梨质地的激光多普勒测振无损检测方法

发布时间：2018-05-22 14:27

本文选题：激光多普勒 + 梨　；参考：《浙江大学》2016年博士论文

【摘要】：梨果产业是我国的第三大果品产业,随着科技进步和人民生活水平不断提高,人们对于水果的品质要求越来越高。水果质地,反应了水果流变学和结构学上的属性,是评价水果品质一项关键指标。梨是一种呼吸跃变型水果,在成熟过程中不断软化。准确地检测梨的质地,对于确定适宜的收获期、采后保存、评估最佳食用期、产品分级以及相关果品加工机械设计等都具有重要意义。水果质地的感官评价存在主观性过强、人与人之间差异大、检测有损、人力成本较高等缺点。传统的仪器检测方法也存在检测有损、检验率低、费时费力并造成大量浪费等缺点。因此,有必要找到一种无损的检测方法实现梨的质地检测。现有的水果质地无损检测研究中,声学振动检测法是最为常用的检测方法。传统的振动测量方法需要把加速度传感器附着于待测物体表面,这种接触式的安装方式会破坏原有的振动状态,甚至在许多场合无法应用,因此限制了它的应用范围。而激光多普勒测振技术作为一种非接触式测量方法,具有精度高、动态响应快、测量范围大、不受环境噪声影响等优点,对检测振幅微小的振动也很有效,其特点满足了水果振动测量的需要。本研究综合利用机械振动学、传感技术、信号分析处理技术和化学计量学等多领域技术手段,以梨为研究对象,研究激光多普勒测振技术用于梨的质地无损检测方法。首先,本研究搭建了一套基于激光多普勒测振技术的水果振动特性检测平台。其次,进行梨的动力学分析,采用有限元谐响应分析模拟梨在振动台上的振动情况,为激光多普勒测振检测法测试水果的振动特性提供理论依据。然后,基于该检测平台,分别研究在扫频振动方式和冲击振动方式下梨质地的无损检测方法,并比较了不同建模方式下预测模型的结果。本研究的目的在于证明利用激光多普勒测振技术用于梨质地无损检测的可行性,建立基于激光多普勒测振技术的梨质地无损检测方法体系,为研发具有自主知识产权的梨质地无损检测仪器和检测线提供方法依据。主要研究结果和结论为：(1)搭建了基于激光多普勒测振技术的水果振动特性检测平台,并对扫频振动方式下的检测参数进行优化。该检测平台包括振动控制系统、振动信号采集系统和信号分析软件3部分组成。结果表明：1)不同扫频方式、扫频速率和定振加速度幅值下频率响应曲线的可重复性较好,但是在不同摆放姿态和激光多普勒测振仪检测点下频率响应曲线的可重复性较差；2)无论何种检测条件,第二共振频率五具有良好的可重复性,800、1200和1600 Hz处相移P800、P1200和P1600具有较好的可重复性；3)通过带交互作用的三因素三水平正交试验,最终确定较优的检测条件组合是：线性扫频方式,扫频速率1400 Hz/min,定振加速度幅值1g和摆放姿态正放。(2)采用基于图像处理的方式建立的梨的三维几何模型,并采用有限元谐响应分析,进行了梨的动力学分析。结果表明：1)基于模态频率的有限元模型修正方法得到的共振频率与用激光多普勒测振检测法得到的实测结果基本一致,说明有限元谐响应分析用于模拟梨在振动台上的受迫振动是可行的；2)梨在第二到第五阶共振频率正巧对应4个振型均没有发生扭转变形,发生的变形均是在水平和竖直方向上的拉压变形,并且最大形变出现在果梗端和花萼端,因此利用激光多普勒测振检测法检测这4个模态时,果梗端和花萼端是检测的较优位置；3)梨的共振频率的影响因素中,弹性模量与共振频率呈正相关关系,密度、泊松比、质量和高径比与共振频率呈负相关关系,其中密度和泊松比对共振频率的影响较小。(3)研究了梨在贮藏过程中扫频振动方式下的振动特性和质地特性的变化情况。结果表明：1)振动特征参数中,弹性系数EI随贮藏时间逐渐减小,相移随贮藏时间逐渐增大,而共振频率处幅值无明显变化规律；2)质地指标中,MT硬度、果肉硬度和果肉弹性率在贮藏过程中总体上均随时间逐渐降低,但果肉弹性率变化更为平稳且灵敏性更优。研究了不同品种梨的振动特征参数与质地指标的相关性。结果表明：1)去皮和带皮两种测试情况下得到的质地指标与振动特征参数的相关系数并没有明显的差别和规律；2)质地指标中,果肉弹性率与振动特征参数的相关性最好,而MT硬度与振动特征参数的相关性受到梨品种的影响；3)振动特征参数中,弹性系数EI与果肉弹性率相关性最好,相移次之,而共振频率处幅值与质地指标基本不存在相关性。比较了激光多普勒测振检测法与传统的M-T戳穿试验法。结果表明：激光多普勒测振检测法不仅能反映梨的质地特性,而且比传统方法的可重复性和灵敏性更好。(4)对果肉弹性率进行了定量分析。在剔除振动信息异常样本和浓度异常样本的基础上,采用一元线性回归和逐步多元线性回归建立了梨的质地预测模型,并提出一种基于果形修正的建模方法,结果表明：1)逐步多元线性回归模型优于一元线性回归模型；2)引入描述果形的指标—高径比,可以提高模型的精度和预测能力,逐步多元线性回归模型引入高径比修正后,校正集相关系数(rc)从0.850提升至0.862,校正均方根误差(Root mean square error of calibration, RMSEC)从0.811 N/mm降低至0.781 N/mm,而验证集相关系数(rp)从0.760提升至0.780,预测均方根误差(Root mean square error of prediction, RMSEP)从0.954 N/mm降低至0.914 N/mm。结果表明,在扫频振动方式下激光多普勒测振技术可用于梨的质地无损检测。此外,初步分析了激光多普勒测振检测法用于玉冠、黄金和丰水3品种梨和用于2013-2015年份的丰水梨,结果表明,激光多普勒测振检测法对于不同品种和不同年份的梨都具有较好的普适性。(5)初步研究了冲击振动方式下激光多普勒测振技术用于梨的质地和新鲜度检测。结果表明：1)采用半正弦脉冲信号对梨进行激励,当冲击加速度幅值在4g以内时都不会对梨造成损伤；2)从梨的冲击振动响应信号提取的15个时域特征参数和8个频域特征参数具有良好的可重复性；3)在剔除振动信息异常样本和浓度异常样本的基础上,采用逐步多元线性回归、前馈反向传播(Back-propagation, BP)神经网络和主成分—BP神经网络3种方式对质地指标进行定量分析,结果表明：主成分—BP神经网络的建模结果较优,MT硬度、果肉硬度和果肉弹性率预测模型的rp分别为0.625,0.840和0.577,RMSEP分别为1.892 N,0.959 N和0.838 N/mmm,说明冲击振动方式下激光多普勒测振技术用于梨的质地检测是可行性的；4)在剔除振动信息异常样本和浓度异常样本的基础上,采用Fisher判别、BP神经网络和主成分—BP神经网络3种方式对梨的不同新鲜度进行判别分析,结果表明：主成分—BP神经网络的判别正确率最高,校正集的判别正确率为89.0%,验证集的判别正确率为83.3%,说明冲击振动方式下激光多普勒测振技术用于梨的新鲜度判别是可行的。
[Abstract]:In this paper , it is necessary to find a nondestructive testing method to test the texture of pear . The repeatability of the frequency response curve is good at the sweep rate and the constant vibration acceleration amplitude , but the repeatability of the frequency response curve is poor under different placement attitude and the detection point of the laser Doppler vibrometer ;
2 ) No matter what detection conditions , the second resonant frequency 5 has good repeatability , and the phase shift P800 , P1200 and P1600 at 800 , 1200 and 1600 Hz have good repeatability ;
