Intelligent adaptive control method for key parameters of vibration rolling during high-speed railway subgrade compaction

Abstract

To improve the intelligent construction of high-speed railway foundations using continuous subgrade compaction control techniques, a large amount of measured data was accumulated through rolling tests at four sites along the Tianjin Beichen Station high-speed railway foundation. An intelligent adaptive control program for key technical parameters of vibration rolling was then proposed using a long short-term memory neural network. The accuracy of the proposed method was verified, the corresponding calculation process presented, and case validation carried out. The results show that between the measured and predicted values of the compaction quality control index, the average absolute value percentage error (APE) was limited to within 10% with a minimum of 4%, and the mean APE was limited to within 5.5% with a minimum of 3.397%; and the root mean square error was limited to within 2.5 with a minimum of 1.135. These results demonstrate that the proposed model was successfully trained to store the non-linear function describing the relationship between the data in the system and that its learning effect was reasonable, meeting the accuracy requirements of the project. The proposed method can fulfil the requirements for real-time precise intelligent compaction by determining the key parameters of the vibratory roller. It can thereby realize modular management and precise pressure compensation to ensure the efficiency and quality of high-speed railway subgrade fill compaction by vibratory rolling.