本文采用压电梁式传感器对香梨进行激励感测，分析了该装置采集的响应信号的稳定性，研究了频带幅值参数与香梨硬度的相关性，并采用频带幅值参数对其硬度进行主成分回归（PCR）分析，建立香梨硬度的检测模型。研究结果表明：香梨声振检测装置的时域和频域信号稳定，在320~880 Hz波段可提取8个频带幅值参数，它们与M-T检测硬度的相关系数最高仅为0.63，采用这些频带幅值PCR法构建的香梨硬度检测模型的校正集和验证集相关系数rc与rv分别为0.77和0.66，均方根误差RMSEC和RMSEV分别为1.23和1.59，对香梨硬度的检测敏感度仅为51.56%，模型检测性能较低、稳定性较差；引入香梨质量对模型修正，rc和rv分别升高到0.81和0.72，RMSEC和RMSEV分别降低为1.13和1.44，模型的检测性能和稳定性得到明显提高，对香梨硬度变化的敏感度提高至58.28%，与M-T检测法敏感度的差异仅为6.21%，可以适用于香梨硬度的无损检测。

A stable non-destructive detection method based on the band magnitude features of acoustic vibration response was investigated for evaluating the stiffness of Korla pear. An experimental setup using piezoelectric beam sensors was developed and built for obtaining band magnitude features of Korla pear. The stability of the experimental system was analyzed. The band magnitude parameters were acquired by the normalized spectrum magnitude of response signal between 320 and 880 Hz. The correlation between eight band magnitude parameters and the stiffness measured with M-T method was studied. The stiffness detection model was acquired by principal component regression (PCR) analysis of the relationships between different band magnitude parameters and stiffness. The results of repeated excitation at the same point showed that the experimental system was stable and reliable. A low correlation was found between every band magnitude parameter and stiffness and the maximum correlation coefficient was only 0.63. The stiffness detection PCR model showed a good correlation. The correlation coefficient for calibration and validation of the model were 0.77 and 0.66, respectively, while the root mean square error of calibration and validation was 1.23 and 1.59, respectively. The detection sensitivity of the model was 51.56%. The modified model with mass had higher correlation coefficients and lower root mean square errors. The detection sensitivity (58.28%) of the modified model was more similar to that of M-T method for pear stiffness measurement. Therefore, this model can be used for non-destructive measurement of Korla pear internal quality.

国家自然科学基金资助项目（31560476）；新疆兵团博士资金项目（2013BB11）