Process parameters and base material thermal properties strongly affect peak temperature evolution in friction stir welding (FSW). This work develops a semi-empirical model for peak temperature estimation during FSW of aluminium alloys. Several bead-on-plate runs were conducted on AA5083-H111, AA6082-T6 and AA7075-T6. Different rotational and welding speeds were tested while measuring the temperature with a thermocouple embedded in the tool. A polynomial model was proposed with the model's coefficients strongly correlating to the material thermal properties. Additionally, a threshold peak temperature was defined to avoid cold defects by combining temperature measurements with joint quality information obtained from X-ray scans. The efficiency of the model and the defined threshold are proven by applying them to the literature.