Method for evaluating rheological properties of puree based on machine learning in combination with optical properties and Monte Carlo simulation

Through the single integral sphere system and Monte Carlo simulation combined with support vector machine algorithm, the rheological characteristics of apple fruit puree are quickly detected, solving the expensive and time-consuming problems of traditional detection methods and achieving precise control of fruit puree quality.

CN118937154APending Publication Date: 2024-11-12NANJING AGRICULTURAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410961460.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing technology cannot quickly and accurately detect the rheological characteristics of apple fruit puree, which makes it difficult to achieve accurate quality control of fruit puree. Traditional equipment is expensive and has a long test time, so it cannot meet the production needs of high-quality fruit puree.

Method used

The single integral sphere system combined with the reverse doubling integration method is used to obtain the optical characteristic signal of apple fruit puree. Through Monte Carlo simulation and support vector machine algorithm analysis, the detection method of apple fruit puree rheology characteristics is constructed, the detection steps are simplified, and the rapid prediction is achieved. Rheological characteristics.

Benefits of technology

The mechanism of influence of particle concentration and size of fruit mud tissue on rheological characteristics is clarified, and the theoretical basis for optical detection technology is provided, which realizes rapid and stable rheological characteristics detection of fruit mud, which reduces detection cost and time and improves the quality of fruit mud. Control level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118937154A_ABST
    Figure CN118937154A_ABST
Patent Text Reader

Abstract

The invention relates to a method for evaluating rheological properties of puree based on machine learning in combination with optical properties and Monte Carlo simulation, and the method comprises the following steps: obtaining light absorption and light scattering signals of apple puree in a short-wave near-infrared band under different tissue particle structures (particle concentration and particle size) and Monte Carlo simulation results under 1050nm; after collected optical signals are preprocessed through a standard normal transformation method, an evaluation model of apple rheological characteristics (consistency coefficient, apparent viscosity, elastic modulus, viscous modulus, yield stress and loss factor) is constructed based on a support vector machine algorithm, and effective prediction of the apple rheological characteristics is achieved only through combination of optical characteristics and Monte Carlo simulation. According to the method, the rheological characteristics of the puree in the hot processing process can be evaluated by combining light absorption and scattered signal acquisition of the puree with Monte Carlo simulation, and technical support is provided for a puree processing enterprise to realize rapid, efficient and stable product texture quality monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of spectral analysis and relates to the detection of apple puree quality by applying short-wave infrared technology. Background Art

[0002] Fruit puree is a healthy complementary food for special groups such as infants, the elderly, etc. The rheological properties are of extremely important practical significance in the processing and manufacturing of fruit puree. They not only affect the processing and filling efficiency, but also directly affect the sensory quality and quality safety of the product. The main evaluation parameters of the rheological properties of fruit puree are apparent viscosity, elastic modulus G′ and viscous modulus G″. In essence, the morphology of pulp tissue particles (size, volume, surface area, particle size distribution) and tissue particle concentration have a significant influence on the rheological properties of fruit puree and are the primary determining factors. Although the dynamic shear rheometer can accurately measure the rheological parameters of fruit puree, the equipment is expensive, the test time is long, and it cannot achieve rapid real-time detection, which cannot meet the needs of precise production of high-quality fruit puree. Therefore, real-time, rapid and accurate detection of the viscoelasticity of fruit puree has important scientific research value and application prospects.

