An Artificial Neural Network Model Based on Experimental Measurements for Estimating the Grounding Resistance

In grounding systems established in rocky and sandy lands where contact resistance with metal electrodes is high, contact resistance is generally the most critical parameter that changes the total grounding resistance value. Therefore, the nonlinear variation of the earth contact resistance according to the soil type cannot be taken into account in determining the grounding resistance with the traditional mathematical formulas given theoretically. This reduces the accuracy of grounding resistance determination. In this study, experimental measurements were made according to soil types and a data set was created. Then, to estimate the total grounding resistance of complex grounding systems, a classification was made using the multi-layer sensor (MLP) type ANN algorithm and the successful results were reported. Thus, according to the data set prepared based on experimental measurements, the proposed general classification algorithm approach can be applied to any grounding system. It presents a different technique from the previous literature as a pre-feasibility study for estimating the grounding resistance, especially before the grounding system installation, which is an early stage of the design process.