Intelligent milk powder production monitoring and control method and system

Abstract

This application relates to the field of milk powder production management technology, and in particular to an intelligent milk powder production monitoring and control method and system. The method includes: decomposing the Raman spectrum of a milk powder sample into multiple IMFs using a modal decomposition algorithm; classifying the multiple IMFs into noise IMFs and non-noise IMFs based on the correlation between each IMF and the Raman spectrum, and denoising the noise IMFs; adding the non-noise IMFs and the denoised noise IMFs to obtain a denoised Raman spectrum; constructing the component characteristics of the milk powder sample; and comparing the component characteristics with standard component characteristics to obtain the quality monitoring results of the milk powder sample. The technical solution of this application can obtain accurate milk powder quality monitoring results.

Description

Technical Field

[0001] The present application relates to the technical field of milk powder production management, and in particular to an intelligent milk powder production monitoring and control method and system. Background Art

[0002] The dairy industry represented by infant formula powder is one of the fastest growing food industries in my country and is closely related to people's daily lives. As an important source of nutrition for infants and young children, the production quality of milk powder directly affects the life and health of infants and young children. Therefore, monitoring the quality of milk powder during the production process is an extremely important task.

[0003] The modal decomposition algorithm can decompose a signal into several IMF components. Each IMF component represents the local characteristics of the signal. The signal can be reconstructed by adding all the IMF components. The modal decomposition algorithm is often used for signal denoising.

[0004] In addition, Raman spectroscopy, as a non-destructive analytical technology, is generated based on the interaction between light and chemical bonds in materials, and can provide detailed information on the chemical structure, crystallinity and molecular interactions in samples. The patent application document with application publication number CN106770152A discloses a method for rapid identification of milk powder brands based on characteristic peaks and algorithm parameter selection, the method comprising: collecting Raman spectrum spectra of milk powder samples; pre-processing the Raman spectrum spectra; using SVM classifier to analyze the Raman spectrum spectra data of milk powder samples, establish the best characteristic peaks for milk powder brand identification, and obtain the characteristic band combination corresponding to the best recognition rate, that is, the best characteristic band combination; perform 100 grid searches on the best characteristic band combination to find the best penalty function parameter c and kernel function parameter g corresponding to the characteristic band combination, perform SVM classification and recognition test, and obtain the penalty function parameter and kernel function parameter with the highest frequency under the test with a recognition rate of 100%, that is, the best algorithm parameter; perform milk powder brand recognition under the best characteristic band combination and the best algorithm parameter.

[0005] The above method determines the content of each component in the milk powder sample according to the characteristic bands in the Raman spectrum, and then realizes milk powder brand identification based on the content of each component; however, the noise information in the Raman spectrum will lead to inaccurate content of each component in the milk powder sample, and it is impossible to accurately confirm the nutritional components in the milk powder, which will lead to inaccurate milk powder quality monitoring results. Summary of the invention

[0006] In order to solve the technical problem of inaccurate milk powder quality monitoring results, the present application provides an intelligent milk powder production monitoring and control method and system, which can obtain accurate milk powder quality monitoring results.

[0007] In a first aspect of the present application, an intelligent milk powder production monitoring and control method is provided, the monitoring and control method comprising: decomposing the Raman spectrum of the milk powder sample into multiple IMFs using a modal decomposition algorithm; dividing the multiple IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum, and denoising the noise IMF; adding the non-noise IMF and the noise IMF after denoising to obtain a denoised Raman spectrum; constructing a component feature of the milk powder sample, the component feature comprising a neighborhood feature of a key point in the denoised Raman spectrum, the key point comprising an extreme point and an inflection point; comparing the component feature with a component standard feature to obtain a quality monitoring result of the milk powder sample.

[0008] In one embodiment, the modal decomposition algorithm includes empirical mode decomposition, variational mode decomposition, ensemble empirical mode decomposition, or complementary ensemble empirical mode decomposition.

[0009] In one embodiment, dividing the multiple IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum includes: calculating the correlation between an IMF and the Raman spectrum, and in response to the correlation being greater than a correlation threshold, marking the IMF as a non-noise IMF; in response to the correlation being not greater than the correlation threshold, marking the IMF as a noise IMF.

[0010] According to the correlation between IMF and Raman spectrum, all IMFs are divided into noise IMF and non-noise IMF. Subsequently, only the noise IMF needs to be denoised, which reduces the amount of noise reduction calculation while ensuring the noise reduction effect.

