近红外光谱对鱼糜凝胶3D打印特性和流变特性预测的研究

范东翠1,刘文超1,陈慧芝1*,张慜1,2*

1(江南大学 食品学院, 江苏 无锡,214122)2(食品科学与技术国家重点实验室(江南大学),江苏 无锡,214122)

摘 要 以不同水分含量和盐含量的鱼糜凝胶体系构建不同打印特性的实验样本,测定样本的流变特性和近红外光谱, 对比不同光谱的预处理方式,通过近红外光谱参数,建立鱼糜凝胶体系的3D打印特性定性分类预测模型和流变参数定量预测模型, 以期建立快速、无损评价食材打印特性的新型检测方法。结果表明:多元散射校正(multiplicative scatter correction,MSC)和标准正态变量变换(standard normal variate,SNV)预处理适合用于处理鱼糜凝胶体系红外光谱,以通过建立主成分分析结合线性判别,分析算法模型,从而进行其打印特性等级的判定,且判定正确率为100%;均值中心化(mean center,MC)预处理最合适应用于对鱼糜凝胶体系的流变参数K的预测,且相应偏最小二乘回归模型测试集的决定系数为0.944 6,测试均方根误差RMSEP为67.43;经过MSC联合MC和SNV联合MC预处理得到的光谱数据对流变参数G′-10、G″-10、G*-10的建模效果类似,均优于其他预处理方式。

关键词 鱼糜凝胶;近红外光谱;流变;预测模型

近红外光谱(near infrared spectrum,NIR)是介于可见光和中红外之间的电磁波,波长范围为780~2 526 nm[1]。NIR无损分析技术是通过NIR测量技术结合光学计量学建立校正模型,对未知样品进行定性或定量分析,已广泛应用于食品领域,具有技术成本低、检测速度快、不破坏样品及无需预处理等优点。便携式NIR仪器在工作原理上与传统NIR相同,但其将传统的光、机、电元件集装在芯片内部,具有微小、易携带、集成度高、智能化和成本低等特点[2]

基于食材的流变特性对挤出型3D打印特性有重要影响,然而传统流变仪法对材料流变特性表征所需时间长,样品处理复杂。本研究拟对便携式NIR技术检测鱼糜凝胶体系流变特性和打印适用性进行研究,以期建立快速、无损评价食材打印特性的新型检测方法。以不同水分含量和盐含量的鱼糜凝胶体系构建不同打印特性的实验样本,测定样本的流变特性和近红外光谱;对比不同光谱的预处理方式,通过NIR参数,建立鱼糜凝胶体系的3D打印特性定性分类预测模型和流变参数定量预测模型。

1 材料和方法

1.1 材料与试剂

冷冻鲢鱼鱼糜,中国湖北洪湖井力水产食品有限公司。

1.2 仪器与设备

DISCOVERY HR-3流变仪,美国TA仪器公司;IAS-3100便携式近红外光谱仪,无锡迅杰光远科技有限公司;美菱BCD-200MCX电冰箱,合肥美菱股份有限公司。

1.3 实验方法

1.3.1 鱼糜凝胶体系的制备

冷冻鲢鱼鱼糜,低温运输,-18 ℃冰箱保存。水分含量为(75.0±0.5) g/100 g,粗蛋白含量为(15.8±0.3) g/100 g。取解冻后的鱼糜置于斩拌机中空斩2 min,再分别添加0、1、2、3 g/100 g的食盐(以鱼糜含量计),斩拌5 min。同时加入0 ℃水,保持低温斩拌,并分别调节不同含盐量的鱼糜溶胶最终水分含量至76.5、80.0、83.5、87.0 g/100 g。最后将鱼糜溶胶装入塑料容器中,4 ℃静置12 h后用于后续打印和测试。

1.3.2 鱼糜凝胶体系流变性质的测定

使用流变仪测量鱼糜凝胶的流变性质,平板直径20 mm,测试间隙1 000 μm,测试温度25 ℃[3]。测试前,将样品加载至传感器探头内,平衡2 min以达到稳定状态[4]。在动态流变测试中,设置应变为0.3%(处于线性黏弹范围内),扫描频率为0.1~100 rad/s。记录储能模量(G′)、损耗模量(G″)、复数模量和损失正切(tanδ=G/G′)。