[0003] Near-infrared (NIR) spectroscopy is an important means for rapid quality identification of agricultural products and processed foods. At present, rapid and efficient quality prediction of the chemical quality characteristics of apple puree (soluble solids, dry matter, and titratable acidity) has been successfully achieved based on NIR spectroscopy. However, the precise detection of the rheological properties of puree (apparent viscosity, viscoelasticity) has not yet been effectively achieved based on NIR spectroscopy. The fundamental reason is the lack of in-depth exploration of the detection mechanism for the intrinsic relationship between the internal physical properties of puree and the optical response, resulting in poor prediction accuracy and stability. This is a technical bottleneck for achieving stable and precise processing control of puree texture characteristics. When light enters the puree sample, it mainly undergoes absorption and multiple scattering, among which the light absorption coefficient (μ a ) can reflect the change of chemical composition, while the light scattering coefficient (μ' s ) can characterize tissue morphology and structural characteristics. This project constructs a fruit puree component simulation system to analyze the rheological properties of fruit puree and the optical properties of tissue particle size and tissue particle concentration. It also uses Monte Carlo simulation to accurately obtain the energy distribution parameters of light in the process of fruit puree tissue transmission. It develops a new method for rapid and accurate quantitative detection of rheological properties of fruit puree based on machine learning algorithms, which is of great significance to improving the level of precision control of fruit puree quality in my country. Summary of the invention

[0004] The purpose of the present invention is to provide a novel method for rapidly detecting rheological properties of apple puree. The optical property signal of apple puree under tissue particle size and concentration is obtained by using a single integrating sphere system combined with an inverse doubling integration method (IAD), and rheological parameters of puree tissue (consistency coefficient, non-Newtonian index, apparent viscosity, elastic modulus, viscous modulus, yield stress, loss factor) are collected, and preprocessed by standard normal transformation (SNV), combined with the energy distribution results of Monte Carlo simulation, and an optical correlation parameter system of apple puree flow characteristics is constructed by using a support vector machine (SVM) algorithm to analyze and construct, so as to realize the detection of rheological properties of apple puree by combining the optical properties of apple puree with Monte Carlo simulation.

[0005] The benefits of the present invention are as follows: (1) the mechanism by which the concentration and size of apple puree tissue particles affect the rheological properties of the puree is clarified, providing an important theoretical basis for controlling the quality of the puree; (2) the mechanism by which the concentration and size of apple puree tissue particles affect the optical properties of the puree is clarified, providing an important theoretical basis for the rapid determination of the rheological properties of the puree by optical detection technology; (3) a method for rapidly predicting the rheological properties of apple puree is developed, which simplifies the detection steps of the rheological parameters of apple puree. Compared with the traditional determination of the rheological parameters of the puree, the present invention saves time, reduces labor, and significantly reduces the determination cost; (4) the rheological properties of the puree are predicted only by rapid optical property characterization of the apple puree and combining it with Monte Carlo simulation, which enables food processing companies to perform rapid, efficient, and stable quality control in the apple puree processing link. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 : Abstract with attached pictures

[0007] Figure 2 : Schematic diagram of the construction of a fruit puree component simulator

[0008] Figure 3 :(a) Appearance of fruit puree with different tissue particle concentrations; (b) Appearance of fruit puree with different tissue particle sizes

[0009] Figure 4 :Changes in particle structure of puree tissue under different heating times

[0010] Figure 5 :Changes in apparent viscosity, elastic modulus, and viscous modulus of purees with different tissue particle concentrations

[0011] Figure 6 :Changes in apparent viscosity, elastic modulus, and viscous modulus of purees with different particle sizes DETAILED DESCRIPTION

[0012] A method for evaluating the rheological properties of fruit puree based on machine learning combined with optical properties and Monte Carlo simulation, the specific implementation method is as follows:

[0013] 1. Experimental Materials and Methods

[0014] Construction of fruit puree component simulator: Fresh "Golden Crown" apples were selected. After each apple was sorted and cleaned, 100 g of apple pulp was weighed and heated at 100°C for T1 (10 min), T2 (15 min), T3 (20 min), T4 (25 min), and T5 (30 min), respectively. The puree samples were prepared by thermal crushing at a mechanical crushing degree of 3000 rpm / min for 30 s. Subsequently, the purees under different thermal processing conditions were centrifuged at 7000 rpm / min for 30 min to obtain pulp (P1, P2, P3, P4, P5) and pulp (S1, S2, S3, S4, S5) with different tissue particle sizes, and the fruit puree component simulator was constructed, such as Figure 2 A total of 150 fruit puree component simulations were obtained, which can be divided into two groups: Group A (fruit puree component simulations with different tissue particle concentrations): the pulp and pulp were mixed at different proportions under the same heating time to obtain 5 fruit puree component simulations with different tissue particle concentrations, including C1 (25%), C2 (30%), C3 (35%), C4 (40%), and C5 (45%), a total of 75 samples = 5 tissue particle concentrations × 5 heat processing times × 3 repetitions. The appearance of the obtained fruit puree component simulation samples is as follows Figure 3 (a) as shown; Group B (simulated fruit puree components with different tissue particle sizes): at the same particle concentration (35%), pulp with different tissue particle sizes was mixed with fruit pulp under a single heat-processing condition to obtain 5 simulated fruit puree components with different tissue particle sizes, a total of 75 samples = 5 tissue particle sizes × 5 fruit pulp × 3 repetitions. The appearance of the prepared fruit puree samples is as follows Figure 3 (b) as shown.

