Determination of residual stress by artificial neural network in Hsla-100 steel weldments

Küçük Resim Yok

Tarih

2009

Dergi Başlığı

Dergi ISSN

Cilt Başlığı

Yayıncı

Maltepe Üniversitesi

Erişim Hakkı

CC0 1.0 Universal
info:eu-repo/semantics/openAccess

Araştırma projeleri

Organizasyon Birimleri

Dergi sayısı

Özet

The residual stress fields near the weld bead in HSLA-100 steel weldments were examined in detail by means of neural network. Many different specimens that were subjected to different conditions were studied. At first, the residual stress by x- ray diffraction is calculated. Then a neural network is created. The input of this network is heat input, preheat, and yield strength, temperature of age and measurement direction. Residual stresses were determined without any calculating or experiment on the surface by using this network under any condition of problem. However, accurate predictions of residual stress could not be obtained without a large number of time and money by experimentally method. An application of the back-propagation neural net work using short term measuring data is presented in this paper. In this study experimental and numerical methods are combined to determination of the residual stress. Some advantage of this numerical method is saving in time and money. The Artificial Neural Network (ANN) is superior to existing experimental techniques. In this study, the neural networks have been employed as a general approximation tool for estimation of the residual stresses in welding of steel. For this purpose used of three functions such as newelm, newff and newcf and by using of MATLAB software, the network is created. The Levenberg-Marquardt algorithm is chosen to perform the training of the networks. A number of samples are analyzed with ANN for parameters of residual stress and the results are compared with experimental method. Back Propagation Neural network (BPN) is used to approximation of residual stress. Resultant low relative error value of the test indicates the usability of the BPN in this area. The results show that, estimation by newelm function is better than newff and newcf functions because it has less error than another function. Also specimens which are subjected to different welding heat input have similar distributions of residual stress on the surface, but the magnitudes of stresses are different. Higher welding heat input generates smaller stress.

Açıklama

Anahtar Kelimeler

Kaynak

International Conference of Mathematical Sciences

WoS Q Değeri

Scopus Q Değeri

Cilt

Sayı

Künye

Heidari, M. (2009). Determination of residual stress by artificial neural network in Hsla-100 steel weldments. Maltepe Üniversitesi. s. 274.