Experimental Characterization And Neural Network Prediction Of Dynamic Behavior Of Zta With Srco3 And Mgo

• Main Author: Arab, Ali
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: T Technology; TJ1-1570 Mechanical engineering and machinery
• Online Access: http://eprints.usm.my/46977/1/Experimental%20Characterization%20And%20Neural%20Network%20Prediction%20Of%20Dynamic%20Behavior%20Of%20Zta%20With%20Srco3%20And%20Mgo.pdf; http://eprints.usm.my/46977/

Ceramics materials are extensively used in armor applications for their attractive properties such as high hardness, low density and high compressive strength. However for designing and selection for appropriate ceramic armor material, a deep knowledge about the dynamic behavior of ceramic is necessary. A number of research has been done on dynamic behavior of ceramic, unfortunately most of work focused on the conventional and limited ceramics (such as Al2O3 , B4C, SiC). For this reason prediction of the dynamic behavior of the new composition of ceramics is difficult and some time is impossible. In this work, mechanical properties and dynamic behavior of ZTA are being investigated. For studying the dynamic behavior of the ZTA, SHPB apparatus is modified (using pulse shaper and sandwich the sample with WC platen) and used. Effect of different amount of YSZ (10-40wt.%) on their properties of ZTA is also investigated dynamically using SHPB. ZTA with 20 wt.% YSZ shows the optimum properties and also their dynamic behavior. Effect of SrCO3 (1-5wt.% ) added to the ZTA with 20 wt.% YSZ and the formation of new phase (SrAl12O19) on porosity and fracture toughness is of interest. The formation of this phase increases the porosity and hence decreases the dynamic performance of the composite. An addition of MgO (0.2-0.9wt.%) to ZTA with 20 wt.% YSZ resulted a reduction in grain size and consequently increase the hardness. Further investigation on different dynamic loading condition on ZTA with 20 wt.% YSZ and 0.2wt.% MgO were also conducted. The dynamic behavior of representative ZTA is predicted by three different machine learning methods (Multilayer Perceptron (MLP), Time Series and Supporting Vector Regression (SVR)). The predictions are compared to each other and the time series neural networks shows the best agreement with the experimental data.