The world estimate for plastic pollution is expected to rise above 300 million tons annually. This prompts the need for alternative chemical recycling solutions, such as pyrolysis. Pyrolysis, high temperature, high pressure reactions, could prove to be a sustainable method to recycle plastics and produce fuel oil. Collecting waste plastic to convert it to fuel via pyrolysis could help significantly reduce the number of waste plastics, though accurately predicting the oil yield remains a challenge. One way to predict the outcomes of these reactions is through machine learning. In this work, 310 datapoints were collected of plastic pyrolysis data already existing in the literature to create models that accurately predict the oil yield of a reaction based on the reaction conditions. These models were created using Scikit-learn’s random forest regression and classification methods. Due to the modest size of the compiled literature data set, emphasis was placed upon incrementally improving the methods over iterations by variable selection, constraining input variables and the output oil yield, and by selectively removing error-prone, outlier data. From the models’ results, it was concluded that machine learning methods could provide a viable way to predict pyrolysis oil yields.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Ferrara, Owen G.
• Skangos, Christopher A.
• Himebaugh, Eric T.
• Rando, Matthew P.

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-050621-090537
• 22696

Keyword

• Recycling
• Pyrolysis
• Data Science
• Machine Learning
• Chemical Engineering
• Sustainability

Advisor

• Timko, Michael T.

Year

• 2021

Date created

• 2021-05-06

Resource type

• Major Qualifying Project

Major

• Chemical Engineering

Rights statement

• In Copyright