在静态流变测试中,剪切速率设置为0.1~100 s-1[5],当样品从平板间隙逸出即停止实验。剪切速率-黏度曲线采用幂律流体模型拟合[6],如公式(1)所示:

(1)

式中:η,特定剪切速率下的黏度,Pa·s;K,稠度系数,剪切速率,1/s;n,描述材料剪切行为的幂律指数(即非牛顿指数)。当0<n<1时,表明材料是具有剪切变稀行为的假塑性流体[7]

1.3.3 鱼糜凝胶体系的NIR检测

便携式NIR仪设备采取下照漫反射的采样方式。该仪器的有效检测波长范围为950~1 650 nm,采样间隔为1 nm。测试前,通过检测空载圆形样品池进行校准。取10 g左右的鱼糜凝胶体系平铺于圆形样品池中进行检测。

1.4 数据分析

使用SPSS 22进行判别分析(Fisher判别法)。使用Unscrambler X 10.4进行光谱预处理和偏最小二乘法(partial least squares regression,PLSR)建模,训练(建模)样本个数∶预测样本个数为2∶1。建模时选择留一法作为交叉验证方法[8]

2 结果与讨论

2.1 数据集划分

本文共获取112个有效样本(16组配方×7组重复),不同盐含量和水分含量的鱼糜凝胶体系4类打印特性样本各28个。采用Kennard-Stone方法[9]对112个有效样本划分训练集(含交叉验证集)和测试集,选取76个样本作为训练集,余下36个样本作为测试集。由于本研究采用交叉验证的方式,所以交叉验证集的划分不在此体现。在正常情况下,测试集的数据范围应大于预测集的数据范围[8]。具体样本数据集情况如表1所示。测试集中nKG′-10、G″-10、G*-10范围均在训练集范围内,所选训练集和测试集均合理。

2.2 NIR响应信号分析

由图1可知,112个鱼糜凝胶样本的NIR曲线趋势整体一致。在970、1 170、1 450 nm附近存在光谱吸收峰。LIU等[10]和CHEN等[11]分别使用相同的便携式NIR仪采集不同干燥条件下的蓝莓和紫薯泥的NIR图在类似的3个波长附近存在光谱吸收峰。其中970 nm附近的吸收峰与O—H伸缩振动的二级倍频相关[12],1 170 nm附近的吸收峰与C—H伸缩振动的二级倍频以及C—H伸缩振动和变形的合频相关[13-14],1 450 nm附近的吸收峰与O—H伸缩振动的一级倍频相关[9]。从NIR图中虽然可以得到光谱波段响应信息,但不同于中红外光谱的“指纹峰”,NIR重叠严重,光谱的解释性差,需要结合化学计量学的方法信息定性或定量分析[15]

表1 各数据集样本的主要流变数据分布情况
Table 1 Distribution of main rheological data in each data set

流变指标分组最大值最小值平均值标准偏差n训练集0.150 500.000 080.046 220.036 89测试集0.131 860.002 070.051 300.039 89K训练集1 262.97139.76673.97339.91测试集1 113.03141.72602.68290.59G′-10训练集21 293.402 123.5210 064.745 680.58测试集20 416.202 606.209 257.945 231.26G″-10训练集2 937.76233.871 390.41828.37测试集2 931.96278.511 448.10812.70G∗-10训练集21 494.302 136.3610 511.875 850.05测试集18 857.102 621.048 623.804 822.59

注:G′-10, G″-10, G*-10为角频率10 rad/s条件下的值(下同)

图1 鱼糜凝胶样品的NIR图
Fig.1 NIR spectra of surimi gel samples

2.3 NIR预处理

在采集NIR的过程中,受到环境背景引起的光散射、仪器随机噪声等非人为因素的影响,采集的光谱数据除样品自身的信息外,常常伴随着背景干扰和噪声等等一些无用冗余的信息。因此,需要对原始光谱信息进行预处理,以减少干扰信息对光谱分析产生的影响,提高所建模型的准确度和精准度。本次使用了多元散射校正(multiplicative scatter correction,MSC)、标准正态变量变换(standard normal variable transformation,SNV)、均值中心化(mean centralization,MC)、基线校正(baseline)等常用的光谱预处理方法[16-18]

