Method for intelligently judging freeze-drying sublimation / analysis transformation point of fruits and vegetables

Abstract

The invention discloses a method for intelligently judging freeze-drying sublimation / analysis transformation points of fruits and vegetables, which comprises the following steps of: respectively establishing a relation model between nuclear magnetic imaging spectrum information and fruit and vegetable moisture content and a relation model between a near infrared spectrum and fruit and vegetable moisture content, predicting freeze-drying fruit and vegetable samples with unknown moisture content by utilizing the established relation models, and comparing output values of the two models to determine the freeze-drying sublimation / analysis transformation points of the fruits and vegetables. And a sublimation / analysis transformation point in the fruit and vegetable freeze-drying process is judged, a judgment result is intelligently fed back to a freeze-drying parameter control center, and intelligent control over drying parameters is carried out. According to the method, the low-frequency nuclear magnetic resonance and the near infrared spectrum are combined, so that the prediction precision of the freeze-drying sublimation / analysis transformation point of the fruits and vegetables is improved.

Description

Technical Field

[0001] The invention belongs to the field of intelligent drying control, and in particular relates to a method for determining a sublimation / analysis transformation point of freeze-dried fruits and vegetables. Background Art

[0002] Freeze drying is currently considered to be the best dehydration strategy for maintaining food quality characteristics, such as nutritional content, appearance, shape and product texture. Because the low pressure environment of the drying chamber and the low temperature of the cold trap need to be maintained for a long time, freeze drying is a time-consuming and energy-consuming process. A considerable part of the current research related to freeze drying of fruits and vegetables is devoted to the optimization of the sublimation / desorption drying stage. In fact, the control of the sublimation / desorption drying transition point (TPS-A) is crucial for optimizing drying time and maintaining product quality. Traditionally, the point at which the drying rate drops significantly is determined as the starting point of the desorption drying stage of freeze drying. This method of determining TPS-A can only be obtained by analyzing the entire data after the entire drying process is completed, which may be suitable for laboratory research, but cannot be applied to industrial production. In addition, if the drying end point is not found in time, the material will be over-dried, resulting in the degradation of heat-sensitive components and increasing unnecessary drying energy consumption. Therefore, it is very important to establish a model that can intelligently determine the drying end point of the material online during the freeze drying process.

[0003] Low-field NMR can quickly and accurately explain the moisture state inside the material. Zang et al. studied the moisture distribution of sea cucumbers during the freezing process based on low-frequency NMR and found that as the freezing time increased, the time for the complete signal decay (relaxation time) of the ready-to-eat sea cucumbers gradually shortened and the decay rate gradually decreased. This result indicates that low-frequency NMR did not detect the signal of free water in the frozen samples.

[0004] Cui Li et al. disclosed a nondestructive detection method for moisture distribution in the drying process of Salvia miltiorrhiza. The method uses raw tape to seal and coat Salvia miltiorrhiza samples with different drying times and drying temperatures, and places the coated Salvia miltiorrhiza samples at room temperature for 3 to 5 minutes. Low-field nuclear magnetic resonance technology is used to obtain T2 spectra and proton density weighted images. 10ms and 100ms in the T2 spectrum are used as boundaries, 0.1 to 10ms is bound water, 10 to 100ms is non-flowing water, and 100 to 1000ms is free water. The peak integrated area is used as the relative content of water, thereby obtaining the relative contents of bound water, non-flowing water and free water. The distribution position of moisture in the Salvia miltiorrhiza sample is obtained according to the proton density weighted image, thereby obtaining the distribution states of bound water, non-flowing water and free water in the Salvia miltiorrhiza samples with different drying times and different drying temperatures, and then obtaining the moisture distribution in the drying process of Salvia miltiorrhiza. However, this method only uses low-field nuclear magnetic resonance technology and cannot determine the frozen water state, and has a narrow application range. Summary of the invention

[0005] In order to solve the technical problems mentioned in the above background technology, the present invention proposes a method for intelligently determining the sublimation / desorption transformation point of freeze-drying of fruits and vegetables, which combines low-frequency nuclear magnetic resonance and near-infrared spectroscopy to online determine the sublimation / desorption drying behavior of fruit and vegetable materials during the freeze-drying process, and applies it to the mechanism of intelligent feedback regulation of the drying process.

