SOIL THERMAL CONDUCTIVTY PREDICTION WITH VEGETATION COVER USING ARTIFICIAL NEURAL NETWORK

• Main Author: KUMALA WARDANI (NIM : 90214004), AFNI
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/25087
• Institution: Institut Teknologi Bandung

Thermal conductivity is one of the most important thermal properties of soil to control heat flow in soil. It is affected by several factors such as soil physical properties, water flow and vegetation cover. Soil and plant as a complex system sustain ecosystem stability and microclimate changes. Each soil type has different thermal conductivity. Recent study of soil thermal conductivity requires a long time, then it needs a method to predict the soil thermal conductivity faster. The most popular soft computing methods to predict a certain value is Artificial Neural Networks (ANN). This study’s aim is to predict soil thermal conductivity that involves plant using ANN. Input data was generated from experiment of soil thermal conductivity using dual-probe sensor. This sensor has dual probes to measure voltage and current during heating process and two thermocouples to measure temperature changes. This experiment use four soil types such as cultivate soil, mud, clay and sandy soil. The measurement was applied to bare soil and soil with green seed to investigate vegetation effect to conduction rate on soil. ANN prediction use multi-layer feed forward neural network which trained using back propagation algorithm on 200 training data and 40 testing data by manipulating number of hidden layer. The eight input data are heating time, temperature from thermocouple 1, temperature from thermocouple 2, delta temperature from thermocouple 1 and 2, voltage, current, conduction rate and gradient of delta temperature vs. time to predict thermal conductivity of soil. The results present Mean Square Error (MSE) to bare soil are MSE-training, MSE-validation, and MSE-testing 4,24×10−9, 1,34×10−8 , and 2,03×10−8 respectively. The soils with green seed demonstrate MSE-training, MSE-validation, and MSE-testing3,40×10−9,1,50×10−8, dan 5,11×10−8 respectively. These three result show conformity between the experiment and prediction result using ANN. ANN can use to predict thermal conductivity of soil with vegetation cover. The largest thermal conductivity of bare soil is clay soil with 3.17 Wm-1K-1, and for soil with green seed is sandy soil with 2.41 Wm-1K-1.