Abstract: In order to solve the problem that there are differences between the analytical and experimental results of ultrasonic testing of carbon fiber reinforced plastics (CFRP) porosity by using ultrasonic attenuation method, a mathematical characterization for pores in CFRP is proposed on the basis of random medium theory. Random medium with ellipsoidal autocorrelation functions is improved through minimum searching method. The effects of building parameters are found with orthogonal test. And parameters such as autocorrelation length a and b, ratio of autocorrelation length r0=a/b and side length of sub-zone D, are optimized according to a large number of CFRP pores morphology and distribution statistical data. Finally, random pores model for CFRP is established. The results show that the morphology and distribution of pores can be characterized by random pores model based on actual statistical data of pores, which provides the theoretical foundations for the analysis of ultrasonic scattering from CFRP pores. For the relation between ultrasonic attenuation coefficient and porosity, the result of numerical calculation accords with that of experimental test. The result shows that random pores model can characterize global elastic properties and partial elastic disturbance from random pores of CFRP

Key words: Carbon fiber reinforced plastics, Porosity, Random medium, Random pores model