(1)多元散射校正(MSC)。图2为112个鱼糜凝胶样品的NIR的MSC预处理结果。

图2 112个鱼糜凝胶样品的NIR的MSC预处理结果
Fig.2 MSC pretreatment results of NIR spectra of 112 surimi gel samples

MSC的基本原理是通过构建一个所测样本的理想光谱作为标准,对所有测量的光谱进行基线平移和偏移校正。MSC可以有效的消除由于散射水平不同带来的光谱差异,从而增强光谱与数据之间的相关性。具体算法如下,首先,计算所有光谱数据的平均值作为理想光谱,如公式(2)所示:

(2)

其次,对每个样本的光谱与平均光谱进行一元线性回归,如公式(3)所示:

(3)

最后,对每个样本的光谱进行校正,如公式(4)所示:

(4)

式中:x,光谱数据的光谱值;n,样本个数;xi,第i个样本的光谱;aibi分别表示为样本i的与平均光谱进行一元线性回归后得到的相对偏移系数和平移量。

(2)标准正态变量变换(SNV)。图3为112个鱼糜凝胶样品的NIR的SNV预处理结果。与MSC一样,SNV的实际结果是它从光谱数据中消除了散射和粒度效应的干扰。

图3 鱼糜凝胶样品的NIR的SNV预处理结果
Fig.3 SNV pretreatment results of NIR spectra of surimi gel samples

SNV的原理是对所有光谱数据进行标准化分析(基于光谱阵的行),用原始光谱值减去所有光谱的平均值,然后用得到的数据除以它们的标准差。具体算法如公式(5)所示:

(5)

式中:i,样本个数,k=1,2,3…;m,波点数;xik,第i个样品在第k个波点数处的光谱数据;i个样品的光谱在全部波点数(m)处光谱数据的平均值。

(3)基线校正(Baseline)。图4为112个鱼糜凝胶样品的NIR的Baseline预处理结果。基线校正的理论假设是当没有有效信号时,检测器获取的信号值应当为0(所谓基线),但实际上由于仪器硬件运行不可避免会带来检测器响应,进而影响目标光谱的获取,所以通过基线校正来进行噪音的近似剔除。本章用原始光谱值减去所有光谱的最小值进行基线校正。

图4 112个鱼糜凝胶样品的NIR的Baseline预处理结果
Fig.4 Baseline pretreatment results of NIR spectra of 112 surimi gel samples

(4)均值中心化(MC)。图5为112个鱼糜凝胶样品的NIR的MC预处理结果。MC的原理是每个光谱减去样品光谱的平均值,使所有光谱数据都分布在零点两侧。该预处理后的数据能更充分反映光谱的变化信息。

图5 鱼糜凝胶样品的NIR的MC预处理结果
Fig.5 MC pretreatment results of NIR spectra of surimi gel samples

2.4 基于NIR参数对鱼糜凝胶体系打印特性的判别分析

线性判别分析(linear discriminant analysis,LDA)是将高维的模式样本投影到最佳鉴别矢量空间,使样本在新空间的可分离性最大化,以达到抽取分类信息的效果。当数据集包含的变量多于样本(即光谱数据),则可以选择主成分分析结合线性判别分析(principal component analysis-linear discriminant analysis,PCA-LDA)算法。PCA是通过寻找多维数据方差最大的线性组合而实现数据降维。PCA-LDA可以相互取长补短,在高维度数据矩阵处理与分类领域中具有程序简单、效率高、结果直观等优势,广泛应用于多种样本的识别[19]。经过不同光谱预处理方法处理后的NIR数据建立的PCA-LDA模型的预测结果见表2。MC处理与原始光谱的判别正确率相同均为91.96%,而经过Baseline处理的判别正确率仅为88.39%,这可能是因为预处理在减少噪音干扰的同时也剔除了部分有效光谱信息。MSC处理和SNV处理的结果最好,判别正确率均为100%。

