Design and Vision-Based Calibration of a Five-Axis Precision Dispensing Machine

Five-axis precision dispensing machines are employed for semiconductor packaging. The dispensing accuracy is significantly affected by multiple geometric errors among the five axes. This paper proposes a vision-based measurement (VBM) system for identifying geometric errors and calibrating kinematics. The VBM system is also employed to complete the detection of the workpiece. A kinematic model of the machine was established using a local product-of-exponential formulation of screw theory. A geometric error identification algorithm was designed. Eight position-independent geometric errors (PIGEs) and position-dependent geometric errors (PDGEs) were involved. The system of overdetermined equations was solved. Combining the singular value decomposition and regularization, eight PIGEs in the A and C axes were identified. Comprehensive error measurement results verified the proposed approach. The VBM system measured a mean spatial position error of approximately 59.9 μm and a mean orientation error of about 160 arcsec for the end-effector, reflecting the geometric error level of the prototype machine. The proposed approach provides a feasible and automated calibration solution for five-axis precision dispensing machines.