Embedded Monitoring System And Modeling Analysis Of Subgrade Settlement

Subgrade settlement plays a vital role in the safe operation and maintenance of highways.Therefore,a set of automated monitoring equipment is needed for subgrade settlement observation.The STM32F103 series chip with Cortex-M3 core is used as the control core,and the peripheral circuits required for settlement monitoring are designed around its minimum system.Among them,the lower computer part includes power supply circuit,SPI Flash circuit,SD card circuit,liquid crystal display circuit,RS485 circuit,DTU,Ethernet port circuit,etc.The sensor focuses on the principle of LVDT and designs the corresponding drive circuit and AD acquisition circuit on the basis of it.In order to realize the real-time response of the system,the open source Free RTOS operating system was transplanted to the system,and the concurrent execution of the process was realized.The software design part elaborated the software execution flow of each module function of the roadbed settlement monitoring system,and tested it accordingly.In order to simplify the displacement calculation method of the settlement sensor,the output voltage value and displacement value of the AD598 drive circuit are calibrated,and the linearity is found to be high by fitting analysis,so the linear function equation can be used as the basis for displacement calculation.At the same time,the sensor uses Kalman filtering algorithm to digitally filter the voltage sampling value to reduce noise interference.The lower computer is essentially a data forwarding station,which uses DTU wireless communication to remotely transmit the settlement data to the upper computer,and RS485 as the main communication method with the sensor.Because the lower computer needs to perform data forwarding and sensor fault recognition,a simple communication protocol is designed for the two.Data transmission experiments show that the protocol between the sensor and the lower computer is accurately recognized and data is transparently transmitted.Through field installation and debugging of the monitoring system,the 336-day non-equal-time settlement data was collected.The smoother the curve,the more accurate the settlement prediction result fitted by the algorithm model.Therefore,the BP neural network optimized by the LM algorithm is used as the method for constructing the isochronous data to process the original data.Based on the above,Adaboost algorithm is used to predict the subgrade settlement.The weak predictors are GM(1,1)model,Logistics model and Compertz model.Taking MRE and MSE as evaluation indexes,the calculation results show that the fitting results of the prediction model based on Adaboost algorithm are better than any of the above weak predictor models.