Résumé: In the preparation of high-performance polyurethane (PU) modified bitumen, due to the different kinds of PU modifiers, the design parameters of the preparation process are numerous, and indexes of the performance response need to be selected. As a result, the preparation process of PU-modified bitumen is not universally applicable. Therefore, according to different application environments, how determining the process parameters of the PU-modified bitumen accurately and efficiently is a key problem to be solved urgently. Based on fthe Kriging-PSO hybrid optimization algorithm, this paper proposed a novel design method for the preparation process for the PU-modified bitumen. The response indicators with high relative sensitivity (softening point, rutting factor, Brookfield viscosity, and dispersion coefficient) were screened by using range and variance analysis to improve the fitting accuracy of the Kriging-PSO model after training. Among them, the dispersion coefficient was evaluated by fluorescence microscopy test using the Christiansen coefficient method to evaluate the uniformity of the dispersed phase of the PU modifier. Through the Kriging-PSO algorithm, the main process parameters for preparing PU-modified bitumen in the laboratory were determined as follows: shearing time 86 min, shearing speed 2450 rpm, shearing temperature 148, and PU content 18.6%. The prepared PU-modified bitumen was placed in an oven at 100 for 2 h. The performance indicators of PU modified bitumen were: softening point 90, rutting factor 30 kPa, Brookfield viscosity 80,000 Pa·s, and dispersion coefficient 0.92. The PU-modified bitumen prepared by this optimal process met the expected performance indicators. The results of this paper showed that the Kriging-PSO algorithm provided a new idea for the design of a modified bitumen preparation process and achieve the purpose of designing the optimum process parameters of PU modified bitumen efficiently using fewer samples. Meanwhile, it created a new way for the application of machine deep learning algorithms in the civil engineering field.

Langue: Anglais