表2 基于NIR参数对鱼糜凝胶体系打印特性的判别分析
Table 2 Classification of printability of surimi gels using discriminant function based on the NIR parameters

光谱预处理方式潜在变量样品编号打印特性ⅠⅡⅢⅣ合计识别率/%原始光谱5Samples A0-32352882.14Samples B0-34242885.71Samples C0-32828100.00Samples D0-32428100.00合计11291.96MSC3Samples A0-32828100.00Samples B0-32828100.00Samples C0-32828100.00Samples D0-32828100.00合计112100.00SNV3Samples A0-32828100.00Samples B0-32828100.00Samples C0-32828100.00Samples D0-32828100.00合计112100.00Baseline5Samples A0-32352882.14Samples B0-32452885.71Samples C0-32542889.29Samples D0-31272896.43合计11288.39MC5Samples A0-32352882.14Samples B0-34242885.71Samples C0-32828100.00Samples D0-32428100.00合计11291.96

2.5 基于NIR参数对鱼糜凝胶体系主要流变参数的预测模型

对于NIR数据,每个样本含有上千个自变量,PLSR结合了主成分分析和多元线性回归两种算法的优势,能够提取自变量中与预测指标相关性较高的变量信息建立模型,使模型更简洁。本文使用NIR参数(950~1 650 nm波长范围的吸光度)建立鱼糜凝胶体系主要流变参数PLSR预测模型,通过光谱预处理方式对原始光谱进行校正,以期提高模型的预测精度。表3为MSC、SNV、基线校正(Baseline)、MC、MSC+MC、SNV+MC、Baseline+MC等常用的光谱预处理方法处理后,建立的PLSR模型结果。

表3 基于NIR参数对鱼糜凝胶体系主要流变参数的PLSR预测模型
Table 3 Calibration and prediction results for main rheological properties (G′, G″ and G* at 1 and 10 rad/s, K) of surimi gels by PLSR models using the NIR parameters

光谱预处理方式潜在变量R2cRMSECR2cvRMSECVR2pRMSEP原始光谱5n0.731 20.032 40.689 00.033 10.294 00.033 05K0.985 391.430.982 898.690.900 390.485G′-100.978 51 690.320.973 51 875.470.887 61 729.605G″-100.975 5252.990.970 8272.960.852 4307.885G∗-100.978 01 781.020.974 61 882.800.870 11 714.01MSC5n0.754 20.029 20.698 10.032 70.282 00.033 33K0.985 092.450.984 893.500.873 4101.964G′-100.985 81 372.740.983 51 481.880.926 11 402.193G″-100.975 8251.160.973 0364.800.830 0330.393G∗-100.982 11 605.160.977 01 828.530.906 01 450.24SNV4n0.718 60.709 10.032 40.709 10.267 00.033 73K0.985 192.130.984 693.730.874 0101.714G′-100.985 91 372.160.983 81 463.710.926 21 404.224G″-100.983 1210.080.980 7224.850.899 2254.413G∗-100.982 11 604.780.979 01 760.720.906 01 457.98Baseline5n0.752 70.029 30.703 40.032 40.282 00.033 34K0.987 285.180.985 191.860.902 689.404G′-100.981 91 553.830.979 01 670.910.905 715 999.444G″-100.978 4237.350.975 1256.800.885 1271.603G∗-100.978 41 763.630.970 02 091.850.891 51 566.61MC4n0.312 40.030 40.221 40.033 00.313 90.032 66K0.957 369.740.945 680.180.944 667.436G′-100.944 91 324.450.921 71 540.470.919 51 463.806G″-100.942 1198.080.920 9234.640.876 6281.536G∗-100.945 41 357.910.923 71 629.590.923 21 318.04MSC+MC4n0.371 50.029 10.257 20.032 30.278 80.033 43K0.941 981.420.933 986.650.896 292.313G′-100.940 91 372.340.933 71 479.070.926 11 402.143G″-100.934 9209.930.927 8222.120.899 2254.383G∗-100.937 31 454.740.931 01 528.720.924 81 303.78SNV+MC5n0.404 20.028 30.291 80.031 70.307 50.032 73K0.941 981.420.935 387.290.896 292.323G′-100.940 91 372.360.932 31 477.490.926 11 402.043G″-100.934 9209.930.926 5226.820.899 2254.403G∗-100.937 31 454.750.932 41 538.020.924 81 303.69Baseline+MC5n0.432 80.027 60.204 30.033 00.291 80.033 15K0.957 369.760.946 180.430.942 368.825G′-100.940 41 377.600.923 31 575.440.922 31 438.155G″-100.941 2199.620.927 0231.620.892 0263.355G∗-100.937 81 449.790.917 31 675.800.908 61 437.65

