Design of a Soil Moisture Sensor with Temperature Compensation Using a Backpropagation Neural Network

Abstract

This paper presents the design and construction of a soil moisture sensor (ITM-01) with temperature compensation using a backpropagation neural network. To validate the sensor measurements, a series of experiments were conducted near the city of Atecuaro, Michoacan, Mexico (19deg35psila N, 101deg11psila W), in two different test sites. The soil contents were 67% clay, 17% lime and 16% sand. Measurements of the soil water content were obtained using a TDR soil moisture sensor, (6050X1 Trase System), and the sensor described in this work. The neural network was trained using data obtained from the gravimetric method. The parameter chosen to evaluate the performance of the sensor was the Sum of Squared Error (SSE) of the measurements compared to the gravimetric method. The results showed that the sensor ITM-01, yields volumetric water content measurements in agreement with gravimetric and TDR measurements. In particular for the data reported in this work, the ITM-01 sensor delivered measurements closer to gravimetric data compared to data obtained using the TDR sensor. Corn Crop: SSE TDR= 926, SSE ITM-01=372. Flat terrain: SSE TDR= 410, SSE ITM-01= 78.