Application of artificial neural networks and finite element method for determining the parameters of elliptic laser beam treatment of quartz sol-gel glasses

Modeling of the process of laser splitting of quartz glasses obtained by the sol-gel method using artificial neural networks and the finite element method was carried out. To form a training data set and data for testing neural networks, calculations of temperature fields and fields of thermoelastic stresses were performed using the finite element method in the ANSYS program. Calculations were completed for 875 variants of input parameters, 800 of which were used for training neural networks. The influence of the architecture of the neural network, the size of the training data array, and the training time on the accuracy of determining thermoelastic stresses and temperatures in the zone of laser processing of quartz sol-gel glasses were investigated.