This paper presents an Artificial Neural Network (ANN) model to determine the permittivity value of glucose solution at different concentrations predicted using the reflection coefficient value (S11). An open-ended probe connected to a vector network analyzer (VNA) was used to measure the complex permittivity value of glucose solutions at different concentrations. The S11 values and permittivity of these samples were analyzed over a frequency range from 500MHz to 5GHz. 11 glucose solution samples are prepared from 0g/mL to 1g/mL or 0% to 100%. By referring to the difference in frequency, concentration, and dielectric properties, the behavior of the dielectric constant and loss factor are analyzed. To develop the ANN model, 132 data points of S11 values are used as input data and 132 data points of permittivity values are used as target data. To achieve the target accuracy, the model consists of a data set that has unbiased ANN design parameters such as network type as feed-forward back propagation, transfer function as Tan-sigmoid, number of 10 neurons in hidden layer and training algorithm as Levenberg-Marquardt Backpropagation. Validation is carried out through MATLAB software by comparing the measured value and the calculated value, in terms of accuracy the equality of values has reached more than 99% accuracy.
