Estimation of sand ratio at soil by LDR sensors and intelligent algorithms

In this study, various estimations and analyses are used on the basis of sand amount in soil texture using popular artificial intelligence methods like linear regression, artificial neural network and adaptive neuro fuzzy inference system. Clay, silt and sand, which form the inorganic texture of the soil, are settled in the water with different speeds due to their density differences. Based on this point, using the density difference of moving particles through a mixed soil and water suspension, estimations are made on specific soil samples' data. The sand particle is the biggest particle among the others in terms of volume, it can be measured by simple tools and can produce more scalable results. The accuracy of the estimations, which are made with frequently used methods in machine learning like linear regression, artificial neural network and adaptive neuro fuzzy inference system, are inspected in details using the magnitude change of the light passing through the beaker due to differentiation of density by time. Details of the analyses and estimations, which can be used to determine the approximate sand amount in soil and in situations that don't require sensitive measuring, are compared with each other in details. Also the possibility of making a mobile measurement device using this setup which is built with cheap and easy-to-find equipment is discussed.