Banca de DEFESA: BRENDA NOVAIS VIANA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : BRENDA NOVAIS VIANA
DATA : 18/12/2019
HORA: 08:30
LOCAL: Sala de Congregação da Escola Politécnica da UFBA
TÍTULO:

Investigation and Prediction of Eucalyptus wood extractives content based on Principal Component Analysis and Artificial Neural Networks

PALAVRAS-CHAVES:

Prediction; Eucalyptus wood extractives;Artificial Neural.

PÁGINAS: 94
GRANDE ÁREA: Engenharias
ÁREA: Engenharia de Produção
RESUMO:

In the pulp and paper production process, wood extractives are low molecular weight organic compounds that cause operational problems, environmental damage and loss of product quality. There is some research on the impact of extractives on biota, the characterization and removal of pitch, however, there is still a lack of studies investigating the cause of extractives content variability in eucalyptus wood. This research aimed to develop a model for predicting the extractives content in eucalyptus wood clones. Principal Component Analysis (PCA) was applied to assess the impact of the various variables on extractives content: planting region; soil type, amount of sand and clay, organic and inorganic matter, pH and soil management, age, basic density, lignin content and wood genetic material. An empirical neural network model was identified from experimental data, with the main components as input to monitor and predict extractives content, as laboratory measurements may take several days and become available only after wood. have already been processed. Experimental data were provided by a pulp company and contained information on eighteen eucalyptus clone species from five regions in the extreme south of Bahia, Brazil. After initial data screening, a set of 119 samples were collected and analyzed using Principal Component Analysis. The variability of the data was represented by eight main components, which indicated that the potential acidity, iron, aluminum saturation, magnesium, pH, sum of bases, remaining phosphorus, zinc, manganese and copper were the variables that most impacted the acid content. eucalyptus wood extractives. Thus, two neural network models were developed
xvi
whose inputs were these most important variables for the extractives and the eight main components. The neural network models were compared to identify the model with the best performance and viability to be applied at industrial level. The effectiveness of the model was verified by statistical parameters, which indicated the reliability, with good quality of fit to the experimental data. Both models performed significantly by providing a systematic tool for predicting and monitoring the contents of eucalyptus wood extractives before low quality wood affects the process. The artificial neural network whose inputs were the ten significant variables obtained by the PCA technique enabled better quality of network adjustment to the experimental data and better viability as the industrial applicability. The approach developed here can help to monitor product quality as well as to prevent damage to the environment and equipment.

MEMBROS DA BANCA:
Interno - 2199115 - CRISTIANO HORA DE OLIVEIRA FONTES
Externo à Instituição - FERNANDO JOSÉ BORGES GOMES
Presidente - 3495808 - KAREN VALVERDE PONTES