[0015] Structural index determination: A high-resolution digital camera (EOS 600D, Canon, Japan, resolution: 5344 × 3516 pixels) was used to capture two-dimensional images of the fruit puree (diluted 100 times) evenly dispersed in the culture dish. After automatic thresholding and binarization, the area and perimeter of the fruit puree tissue particles were measured using Image J software to determine the tissue particle size. Figure 4 shown.

[0016] Apparent viscosity parameter determination: At 22.5°C, the apparent viscosity curve of each puree sample was measured using a viscometer equipped with a VT-31-50 rotor (RST-CC, Brookfield, USA). -1 The viscosity of the puree is measured as a function of shear rate within the shear rate range. The shear rate is recorded as 50s-1 and 100s -1 The apparent viscosity value η 50 , η 100 Based on the apparent viscosity curve, the consistency coefficient (K) and non-Newtonian index (n) were calculated according to the power law equation (1).

[0017]

[0018] Where η, K, n represents the apparent viscosity value, consistency coefficient, shear rate and non-Newtonian index respectively.

[0019] Determination of viscoelastic parameters: At 22.5°C, the elastic modulus, viscous modulus, yield stress, and loss factor of each puree sample were measured using a dynamic shear rheometer (MCR302e, Anton Paar, Austria) equipped with PP50 parallel plates. The puree was subjected to an amplitude sweep test at an angular frequency of 10 rad / s and a strain range of 0.01%-100% to obtain the linear viscoelastic range of the puree, and then the average values ​​of the elastic modulus G′ and the viscous modulus G″ within the linear viscoelastic range were calculated as the elastic modulus G′ and the viscous modulus G″ of the puree. In addition, the intersection of the elastic modulus curve and the viscous modulus curve was obtained as the yield stress point of the puree, and the loss factor of the puree was calculated according to Formula 2.

[0020]

[0021] Where tanδ, G′, and G″ are loss factor, elastic modulus, and viscous modulus, respectively.

[0022] Significance and correlation analysis: First, the mean and standard deviation of the same sample group were calculated to indicate the trend and dispersion of the data. Then, IBM SPSS 20 (The SPSS, USA) was used to analyze K, n, η at different tissue particle concentrations and sizes. 50 , η 100 One-way ANOVA was performed to determine the significance of the effects of different tissue particle sizes and concentrations on the rheological properties of the puree. Different letters after the values ​​in the same row indicate that the indicators have significant differences under different heat treatment conditions, p < 0.05.

[0023] Spectra and spectral data acquisition: A single integrating sphere system was used to measure the optical properties of all fruit puree component simulated body samples in the short-wave near-infrared band (900-1650nm). The fruit puree of each sample was placed in a quartz dish with an inner diameter of 40mm long, 40mm high, and 1.9mm wide under light-proof conditions. The absorption coefficient (μ) of the fruit puree was calculated using the reverse doubling integration method by adjusting the position of the light source and the sample and enabling the system to switch between reflection mode and transmission mode.a ) and the reduced scattering coefficient (μ s ′). Each sample was measured three times. The calculation of reflectance (R) and transmittance (T) needs to include reference noise and dark noise, as shown in equations (1) and (2).