[0006] In order to achieve the above technical objectives, the technical solution of the present invention is:

[0007] A method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables comprises the following steps:

[0008] S1. Establish the relationship models between NMR spectral information and moisture content of fruits and vegetables and between NIR spectrum and moisture content of fruits and vegetables respectively;

[0009] The method for establishing the relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables is as follows: taking the nuclear magnetic resonance imaging spectrum information of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, and determining the optimal relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables;

[0010] The method for obtaining the nuclear magnetic resonance imaging spectrum information of the training samples is as follows: microwave vacuum dried samples with different moisture contents are placed in a nuclear magnetic resonance detection tube for nuclear magnetic resonance imaging information sampling and detection. After image acquisition and correction, a region of interest is selected, and the reflection spectrum curve of the nuclear magnetic resonance imaging picture of the region of interest is extracted. The spectrum data obtained from the reflection spectrum curve is averaged as the average spectrum value of the sample.

[0011] The method for establishing the relationship model between near infrared spectrum and moisture content of fruits and vegetables is as follows: taking the absorbance value of the near infrared spectrum of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, and determining the optimal relationship model between the near infrared spectrum and moisture content of fruits and vegetables;

[0012] The method for obtaining the near infrared spectrum absorbance value of the training sample is as follows: taking a microwave vacuum dried sample and placing it on an infrared detection platform for sampling and detection;

[0013] S2. Predicting freeze-dried fruit and vegetable samples with unknown moisture content using the relationship model established in step S1;

[0014] S3. By comparing the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables obtained in step S2 and the output value of the relationship model between the near infrared spectrum and the moisture content of fruits and vegetables, the sublimation / analysis transformation point in the freeze-drying process of fruits and vegetables is determined;

[0015] Specifically, the nuclear magnetic resonance imaging spectrum information of the freeze-dried sample is used as the input parameter of the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables, and the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample is output; the near-infrared spectrum absorbance value of the freeze-dried sample is used as the input parameter of the relationship model between the near-infrared spectrum and the moisture content of fruits and vegetables, and the near-infrared moisture content prediction result of the freeze-dried sample is output; the coincidence moment of the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample and the near-infrared moisture content prediction result of the freeze-dried sample is used as the freeze-drying sublimation / analysis transformation point;

[0016] S4. Intelligently feed back the judgment result of step S3 to the freeze-drying parameter control center to perform intelligent control of the drying parameters.

[0017] Preferably, in step S1, the number of neurons in the hidden layer is 6-15.

[0018] Preferably, in step S1, the method for obtaining the microwave vacuum dried sample is:

[0019] S11. Sample processing: After washing the fruit and vegetable samples, filter them with a vacuum filter;

[0020] S12. Determination of initial moisture content of the sample: Take the sample obtained in step S11 and place it in a 105°C oven to dry to constant weight, and weigh it to obtain the initial moisture content of the sample;

[0021] S13. Obtaining samples with different moisture contents: placing the samples obtained in step S11 into a microwave vacuum drying chamber, taking samples once at the same interval, weighing them, and calculating their moisture contents to obtain microwave vacuum dried samples with different moisture contents.

[0022] Preferably, in step S11, the vacuum filtration time is 2 to 3 minutes.

[0023] Preferably, in step S12, 10-20 g of the sample obtained in step S11 is put into an oven at 105°C, and the constant weight means that the weight difference of the sample before and after is less than 0.02 g.

[0024] Preferably, in step S13, for obtaining microwave vacuum dried samples with different moisture contents, the microwave power is controlled at 2-3 W / g, the sampling interval is 10-15 min, and the number of microwave vacuum dried samples obtained is 100-150.

[0025] Preferably, in step S1, for acquiring the sample nuclear magnetic resonance image, the required sample mass is 3-5 g, the nuclear magnetic resonance frequency is 25 MHz, the region of interest is selected using ENVI software, and the spectral data is extracted using MATLAB software.

[0026] Preferably, in step S1, for near-infrared detection of the sample, the mass of the sample required is 20-25 g, the starting wavelength of the near-infrared detection is 900 nm, the ending wavelength is 1800 nm, and the number of detection points is 20801.

