Una red neuronal binaria para la identificación de mecanismos isomorfos. // A binary Neural network for identifying isomorphic mechanisms.
Abstract
Un problema de importancia primordial en el diseño mecánico es identificar los mecanismos isomorfos, puesto que los isomorfismos no
detectados generan soluciones duplicadas y por tanto suponen un esfuerzo innecesario en el proceso de diseño. Desde 1960, una gran
cantidad de métodos han sido propuestos para la detección de mecanismos isomorfos. Sin embargo, diversos trabajos han mostrado que
aunque pueden existir algoritmos eficaces para casos particulares, en el caso general los métodos tradicionales para la detección de
isomorfismos en cadenas cinemáticas no proporcionan usualmente soluciones eficientes para este problema, que ha sido clasificado como
NP-duro. Un eficaz método alternativo para la resolución de problemas NP-duros ha surgido recientemente con las redes neuronales. En
este trabajo proponemos un nuevo modelo de red neuronal binaria diseñado para la resolución del problema de detección de mecanismos
isomorfos. El modelo propuesto se halla basado en unas dinámicas discretas que garantizan una rápida y correcta convergencia de la red
hacia soluciones aceptables. Las simulaciones numéricas muestran en los mecanismos analizados que la red neuronal propuesta
proporciona excelentes resultados, mostrándose además muy superior a la red de Hopfield continua tradicional en lo que respecta al tiempo
de computación y en la facilidad de su implementación.
Palabras claves: Mecanismos isomorfos, síntesis de mecanismos, isomorfismo de grafos, red neuronal binaria,
redes de Hopfield.
____________________________________________________________________________
Abstract
An important step in the kinematic mechanism synthesis process is to identify graph isomorphism while
generating new mechanisms. Undetected isomorphisms result in duplicate solutions and unnecessary effort.
Since 1960, a lot of methods have been proposed for the graph isomorphism identification. Some authors have
concluded that, in the general case, the traditional methods of detecting kinematic chain isomorphism have
been not found to be an efficient solution of the kinematic chain isomorphism problem, classified as NP-hard.
This has motivated to attempt a new direction of approach based on neural networks. In this paper we present
a new binary neural network designed for solving this problem. The model is based on appropriate dynamics
for a binary network in order to always generate fast and correct solutions. Simulation runs for the selected
mechanisms show that our network provides fast and good quality solutions and performs better than the
traditional continuous Hopfield network, because of its easier implementation and smaller computation time.
Keywords: isomorphic mechanisms, synthesis of mechanisms, graph isomorphism, binary neural
network, Hopfield networks.
Downloads
How to Cite
Issue
Section
License
Authors who publish with this journal agree to the following terms:a. Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
b. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
c. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).