As the core component of the power transmission system, the Dual Clutch Transmission (DCT) presents significant advantages with its superior combined transmission efficiency, excellent manufacturing compatibility and high adaptability to hybrid systems. However, the dynamic characteristics of DCT system are complex, and the friction coefficient of clutch has the characteristics of strong nonlinear and parameter time-varying, so it is difficult to establish the dynamic model of DCT system accurately by traditional mechanism modeling methods. Therefore, this article proposes a DCT system dynamics modeling method that combines mechanism data dual drive to address the complexity of DCT system dynamics. Firstly, a dynamic torque model for the engine and a torque transmission model for the clutch were constructed, as well as a DCT start/shift dynamics model based on ideal assumptions. Secondly, six factors affecting the friction coefficient of the clutch were analyzed, and a CNN-LSTM deep learning model was constructed using Convolutional Neural Network (CNN) and Long Short- Term Memory (LSTM) networks to predict the friction coefficient. The model takes parameters such as lubricating oil flow rate, temperature, and pressure as inputs and friction coefficient as output. Finally, the CNN-LSTM model was combined with the mechanism model to form a mechanism data dual driven DCT system dynamics model. Through MATLAB/Simulink simulation verification, the model has high accuracy, laying the foundation for the optimization and improvement of DCT vehicle intelligent control strategy.