Due to the large number of materials and colors of waste plastic bottles, the sorting process is very complicated. In order to improve the rejection rate of GLP4 plastic infrared sorter, three plastic bottles made of PET were selected as identification targets in the visible light+infrared light dual-optical path excitation environment. The proportion of mixed materials is controlled as 80% conventional blue-white-green (off-label+non-off-label), 10% twisted blue-white-green bottles, and 10% 1.2~1.5 L bottles, and the sorting test is carried out under visible light and infrared light. Firstly, fluent was used to simulate the air jet conditions under different nozzle diameters to determine the appropriate nozzle, and then the sample rejection rate under different working conditions was studied. The results show that when the nozzle diameter is 4.0 mm, the high-pressure gas can act on the material positioned at 100 mm, and the internal pressure of the nozzle is not completely consumed. The rejection rate of the sample is 99.36% when the new separator distance is 72 mm, the rejection rate is 99.08% at the angle of 0°, and the rejection rate is 99.48% when the nozzle diameter is 4.0 mm, which verifies the accuracy of the simulation results.