masterThesis
Modelo de reconhecimento de padrão aplicado no processo de esterilização a vapor com enfâse em ciclos falhos
Fecha
2021-08-30Registro en:
PELISSARI, Thiago José. Modelo de reconhecimento de padrão aplicado no processo de esterilização a vapor com enfâse em ciclos falhos. 2021. Dissertação (Mestrado em Inovações Tecnológicas) - Universidade Tecnológica Federal do Paraná, Campo Mourão, 2021.
Autor
Pelissari, Thiago José
Resumen
Data mining presents itself as a powerful tool to achieve the standardization of
microbial behavioral parameters, so it can help positively in the analysis of the step
monitoring in the steam sterilization process. Therefore, mathematical models can help
the understanding of the microorganisms’ survival probability present in the culture
media of biological indicators, used in the validation and monitoring of sterilization
processes. In addition, statistical techniques such as discriminant analysis and logistic
regression analysis, allows the search of the bacterial activity behavior correlating the
failed cycles in the steam sterilization process. Therefore, a quick and effective
predictive evaluation of the monitoring step of the failed cycles in the sterilization
process, carried out in up to 3 hours when compared to the conventional method 48
hours, implies a reduction in the result time, as well as in decision-making based on
this result. This possibility allows the manufacturing industry of biological indicators to
have a differential focus on its performance and competitive capacity in an existing
market segment, but currently unexplored. Hereupon, the present study aimed to
develop a model for pattern recognition, allowing the definition and evaluation of the
efficiency of a steam sterilization process, when subjected to failed cycles, through
observation of the monitoring and validation test result in up to 3 hours. The application
of the factorial analysis method, logistic regression method, and quadratic score
method, were the prediction model effectiveness. The behavior of the predicted pattern
recognition model linked to the logistic regression method presented at least 97% of
acceptability for data related to the result of the monitoring and validation test in up to
3 hours for failed cycles. Thus, it can be concluded that the results observed in this
work are consistent and meets the given requirements.