| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.66 MB | Adobe PDF |
Authors
Abstract(s)
O advento da Indústria 5.0, introduz na Europa uma abordagem centrada no ser humano, que pretende garantir o bem-estar dos trabalhadores nos processos produtivos, incluindo uma especial preocupação com a monitorização e gestão da sua saúde física e mental, recorrendo a novas tecnologias como as de interação homem máquina, os Digital Twins e a inteligência artificial. A literatura indica que se um trabalhador estiver cansado, é mais suscetível a cometer erros e tem uma pior performance, assim é importante avaliar o seu estado de fadiga. Apesar de existirem várias investigações relacionando a fadiga com os tempos de reação em testes Go/NoGo, não existem detetores baseados neste sistema. Este trabalho reporta a criação de um detetor de fadiga baseado neste sistema, recorrendo à aprendizagem automática. Os dados recolhidos indicam uma maior correlação da tarefa NoGo com a fadiga, notando-se uma redução do tempo de reação para níveis de fadiga elevados. O modelo final, de previsão da fadiga em três classes apresenta uma taxa de acerto de 57%. Mas um modelo alternativo de deteção de fadiga muito elevada apresenta uma taxa de acerto de 84%, revelando uma adequação do teste Go/NoGo a este propósito.
The advent of Industry 5.0 introduces a human-centered approach in Europe, which aims to ensure the well-being of workers in production processes, including a special concern with the monitoring and management of their physical and mental health. Using new technologies such as human-machine interaction, Digital Twins and artificial intelligence. The literature indicates that if a worker is tired, he is more susceptible to making mistakes and has a worse performance, so it is important to assess his state of fatigue. Although there are several investigations relating fatigue with reaction times in Go/NoGo tests, there are no detectors based on this system. This work reports the creation of a fatigue detector based on this system, using machine learning. The collected data indicates a greater correlation of the NoGo task with fatigue than the reaction time of the Go task, and a reduction in the reaction time for high levels of fatigue. The final model, predicting fatigue in three classes, has an accuracy of 57%. But an alternative model of very high fatigue detection has an accuracy of 84%, revealing the suitability of the Go/NoGo test for this purpose.
The advent of Industry 5.0 introduces a human-centered approach in Europe, which aims to ensure the well-being of workers in production processes, including a special concern with the monitoring and management of their physical and mental health. Using new technologies such as human-machine interaction, Digital Twins and artificial intelligence. The literature indicates that if a worker is tired, he is more susceptible to making mistakes and has a worse performance, so it is important to assess his state of fatigue. Although there are several investigations relating fatigue with reaction times in Go/NoGo tests, there are no detectors based on this system. This work reports the creation of a fatigue detector based on this system, using machine learning. The collected data indicates a greater correlation of the NoGo task with fatigue than the reaction time of the Go task, and a reduction in the reaction time for high levels of fatigue. The final model, predicting fatigue in three classes, has an accuracy of 57%. But an alternative model of very high fatigue detection has an accuracy of 84%, revealing the suitability of the Go/NoGo test for this purpose.
Description
Keywords
Deteção de fadiga Teste Go/NoGo Aprendizagem automática Fatigue Detection Go/NoGo testing Machine Learning