Abstract <p>The goal of this work is to reveal the possibilities of applying deep machine learning methods to solving transient problems of thermoelasticity. The results of numerical solutions obtained using deep machine learning methods are compared with analytical solutions constructed by method of separation of variables and with numerical solutions constructed by finite difference method. It is shown that physically informed neural networks are capable of approximating with a sufficient degree of accuracy solutions to various transient problems in mechanics of deformable solid. The application of deep machine learning methods to solving physical and mathematical problems can become a promising tool for solving more complex problems, including inverse and contact problems in mechanics of deformable solid.</p>