The extreme impact conditions faced by tunneling boring machines during construction in complex geological environments cause significant damage to key components such as the main drive, and such impact loads are difficult to capture through the commonly used low-frequency hydraulic cylinder pressure parameters. This article proposes a strain-load identification method for tunneling boring machines. By establishing a simulation model of the main drive structure, the sensitivity is calculated to determine the sensitive measurement points. Furthermore, a load identification model based on one dimensional convolutional neural network model is constructed, and high-frequency impact load such as thrust and torque of the tunneling machine are solved using strain monitoring signals. The method proposed in this article was verified in the shield tunnel contact channel site of the Haizhu Bay Tunnel Project in Guangzhou. The average relative errors of identified thrust and torque are 1.48% and 5.29%, respectively, which satisfied the engineering accuracy requirements. This provides a new method for design optimization of the tunneling machine and over limit condition warning.