Abstract:
This study used finite-element calculations to develop regression and neural network models for the double-beam
laser cleaving process of tubular-shaped glass products. A numerical experiment was performed using the central composite
design, where the rotation speed of the glass tube, the geometric parameters of an elliptical laser beam with a wavelength of
10.6 $\mu$m, and the powers of laser beams with wavelengths of 10.6 $\mu$m and 1.06 $\mu$m were used as variable factors. The maximum
temperatures and maximum thermoelastic tensile stresses in the zone of double-beam processing of glass tubes were determined
as responses using finite element modelling with APDL (Ansys Parametric Design Language). The effective architectures for
artificial neural networks have been established with TensorFlow in order to determine the maximum temperatures and
thermoelastic stresses in the laser-treated area. The comparative analysis of neural network and regression models was carried
out. A multi-criteria optimization of the parameters of double-beam laser cleaving of tubular-shaped glass products was
performed using MOGA (multi-objective genetic algorithm).
Fulltext:
PDF file (1443 kB)