Abstract: To address the challenges of low efficiency and heavy reliance on manual expertise in machining process optimization, an autonomous process optimization framework for smart machine tools based on a geometry-physics-motion collaborative simulation approach is presented. Firstly, a geometric model is established to digitally represent the workpiece and tool, enabling the reconstruction of the material removal process during cutting. Secondly, a physical model incorporating a neural network is developed to analyze the thermomechanical loads on the cutting edge, facilitating precise prediction of cutting forces and tool wear. Thirdly, a machine tool motion model is constructed to analyze the kinematic chain and simulate the five-axis simultaneous machining process, constraining the tool center point (TCP) position and tool orientation variation. Building upon these integrated simulation models, an autonomous process optimization method is introduced. This method employs a cyclic optimization strategy coupled with a gradient descent algorithm to autonomously iterate and refine process parameters. The proposed framework has been validated through experimental machining of aerospace engine blades. Results demonstrate that the method effectively achieves process optimization for diverse machining objectives, exhibiting both broad applicability and significant optimization efficiency.