Using a Gaussian based model to upcycle seaweed waste into bioplastics with tunable mechanical and material properties

Open Access
- Author:
- Cummings, Christine
- Graduate Program:
- Mechanical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 14, 2022
- Committee Members:
- Tak Sing Wong, Thesis Advisor/Co-Advisor
Xiang Yang, Committee Member
Daniel Haworth, Professor in Charge/Director of Graduate Studies - Keywords:
- bioplastic
gaussian process
regression
optimization
seaweed
mechanical property
water absorption
elastic modulus - Abstract:
- The world has a plastics problem. Currently, up to 40% of plastic waste becomes air, water, or land pollution and this problem is only expected to grow. Bioplastics are a potential, renewable solution, but current bioplastics tend to be high cost, have poor mechanical properties, and use land and water resources that could be dedicated to food production. To address these concerns, this research investigates the use of seaweed as a low-cost and non-resource-intensive biomaterial to serve as the basis for bioplastics. Seaweed is a renewable material that blooms to excess, making it a low-cost bioplastic feedstock that will not require additional farmland or water resources to produce. To improve the bioplastic properties, this research aims to assess the use of a combination of linear Gaussian regression and Gaussian process regression to predict the properties of bioplastics over a large range of combinations of different seaweeds and biomaterials. These predictions can be used to efficiently optimize multiple mechanical and material properties simultaneously in order to best match the desired properties of a synthetic plastic application. These experiments serve as a proof-of-concept and use of a simple Gaussian-based model for the prediction and efficient optimization of elastic modulus and water absorption by adjusting the ratio of three inputs: sargassum seaweed, dulse seaweed, and microcrystalline cellulose. Considering all the possible combinations of these three ingredients in 5% intervals yields 231 potential bioplastics formulations. This project considers an initial sample of 8 bioplastics and a second iteration of 5 bioplastics yielding a bioplastic with an optimized elastic modulus and water absorption. The combinations of sargassum, dulse, and cellulose for the second iteration were chosen to maximize knowledge about the search-space in order to achieve a target elastic modulus of 0.5 GPa and water absorption of 50%. The model’s prediction based on the initial eight bioplastic samples was accurate to within two standard deviations for all of the five bioplastics created in the second iteration. Results were within one standard deviation for four out of five of the elastic modulus predictions and three out of five of the water absorption predictions. Using the Gaussian model instead of a brute force approach reduced the number of tests required by 94% while achieving the target mechanical and materials performances by within two standard deviations. These results indicate that this method can be expanded to optimize additional mechanical and material properties across a wider input of bioplastic ingredients.