经过MC方式处理得到的光谱数据对流变参数K的建模效果最好,交叉验证集的决定系数为0.945 6,交叉验证均方根误差RMSECV为80.18,测试集的决定系数为0.944 6,测试均方根误差RMSEP为67.43。经过MSC+MC方式和SNV+MC方式处理得到的光谱数据对流变参数G′-10、G″-10、G*-10的建模效果类似,均优于其他预处理方式。经过MSC+MC方式和SNV+MC方式处理得到的光谱数据对流变参数G′-10建模的交叉验证集的决定系数分别为0.933 7和0.932 3,交叉验证均方根误差RMSECV分别为1 479.07和1 477.49,测试集的决定系数Rp2分别为0.926 1和0.926 1,测试均方根误差RMSEP分别为1 402.14和1 402.04;经过MSC+MC方式和SNV+MC方式处理得到的光谱数据对流变参数G″-10建模的交叉验证集的决定系数分别为0.927 8和0.926 5,交叉验证均方根误差RMSECV分别为222.12和226.82,测试集的决定系数分别为0.899 2和0.899 2,测试均方根误差RMSEP分别为254.38和254.40;经过MSC+MC方式和SNV+MC方式处理得到的光谱数据对流变参数G*-10建模的交叉验证集的决定系数分别为0.931 0和0.932 4,交叉验证均方根误差RMSECV分别为1 528.72和1 538.02,测试集的决定系数分别为0.924 8和0.924 8,测试均方根误差RMSEP分别为1 303.78和1 303.69。对于流变参数n,原始光谱数据和所有预处理后的光谱数据的预测效果均不佳。本研究团队前期也建立了基于NIR参数的模型来预测紫薯泥的流变参数[11],从而实现土豆泥3D打印特性的间接预测。

3 结论

本文主要研究了利用NIR检测鱼糜凝胶体系的打印特性的分类识别和主要流变参数的预测效果,对比了不同光谱预处理对模型的预测效果。结果表明,经过MSC和SNV预处理后,建立的PCA-LDA模型对鱼糜凝胶体系的打印特性等级进行判别,判别效果最好,判别正确率均为100%。经过MC预处理后,建立的PLSR模型对鱼糜凝胶体系的流变参数K预测效果最好,测试集的决定系数为0.944 6,测试均方根误差RMSEP为67.43;经过MSC联合MC方式和SNV联合MC方式预处理得到的光谱数据对流变参数G′-10、G″-10、G*-10的建模效果类似,均优于其他预处理方式。

参考文献

[1] 徐广通, 袁洪福, 陆婉珍.现代近红外光谱技术及应用进展[J].光谱学与光谱分析, 2000, 20(2):134-142.

XU G T, YUAN H F, LU W Z.Development of modern near infrared spectroscopic techniques and its applications[J].Spectroscopy and Spectral Analysis, 2000, 20(2):134-142.

[2] 刘建学, 尹晓慧, 韩四海, 等.便携式近红外光谱仪研究进展[J].河南农业大学学报, 2019, 53(4):662-670.

LIU J X, YIN X H, HAN S H, et al.Review of portable near-infrared spectrometers[J].Journal of Henan Agricultural University, 2019, 53(4):662-670.

[3] LIU Z B, ZHANG M, YANG C H.Dual extrusion 3D printing of mashed potatoes/strawberry juice gel[J].LWT, 2018, 96:589-596.

[4] LIU Z B, ZHANG M, BHANDARI B.Effect of gums on the rheological, microstructural and extrusion printing characteristics of mashed potatoes[J].International Journal of Biological Macromolecules, 2018, 117:1 179-1 187.