[0027] Preferably, in step S3, the method for obtaining the freeze-dried sample is: placing the washed and filtered fruit and vegetable samples in a -40°C to -80°C refrigerator and freezing them for 10 to 20 hours, taking out the frozen samples and placing them in a freeze-drying chamber for freeze-drying and dehydration, taking samples every 1 hour to determine the moisture content of the samples, and stopping sampling until the drying is completed to obtain freeze-dried samples with different moisture contents.

[0028] Preferably, the freeze-drying dehydration process is: cold trap temperature -40 to -60°C, system pressure 80 to 120 Pa, heating plate temperature 40 to 100°C, drying for 6 to 25 hours, until the moisture content of the sample reaches 3 to 8%.

[0029] The beneficial effects brought by adopting the above technical solution are:

[0030] Compared with the traditional drying process, the advantage of the present invention is that it overcomes the problem that low-field nuclear magnetic resonance cannot detect frozen water by combining with near infrared, and can monitor the arrival of sublimation / analysis transformation point in real time, improve the quality of dried products by 30% to 50% through data feedback and adjustment of drying parameters, and reduce the drying time by 5 to 6 hours. Experiments have proved that the overall accuracy of the prediction model of the present invention is good, the MSE of the data set is at the level of 10-4, the R2 is above 0.99, and the accuracy rate is higher than 90%. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0032] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0033] The present invention designs a method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables, such as Figure 1 As shown, the following steps are included:

[0034] S1. Establish the relationship models between NMR spectral information and moisture content of fruits and vegetables and between NIR spectrum and moisture content of fruits and vegetables respectively;

[0035] The method for establishing the relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables is as follows: taking the nuclear magnetic resonance imaging spectrum information of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, wherein the number of hidden layer neurons is 6 to 15, and determining the optimal relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables;

[0036] The method for obtaining the nuclear magnetic resonance imaging spectrum information of the training samples is as follows: microwave vacuum dried samples with different moisture contents are placed in a nuclear magnetic resonance detection tube for nuclear magnetic resonance imaging information sampling and detection. After image acquisition and correction, a region of interest is selected, and the reflection spectrum curve of the nuclear magnetic resonance imaging picture of the region of interest is extracted. The spectrum data obtained from the reflection spectrum curve is averaged as the average spectrum value of the sample.

[0037] For the acquisition of sample NMR images, the required sample mass is 3-5 g, the NMR frequency is 25 MHz, the region of interest is selected using ENVI software, and the spectral data is extracted using MATLAB software;

[0038] The method for establishing the relationship model between the near infrared spectrum and the moisture content of fruits and vegetables is as follows: taking the absorbance value of the near infrared spectrum of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, wherein the number of hidden layer neurons is 6 to 15, and determining the optimal relationship model between the near infrared spectrum and the moisture content of fruits and vegetables;

[0039] The method for obtaining the near infrared spectrum absorbance value of the training sample is as follows: taking a microwave vacuum dried sample and placing it on an infrared detection platform for sampling and detection;

[0040] For sample near-infrared detection, the required sample mass is 20-25 g, the starting wavelength of near-infrared detection is 900 nm, the ending wavelength is 1800 nm, and the number of detection points is 20801;

[0041] The method for obtaining the microwave vacuum dried sample is:

[0042] S11. Sample processing: After washing the fruit and vegetable samples, filter them with a vacuum filter for 2 to 3 minutes;

[0043] S12. Determination of initial moisture content of the sample: 10 to 20 g of the sample obtained in step S11 was placed in an oven at 105 ° C and dried to constant weight, where the constant weight means that the weight difference before and after the sample is less than 0.02 g, and the initial moisture content of the sample was obtained after weighing;

[0044] S13. Obtaining samples with different moisture contents: placing the samples obtained in step S11 into a microwave vacuum drying chamber, sampling once at the same interval, weighing, and calculating the moisture content, to obtain microwave vacuum dried samples with different moisture contents. For obtaining microwave vacuum dried samples with different moisture contents, the microwave power is controlled at 2 to 3 W / g, the sampling interval is 10 to 15 min, and the number of microwave vacuum dried samples obtained is 100 to 150;

[0045] S2. Predicting freeze-dried fruit and vegetable samples with unknown moisture content using the relationship model established in step S1;

