In the automotive field, welded joints are the weak points of the steel parts of the chassis in terms of fatigue. The fatigue strength of arc-welded parts is generally evaluated using fatigue tests under constant amplitude loadings designed according the Palmgren-Miner rule. In this paper, the relevance of this process is tested with an experimental campaign on welded specimens subjected to various variable amplitude loads. In addition, to solve the recurrent issue of lack of fatigue data, a thermographic method is investigated as an alternative for monitoring the damage rate evolution. A theoretical thermal model is proposed to estimate a local heat source from the infrared thermal measurements during constant and variable amplitude loading tests. This heat source is shown to be correlated with fatigue damage, which provides a further insight on the evolution of damage during nonconstant amplitude fatigue tests.