Cracked crankshaft failure is a serious fault that reduces the operating efficiency of diesel engines and increases vibration and noise. Therefore, it is of great significance to monitor and diagnose the crack fault. However, the single-signal crankshaft crack diagnosis method is easily affected by the fluctuation of working conditions and leads to misjudgment. Based on this, the vibration characteristics and fault diagnosis of crack-containing crankshafts are investigated, and a multi-signal fusion crack fault diagnosis method is established in this study. Firstly, a bending-longitudinal-torsional coupling dynamics model of the crankshaft containing breathing cracks was established for a four-crank eight-cylinder diesel engine by comprehensively utilizing the transfer matrix method, Timoshenko beam theory, continuous stiffness method, and Lagrange method. The model was also used to obtain the acceleration and strain characteristics of the main bearing nearest to the crack and the rotational speed characteristics at the flywheel. Next, the acceleration signals, strain signals and rotational speed signals at the corresponding positions were also obtained using sensors. Finally, by combining the simulation results with the experimental data, the vibration characteristics of the crankshaft under the action of the crack were obtained, and the connection between multiple types of signals and the crack was revealed, which led to the establishment of a crack fault diagnosis method that synthesizes multiple types of signals.