[0024]

[0025]

[0026] Where Rs, Rd, and Rt are the reflection intensities of the sample, dark noise, and reference noise, respectively; similarly, Ts, Td, and Tt represent the transmission intensities of the sample, dark noise, and reference noise, respectively; R and T are then used in IAD 3.9.10 to calculate the optical parameters (μ a and μ s ′).

[0027] Monte Carlo simulation: The open source Monte Carlo simulation algorithm (MCXLab, USA) was used to simulate the light transmission process of measuring the optical properties of puree using a single integrating sphere to verify the feasibility of Monte Carlo simulation. Then, the light transmission energy distribution (absorption, transmission, diffuse reflection) at 1050 nm was obtained according to the various parameters of the Monte Carlo simulation set in Table 1.

[0028] Table 1 Monte Carlo simulation parameter settings

[0029]

[0030] Spectral data set: 900 spectra in total (150 puree samples × 2 optical properties × 3 replicates).

[0031] 2. Spectral preprocessing and model construction

[0032] use R2023a (The MathWorks, USA) was used to build the support vector machine (SVM) model. Before building the model, 900 spectral data were preprocessed by standard normalization (SNV). For the calibration set of the model: 4 tissue particle concentrations (C1, C2, C4, C5) and 4 tissue particle sizes (P1, P2, P4, P5) of fruit puree component simulation bodies, including 360 μ a and 360 μ s ′. For the validation set of the model: the fruit puree component simulation body at C3 tissue particle concentration and P3 tissue particle concentration, which contains 90 μ a and 90 μ s In addition, based on the combined absorption coefficient and the reduced scattering coefficient (μ a +μ s′) to build the SVM model, and based on the combined absorption coefficient, reduced scattering coefficient and Monte Carlo simulation results at 1050nm (OPs+MCX 1050nm ) to build the SVM model. The root mean square error (RMSEV) of the validation set and the prediction accuracy (R c 2 ) and the prediction accuracy of the validation set (R v 2 The determination coefficient and relative prediction deviation (RPD) were used to evaluate the prediction performance of the model.

[0033] 3. Analysis of the changing rules of puree structure and rheological parameters during thermal processing

[0034] like Figure 3 As shown in the figure, during the heating period of 10-30 minutes, the appearance of the puree did not change significantly. However, its specific tissue particle structure changed significantly ( Figure 4 As shown in Table 2, with the increase of heating time, the particle area of ​​the puree tissue increased from 0.059 mm 2 Down to 0.028mm 2 The circumference of tissue particles decreased from 0.760 to 0.668 mm, and the change was significant (p < 0.05).

[0035] Table 2 Changes in tissue size of apple puree under different heat treatment conditions

[0036]

[0037] Note: Different letters after the values ​​in the same row indicate that the index has significant difference under different heat treatment conditions, p < 0.05.

[0038] like Figure 5 As shown in the figure, the apparent viscosity of the puree increases with the increase of tissue particle concentration. Through power law equation fitting, it was found that the consistency coefficient (K) was significantly different (p < 0.05) when the tissue particle concentration was 25%, 35%, 40%, and 45%, and increased with the increase of tissue particle concentration, while the non-Newtonian index (n) showed a downward trend with the increase of tissue particle concentration. Consistent with the change of consistency coefficient, the shear rate was 50s -1 and 100s -1 The apparent viscosity value η under 50 and η 100It increased significantly from 0.87Pa and 0.54Pa to 2.95Pa and 1.61Pa, respectively (p < 0.05). For the viscoelasticity of the puree, the increase in the concentration of tissue particles caused the elastic modulus (G′) and viscosity modulus (G″) of the puree to increase significantly from 246.6Pa and 103.6Pa to 3046.5Pa and 708.2Pa, respectively (p < 0.05). In addition, when the concentration of puree tissue particles increased from 25% to 45%, the yield stress had significant differences in each concentration gradient (p < 0.05), and increased from 1.46Pa to 25.67Pa. However, the puree loss factor (tanδ) was significantly different only at a lower tissue particle concentration (25%) from other concentrations, and showed an overall downward trend.

