Cássio Dener Noronha Vinhal

Cássio Dener Noronha Vinhal

Auto-generated profile, based on publicly available information

Cássio Dener Noronha Vinhal

School of Electrical, Mechanical, and Computer Engineering (EMC), Federal University of Gois (UFG), Goinia, Brazil

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 May 2026 in RECIIS
Verified
Journal Article
Open access
AJAntonio Lomanto Pires Santos Júnior
Antonio Lomanto Pires Santos Júnior
TNTiago Do Carmo Nogueira
Tiago Do Carmo Nogueira
GJGelson da Cruz Júnior
Gelson da Cruz Júnior
CVCássio Dener Noronha Vinhal
Cássio Dener Noronha Vinhal
Este estudo aplicou o Faster R-CNN para a detecção e a classificação de tumores cerebrais primários (glioma e meningioma) utilizando imagens de ressonância magnética. O modelo utilizou as arquiteturas VGG-19 e ResNet-101, alinhando uma Rede de Proposta de Região capaz de classificar os objetos e planos de fundo. O conjunto de dados contém 6.307 imagens de gliomas, 6.391 de meningiomas e 3.066 sadias, dividido em 60% para treino, 20% para teste e 20% para validação. As métricas de avaliação incluem Acurácia, F1-Score, Precisão Média, Média de Precisão Média e Interseção sobre União. Nesse sentido, a VGG-19 obteve melhor desempenho para meningioma (Acurácia 0,910, F1-Score 0,900, IoU 0,830) e glioma (Acurácia 0,895, IoU 0,810), superando a ResNet-101 (meningioma: Acurácia 0,880, F1-Score 0,822; glioma: Acurácia 0,870, IoU 0,740).

12 June 2026 in IEEE Access
Verified
Journal Article
Open access
RCRoberto C. P. Chaves
Roberto C. P. Chaves
GCGélson Da Cruz
Gélson Da Cruz
CVCássio D. N. Vinhal
Cássio D. N. Vinhal
WCWesley P. Calixto
Wesley P. Calixto
Electric energy waste in water pumping stations is often associated with empirical operations, to the detriment of technical methods focused on efficiency. Even in modern systems, control criteria are typically limited to reservoir-level monitoring, neglecting essential variables, such as pump efficiency and the variation of energy tariffs. This approach compromises system performance, leading to premature equipment wear, increased operational costs, and inefficient use of water resources.In view of this, this study proposes a novel, comprehensive, and integrated methodology, validated with operational data from a real pumping station, overcoming the limitations of previous approaches by structuring, three complementary stages: prediction of daily water consumption curves using long short-term memory (LSTM) recurrent neural networks; automated delineation of the operational zones with highest energy efficiency under different hydraulic conditions, whose integration with the optimal scheduling stage is enabled by a novel three-dimensional hydraulic performance tensor, responsible for converting the zoning results into a representation compatible with the optimization process, with or without variable frequency drives (VFDs); and determination of the optimal operational schedule, taking into account the hourly energy tariff structure. The latter two stages employ a particle swarm optimization (PSO) algorithm. The results demonstrate the effectiveness of the methodology, with an average energy cost saving of 5.57%, even in a station with a high utilization factor and limited operational flexibility. It also provides greater operational predictability, as evidenced by the good generalization between predicted and observed consumption curves. Furthermore, restricting the operation to the highest-efficiency ranges of pumps promotes more stable and safer operating conditions. These results highlight the distinctiveness of the proposed approach compared with previous studies, offering a holistic, technically robust, and practically applicable solution for real-world water utility contexts.