Biblioteca Digital do IPB
Publications Repository of the Polytechnic Institute of Bragança
Recent Submissions
A machine learning approach for enhanced glucose prediction in biosensors
Publication . Abreu, António; Oliveira, Daniela dos Santos; Vinagre, Inês; Cavouras, Dionisios; Alves, Joaquim A.; Pereira, Ana I.; Lima, Jose; Moreira, Felismina T. C.
The detection of glucose is crucial for diagnosing diseases such as diabetes and enables timely medical intervention. In this study, a disposable enzymatic screen-printed electrode electrochemical biosensor enhanced with machine learning (ML) for quantifying glucose in serum is presented. The platinum working surface was modified by chemical adsorption with biographene (BGr) and glucose oxidase, and the enzyme was encapsulated in polydopamine (PDP) by electropolymerisation. Electrochemical characterisation and morphological analysis (scanning and transmission electron microscopy) confirmed the modifications. Calibration curves in Cormay serum (CS) and selectivity tests with chronoamperometry were used to evaluate the biosensor’s performance. Non-linear ML regression algorithms for modelling glucose concentration and calibration parameters were tested to find the best-fit model for accurate predictions. The biosensor with BGr and enzyme encapsulation showed excellent performance with a linear range of 0.75–40 mM, a correlation of 0.988, and a detection limit of 0.078 mM. Of the algorithms tested, the decision tree accurately predicted calibration parameters and achieved a coefficient of determination above 0.9 for most metrics. Multilayer perceptron models effectively predicted glucose concentration with a coefficient of determination of 0.828, demonstrating the synergy of biosensor technology and ML for reliable glucose detection.
Airline stock market reaction to CrowdStrike IT outage: an event study analysis
Publication . Costa, João; Cró, Susana; Moutinho, Nuno; Martins, António Miguel
This study investigates the short-term market effect of CrowdStrike IT outage in the airline industry. Using an event study methodology, we evidence that airline stocks respond significantly negatively to the technology disruption within two days before and after the event day. IT disruptions, by creating friction in daily operations, such as broken schedules, delayed or cancelled flights, negative externalities, and customer dissatisfaction, lead to loss of value for airlines. The results also show that the most affected airlines are those from main CrowdStrike customers countries (mainly non-Asian countries) and an irrelevance of the business model. Finally, the extent of the stock market’s response to the CrowdStrike IT outage is influenced by other airline characteristics such as profitability, size, leverage, and cyber risk rating.
Collaborative fault tolerance for cyber-physical systems: the diagnosis stage
Publication . Piardi, Luís; Costa, Pedro; Oliveira, André Schneider de; Leitão, Paulo
The reliability and robustness of cyber-physical systems (CPS) are critical aspects of the current industrial landscape. The high level of autonomous and distributed components associated with a large number of devices makes CPS prone to faults. Despite their importance and benefits, traditional fault tolerance methodologies, namely local and/or centralized, often overlook the potential benefits of collaboration between cyber-physical components. This paper introduces a collaborative fault diagnosis methodology for CPS, integrating self-fault diagnosis capabilities in agents and leveraging collaborative behavior to enhance fault diagnosis. The contribution of this paper relay in propose a methodology for fault diagnosis for CPS, based on multi-agent system (MAS) technology as a backbone of infra-structure, highlighting the components, agent behavior, functionalities, and interaction protocols, to explore the benefits of communication and collaboration between agents. The proposed methodology enhance the accuracy of fault diagnosis when compared with local approach. A case study was conducted in a laboratory- scale warehouse, focusing on diagnosing drift, bias, and precision faults in temperature and humidity sensors.
Experimental results reveal that the collaborative methodology significantly outperforms the local approach in fault diagnosis, as evidenced by performance improvements in diagnosis classification. The statistical significance of these results was validated using the Wilcoxon signed-ranks test for paired samples.
