Abstract: A thermal error modeling method based on the fuzzy C-means (FCM) clustering algorithm and multiple linear regression (MLR) is proposed to address the lack of robustness in thermal error models caused by multi-component thermal deformation coupling in inclined bed precision CNC lathes. The FCM algorithm was used to screen temperature measurement points, and multiple linear regression (MLR) models were constructed for the deflection and linear thermal error of two key components, namely the turret and the spindle. The two submodels were then integrated to form the overall thermal error model of the CNC lathe. The results indicate that the key components and overall thermal error prediction model have good prediction accuracy and generalization ability under multiple operating conditions, meeting the requirements of CNC lathe thermal error compensation.