Abstract: Ultrasonic thickness measurement is an important nondestructive testing technology for maintaining pipeline safety. Its measurement principle is to use the ultrasonic pulse reflection method to monitor the thickness and corrosion of oil pipelines online.The accuracy of ultrasonic pulse reflection thickness measurement mainly depends on the speed of sound, whereas the latter is greatly affected by the ambient temperature. Therefore, the author has performed a lot of experimental studies on the relationship between the speed of propagation of ultrasonic waves in solids and different temperature. Based on these studies, a neural network error compensation model and a linear regression error compensation model for temperature and sound velocity are proposed. The two models are integrated into the ultrasonic pulse reflection method, and the error compensation effects of the two models are compared and analyzed. An ultrasonic sound velocity self-correction error compensation online monitoring system is developed, which eliminates the error caused by temperature influences on the ultrasonic thickness measurement to a certain extent.