An acoustic method for flow rate estimation in agricultural sprayer nozzles

Abstract

The cost of current flow rate measurement devices is quite high compared to the cost of low-end microphones. This circumstance, together with the fact that common agricultural sprayers have more than 50 nozzles, makes the use of current flow rate measurement devices cost-prohibitive. That considered, this article examines, by proposing one particular method, the feasibility of using microphones as flowmeters for nozzle tips in agricultural sprayers. The proposed method consists of the following stages: (i) acquisition of the digital acoustic data sequence, (ii) signal preprocessing, (iii) frequency domain transformation using Fast Fourier Transform (FFT) analysis, (iv) in-band power calculation, (v) power normalization, and (vi) regression or curve fitting. This method was assessed in an in-lab sprayer test bench employing 11 commercial nozzle tips at several operating flow rates within or close to those recommended by the manufacturers. The experimental results yielded, for all the tested nozzle tips, average absolute and relative Root Mean Square Error (RMSE) values always below 0.08 liters per minute (lpm) and 5%, respectively, while the overall mean absolute and relative RMSE values were lower than 0.05 lpm and 2.5%. Furthermore, for each tested nozzle tip, the Maximum Absolute Error (MAE) was always bounded below 0.3 lpm, being the absolute error lower than 0.15 lpm for 95% of the time. The accuracies when employing a high-end microphone instead of a low-end one presented no statistically significant differences. These results provide strong evidence of the feasibility of accurately estimating the nozzle tip flow rate in real time based on acoustic signals. Moreover, no significant improvements are to be expected by using a high-end microphone instead of a low-end one. However, there are still some issues that should be tackled in order to enable the application of this method in real agricultural settings.