As an essential means of energy transportation, pipelines have been widely used in various fields. However, many external factors such as vibration and corrosion can cause damage at the flange part, which seriously affects the safety of pipeline transportation. Quite a number of methods for troubleshooting at pipeline flanges have been continuously proposed, yet there is little research on diagnostic methods for the stabilizer at the flange. Therefore, in this paper, we focus on the stabilizer of the flange and a method that combines traditional detection and machine learning with each other to detect stabilizer faults is proposed. At first, we can obtain a stable and reliable diagnostic data by combining the advantages of the preload of the bolt and the acoustic signal. Subsequently, the optimized N-Beats model is trained based on the measured bolt preload data to predict the service state of the stabilizer. Finally, the data measured by the sensors as well as the predicted data are analyzed by a simplified classification algorithm to determine whether a fault has occurred and to classify the fault. The fault detection method used in this paper not only improves the accuracy of detection and shortens the fault detection time, but also improves the automation level of pipeline inspection. Hence, the work done in this paper has far-reaching practical significance for ensuring the safe and stable operation of pipelines.