Machine learning-incorporated membrane performance prediction for membrane processes
This research project aimed to develop a machine learning model for predicting membrane performance in membrane processes. The study analysed low and multi-dimensional data using Jupyter as coding environment with Python language. Inputs such as membrane type, pressure, temperature, pressure, and so...
Saved in:
| Main Author: | |
|---|---|
| Format: | Undergraduates Project Papers |
| Language: | en |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://umpir.ump.edu.my/id/eprint/47137/1/Machine%20learning-incorporated%20membrane%20performance%20prediction%20for%20membrane%20processes.pdf https://umpir.ump.edu.my/id/eprint/47137/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This research project aimed to develop a machine learning model for predicting membrane performance in membrane processes. The study analysed low and multi-dimensional data using Jupyter as coding environment with Python language. Inputs such as membrane type, pressure, temperature, pressure, and solute concentration were considered. The multilayer perceptron, a fully connected class of feedforward artificial neural network (ANN), was chosen and implemented in Jupyter Notebook as the machine learning technique. The project aimed to overcome the limitations of traditional methods, offering a more efficient and cost-effective approach to predicting membrane performance. The study included examining prediction of permeance and rejection of the membrane, mostly focused on silica-coated membrane. The dataset comprised 361 samples with essential variables and underwent pre-processing steps. Performance evaluation employed mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE) and coefficient of determination (R2). In exploring the vital parameters affecting membrane performance, a comprehensive principal component analysis (PCA) revealed surprising similarities in factors influencing both permeance and rejection. Employing AI models resulted in unprecedented accuracy, reaching as high as 89% for permeance and 90% for rejection prediction. Therefore, this research successfully developed a machine learning model for predicting membrane performance, focusing on silica-coated membranes. In summary, through data analysis, model development, and validation, the project sought to provide a more efficient and cost-effective approach to predict membrane performance, contributing to sustainable membrane processes. |
|---|