To improve the trajectory tracking control performance of hybrid mechanism systems under the influence of parameter perturbations, time-varying friction, external disturbances, while also considering the impact of the discontinuity of friction forces on the trajectory tracking performance, a convolutional neural network (CNN) based predefined-time sliding mode control method is proposed. A CNN is designed to estimate the lumped uncertainties of the hybrid mechanism system, including discontinuous friction forces. On this basis, to address the issue that the settling time of conventional sliding mode control is typically difficult to preset and challenging to achieve fast convergence in practical engineering applications, a CNN-based predefined-time sliding mode control algorithm is designed. This approach aims to suppress the impact of uncertainties, including discontinuous friction forces, and ensures that the system achieves rapid stability within a predefined time. Matlab simulations and experiments on a hybrid mechanism prototype for an automotive electrocoating conveyor system demonstrate the effectiveness and superiority of the proposed control method.