Atmospheric residue desulfurization (ARDS) is a key catalytic process for removing impurities from high-sulfur atmospheric residues (HSAR). During long-term operation, catalyst deactivation progressively occurs, requiring a gradual increase in reactor temperature to maintain product quality. In industrial practice, reactor temperature adjustment is largely dependent on operator experience, which often results in suboptimal energy utilization. To address this issue, this study proposes an optimization strategy that integrates an artificial neural network (ANN) model with a genetic algorithm (GA). The ANN framework incorporates an impurity-removal prediction model developed in our previous work, together with a newly established model for predicting the treated-atmospheric residue (T-AR) flow rate. By coupling these ANN models with a GA-based optimizer, the optimal daily reactor temperature profiles are determined that satisfy impurity specifications in the T-AR product while minimizing energy consumption. Compared with conventional operating strategies based on operator experience, the proposed method achieves a 3.03% reduction in cumulative energy consumption. Furthermore, when operating cases with identical operating temperature conditions are excluded, the energy-saving potential increases to 6.50%, demonstrating the effectiveness of the proposed optimization framework for improving energy efficiency in ARDS operations.
With the rapid advancement of flexible electronics and artificial intelligence, high-performance electronic skin (e-skin) has emerged as a vital platform for real-time health monitoring and intelligent human-machine interfaces. Given the increasing complexity of future application scenarios, the development of next-generation e-skin technologies places higher demands on the multifunctionality, biocompatibility, and biodegradability of sensing units. In this work, a silk fibroin e-skin (SFES) that integrates a capacitive e-skin (CPES) and an electromyographic e-skin (EMES) is proposed. The CPES, composed exclusively of silk fibroin ionic aerogel (SFIA) and a thin silver layer, features a minimalist architecture and a high specific surface area, achieving a remarkable tactile sensitivity (14.19 kPa-1), a broad detection range (0.6 Pa-40 kPa), and an ultrafast response time (5.6 ms). Meanwhile, the EMES module, which is composed of SFIA and silver electrodes, can operate independently when attached to the skin, enabling precise acquisition of physiological electrical signals. By integrating CPES and EMES, the SFES achieves multifunctional sensing and adaptive performance across diverse application scenarios. When further combined with a 1D convolutional neural network (1D-CNN) and a human-machine interaction interface, the SFES enables an intelligent bidirectional interaction system capable of gesture recognition, real-time control of virtual avatars, and immediate feedback to the user. Moreover, cell culture and degradation experiments confirm the outstanding biocompatibility and degradability of the SFES materials. This work thus provides new insights into the development of sustainable e-skin technologies and offers a promising strategy for next-generation intelligent human-machine interfaces. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
This study employs both quantitative topic modeling using BERTopic and qualitative thematic analysis to investigate how themes in Governance evolved between 1990 and 2023. During that time Governance has become a leading journal in public administration and public governance, and understanding how those fields of scholarship appear in the journal can say a good deal about the development of those fields. Governance has both reflected and influenced changes in public administration by theoretically and empirically exploring executive politics along with administrative reforms and regulatory governance as well as democratic mechanisms. The analysis reveals four main conceptual dimensions, which are Structure, Mechanisms, Process, and Democracy, as each dimension represents separate yet interconnected academic discourses. These developments together demonstrate a larger intellectual movement in public administration which moves away from structural determinism to embrace detailed analyses of governance procedures and democratic practices while responding to global issues like populism and democratic decline. This research delivers a historical and intellectual evaluation of governance studies as it appears in a major journal, while delineating trajectories that will guide future research agendas in governance, public administration and policy.
Island destinations attract international tourists seeking unique natural, cultural, and lifestyle experiences. However, due to geographical isolation and infrastructural limitations, they often become overdependent on tourism development, challenging sustainable competitiveness. Integrating push-pull motivation theory and the resource-based view, this study explores how text mining of tourist experiences can identify key attributes shaping island tourism competitiveness between Bali and Lombok. Results show Bali's competitiveness is driven by human services, cultural attractions, and food quality, while Lombok's strengths lie in natural serenity and marine tourism. Yet, Lombok faces infrastructural and environmental issues, especially cleanliness. The findings stress the value of regional collaboration through resource sharing and coordinated planning and suggest benefits of text mining analytics. As a result, this study contributes to tourism theory and offers actionable insights for enhancing the sustainability and competitiveness of island destinations.
Precise six-degree-of-freedom (6DoF) head pose estimation is crucial for safety-critical applications and human-computer interaction scenarios, yet existing monocular methods still struggle with robust pose estimation. We revisit this problem by introducing TRGv2, a lightweight extension of our previous Translation, Rotation, and Geometry (TRG) network, which explicitly models the bidirectional interaction between facial geometry and head pose. TRGv2 jointly infers facial landmarks and 6DoF pose through an iterative refinement loop with landmark-to-image projection, ensuring metric consistency among face size, rotation, and depth. To further improve generalization to out-of-distribution data, TRGv2 regresses correction parameters instead of directly predicting translation, combining them with a pinhole camera model for analytic depth estimation. In addition, we identify a previously overlooked source of bias in cross-dataset evaluations due to inconsistent head center definitions across different datasets. To address this, we propose a reference system alignment strategy that, given the availability of 3D face geometry labels, quantifies and corrects translation bias to enable fair comparisons across datasets. Extensive experiments on ARKitFace, BIWI, and the challenging DD-Pose benchmarks demonstrate that TRGv2 outperforms state-of-the-art methods in both accuracy and efficiency. Code and newly annotated landmarks for DD-Pose will be publicly available.
