The advent of 3D printing technologies has enabled the development of low-cost prosthetic underactuated hands, with cables working as tendons for flexion. Despite the particular relevance to human grasp, its conception in prosthetics is based on vague intuitions of the designers due to the lack of studies on its relevance to the functionality and performance of the device. In this work, some criteria for designers are provided regarding the carpometacarpal joint of the thumb in these devices. To this end, we studied four prosthetic hands of similar characteristics with the motion of abduction/adduction of the thumb resolved in three different ways: fixed at a certain abduction, coupled with the motion of flexion/extension, and actuated independently of the flexion/extension. The functionality and performance of the hands were assessed for the basic grasps using the Anthropomorphic Hand Assessment Protocol (AHAP) and a reduced version of the Southampton Hand Assessment Procedure (SHAP). As a general rule, it seems desirable that thumb adduction/abduction is performed independently of flexion/extension, although this adds one degree of control. If having this additional degree of control is beyond debate, coupled flexion/extension and adduction/abduction should be avoided in favour of the thumb having a fixed slight palmar abduction.

Stroke often causes neuromotor disabilities, impacting index finger function in daily activities. Due to the role of repetitive, even passive, finger movements in neuromuscular re-education and spasticity control, this study aims to design a rehabilitation exoskeleton based on the pulp pinch movement. The exoskeleton uses an underactuated RML topology with a single degree of mobility, customized from 3D scans of the patient’s hand. It consists of eight links, incorporating two consecutive four-bar mechanisms and the third inversion of a crank–slider. A two-stage genetic optimization was applied, first to the location of the intermediate joint between the two four-bar mechanisms and later to the remaining dimensions. A targeted genetic optimization process monitored two quality metrics: average mechanical advantage from extension to flexion, and its variability. By analyzing the relationship between these metrics and key parameters at different synthesis stages, the population evaluated is reduced by up to 96.2%, compared to previous studies for the same problem. This custom-fit exoskeleton uses a small linear actuator to deliver a stable 12.45 N force to the fingertip with near-constant mechanical advantage during flexion. It enables repetitive pulp pinch movements in a flaccid finger, improving rehabilitation consistency and facilitating home-based therapy.

Endodontic rotary file geometries are often obtained using computed tomography (CT) scans, which produce 3D models comprising point clouds and triangulated surfaces. Despite being widely used, this approach has significant limitations; scanned geometries may deviate from the theoretical design due to physical deformation during manipulation, inaccuracies due to the scanner resolution, and the non-parametric nature of the resulting mesh, preventing design modification or parameter extraction. This study proposes a methodology to overcome these limitations by recognizing and reconstructing the scanned geometry. This process involves correcting deformation caused by flexion, applying filtering techniques to minimize scan-induced noise, and identifying key geometric parameters. This enables the generation of a manipulable and accurate CAD model which not only preserves the original design intention but also allows for parametric modifications and advanced finite element analysis. The proposed method bridges the gap between real geometry acquisition and design-based simulation, providing a powerful tool for endodontic instrument evaluation and optimization.