Simulação numérica de vigas de aço inoxidável sob incêndio e temperaturas elevadas
Publication . Zanoni, Andre Luiz; Piloto, P.A.G.; Mesquita, L.M.R.; Rossetto, Diego
Incêndios em áreas urbanas representam um risco significativo à vida humana, destacando a importância de analisar a resistência estrutural em altas temperaturas para prevenir colapsos. O aço inoxidável, amplamente utilizado em edificações por sua durabilidade e resistência à corrosão, pode apresentar colapso sob calor intenso. Para prever seu comportamento em situações de incêndio, a simulação numérica, especialmente o método dos elementos finitos (FEM), é uma ferramenta essencial. Este estudo utiliza o programa ANSYS para modelar as alterações nas propriedades mecânicas do aço inoxidável em altas temperaturas, contribuindo para o aprimoramento da segurança estrutural. A partir dos ensaios realizados nas amostras B1-B6 descritas por Fan et al. (2016), foram obtidos resultados da capacidade portante, resultados do campo de temperatura e resultados do tempo de resistência ao fogo. Na análise da capacidade portante, o gráfico de força versus deslocamento mostrou alta precisão do modelo numérico em comparação com o modelo experimental, com um erro médio inferior a 4%. Na validação térmica, as curvas numéricas apresentaram valores superiores aos experimentais devido a diferenças no isolamento térmico e na aplicação das condições fronteira em relação às condições reais dos ensaios. Para a validação do tempo de resistência (análise termomecânica), a curva de deslocamento em função do tempo revelou diferenças significativas nas amostras B3 à B6, atribuídas às discrepâncias térmicas previamente demonstradas. A amostra B2 mostrou boa aproximação com os resultados experimentais. Todas as amostras falharam por flexão no plano devido à presença de enrijecedores. No estudo paramétrico, foi analisada a influência das dimensões geométricas e dos subtipos de materiais na capacidade resistente das vigas em aço inoxidável. O aço austenítico 1.4301 apresentou a menor resistência, enquanto o duplex 1.4462 demonstrou possuir maior capacidade, embora com menor ductilidade. Nas análises térmicas, os materiais austeníticos e duplex tiveram comportamento semelhante, enquanto o ferrítico exibiu maior condução de calor. A espessura e a largura influenciaram mais a distribuição de temperatura do que a altura. Nas análises termomecânicas, foram aplicados níveis de carga de 0.2, 0.4 e 0.6 da carga máxima para avaliar o tempo de resistência ao fogo e a temperatura crítica de resistência das vigas. O aço austenítico 1.4571 apresentou a maior resistência, devido ao alto teor de níquel, enquanto o ferrítico 1.4016 mostrou menor desempenho (mais parecido com aço ao carbono). Os resultados destacam que a escolha do material e das dimensões geométricas influenciam diretamente a resistência ao fogo das vigas de aço inoxidável sob condições de incêndio, contribuindo para o desenvolvimento de estruturas mais seguras e eficientes.
Comparison of neural network architectures for diabetes prediction
Publication . Guerreiro, Nathan Antonio; Nijo, Rui; Teixeira, João Paulo
Diabetes represents a significant global health challenge, with millions of individuals affected and substantial impacts on healthcare systems. In this study, we compare two neural network architectures for diabetes prediction: the Feedforward Neural Network (FFNN) and the Cascade-Forward Backpropagation Neural Network (CFBPNN). Utilizing the Diabetes Prediction Dataset, comprising 100,000 samples, and after a balanced result, 17,000 samples were obtained. The networks are trained using the Levenberg-Marquardt and Resilient Backpropagation algorithms, and performance metrics, including precision, sensitivity, specificity, accuracy, F1-score, and computational time, are evaluated. Results indicate that the FFNN architecture paired with the Levenberg-Marquardt algorithm demonstrates superior diagnostic prediction accuracy with 91,10%. However, this comes at the cost of longer computational time compared to the CFBPNN.