Banca de QUALIFICAÇÃO: EGON PIRAGIBE BARROS SILVA BORGES

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : EGON PIRAGIBE BARROS SILVA BORGES
DATE: 01/12/2023
TIME: 14:00
LOCAL: Online
TITLE:

Reusability of nanocellulose and natural rubber latex cryogels: Characterization by X-ray microtomography and segmentation of low contrast images by deep learning

PAGES: 75
BIG AREA: Ciências Exatas e da Terra
AREA: Química
SUBÁREA: Físico-Química
SUMMARY:

The impact of oil spills and industrial effluent leaks on aquatic ecosystems has been significant. In the pursuit of effective environmental remediation strategies, several studies have focused on the development of new absorbent and renewable materials for the efficient capture of organic contaminants. Nanocellulose has emerged as a promising material in this context due to its ability to self-organize into highly porous structures with a large surface area. However, the hydrophilic nature of nanocellulose limits its application in absorbing hydrophobic contaminants in aqueous environments.  Recent findings have shown that adding natural rubber latex (NRL) to nanofibrillated cellulose (CNF) imparts hydrophobic properties and structural resilience to the resulting cryogel (CNF@NRL), enabling its use in oil absorption. Based on this background, this work aimed to assess the reusability of CNF@NRL cryogel for oil capture using a novel solvent-free centrifugation approach (for contaminant removal). It also included a detailed morphological characterization throughout reuse cycles using X-ray microtomography (μCT) tomograms. The μCT results were used to analyze structural changes in the cryogel and understand the underlying reasons for performance decline over reuse cycles. However, tomograms of CNF@NRL cryogel with oil exhibited low contrast CNF@NRL cryogel with oil exhibited low contrast, making image segmentation difficult. To address this issue, an innovative approach was proposed, involving deep neural networks (DNN) for μCT image segmentation. A self-supervised method was developed for the SA-Unet network, using oil-free cryogel images as true labels. Textures were added to the images based on these labels to simulate the characteristics of oil-impregnated cryogel tomograms. This approach enabled the DNN to effectively segment low-contrast images of oil-impregnated cryogels. Results from reuse tests indicated that the cryogels retained up to 60% of their initial absorption capacity, with a maximum 25% reduction in porosity after multiple absorption-desorption cycles, since after 30 cicles of reuse. Once trained using the developed method, the SA-Unet network successfully segmented real images of cryogels containing oil. Moreover, the innovative approach of automating the segmentation process using deep neural networks significantly optimized data analysis and processing time, making it possible to segment these images within minutes. This research contributes to the development of more efficient and eco-friendly solutions for addressing the challenges posed by oil contamination in aquatic ecosystems. 

COMMITTEE MEMBERS:
Presidente - Interno ao Programa - 052.991.506-52 - JULIANA DA SILVA BERNARDES - UFABC
Membro Titular - Examinador(a) Externo ao Programa - 1600876 - FRANCISCO DE ASSIS ZAMPIROLLI
Membro Titular - Examinador(a) Externo à Instituição - AUGUSTA CERCEAU ISAAC NETA - UFMG
Membro Suplente - Examinador(a) Interno ao Programa - 050.966.059-22 - MURILO SANTHIAGO - CNPEM
Membro Suplente - Examinador(a) Externo à Instituição - ELISA SILVA FERREIRA
Membro Suplente - Examinador(a) Externo à Instituição - VICTOR RAMÓN MARTINEZ ZELAYA