MeterFTX: Shuffled Frog Leap-Optimized Feed Forward Neural Network Applied on Transmitter Monitoring System for Multi-array Antenna Subsurface Imaging

Capacitive resistivity subsurface imaging utilizes a transceiver system for non-intrusive survey mapping of underground structures. To explore this technique more, a newly developed transmitter design was considered where electrical properties such as frequency, output voltage, power, and waveform could be altered. And to monitor its output for proper observation during experiments, a measuring circuit was designed to measure its operating frequency, current, and voltage simultaneously. MeterFTX was made possible by integrating a microcomputer into the measuring circuit. From the circuit, the microcomputer will have inputs of a DC signal and pulse signal for current amplitude and frequency measurement, respectively, which will be used for predicting the current with an artificial neural network (ANN) and a cascaded ANN for voltage prediction from the resulting current prediction. With feed forward (FFNN) and recurrent (RNN) neural networks for predictions, the FFNN models elicited a lower MSE. These FFNN models were optimized using Ant Lion Optimizer, Black-Hole-based Optimization, and Shuffled Frog Leap Algorithm (SFLA). SFLA-optimized FFNN produced the most acceptable accuracy with negligible MSE of 9.9971e-
⁸
and 2.7932e-
⁵
for current and voltage predictions, respectively. This makes MeterFTX a feasible monitoring device for its intended application avoiding multiple connections and bulky highvoltage components.