[0039] Table 3 Changes of rheological parameters of fruit puree at different tissue particle concentrations

[0040]

[0041]

[0042] Note: Different letters after the values ​​in the same row indicate that the index has significant difference under different heat treatment conditions, p < 0.05.

[0043] As the tissue particles become smaller, the apparent viscosity and viscoelasticity of the puree both show a downward trend ( Figure 5 ). For the fitting results of the apparent viscosity curve, the K value ranges from 106.70 Pa·s n Down to 24.80Pa·s n There were significant differences among the P1, P2, P3, and P4 / P5 groups (p < 0.05), while the n value showed an upward trend and had significant differences among the P1 / P2 and P4 / P5 groups (p < 0.05). 50 , η 100 They decreased from 4.59Pa·s and 2.63Pa·s to 1.24Pa·s and 0.73Pa·s, respectively, and there were significant differences among the P1, P2 / P3, and P4 / P5 groups (p < 0.05). The viscoelastic parameters of the puree (G′, G″, yield stress) decreased with the decrease of tissue particle size, from 3281.9Pa, 712.8Pa, and 22.91Pa to 880.8Pa, 217.3Pa, and 6.97Pa, respectively. However, its tanδ did not change significantly with the change of tissue particle size. Therefore, compared with the change of tissue particle size, the concentration of tissue particles has a greater effect on the rheological properties of the puree.

[0044] Table 4 Changes in rheological parameters of fruit puree at different tissue particle sizes

[0045]

[0046] Note: Different letters after the value in the same row indicate that the index has significant differences under different heat treatment conditions, p < 0.05. 4. Comparative analysis of the results of predicting rheological properties of fruit puree based on machine learning combined with optical properties and Monte Carlo simulation

[0047] The results of predicting the rheological properties of fruit puree based on the SVM algorithm combined with optical properties and Monte Carlo simulation are shown in Table 5, where the data range represents the numerical distribution of the index in the prediction set, R c 2 It indicates the prediction accuracy of the model for the calibration set, R v 2 R represents the prediction accuracy of the model for the validation set, RMSEV represents the average prediction deviation of the model for the validation set, and RPD represents the relative deviation between the model's predicted value and the actual value. c 2 With R v 2 The larger the value is, and the smaller the RMSEV is, the better the prediction effect of the model is. The larger the RPD is, the higher the prediction accuracy of the model is. s The SVM model based on μ′ was significantly better than the one based on μ a In addition, compared with μ a +μ s The constructed SVM model is based on OPs+MCX 1050nm The constructed SVM model improves the performance of the SVM model in predicting the rheological parameters of fruit puree. s The SVM model with ′ can best predict the K(R v 2 =0.90, RMSEV=9.40, RPD=3.12), η 50 (R v 2 =0.84, RMSEV=0.35, RPD=2.53), η 100 (R v 2 =0.82, RMSEV=0.21, RPD=2.56) The results are satisfactory and have commercial value. 1050nm The constructed SVM model is effective in evaluating the fruit puree n(R v 2 =0.73, RMSEV=0.02, RPD=1.93) and obtained a relatively satisfactory result.

[0048] For the viscoelastic parameters of puree, in different SVM models, based on OPs+MCX1050nm The constructed SVM model is effective in evaluating the fruit puree G′(R v 2 =0.84, RMSEV=353.44, RPD=2.56), G″(R v 2 =0.83, RMSEV = 75.89, RPD = 2.47), yield stress (R v 2 =0.80, RMSEV=3.33, RPD=2.25) and tanδ(R v 2 =0.81, RMSEV=0.02, RPD=2.29) all obtained satisfactory results, indicating that machine learning combined with optical properties and Monte Carlo simulation has the potential to evaluate the viscoelasticity of puree.

[0049] Table 5 Prediction of rheological parameters of puree based on SVM algorithm

[0050]

[0051] In summary, the SVM model evaluation results based on optical properties and Monte Carlo simulation show that only μ a , μ s ′ characterizes the light absorption and scattering properties, and through Monte Carlo simulation of the light energy distribution in the puree, the rheological properties of apple puree can be accurately predicted.

