Study On Acoustic Vibration Characteristics And Non-destruction Detection Of Internal Quality Of Korla Pear

Korla pear is one of the characteristic fruits of Xinjiang. Because of the difference of internal quality and rough pears caused by the producing area of pear expansion, pear grade and commodity rate had been a serious decline. Therefore, there is a highly required to solve the key scientific problems in the non-destructive detection of internal quality of Kolar pear. For this situation, a setup was put up based on the acoustic impulse response of Korla pear for non-destructive determining its stiffness. Two piezo beam transducers (PBT) were used for signal excitation and sensing. Multiple acoustic response parameters were extracted. The stability of testing system and optimum testing conditions was studied. Finally, the multiple stiffness detection models were acquired based on different response parameters. This would provide research foundation for automatic online non-destructive measurement of internal quality of Korla pear. The main contributions are as follows:(1) The original excitation signal with 2.5 V peak voltage was produced through the simulation of a rubber head hammer impacted pear at a 15 N force. In order to obtain the excitation signal, the peak voltage of original excitation signal was amplified to 80 V. The results of repeated excitation at the same point showed that the acquired response signal from the system was stable and reliable. The differences of the resonance frequency and sound propagation velocity detected from different sensing points at the pear equator were insignificant (P>0.05).(2) Three kinds of stiffness detection model were constructed based on resonance frequency detection stiffness(Sf) and sound propagation velocity detection stiffness(Sv). All the stiffness detection models showed good correlation with high coefficient. The correlation coefficient(r) of model based on Sf was 0.841 while it was 0.877,0.938 for model based on Sv and Sf-Sv, respectively. The root mean square error(RMSE) for models based on Sf, Sv and SfSv was 1.22 N/mm,1.09 N/mm and 0.80 N/mm, respectively. The sensitivity of these models based on Sf, Sv and Sf-Sv was 53.53%、56.09% and 59.43%, respectively. When different classes pear were detected, the discrimination rate for the models based on Sf,Sv and Sf-Sv was 80.0%,84.4% and 86.7%, respectively. It was found that the detection performance of model based on Sf-Sv was best among three models. Since the sensitivity and discrimination rate of the model based on Sv were close to model based on Sf-Sv, the model based on Sv was more suitable for practical commercial development.(3) Eight band magnitude parameters could be extracted in 320±880 Hz of frequency domain response signal using 73.75 Hz band division interval. Pear stiffness detection model based on eight band magnitude parameters and mass of pear was established using partial least squares regression (PLSR) method. The correlation coefficient for calibration set(rc) and validation set(rv) of model were 0.83 and 0.74, respectively. The root mean square error of calibration set(RMSEC) and validation set(RMSEV) were 1.07 N/mm andl.40 N/mm, respectively. The detection sensitivity and discrimination rate for detection different classes pear of PLSR model were 55.36% and 82.2%, respectively, which were similar to that of the model based on Sv. The discrimination rate of the PLSR model was low for rigid pears, but for soft pears, it was significantly higher than the model based on Sv.