Inverse calibration of a thermal model for Directed Energy Deposition from thermocouple and infrared thermography data

Abstract

Thermal modelling of metal Directed Energy Deposition (DED) relies on the calibration of parameters that are not directly measurable and often interdependent. Thermocouple measurements are widely used, but their capability to uniquely determine the thermal field remains challenging. In this context, a staged inverse calibration strategy based on Bayesian optimisation and automated Finite Element model generation with separated heating and cooling phases is investigated. Thereby, thermocouple-based optimisation is evaluated against infrared measurements and dilution geometry.

Thermocouple measurements enable identification of global heat input during heating as well as heat losses during cooling, but provide insufficient information about near-field temperature distributions. In the heating phase, conflicting objectives arise when comparing thermocouple and infrared data, but a correlation analysis identifies the peak-range thermocouple error as suitable measure for heat input calibration. In the cooling phase, both observables are governed by different parameters without strong competition, where the thermocouple error is primarily associated with heat losses, while the thermography data is primarily sensitive to the effective bead-substrate interface conductance. The results highlight that thermocouple-based calibration alone does not fully constrain the thermal field; identifiability can be improved by staged calibration with physically distinct observables, as investigated in this work.

The source code and calibration data are available at https://github.com/InstituteOfMechanics/DEPOQUS_Parameter_Identification.