3 ) Through three - factor three - level orthogonal test with interaction , it is determined that the better combination of detection conditions is : linear sweep mode , frequency sweep rate 1400 Hz / min , constant vibration acceleration amplitude 1g and placing posture are positive . ( 2 ) The dynamic analysis of pear is carried out by using finite element harmonic response analysis .
2 ) When the resonance frequency of the second to fifth terrace happens to correspond to four modes , the torsion deformation is not generated , and the deformation occurs in the horizontal and vertical direction , and the maximum deformation occurs at the end of the fruit stalk and the end of the flower . Therefore , when the four modes are detected by using the laser Doppler vibration measurement method , the fruit stalk end and the end of the flower are the better position to be detected ;
3 ) The relation between the elastic modulus and the resonance frequency is positively correlated with the resonance frequency , and the density , Poisson ' s ratio , mass and high - diameter ratio are inversely related to the resonance frequency .
2 ) In texture index , the hardness of MT , the hardness of pulp and the elastic rate of pulp decreased gradually over time during storage , but the change of fruit flesh elasticity was more stable and more sensitive . The correlation between vibration characteristic parameters and texture index of different varieties of pear was studied .
2 ) In texture index , the correlation between the elastic rate of pulp and the parameters of vibration characteristic is the best , and the correlation between MT hardness and vibration characteristic parameter is affected by pear variety ;
3 ) In the vibration characteristic parameters , the correlation between the elastic modulus EI and the flesh elastic rate is the best , the phase shift is the second , and the amplitude and the texture index of the resonant frequency are not correlated with the texture index . The results show that the laser Doppler vibration measurement method not only can reflect the texture characteristic of the pear , but also has better repeatability and sensitivity than the traditional method .
The results show that the laser Doppler vibration measurement method can be used to detect the texture and freshness of the pear . The results show that the laser Doppler vibration measurement method can be used to detect the texture and freshness of the pear from 0.760 to 0.78N / mm .
2 ) 15 time domain characteristic parameters and 8 frequency domain characteristic parameters extracted from the impact vibration response signal of the pear have good repeatability ;
3 ) Quantitative analysis of texture indexes by stepwise multiple linear regression , forward - propagation ( BP ) neural network and main component - BP neural network was adopted on the basis of rejecting abnormal samples and concentration abnormal samples of vibration information . The results show that the model results of the main component - BP neural network are 0 . 625 , 0 . 840 and 0.577 respectively , and the RMSEP is 1.892 N , 0.959 N and 0.838 N / mm , respectively .
4 ) On the basis of rejecting abnormal sample and concentration abnormal sample of vibration information , discriminant analysis of different freshness degree of pear was carried out by means of Fisher ' s judgment , BP neural network and main component - BP neural network . The results show that the correct rate of the main component - BP neural network is the highest , the correct rate of the correction set is 89.0 % , the correct rate of the verification set is 83.3 % , and it is feasible to judge the freshness of the pear by the laser Doppler vibration measurement technique in the mode of impact vibration .
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：S661.2

【相似文献】