Abstract: A thickness determination method of fluid thin layer between casing and cement is put forward by using pulse-echo low frequency ultrasonic waves emitting into a cased well vertically. Sensitivity function shows that comparing with the resonant frequencies of the reflection model the phase jump values at these frequencies are more suitable for inversion. The 3^rd~10^th phase jump values are selected to be the inversion characteristics by using error function analysis. A back propagation neural network is applied to inverse different thin layer thicknesses which are between 0.1 mm and 1 mm. The average relative error between the inversion value and the actual value is 4.6% and the minimum inversion thickness is 18% of the wavelength. The inversion characteristics mentioned can also be used in thickness determination of thin layer beneath layered media with high acoustic impedance.