Improved fiber optic sensing technology of soil moisture based on neural network

Accurate monitoring of temporal and spatial distribution of soil moisture is of great significance to engineering geological assessment and geo-hazard prevention. A large gradient of moisture content of soil has a relatively great influence on the measurement precision of the actively heated fiber Bragg grating (AH-FBG) method. To analyze the source of measurement errors and its distribution along the depth, three sets of laboratory soil column tests are designed and carried out. A joint analysis method based on the artificial neural network (ANN) algorithm is further proposed to improve the analysis method of AH-FBG moisture sensing technology. The results show that when the AH-FBG method is applied to the soil with a large gradient of moisture content, the longitudinal heat transfer of the sensor and soil will both occur during the heating process simultaneously, and the longitudinal heat transfer of the sensor is dominant. This effect reduces the monitoring accuracy of moisture content, and the related errors cannot be decreased by reducing heating time. The data from laboratory tests and field monitoring indicate that compared with the traditional maximum heating value method, the joint analysis method considers the heat transition and drag effect, and therefore the monitoring accuracy of moisture content is noticeably improved, which proves the superiority of the method.