The density of engineering plastics is a key parameter for ensuring their safety and reliability. In order to achieve rapid and high-precision on-site detection, a method based on the acoustic pressure reflection coefficient is proposed. First, finite element simulation analysis was conducted to obtain the acoustic field distribution during ultrasound propagation under water immersion conditions. The correlation between interface echo intensity and material density was determined. Optimal detection parameters were designed to reduce measurement errors caused by beam overlap and diffusion attenuation. A water immersion ultrasonic experimental system was constructed, and the measurement accuracy of the method was tested using chlorinated polyvinyl chloride pipes. The results show that, compared to the measurement results of the Archimedean drainage method, the maximum error of ultrasonic measurements does not exceed 1.7%, and the overall variance is less than 1.2%. The measurement accuracy of this method is compared with the regression results of different machine learning models. It is demonstrated that, compared to regression methods based on variable correlation, this method retains the advantages of high efficiency and low cost in ultrasonic density measurement, while achieving higher measurement accuracy. Additionally, it does not require a dataset for training support, making it promising and valuable for practical applications.