[0046] S3. By comparing the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables obtained in step S2 and the output value of the relationship model between the near infrared spectrum and the moisture content of fruits and vegetables, the sublimation / analysis transformation point in the freeze-drying process of fruits and vegetables is determined;

[0047] Specifically, the nuclear magnetic resonance imaging spectrum information of the freeze-dried sample is used as the input parameter of the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables, and the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample is output; the near-infrared spectrum absorbance value of the freeze-dried sample is used as the input parameter of the relationship model between the near-infrared spectrum and the moisture content of fruits and vegetables, and the near-infrared moisture content prediction result of the freeze-dried sample is output; the coincidence moment of the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample and the near-infrared moisture content prediction result of the freeze-dried sample is used as the freeze-drying sublimation / analysis transformation point;

[0048] The freeze-dried sample is obtained by placing the washed and filtered fruit and vegetable samples in a -40°C

[0049] Freeze in a refrigerator at -80℃ for 10-20h, take out the frozen sample and place it in a freeze-drying chamber for freeze-drying and dehydration, take samples every 1h to measure the moisture content of the sample, and stop sampling until the drying is completed to obtain freeze-dried samples with different moisture contents;

[0050] The freeze-drying dehydration process is: cold trap temperature -40 to -60°C, system pressure 80 to 120 Pa, heating plate temperature 40 to 100°C, drying for 6 to 25 hours until the water content of the sample reaches 3 to 8%;

[0051] S4. Intelligently feed back the judgment result of step S3 to the freeze-drying parameter control center to perform intelligent control of the drying parameters.

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Example 1: Using the present invention to intelligently determine the sublimation / analysis transformation point of blueberry freeze-dried

[0054] After the blueberries are cleaned, they are filtered with a vacuum filter for 3 minutes; 10g of the washed sample is placed in a 105℃ oven until the weight difference between the sample and the sample is less than 0.02g, and the weighing is stopped to obtain the initial moisture content of the blueberries; the washed blueberries are placed in a microwave vacuum drying chamber, and samples are taken every 10 minutes at a microwave power of 3W / g, weighed, and the moisture content is calculated. Repeat the test to make the total number of sampling times 90 times; the washed blueberry samples are placed in a -80℃ refrigerator and frozen for 20 hours, and the samples are taken out and placed in a freeze-drying chamber for dehydration. The freeze-drying dehydration process is as follows: cold trap temperature -40℃, system pressure 80Pa, heating plate temperature 40℃, drying for 20 hours, until the sample moisture content reaches 3%. During the drying process, the sample moisture content is measured every 1 hour until the drying is completed and sampling is stopped; 3g of the obtained blueberry samples with different moisture contents (including microwave vacuum dried samples and freeze-dried samples) are placed in a nuclear magnetic detection tube for nuclear magnetic imaging sampling and detection in a 25MHz nuclear magnetic resonance. After image acquisition and correction, the blueberry sample part was manually segmented and selected as the region of interest using ENVI software, and the reflectance spectrum curve within the region of interest was calculated and extracted, and the spectral data obtained from the reflectance spectrum curve was averaged as the average spectral value of the sample. The software ENVI 4.7 and MATLAB R2019a were used for region of interest identification, spectral data extraction and calculation. 20 g of the obtained blueberry samples with different moisture contents (including microwave vacuum dried samples and freeze-dried samples) were placed on an infrared detection platform for sampling and detection, and the detection parameters were as follows: the starting wavelength of the detector was 900 nm, the ending wavelength was 1800 nm, and the number of detection points was 20801; the nuclear magnetic resonance imaging spectrum information of the microwave vacuum blueberry samples with different moisture contents determined by nuclear magnetic resonance imaging spectrum information was taken as the neural network input parameter, and the true moisture content of the corresponding blueberries was taken as the output value; the number of hidden layer neurons of the BP-ANN neural network was optimized to be 10, the transfer function, and the training function, and the optimal low-field nuclear magnetic resonance-moisture content model was determined; the absorbance value of the near-infrared spectrum of the microwave vacuum blueberry samples with different moisture contents determined by near-infrared was taken as the neural network input parameter; the true moisture content of the corresponding blueberries was taken as the output value. The water content is the output parameter of the neural network, and the number of hidden layer neurons of the BP-ANN neural network is optimized to 10, the transfer function, and the training function to determine the optimal near-infrared-water content model; the spectral information of the freeze-dried blueberry sample extracted by nuclear magnetic resonance imaging is used as the input parameter of the neural network to import it into the established low-field nuclear magnetic resonance-water content model to obtain the low-field nuclear magnetic resonance water content prediction result of the freeze-dried sample; the absorbance value of the near-infrared spectrum of the freeze-dried blueberry samples with different moisture contents measured by near infrared is used as the input parameter to import it into the established near-infrared-water content model to obtain the near-infrared water content prediction result of the freeze-dried sample; the moment when the water content of the freeze-dried blueberry sample obtained by the nuclear magnetic resonance imaging information-water content model coincides with the water content of the blueberry sample obtained by the near-infrared-water content model is determined to be the freeze-drying sublimation / analysis conversion point.

