Objective To analyze the biomechanical characteristics of lumbar fusion by 3 internal fixation methods using three-dimensional (3D) finite element (FE) method. Methods The FE fixation models of physiological L4-5, unfixed fusion L4-5, translaminar facet screw (TLFS), lumbar pedical screw+translaminar facet screw (LPS+TLFS), bilateral pedical screw (BPS) with complete osteotomy or partial osteotomy of facet joint were established, respectively. The biomechanical characteristics of L4 centrum and implants under six motion states (spinal flexion, extension, lateral bending and axial rotation) in L4-5 fusion model and three fixation models were compared by FE analysis. Results The average maximal displacements of L4 centrum in L4-5 unfixed fusion model, TLFS model, TLFS+LPS model, BPS model were 1.410 8, 0.629 8, 0.336 9, 0.252 8 mm (complete osteotomy of facet joint) and 1.296 7, 0.844 9, 0.340 9, 0.273 8 mm (partial osteotomy of facet joint); the average maximal displacements of cage were 0.479 9, 0.319 5, 0.167 6, 0.126 4 mm (complete osteotomy of facet joint) and 0.378 7, 0.348 4, 0.183 5, 0.137 2 mm (partial osteotomy of facet joint)；the average maximum stresses of screws and rods during 6 motions in TLFS model, TLFS+LPS model, BPS model were 178.34, 79.55, 56.33 MPa (complete osteotomy of facet joint) and 142.29, 103.02, 59.69 MPa (partial osteotomy of facet joint). Conclusions In percutaneous transforaminal lumbar interbody fusion, the fixation effect of BPS model was similar to that of LPS+TLFS model. BPS model could achieve the best spinal stability, and LPS+TLFS model was also a good fixation method. The stability of TLFS model alone was relatively poor, but it was still better than that of cage bone graft without internal fixation. In the absence of internal fixation, preservation of the articular process significantly increased stability of the spine.