Optimization of bentonite clay mechano-chemical activation using artificial neural network modeling

详细信息    查看全文

• 作者：Anja Terzić^a ; ^{anja.terzic@institutims.rs} ; Lato Pezo^b ; Ljubi&scaron ; a Andrić^c ; Vladimir B. Pavlović^d ; Vojislav V. Mitić^d
• 关键词：A. Milling ; B. Composites ; B. Electron Microscopy ; C. Chemical ; mechanical and thermal properties ; D. Clays ; E. Structural and thermal applications
• 刊名：Ceramics International
• 出版年：2017
• 出版时间：1 February 2017
• 年：2017
• 卷：43
• 期：2
• 页码：2549-2562
• 全文大小：2267 K
• 卷排序：43

文摘

The properties of seven montmorillonite-rich bentonites of different geological origin were investigated prior and subsequent to mechano-chemical processing in an ultra-centrifugal mill. The objective of the experiment was altering the bentonite types and activation parameters in order to determine the optimal milling conditions that produce material which is physico-mechanically and microstructurally applicable as a binder replacement and sorbent in the construction composites. The efficiency of bentonite activation was assessed by chemometrics and Artificial neural networks mathematical modeling. Principal component analysis and analysis of variance were used in the observation of the influence of input variables (bentonite chemical composition) and process parameters (milling duration, rotor velocity) on the product characteristics: density, specific surface area, grain size and distribution, cation exchange capacity, melting point, compressive strength, shrinkage and porosity. When the ANN models for the observed responses, related to predicted bentonite characteristics and quality, were compared to experimental results, they correctly predicted the responses. The processed data also adequately fitted to the regression second order polynomial models. The SOP models, which showed r2 values from 0.357 to 0.948, and were able to predict the observed responses in a wide range of processing parameters, while ANN models performed high prediction accuracy (0.776–0.901) and can be considered as precise for response variables prediction. The combination of the conducted mathematical analyses showed that that increase/decrease in output values was stabilized after 30 min of activation. Mathematically attained interpretations were correlated with the results of the instrumental analyses (XRD, DTA/TG, SEM) to confirm the adoption of B6 bentonite as a preferable type and 30 min as an optimal milling time for acquiring quality of clay powder that will be used in structural and thermal applications.