[0052] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form, and any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for evaluating the rheological properties of fruit puree based on machine learning combined with optical properties and Monte Carlo simulation, comprising the following steps: S1. Select fresh "Golden Crown" apples, sort and clean each apple, weigh 100g of apple pulp, heat at 100℃ for T1 (10min), T2 (15min), T3 (20min), T4 (25min), T5 (30min), and perform thermal crushing at a mechanical crushing degree of 3000rpm / min for 30s to prepare fruit puree samples. Subsequently, the fruit purees under different thermal processing conditions were centrifuged at 7000rpm / min for 30min to obtain fruit pulp (P1, P2, P3, P4, P5) and pulp (S1, S2, S3, S4, S5) with different tissue particle sizes, and construct fruit puree component simulation bodies: Group A: Fruit puree component simulation bodies with different tissue particle concentrations: The fruit pulp and pulp under the same heating time were mixed in different proportions to obtain 5 groups of fruit puree component simulation bodies with different tissue particle concentrations, including C1 (2 Group B: fruit puree component simulation bodies with different tissue particle sizes: According to the same particle concentration (35%), pulps with different tissue particle sizes were mixed with pulp under a single thermal processing condition to obtain 5 groups of fruit puree component simulation bodies with different tissue particle sizes, with a total of 75 samples = 5 tissue particle sizes × 5 pulps × 3 repetitions. A total of 150 fruit puree component simulation body samples were obtained. S2. Using a single integrating sphere optical measurement system, 150 simulated samples of fruit puree components were placed in quartz cuvettes to measure the absorption coefficient (μ) of the fruit puree samples in the short-wave near-infrared (900-1650nm) band. a ) and the reduced scattering coefficient (μ s ′), and the measurement was repeated 3 times. A total of 900 spectral data (150 fruit puree samples × 2 optical properties (μ a / μ s The Monte Carlo simulation algorithm was used to simulate the diffuse reflection, absorption and transmission energy distribution at 10501 nm. S3. Use a viscometer and a dynamic shear rheometer to perform apparent viscosity curve test and amplitude sweep test to obtain 150 puree samples at a shear rate of 50s -1 and 100s -1 The apparent viscosity value η under 50 , η 100 , as well as the average elastic modulus (G′), the average viscous modulus (G″) and the yield stress value at the intersection in the linear viscoelastic region. The consistency coefficient (K) and non-Newtonian index (n) are obtained by curve fitting, and the loss factor (tanδ) is obtained by calculating the ratio of the elastic modulus to the viscous modulus. S4. Use a high-resolution camera to capture two-dimensional images of the puree tissue particle structure evenly dispersed in the culture dish. After the image is automatically thresholded and binarized, use Image J software to analyze and obtain the characteristic parameters of the puree tissue particle structure (particle area, particle perimeter). S6. Standard normal variable (SNV) preprocessing was performed on 450 absorption coefficient or reduced scattering coefficient spectral data. In the model construction, the fruit puree component simulation bodies with four different tissue particle concentrations (C1, C2, C4, C5) and the fruit puree component simulation bodies with four different tissue particle sizes (P1, P2, P4, P5) were divided into the calibration set of the model, and the fruit puree component simulation body with tissue particle concentration of C3 and the fruit puree component simulation body with tissue particle size of P3 were divided into the validation set of the model. The absorption coefficient and the reduced scattering coefficient were combined (μ a +μ s ′) and the combination of absorption coefficient, reduced scattering coefficient and Monte Carlo simulation results (OPs+MCX 1050nm ). S6. Using Matlab software, the support vector machine (SVM) algorithm was used to analyze the spectral distribution of 4 different spectra (μ a , μ s ′、μ a +μ s ′、OPs+MCX 1050nm ) to build a regression model to predict the rheological properties of fruit puree. The root mean square error (RMSEV) and the prediction accuracy of the corrected data set (R c 2 ) and the prediction accuracy of the validation dataset (R v 2 The prediction performance of the SVM model was evaluated by the coefficient of determination and relative analytical error (RPD).