En los últimos años se han desarrollado considerablemente diversas herramientas de inteligencia artificial (IA) basadas en modelos de procesamiento de lenguaje natural, que son capaces de contestar preguntas y resolver problemas diversos. En el ámbito educativo, falta todavía explorar en detalle las posibilidades que estas herramientas ofrecen como complementos a la labor docente realizada por los profesores y su fiabilidad. En este trabajo se analizan y comparan las capacidades de dos herramientas representativas de este tipo (ChatGPT y Gemini) para resolver cuestiones y problemas en el ámbito de la Teoría de Máquinas y Mecanismos (TMM).Para ello, se generó una batería de cuestiones relacionadas con TMM, clasificadas de acuerdo a la taxonomía de Bloom revisada. El conjunto estaba formado por un total de 48 cuestiones, ocho por cada una de las seis dimensiones de la taxonomía (D1-recordar, D2-comprender, D3-aplicar, D4-analizar, D5-evaluar y D6-crear). Las cuestiones fueron elaboradas por los autores, profesores del área de ingeniería mecánica de la Universitat Jaume I, todos con experiencia en la docencia en TMM, tratando de conseguir un conjunto representativo de los diferentes contenidos incluidos habitualmente en la asignatura. Cada cuestión fue planteada exactamente igual a cada una de las dos IAs por uno de los participantes. Las respuestas anonimizadas de cada una de las IAs fueron luego corregidas por otros dos participantes en este trabajo, según un reparto en el que cada profesor realizó el mismo número de interacciones con las IAs y correcciones en cada nivel de la taxonomía. La corrección se realizó utilizando una rúbrica que tenía en cuenta tres aspectos: A1) la calidad técnica de la respuesta, A2) la precisión y completitud de la misma y A3) su redacción y ortografía, siguiendo una escala de valoración de Likert de 1 a 5 (1-muy mal, 2-necesita mejorar, 3-satisfactoria, 4-buena, 5-excelente). Se analizó la valoración conseguida por cada IA en los diferentes aspectos valorados y en las diferentes dimensiones de la taxonomía de Bloom.Los resultados indican diferencias significativas entre las dos IAs en cuanto a valoración promedio en calidad técnica de la respuesta, situándose ambas entre buena y satisfactoria (ChatGPT 3.8, Gemini 3.3, p<0.05), y en cuanto a precisión y completitud (ChatGPT 4.0, Gemini 3.4, p0.05). Se observó que los resultados en los aspectos A1 y A2 dependen significativamente de la dimensión en la taxonomía de Bloom de la respuesta, siendo mejores los resultados en la dimensión D1 (recordar) que en la D6 (crear), con menores diferencias entre las categorías intermedias (para A1: D1: 4.1, D2: 3.5, D3: 3.4, D4: 3.7, D5: 3.7, D6: 2.8, para A2: D1: 4.0, D2: 3.7, D3: 3.8, D4: 3.8, D5: 4.0, D6: 2.9). Se observaron diferencias significativas entre correctores al evaluar los aspectos A2 y A3, pero no para el A1.

This work presents a modified numerical model for predicting the transverse free vibration of stepped beams with improved accuracy over the classical Timoshenko beam theory (TIM model). While TIM model performs well for beams with constant cross-sections, it fails to capture the dynamic response near abrupt changes in geometry, as is typical in stepped beams. To overcome these limitations, a new formulation is proposed (MOD model) that introduces smooth transitions in cross-sectional properties via a parametrized sigmoid function applied to stiffness and mass matrices. A systematic optimization of the transition parameters is conducted using a genetic algorithm to best approximate high-fidelity finite element results (REF model). The MOD model is validated across 35 beam configurations and outperforms the TIM model in most cases, especially for beams with high geometric discontinuities. Additional case studies involving asymmetric and multi-step beams demonstrate the generalization capability of the proposed approach. The MOD model thus offers a computationally efficient yet accurate tool for dynamic analysis of complex beam geometries.

Background: The rapid development and adoption of generative artificial intelligence (GAI) tools, particularly following ChatGPT's release, have created new opportunities and challenges for engineering education. While several studies have explored AI capabilities in mechanical engineering, systematic analysis across different cognitive skill levels remains limited. Purpose: This study evaluates and compares the performance of five conversational AI tools (ChatGPT, Gemini, Claude, DeepSeek, LeChat) in Theory of Machines and Mechanisms (TMM), analyzing their capabilities across the six cognitive levels of revised Bloom's taxonomy: L1‐Remember, L2‐Understand, L3‐Apply, L4‐Analyze, L5‐Evaluate, and L6‐Create. Methods: A set of 48 TMM questions was developed, with eight questions targeting each of Bloom's taxonomy levels. Four mechanical engineering professors interacted with the five AI tools using standardized prompts. Responses were anonymized and independently evaluated by two graders using a 5‐point Likert scale across three aspects: technical quality, accuracy and completeness, and quality of writing. Statistical analysis was performed through ANOVA, Kruskal–Wallis test, and linear mixed‐effects models, to examine differences among AI tools and cognitive levels. Results: AI tools demonstrated generally good performance, with mean scores ranging from 3.76 to 4.22 for technical quality and 3.79 to 4.27 for accuracy and completeness. Writing quality scores consistently exceeded 4.65. LeChat performed significantly worse than the other tools. Performance was notably weaker at the L6‐Create level compared to all other cognitive levels. Conclusions: GAI tools show promise as educational resources for TMM, particularly at lower cognitive levels, though significant limitations persist in creative tasks requiring higher‐order reasoning and synthesis.