[0055] Example 2: Using the present invention to intelligently determine the sublimation / analysis transformation point of strawberry freeze-dried

[0056] After the strawberries are cleaned, they are filtered with a vacuum filter for 2 minutes; 20g of the washed sample is placed in a 105℃ oven until the weight difference between the sample before and after is less than 0.02g, and the weighing is stopped to obtain the initial moisture content of the strawberries; the washed strawberries are placed in a microwave vacuum drying chamber, and samples are taken every 15 minutes at a microwave power of 2.5W / g, weighed, and the moisture content is calculated. Repeat the test to make the total number of sampling times 100 times; the washed strawberry samples are placed in a -80℃ refrigerator and frozen for 20 hours, and the samples are taken out and placed in a freeze-drying chamber for dehydration. The freeze-drying dehydration process is as follows: cold trap temperature -40℃, system pressure 80Pa, heating plate temperature 60℃, drying for 15 hours, until the sample moisture content reaches 2.5%. During the drying process, the sample moisture content is measured every 1 hour until the drying is completed and sampling is stopped; 3g of the obtained strawberry samples with different moisture contents (including microwave vacuum drying samples and freeze-dried samples) are placed in a nuclear magnetic detection tube for nuclear magnetic imaging sampling and detection in a 25MHz nuclear magnetic resonance. After image acquisition and correction, the strawberry sample part was manually segmented and selected as the region of interest using ENVI software, and the reflectance spectrum curve within the region of interest was calculated and extracted, and the spectral data obtained from the reflectance spectrum curve was averaged as the average spectral value of the sample. The software ENVI 4.7 and MATLAB R2019a were used for region of interest identification, spectral data extraction and calculation; 20g of the obtained strawberry samples with different moisture contents (including microwave vacuum dried samples and freeze-dried samples) were placed on the infrared detection platform for sampling and detection, and the detection parameters were as follows: the starting wavelength of the detector was 900nm, the ending wavelength was 1800nm, and the number of detection points was 20801; the nuclear magnetic imaging spectrum information of microwave vacuum strawberry samples with different moisture contents measured by nuclear magnetic imaging spectrum information was taken as the neural network input parameter, and the true moisture content of the corresponding strawberry was the output value; the number of hidden layer neurons of the BP-ANN neural network was optimized to be 10, the transfer function, and the training function to determine the optimal low-field nuclear magnetic-water content model; the absorbance value of the near-infrared spectrum of microwave vacuum strawberry samples with different moisture contents measured by near-infrared was taken as the neural network The corresponding true moisture content of strawberries is used as the output parameter of the neural network, and the number of hidden layer neurons of the BP-ANN neural network is optimized to be 10, the transfer function, and the training function to determine the optimal near-infrared-moisture content model; the spectral information of the freeze-dried strawberry sample extracted by step nuclear magnetic resonance imaging is imported into the established low-field nuclear magnetic resonance-moisture content model as the input parameter of the neural network to obtain the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample; the absorbance value of the near-infrared spectrum of freeze-dried strawberry samples with different moisture contents measured by near infrared is imported into the established near-infrared-moisture content model as the input parameter to obtain the near-infrared moisture content prediction result of the freeze-dried sample; the moment when the moisture content of the freeze-dried strawberry sample obtained by the nuclear magnetic resonance imaging information-moisture content model coincides with the moisture content result of the strawberry sample obtained by the near-infrared-moisture content model is determined to be the freeze-drying sublimation / analysis conversion point.

