Improved piezoelectric grain cleaning loss sensor based on adaptive neuro-fuzzy inference system

• Published in: Precision agriculture Vol. 23; no. 4; pp. 1174 - 1188
• Main Authors: Jin, Mingzhi, Zhao, Zhan, Chen, Shuren, Chen, Junyi
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-08-2022; Springer Nature B.V
• Subjects: Adaptive systems; Agricultural equipment; Agriculture; Artificial neural networks; Atmospheric Sciences; Biomedical and Life Sciences; Chemistry and Earth Sciences; Circuit design; Cleaning; Combine harvesters; Computer Science; Electric potential; Error analysis; Fuzzy logic; Harvesting; Impact analysis; Inference; Information processing; Life Sciences; Measurement methods; Oligodendrocyte-myelin glycoprotein; Performance indices; Physics; Piezoelectric ceramics; Piezoelectricity; Remote Sensing/Photogrammetry; Sensors; Signal processing; Soil Science & Conservation; Square waves; Statistics for Engineering; Vibration analysis; Voltage; Grain impact; Measurement performance; ANFIS; Piezoelectric effect; Loss sensor
• ISSN: 1385-2256; 1573-1618
• DOI: 10.1007/s11119-022-09879-2

Grain cleaning loss rate is an important performance index of combine harvesters which needs to be measured in real time during the harvesting operation. To improve the measurement accuracy and range, a grain loss sensor based on piezoelectric effect and adaptive neuro-fuzzy inference system (ANFIS) was proposed. A piezoelectric ceramic was fixed on the bottom of a thin sensitive plate to detect grain impact, and the sensitive plate was fixed to a support plate with a piece of shock-absorbing rubber between them to increase the attenuation rate of the vibration generated by grain impact. Based on the analysis of the reasons that restrict the improvement of measurement performance of traditional measurement methods, a novel signal processing circuit was designed. The circuit could simultaneously measure the number and energy of grain impacts, and output the results in the form of square wave voltage and analog voltage, respectively. Variation characteristics of the two output signals under different grain impact frequencies were analyzed. Then, a grain impact frequency prediction method based on ANFIS fusion of the two signals was proposed, and the established ANFIS model was trained through the calibration tests. Finally, measurement tests were carried out, and the results indicated that the measurement errors of grain impact were less than 2.5, 3.9, 4.4, 6.5 and 9.2% with measurement ranges of 100, 200, 600, 1000 and 1500 grain/s, respectively. With increase of MOG/grain mass ratio, the measurement error of the sensor was increased gradually due to the collision interference between MOG and grain. Compared with traditional sensors, the measurement accuracy and range were both improved significantly.