بهینه‌سازی متغیرهای غشاء اولترافیلتراسیون با استفاده از یادگیری ماشThe optimization of ultrafiltration membrane variables using machine learning for separation of metals from waste waters and industrial solutionsین جهت جداسازی فلزات از پساب‌ها و محلول‌های صنعتی

نوع: Type: Thesis

مقطع: Segment: masters

عنوان: Title: بهینه‌سازی متغیرهای غشاء اولترافیلتراسیون با استفاده از یادگیری ماشThe optimization of ultrafiltration membrane variables using machine learning for separation of metals from waste waters and industrial solutionsین جهت جداسازی فلزات از پساب‌ها و محلول‌های صنعتی

ارائه دهنده: Provider: Fatemeh Sarhadi

اساتید راهنما: Supervisors: Dr. Meisam Nouri & Dr. Hamid Esfahani

اساتید مشاور: Advisory Professors: Dr. Moslem Nouri

اساتید ممتحن یا داور: Examining professors or referees: Dr. Minoo Karbasi & Dr. Muharram Mansoorizade

زمان و تاریخ ارائه: Time and date of presentation: 2025

مکان ارائه: Place of presentation: seminar

چکیده: Abstract: In recent years, the development of new technologies, especially membranes synthesized by electrospinning, has become one of the important priorities in the field of environment for the removal of heavy metal pollutants from water sources. In this study, modeling and prediction of the performance of electrospun membranes in the removal of heavy metals from aqueous solutions was investigated using various machine learning algorithms. The main objectives of this study were to predict the percentage of heavy metal removal by electrospun membranes, their fiber diameter and porosity percentage. To predict the percentage of removal, the input variables included the parameters of the electrospinning process, the characteristics of the aqueous solution and the operating conditions of the filtration. However, to predict the fiber diameter and porosity percentage, only the variables of the electrospinning process were used as inputs. For each of the objectives, four algorithms including decision tree, random forest, nearest neighbor and gradient boosting were used. In predicting the percentage of heavy metal removal, the gradient boosting algorithm showed the best performance with an accuracy of 84.3% and an average absolute error percentage of 0.11. Based on the analysis of Shapley values, the flow rate of the aqueous solution was identified as the most important influential variable in predicting the percentage of removal. In predicting the fiber diameter, all algorithms had an accuracy of more than 99% and the applied voltage was identified as the most influential factor. In predicting the percentage of membrane porosity, the nearest neighbor algorithm also performed well with an accuracy of 83.4% and the molecular weight of the polymer was identified as the most effective variable. To evaluate the effect of membrane structural characteristics (fiber diameter and percentage of porosity) on predicting the percentage of heavy metal removal, these two variables were also added to the inputs of the models. The results showed that although the accuracy of the models increased to some extent, this improvement was not significant, which could be due to the overlapping effect of the electrospinning process variables and the structural characteristics of the membrane. In order to validate the results of the developed models, a membrane was produced using the proposed input variables of the model from polyamide-6 with the addition of 9 wt% titanium oxide by electrospinning. The fiber diameter in this membrane was measured to be 108 nm and the porosity percentage was 34.5%. Also, the percentage of heavy metal removal by the fabricated membrane was obtained to be 90.49%; in contrast, the developed model was able to predict the fiber diameter, porosity percentage, and heavy metal removal percentage as 123.2 nm, 32%, and 95.6%, respectively. Finally, two graphical user interfaces (GUI) were designed to facilitate the exploitation of the results. The first interface predicts the percentage of heavy metal removal by receiving the electrospinning process variables, solution characteristics, and filtration conditions; and the second interface estimates the fiber diameter and membrane porosity percentage by receiving the electrospinning process variables. The findings of this study showed that the use of machine learning models along with laboratory validation can be an efficient tool for predicting and optimizing input variables, which is not only efficient in predicting structural parameters such as fiber diameter and porosity percentage, but also used in the design and manufacture of filtration membranes for removing heavy metals from aqueous solutions.

فایل: ّFile: Download فایل