[5] YANG F L, GUO C F, ZHANG M, et al.Improving 3D printing process of lemon juice gel based on fluid flow numerical simulation[J].LWT, 2019, 102:89-99.

[6] LIU Z B, BHANDARI B, PRAKASH S, et al.Linking rheology and printability of a multicomponent gel system of carrageenan-xanthan-starch in extrusion based additive manufacturing[J].Food Hydrocolloids, 2019, 87:413-424.

[7] SWEENEY M, CAMPBELL L L, HANSON J, et al.Characterizing the feasibility of processing wet granular materials to improve rheology for 3D printing[J].Journal of Materials Science, 2017, 52(22):13 040-13 053.

[8] FENG L, ZHANG M, BHANDARI B, et al.Determination of postharvest quality of cucumbers using nuclear magnetic resonance and electronic nose combined with chemometric methods[J].Food and Bioprocess Technology, 2018, 11(12):2 142-2 152.

[9] DE LIMA G F, ANDRADE S C, DA SILVA V H, et al.Multivariate classification of UHT milk as to the presence of lactose using benchtop and portable NIR spectrometers[J].Food Analytical Methods, 2018, 11(10):2 699-2 706.

[10] LIU W C, ZHANG M, BHANDARI B, et al.A novel combination of LF-NMR and NIR to intelligent control in pulse-spouted microwave freeze drying of blueberry[J].LWT, 2021, 137:110455.

[11] CHEN J L, ZHANG M, DEVAHASTIN S, et al.Novel alternative use of near-infrared spectroscopy to indirectly forecast 3D printability of purple sweet potato pastes[J].Journal of Food Engineering, 2021, 296:110464.

[12] HUANG Y P, LU R F, CHEN K J.Prediction of firmness parameters of tomatoes by portable visible and near-infrared spectroscopy[J].Journal of Food Engineering, 2018, 222:185-198.

[13] SILVA L C R, FOLLI G S, SANTOS L P, et al.Quantification of beef, pork, and chicken in ground meat using a portable NIR spectrometer[J].Vibrational Spectroscopy, 2020, 111:103158.

[14] 魏雨晴, 王毓宁, 李绍佳, 等.基于自制便携式近红外光谱仪的枇杷果实可溶性固形物无损检测及年度重复验证[J].浙江大学学报(农业与生命科学版), 2020, 46(1):119-125.

WEI Y Q, WANG Y N, LI S J, et al.Non-destructive detecting and annual duplicate verification of loquat fruit’s total soluble solids based on self-developed portable near-infrared spectrometer[J].Journal of Zhejiang University (Agriculture and Life Sciences), 2020, 46(1):119 -125.

[15] 邹涛. 基于便携式近红外光谱仪快速检测大豆蛋白优化方法的研究[D].镇江:江苏大学, 2019.

ZOU T.Study on optimization method for rapid detection of soybean protein based on portable near infrared spectrometer[D].Zhenjiang:Jiangsu University, 2019.

[16] 李鸿博. 基于近红外光谱的红松子品质检测模型研究[D].沈阳:东北林业大学, 2021.

LI H B.Research on the quality detection model of red pine nuts based on near-infrared spectroscopy[D].Shenyang:Northeast Forestry University, 2021.

[17] 刘泽宇. 基于便携式近红外光谱和嗅觉可视化的啤酒分类与品质检测研究[D].镇江:江苏大学, 2019.

LIU Z Y.Beer classification and qualification detection based on portable near infrared spectroscopy and olfactory visualization[D].Zhenjiang:Jiangsu University, 2019.

[18] 吴磊. 基于计算机视觉和近红外光谱技术的鱼新鲜度检测方法研究[D].镇江:江苏大学, 2013.

WU L.Study on detection method for fish freshness based on computer vision and near infrared spectroscopy[D].Zhengjiang:Jiangsu University, 2013.

[19] ZHANG L, WANG S S, DING Y F, et al.Discrimination of transgenic rice based on near infrared reflectance spectroscopy and partial least squares regression discriminant analysis[J].Rice Science, 2015, 22(5):245-249.

