Abstract: Objectives Accurate monitoring of ice-induced loads on propellers is crucial for ensuring the safety of polar ship navigation. Due to the intense interaction between ice and the propeller, pressure sensors on the blade surface are easily damaged, making it difficult to directly measure the ice-induced loads at the blade. Existing research often adopts indirect inversion methods to avoid this problem, but the inversion results often focus on blade root torque, making it difficult to achieve refined inversion of blade ice loads. To address this shortcoming, this paper proposes an ice-induced load inversion method for propellers based on shaft response. This method utilizes torque and speed time histories at a certain point in the shaft system as inputs to invert and obtain the ice load at a specific location on the blade. Methods Firstly, based on the modal superposition method, a continuum dynamics model of the propeller shaft system is constructed, and the JWH-α numerical integration algorithm is introduced to achieve the inversion of ice-induced torque. Secondly, based on the Green's kernel function method, a transfer function between ice-induced torque and ice-induced load is constructed, and the Tikhonov regularization method is introduced to effectively solve the ill-posed problem in the inversion process. Results The finite element simulation results indicate that under typical operating conditions, the time-domain evolution characteristics of the dynamic ice load obtained through inversion by this method closely match the actual input load, with the peak error controlled within 6.5%. Conclusions The research results verify the accuracy and robustness of this method, providing theoretical basis and technical support for enhancing the monitoring level of navigation safety for polar ships, thereby ensuring the safety of polar navigation.