[0057] Example 3: Using the present invention to intelligently determine the sublimation / analysis transformation point of carrot freeze-dried

[0058] After washing the carrots, use a vacuum filter to filter for 3 minutes; take 15g of the washed sample and put it into a 105℃ oven until the weight difference between the sample and the sample is less than 0.02g, stop weighing, and obtain the initial moisture content of the carrots; put the washed carrots into a microwave vacuum drying chamber, sample once every 10 minutes at a microwave power of 3W / g, weigh, and calculate its moisture content, repeat the test to make the total number of sampling times 100 times; put the washed carrot samples into a -80℃ refrigerator and freeze for 15 hours, take out the samples and place them in a freeze drying chamber for dehydration. The freeze drying and dehydration process is as follows: cold trap temperature -40℃, system pressure 80Pa, heating plate temperature 60℃, dry for 15 hours, until the sample moisture content reaches 1.5%. During the drying process, sample the sample moisture content every 1 hour until the drying is completed and stop sampling; 4g of the obtained carrots with different moisture contents (including microwave vacuum drying samples and freeze-dried samples) are placed in a nuclear magnetic detection tube for nuclear magnetic imaging sampling and detection in 25MHz nuclear magnetic. After image acquisition and correction, the carrot sample part was manually segmented and selected as the region of interest using ENVI software, and the reflectance spectrum curve within the region of interest was calculated and extracted, and the spectral data obtained from the reflectance spectrum curve was averaged as the average spectral value of the sample. The software ENVI 4.7 and MATLAB R2019a were used for region of interest identification, spectral data extraction and calculation; 23g of carrot samples with different moisture contents (including microwave vacuum dried samples and freeze-dried samples) were placed on the infrared detection platform for sampling and detection, and the detection parameters were as follows: the starting wavelength of the detector was 900nm, the ending wavelength was 1800nm, and the number of detection points was 20801; the nuclear magnetic imaging spectrum information of microwave vacuum carrot samples with different moisture contents measured by nuclear magnetic imaging spectrum information was taken as the input parameter of the neural network, and the actual moisture content of the corresponding carrot was the output value; the number of hidden layer neurons of the BP-ANN neural network was optimized to 10, the transfer function, and the training function to determine the optimal low-field nuclear magnetic-moisture content model; the absorbance value of the near-infrared spectrum of microwave vacuum carrot samples with different moisture contents measured by near-infrared was taken as the neural network Input parameters; the corresponding actual moisture content of carrots is the output parameter of the neural network, the number of hidden layer neurons of the BP-ANN neural network is optimized to 10, the transfer function, and the training function, and the optimal near-infrared-moisture content model is determined; the spectral information of the freeze-dried carrot sample extracted by nuclear magnetic resonance imaging is used as the input parameter of the neural network to import the nuclear magnetic resonance imaging information-moisture content model to obtain the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample; the absorbance value of the near-infrared spectrum of freeze-dried carrot samples with different moisture contents measured by near infrared is used as the input parameter to import it into the constructed near-infrared-moisture content model to obtain the near-infrared moisture content prediction result of the freeze-dried sample; the moment when the moisture content of the freeze-dried carrot sample obtained by the nuclear magnetic resonance imaging information-moisture content model coincides with the moisture content of the carrot sample obtained by the near-infrared-moisture content model is determined to be the freeze-drying sublimation / analysis conversion point.

