In:
Materialwissenschaft und Werkstofftechnik, Wiley, Vol. 40, No. 3 ( 2009-03), p. 169-177
Abstract:
Recently, artificial neural networks (ANN) are being used as an interdisciplinary tool in many applications. There are various training algorithms used in neural network applications. The aim of this study is to investigate the effect of various training algorithms on the learning performance of the neural networks in the prediction and modeling of the effect of micron and nano size oxide material additives on reaction sintering and physical properties of Al 2 TiO 5 ‐based ceramics. Aluminum‐titanate (tialite) based ceramics have found widespread applications due to their good thermal shock resistance and low thermal expansion. Eutectoid decomposition in the initial oxides and low mechanical strength limit the well‐known properties of aluminum‐titanate. In the present work, first, good stabilizing behavior was achieved by addition of micron size talc and appropriate properties were obtained by adding nano boehmite and colloidal silica that result in mullite phase formation. Then, the effect of the weight percentages of these different additives and also the temperature on the bulk density was predicted with four different training algorithms using a back‐propagation neural network. The training sets for the neural network were selected from experimental results. After training ANN, a regression analysis was used to check the system accuracy for each training algorithm. In conclusion, Levenberg–Marquardt (LM) learning algorithm gave the best prediction for the bulk density behavior of Al 2 TiO 5 based ceramics. The response surfaces between the response variables, i.e. weight percent of additives and temperature of the tialite and the processing parameter are presented. The trained artificial neural network can be used for optimizing the sintering process of Al 2 TiO 5 ‐based ceramics.
Type of Medium:
Online Resource
ISSN:
0933-5137
,
1521-4052
DOI:
10.1002/mawe.200900423
Language:
English
Publisher:
Wiley
Publication Date:
2009
detail.hit.zdb_id:
200651-0
detail.hit.zdb_id:
2025930-X
detail.hit.zdb_id:
246726-4
Bookmarklink