Publications by authors named "Mengjiao Zhai"
Graphene oxide (GO)-based membranes have demonstrated great potential in water treatment. However, microdefects in the framework of GO membranes induced by the imperfect stacking of GO nanosheets undermine their size-sieving ability and structural stability in aqueous systems. This study proposes a targeted growth approach by growing zeolitic imidazolate framework-8 (ZIF-8) nanocrystals precisely to patch microdefects as well as to cross-link the porous graphene oxide (PGO) flakes coated on the outer surface of the hollow fiber (HF) alumina substrate (named the hybrid PGO/ZIF-8 membrane).
View Article and Find Full Text PDFMetal‒organic frameworks (MOFs) are nanoporous crystalline materials with enormous potential for further development into a new class of high-performance membranes. However, the preparation of defect-free and water-stable MOF membranes with high permselectivity and good structural integrity remains a challenge. Herein, we demonstrate a dual-source seeding (DS) approach to produce high-performance, water-stable MOF-303 membranes with hollow fiber (HF) geometry and preferentially tailored crystallographic orientation.
View Article and Find Full Text PDFArticle Synopsis
- Cancer diagnosis using machine learning, especially with Support Vector Machines (SVM), can struggle due to the complexity and redundancy in gene expression data, leading to poor classification results.
- This paper introduces a hybrid feature selection algorithm called IG-GPSO that effectively ranks and groups features based on their information gain, enabling better data organization for the SVM.
- Experimental findings reveal that IG-GPSO enhances SVM's accuracy to 98.50%, outperforming traditional feature selection methods and demonstrating superior classification performance compared to KNN algorithms.
Article Synopsis
- - The study addresses the issue of sample imbalance in medical datasets, which hinders the accuracy of machine learning models used for clinical diagnosis.
- - A new hybrid sampling algorithm, combining SMOTE (to oversample minority classes) and ENN (to remove noise from majority classes), is proposed to balance datasets for missed abortion and diabetes.
- - The results demonstrate that using this SMOTE-ENN method significantly enhances the classification performance of the Random Forest model, achieving high MCC indices of 95.6% for missed abortion and 90.0% for diabetes.
Article Synopsis
- ZnO is identified as an n-type semiconductor with photocatalytic abilities when exposed to ultraviolet light, prompting research into its structural modifications for improved performance.
- A new method for creating nano-ZnO coatings involves plasma spraying a mixture of liquid precursors with pre-loaded ZnO and Zn to achieve a porous structure with ultrafine grains.
- The developed coatings demonstrated significantly better photocatalytic activity, characterized by a narrower band gap and altered oxygen defects, compared to coatings made from single liquid sources, highlighting an innovative approach to producing functional nanostructured coatings.