Publicación: Aplicação de Machine Learning na otimização multiobjetivo de ventiladores centrífugos
Fecha
2024-10-22
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (España), Universidad de Concepción - Chile. Departamento de Ingeniería Mecánica
Resumen
Este estudo investiga a combinação de técnicas de Machine Learning, Computational Fluid Dynamics e MultiObjective Particle Swarm Optimization para otimizar ventiladores centrífugos. Geometrias e simulações são feitas com o software Ansys para alterar variáveis de entrada e obter variáveis de saída. Com base nesses dados, são desenvolvidos modelos de regressão utilizando Random Forest, Decision Tree e Support Vector Machines. A otimização dos hiperparâmetros é conduzida por Grid Search e a avaliação dos modelos é feita com métricas de precisão, acurácia, recall e F1-Score. Após a validação dos modelos, é aplicada uma otimização multiobjetivo para maximizar o Índice de Desempenho e minimizar o Custo do Material. O ventilador estudado é fabricado em aço inoxidável AISI 310. A eficácia da abordagem é confirmada por simulações fluidodinâmicas adicionais com os valores ótimos encontrados, em que o Random Forest apresentou a melhor performance com R² de 0,966 para o Índice de Desempenho e 0,946 para o Custo do Material. As simulações fluidodinâmicas apresentaram uma diferença inferior a 10% em relação aos resultados previstos pelo modelo, evidenciando a precisão do método proposto.
This study investigates the combination of Machine Learning, Computational Fluid Dynamics, and MultiObjective Particle Swarm Optimization techniques to optimize centrifugal fans. Geometries and simulations are performed using Ansys software to alter input variables and obtain output variables. Based on these data, regression models are developed using Random Forest, Decision Tree, and Support Vector Machines. Hyperparameter optimization is conducted via Grid Search, and model evaluation is performed with metrics such as precision, accuracy, recall, and F1-Score. After model validation, a multi-objective optimization is applied to maximize the Performance Index and minimize the Material Cost. The studied fan is manufactured from AISI 310 stainless steel. The effectiveness of the approach is confirmed by additional fluid dynamic simulations with the optimal values found, where Random Forest showed the best performance with an R² of 0.966 for the Performance Index and 0.946 for the Material Cost. The fluid dynamic simulations demonstrated a discrepancy of less than 10% compared to the model's predicted results, highlighting the accuracy of the proposed method.
This study investigates the combination of Machine Learning, Computational Fluid Dynamics, and MultiObjective Particle Swarm Optimization techniques to optimize centrifugal fans. Geometries and simulations are performed using Ansys software to alter input variables and obtain output variables. Based on these data, regression models are developed using Random Forest, Decision Tree, and Support Vector Machines. Hyperparameter optimization is conducted via Grid Search, and model evaluation is performed with metrics such as precision, accuracy, recall, and F1-Score. After model validation, a multi-objective optimization is applied to maximize the Performance Index and minimize the Material Cost. The studied fan is manufactured from AISI 310 stainless steel. The effectiveness of the approach is confirmed by additional fluid dynamic simulations with the optimal values found, where Random Forest showed the best performance with an R² of 0.966 for the Performance Index and 0.946 for the Material Cost. The fluid dynamic simulations demonstrated a discrepancy of less than 10% compared to the model's predicted results, highlighting the accuracy of the proposed method.
Descripción
Organizado y patrocinado por: Federación iberoamericana de Ingeniería Mecánica y Universidad de Concepción - Chile. Departamento de Mecánica, FeIbIm – FeIbEM
Categorías UNESCO
Palabras clave
Computational Fluid Dynamics, Machine Learning, Multiobjective Optimization, Meta-Heurística, Particle Swarm, Centrifugal Fan
Citación
-
Centro
E.T.S. de Ingenieros Industriales
Departamento
Mecánica