Desenvolvimento de um sensor capacitivo para o monitoramento de umidade do solo

Abstract

The development of a capacitive sensor based on dielectric characteristics of soil was carried through this work. The sensor was designed as a probe. A fiberglass circuit board was chosen to simulate a parallel plate capacitor. To simulate a capacitor with characteristics similar to the ones commercially sold, within electrical isolation, the plates were isolated by varnish of the type used for isolation of car engines. A value of 106 W of the insulation was measured by a multimeter. The sensor was part of an oscillator RC provided by an integrated circuit to perform the oscillator, the 74LS122 a multivibrating oscillator which gave better responses. The RC circuit output was the input of a frequency divider. The IC TC4040 and HEF4040 both similar counters were used in order to verify the output wave forms. The second one provided a uniform and square waveform, although the first one provided an output waveform tending to triangular. The circuit was printed on the plate of the sensor in order to minimize the influence inherent wire capacitance. The data acquisition, monitoring and the probe calibration had been acquired daily, four times a day and with three readings for sixty three days. The temperature also was monitored this way for both systems. The first one installed in a 2" PVC pipe filled with air dry soil, the water drainage was monitored by weighting. In another system had been installed three sensors in a box full filled by humid soil. The calibration of the sensor was supplied by the strain gage. The time response was obtained using the PVC system, whose water was drained, the soil submitted to saturation again and data were acquired by a microprocessor-based system of data acquisition supplying an output of seconds. The factors that had influenced the sensor output had been evaluated, concluding that the temperature influences however not in so significant way as moistures. The moisture significance levels had been of up to 0,01 % contrasting with the ones of the temperature 7 %. This was valid for all the sensors and also strain gage. The statistical models that more fitted the sensors output were the multiple regression followed by the polynomial regression.