terrain adjustment and vegetation. Soil moisture dynamics in the 0–60 cm soil layer were measured at 10 cm intervals using the TRIME-PICO64 TDR® device on slopes with similar gradients representing three land use types: transverse ridge tillage (TRT) farmland, terraced fields (TFs) farmland, and pure forest woodland (WL). The results indicate significant variations in soil moisture content and water storage across different land use types in the order of TF &gt; TRT &gt; WL. The study further identified that soil bulk density, porosity, and water-holding indicators were in the order of WL &gt; TF &gt; TRT, inconsistent with the soil moisture results, indicating that soil quality cannot be the sole reason for the differences in moisture. The moisture differences between farmland types (TRT and TF) and WL are substantial, especially during the rainy season. In the rainy season (0–60 cm) and the dry season (30–60 cm), significant differences in moisture content are observed (p &lt; 0.05). Significant differences in moisture content between farmland types are found at 0–40 cm during the rainy season and at 0–10 cm during the dry season. In the rainy season, soil moisture for TRT and TFs first decreases from 26.76% and 30.85% to 22.44% and 25.38%, then slightly increases to 27.01% and 27.07% along the slope. Meanwhile, WL displays the opposite pattern on upper, relatively steep slopes, with soil moisture increasing from 16.66% to 17.81%, and exhibits a pattern of change similar to TRT and TFs on lower, gentler slopes. TFs consistently show higher soil moisture and water storage at all slope positions than TRT and WL. TFs improve soil quality, reduce erosion and sedimentation, and shift the lowest soil moisture content to a lower slope position. During the dry season, soil moisture differences between slope positions for TRT and WL were small. In general, terracing can effectively modulate moisture distribution along slopes, increasing moisture by an average of 0.26~12.43%, while afforestation, despite improving soil quality, leads to an 18.14~31.13% reduction in soil moisture content, with the impact being particularly significant during the rainy season. These findings provide important insights for optimizing land use and ecological construction, including keeping the balance between soil and water conservation, especially for sub-humid slope terrain areas.

An efficient, multidimensional instrument is needed to screen non-optimal prenatal parental representations predictive of postnatal parenting behavior and child attachment. The present work aimed to revise and validate the Prenatal Caregiving Expectations Questionnaire-Revised (PCEQ-R). Survey data from two independent samples of pregnant, primarily Danish, women (N=300/322) were collected to 1) test the factor structure and select items for a 20-item version, and 2) confirm the factor structure, examine internal consistency, and establish initial construct validity. Confirmatory factor analysis supported a three-factor model of helpless-dysregulated, anxious-hyperactivated, and avoidant-deactivated caregiving representations. Internal consistency was acceptable (α>.73). Construct validity analyses showed that higher helpless-dysregulated caregiving was associated with low maternal antenatal attachment quality (rs=-.36) and intensity (rs=-.11), increased risk of perinatal depression (rs =.37), and trait anxiety (rs =.37). Higher anxious-hyperactivated caregiving was associated with better maternal antenatal attachment quality (rs =.20) and higher intensity (rs=.26), while avoidant-deactivated caregiving was not associated with maternal antenatal attachment. These findings support the validity and multidimensional structure of the measure. The homogenous nature of the sample limits generalizability of results. Future studies should examine predictive validity of the PCEQ-R and include clinical samples.