3D printing characteristics and rheological properties of surimi gel based on near infrared spectroscopy

FAN Dongcui1, LIU Wenchao1, CHEN Huizhi1*, ZHANG Min1,2*

1(School of Food Science and Technology, Jiangnan University, Wuxi 214122, China) 2(The State Key Laboratory of Food Science and Technology, Jiangnan University, Wuxi 214122, China)

ABSTRACT 3D printing is a technology based on digital model, which produces pre-designed 3D objects by layering adhesive materials. Different from the traditional manufacturing methods, 3D printing technology has been greatly developed in recent years due to its potential advantages, such as customized geometry, reduced production cost, shortened manufacturing cycle and almost unconstrained appearance complexity, which are considered to be likely to promote the third industrial revolution. Food 3D printing is a new food processing technology. Due to the advantages of functionalization, customization, personalized nutrition design, simplified supply chain and broadening existing food materials, 3D printing has been widely studied in food industry in the past decade. At present, there are four types of 3D printing technologies for food processing: extrusion-based printing, selective laser sintering, adhesive spraying and inkjet printing. Among them, 3D printing based on extrusion is the most widely used, which can be further divided into room temperature extrusion, melt deposition manufacturing and gel forming extrusion according to different printing material states. In the extrusion process, the rheological properties of materials are very important for providing proper extrudability, bonding different food layers together and supporting the weight of the deposited layer. Gels with fast and reversible modulus are suitable for direct writing in 3D printing, because they are easily extruded from the nozzle tip and can maintain enough mechanical integrity to support the next printing layer without deformation. In food printing, soft materials such as dough and meat paste have been used to print 3D objects. Apparent viscosity is an important factor, which should be low enough to allow an easy extrusion process and high enough to adhere to previously deposited layers. Thus, it can be seen that the rheological properties of food have an important influence on the extrusion 3D printing characteristics. However, the traditional rheometer method takes a long time to characterize the rheological properties of materials and the sample processing is complicated. Near infrared spectroscopy (NIR)detection has the advantages of low technical cost, high detection speed, no sample damage and no pretreatment. In the near infrared spectrum, the relative contribution of reflected and absorbed radiation depends on the chemical composition, physical parameters and microstructure of the product. Many reports show that there is a significant correlation between the concentration of chemical components and NIR spectrum estimation. The present study was aimed at establishing a novel testing method for fast and nondestructive evaluation of food printing characteristics of surimi gel system. The ink with different printing characteristics was built with surimi gel systems with different water content and salt content, the rheological properties and near infrared spectra of samples were measured, and the pretreatment methods of different spectra were compared. Through NIR parameters, the 3D qualitative classification prediction model of surimi gel system and the quantitative prediction model of rheological parameters were established. The results show that multivariate scattering correction and standard normal variable transformation pretreatment are suitable for processing the infrared spectrum of surimi gel system to judge its printing characteristic grade by establishing PCA-LDA model, and the correct rate of judgment is 100%. The centralized pretreatment is most suitable for predicting the rheological parameter K of surimi gel system, and the determination coefficient of the corresponding PLSR model test set is 0.9446, and the root mean square error RMSEP is 67.43. The modeling effect of rheological parameters G′-10、G″-10、G*-10 of spectral data preprocessed by means of multivariate scattering correction combined with mean centralization and standard normal variable transformation combined with mean centralization is similar and superior to other preprocessing methods.

Key words surimi gel; near infrared spectroscopy; rheology; prediction model

DOI:10.13995/j.cnki.11-1802/ts.030281

引用格式:范东翠,刘文超,陈慧芝,等.近红外光谱对鱼糜凝胶3D打印特性和流变特性预测的研究[J].食品与发酵工业,2022,48(9):163-169.FAN Dongcui, LIU Wenchao, CHEN Huizhi, et al.3D printing characteristics and rheological properties of surimi gel based on near infrared spectroscopy[J].Food and Fermentation Industries,2022,48(9):163-169.

第一作者:硕士,实验师(陈慧芝博士和张慜教授为共同通信作者,E-mail:chenhuizhiht@163.com;minlichunli@163.com)

基金项目:中国博士后科学基金资助项目(2019M651712);国家自然科学基金项目(32001777)

收稿日期:2021-11-30,改回日期:2021-12-15