PSI - Issue 40

V.S. Kanakin et al. / Procedia Structural Integrity 40 (2022) 194–200 Kanakin V.S. et al. / Structural Integrity Procedia 00 (2022) 000 – 000

196

3

Fig. 1. An EBSD image (a) and a SEM image (b) of the microstructure of the AlMg6/10% SiC metal matrix composite.

Under conditions of high deformation temperatures, relaxation processes occur in metallic materials and composites based on them. This is reflected in the form of flow stress curves (Gourdet and Montheillet, 2000; Jiang et al., 1999; Rollett et al., 2004). For the AlMg6/10% SiC metal matrix composite, the flow stress curves (Fig. 2) can be divided into several sections (stages). At the first stage, the composite undergoes hardening until the peak (maximum) value of strain stress is reached. Moreover, the strain at which there is the peak stress decreases with increasing temperature and increases with increasing strain rate. The hardening stage is followed by the stage corresponding to material softening, where the flow stress value decreases with increasing strain. At the third stage (the steady-state portion), the hardening and softening rates are close, and the flow stress value remains almost unchanged with increasing strain. This rheological behavior of the composite at high temperatures corresponds to dynamic recrystallization occurring in metal materials (Gourdet and Montheillet, 2000; Jiang et al., 1999; Rollett et al., 2004). For the obtained dependences of flow stress on thermomechanical loading conditions (Fig. 2), we built a neural network, the scheme of which is shown in Fig. 3. Since the accuracy of neural network description depends on the number of neurons in the hidden layer, computational experiments were made to determine the number of neurons for the average relative deviation  and the coefficient of variation  . These parameters are calculated by the following formulas:

    

    





N  

S( ) 100    

z i i z i

N 1

% ,



100

% ,







1

i

where N is the total number of points used in the identification (verification) of the neural network; i  and i z are the calculated flow stress and that used in the identification (verification) of the neural network, respectively; S( )  is the standard deviation of the experimentally obtained flow stress values from those described by the neural network.