[0059] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables, characterized in that: The following steps are involved: S1. Establish the relationship models between NMR spectral information and moisture content of fruits and vegetables and between NIR spectrum and moisture content of fruits and vegetables respectively; The method for establishing the relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables is as follows: taking the nuclear magnetic resonance imaging spectrum information of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, and determining the optimal relationship model between nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables; The method for obtaining the nuclear magnetic resonance imaging spectrum information of the training samples is as follows: microwave vacuum dried samples with different moisture contents are placed in a nuclear magnetic resonance detection tube for nuclear magnetic resonance imaging information sampling and detection. After image acquisition and correction, a region of interest is selected, and the reflection spectrum curve of the nuclear magnetic resonance imaging picture of the region of interest is extracted. The spectrum data obtained from the reflection spectrum curve is averaged as the average spectrum value of the sample. The method for establishing the relationship model between near infrared spectrum and moisture content of fruits and vegetables is as follows: taking the absorbance value of the near infrared spectrum of the training sample as the input parameter of the BP-ANN neural network, and the moisture content parameter corresponding to the training sample as the output parameter of the BP-ANN neural network, optimizing the number of hidden layer neurons, transfer function and training function of the BP-ANN neural network, and determining the optimal relationship model between the near infrared spectrum and moisture content of fruits and vegetables; The method for obtaining the near infrared spectrum absorbance value of the training sample is as follows: taking a microwave vacuum dried sample and placing it on an infrared detection platform for sampling and detection; S2. Predicting freeze-dried fruit and vegetable samples with unknown moisture content using the relationship model established in step S1; S3. By comparing the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables obtained in step S2 and the output value of the relationship model between the near infrared spectrum and the moisture content of fruits and vegetables, the sublimation / analysis transformation point in the freeze-drying process of fruits and vegetables is determined; Specifically, the nuclear magnetic resonance imaging spectrum information of the freeze-dried sample is used as the input parameter of the relationship model between the nuclear magnetic resonance imaging spectrum information and the moisture content of fruits and vegetables, and the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample is output; the near-infrared spectrum absorbance value of the freeze-dried sample is used as the input parameter of the relationship model between the near-infrared spectrum and the moisture content of fruits and vegetables, and the near-infrared moisture content prediction result of the freeze-dried sample is output; the coincidence moment of the low-field nuclear magnetic resonance moisture content prediction result of the freeze-dried sample and the near-infrared moisture content prediction result of the freeze-dried sample is used as the freeze-drying sublimation / analysis transformation point; S4. Intelligently feed back the judgment result of step S3 to the freeze-drying parameter control center to perform intelligent control of the drying parameters.

2. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 1, characterized in that: In step S1, the number of neurons in the hidden layer is 6 to 15.

3. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 1, characterized in that: In step S1, the method for obtaining the microwave vacuum dried sample is: S11. Sample processing: After washing the fruit and vegetable samples, filter them with a vacuum filter; S12. Determination of initial moisture content of the sample: Take the sample obtained in step S11 and place it in a 105°C oven to dry to constant weight, and weigh it to obtain the initial moisture content of the sample; S13. Obtaining samples with different moisture contents: placing the samples obtained in step S11 into a microwave vacuum drying chamber, taking samples once at the same interval, weighing them, and calculating their moisture contents to obtain microwave vacuum dried samples with different moisture contents.

4. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 3, characterized in that: In step S11, the vacuum filtration time is 2 to 3 minutes.

5. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 3, characterized in that: In step S12, 10-20 g of the sample obtained in step S11 is put into a 105° C. oven. The constant weight means that the weight difference before and after the sample is less than 0.02 g.

6. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 3, characterized in that: In step S13, for obtaining microwave vacuum dried samples with different moisture contents, the microwave power is controlled at 2-3 W / g, the sampling interval is 10-15 min, and the number of microwave vacuum dried samples obtained is 100-150.

7. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 3, characterized in that: In step S1, for obtaining the sample NMR image, the required sample mass is 3-5 g, the NMR frequency is 25 MHz, the region of interest is selected using ENVI software, and the spectral data is extracted using MATLAB software.

8. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 3, characterized in that: In step S1 , for the near-infrared detection of the sample, the mass of the sample required is 20-25 g, the starting wavelength of the near-infrared detection is 900 nm, the ending wavelength is 1800 nm, and the number of detection points is 20801.

9. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 1, characterized in that: In step S3, the method for obtaining the freeze-dried sample is: placing the washed and filtered fruit and vegetable samples in a -40°C to -80°C refrigerator and freezing them for 10 to 20 hours, taking out the frozen samples and placing them in a freeze-drying chamber for freeze-drying and dehydration, taking samples every 1 hour to measure the moisture content of the samples, and stopping sampling until the drying is completed to obtain freeze-dried samples with different moisture contents.

10. The method for intelligently determining the sublimation / analysis transformation point of freeze-dried fruits and vegetables according to claim 9, characterized in that: The freeze-drying dehydration process is as follows: cold trap temperature -40 to -60°C, system pressure 80 to 120 Pa, heating plate temperature 40 to 100°C, drying for 6 to 25 hours until the water content of the sample reaches 3 to 8%.

Images