Objective Prediction of anterior cruciate ligament (ACL) stress in the left leg of a volleyball player landing on the ball by XCM deep neural network. Method A complete finite element model of the knee joint was established based on MRI and CT images; the kinematic and kinetic data of the volleyball player were synchronously collected by an 8-lens Qualisys motion-capture system and a Kistler 3D force platform; the knee joint moments were computed by the musculoskeletal model in OpenSim, and the joint moments were used as inputs to the finite element model, with ACL stresses as the outputs; kinematics and kinetic data were used as the input of neural network and ACL stress as the output.Result The peak equivalent stress of ACL of volleyball player's dunk landing was 27.7±0.36 MPa; the maximum principal stress was 8.2±0.23 MPa; the maximum shear stress was 14.7±0.32 MPa; and the equivalent strain was 5.7±0.008%, the maximum principal strain was 5.0±0.006%, and the maximum shear strain was 7.6±0.009%. The NRMSE between predicted and calculated values ranged from 5.84% to 7.12%, and the RMSE ranged from 0.251 to 0.282.Conclusion The XCM deep neural network model can predict the ACL stress during the smashing process of volleyball players within a certain range. Provide a new way to obtain biomechanical data of volleyball players; It provides an effective way to help volleyball players prevent ACL injuries.