In this work, an advanced method is proposed for computational recovery of thermo-physical properties in a wide range from room temperature to values above the alloy melting point. This method allows us to improve material databases of CAE (Computer-Aided Engineering) programs for simulation of high-temperature technological processes. In most cases, computation of unsteady temperature fields is an important stage of CAE analysis in mechanical engineering. Moreover, various manufacturing technologies such as casting, welding, surface hardening, coating, heat treatment provide the desired material structure and its strength by controlling temperature field during solidification and subsequent cooling of a machine part. The problem of reliability in computer modeling arises from the fact that CAE programs usually are not equipped with comprehensive material databases. We have solved this problem especially for casting simulation and molding materials, which can differ in composition at different plants and therefore cannot be combined into a common database. To determine unknown thermo-physical properties, the temperature had been measured in several points of solidifying cylindrical sample initially, and the appropriate computer simulation of the test technology was performed. Then the difference between the calculated and experimental temperatures was minimized using the modified Levenberg-Marquardt optimization algorithm.