[0011] In one embodiment, the denoising process of the noise IMF includes: obtaining a spectrum diagram of the noise IMF, and arranging the frequencies in descending order of energy values ​​to obtain a frequency sequence of the noise IMF; calculating the cutoff energy of each frequency in the frequency sequence, and the cutoff energy satisfies the relationship: ,in, is the frequency sequence The energy value of the frequency, is the sum of the energy values ​​of each frequency in the frequency sequence, is the frequency sequence In response to the cutoff energy of a frequency being greater than an energy threshold, the frequency and the frequency before the frequency in the frequency sequence are taken as target frequencies, the energy value of the target frequency is retained and the energy values ​​of other frequencies except the target frequency are set to 0, so as to obtain the noise IMF after denoising; wherein the energy threshold is positively correlated with the correlation of the noise IMF.

[0012] The correlation between the noise IMF and the Raman spectrum can characterize the noise content in the noise IMF, and then the energy threshold can be determined based on the noise IMF; the energy values ​​of different frequencies in the noise IMF are obtained, and the information corresponding to the frequencies with larger energy values ​​is preferentially retained until the sum of the energy values ​​of the retained frequencies is greater than the energy threshold, indicating that the retained frequency energy covers the effective information except for the noise, and the noise information in the noise IMF is accurately removed.

[0013] In one embodiment, the energy threshold of the noise IMF satisfies the relationship:

[0014] ,in is the energy reference value, is the correlation between the noise IMF and the Raman spectrum.

[0015] In one embodiment, the method for obtaining the neighborhood features of the key point includes: for a key point, calculating the distance between other key points and the key point, and selecting K other key points in order of distance from small to large as neighborhood key points; respectively forming key point pairs with the coordinate positions of the key point and each neighborhood key point, and obtaining the hash value of each key point pair; using the hash value of each key point pair as a characterization feature, and using the horizontal coordinate of the key point and the characterization feature as the neighborhood features of the key point.

[0016] The characterization features of the key points are constructed through the K neighborhood key points of the key points, which can accurately reflect the relative relationship between different chemical bonds or molecular structures in the local range of each key point. At the same time, the horizontal coordinate of the key point can characterize the type of chemical bond or molecular structure. The horizontal coordinate and the characterization features are combined to accurately quantify the neighborhood features of the key points.

[0017] In one embodiment, the method for acquiring the neighborhood features of the key point includes: taking the area within a preset distance on both sides of the key point as the neighborhood range, taking the denoised Raman spectrum within the neighborhood range as the characterization feature, and taking the characterization feature and the horizontal coordinate of the key point as the neighborhood features of the key point.

[0018] In one embodiment, the component standard feature includes neighborhood features of multiple standard key points, and the component feature is compared with the component standard feature to obtain the quality monitoring result of the milk powder sample, including: calculating the similarity between the key point in the component feature and each standard key point, if the maximum similarity is greater than the similarity threshold, taking the key point and the standard key point corresponding to the maximum similarity as a matching pair, and taking the maximum similarity as the matching degree of the matching pair; obtaining multiple matching pairs, and calculating the quality monitoring value, the quality monitoring value for: ;in, is the number of matching pairs, is the number of key points in the component features, For the In response to the quality monitoring value being greater than the monitoring threshold, the quality monitoring result is qualified, otherwise, the quality monitoring result is unqualified.

[0019] The quality monitoring value is obtained by comprehensively considering the number of matching pairs and the matching degree of each matching pair. When a matching pair can be obtained for each key point in the ingredient characteristics and the matching degree of all matching pairs is 1, the quality monitoring value is the largest, achieving accurate quantification of milk powder quality.

[0020] In one embodiment, the similarity satisfies the relationship:

[0021] ;in and Key points The horizontal axis and characterization characteristics, and Standard key points The horizontal axis and characterization characteristics, Key points in component characteristics Key points with standards The similarity.

[0022] In a second aspect of the present application, there is also provided an intelligent milk powder production monitoring and control system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent milk powder production monitoring and control method according to the first aspect of the present application is implemented.

[0023] The technical solution of this application has the following beneficial technical effects:

[0024] Firstly, the Raman spectrum of the milk powder sample is collected. In order to avoid the influence of the key points (extreme points and inflection points) caused by the noise information in the Raman spectrum on the quality monitoring results, the Raman spectrum is decomposed into multiple IMFs, and the multiple IMFs are divided into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum. The non-noise IMF contains the effective information of the Raman spectrum, while the noise IMF contains both the noise information and the effective information of the Raman spectrum. Therefore, it is necessary to denoise the noise IMF, and add the non-noise IMF and the noise IMF after denoising to obtain the denoised Raman spectrum. The extreme points and inflection points in the denoised Raman spectrum can accurately reflect the content information of each chemical bond or molecular structure in the milk powder sample; further, the key points in the denoised Raman spectrum are used to construct the component characteristics of the milk powder sample, and accurate quality monitoring results are obtained by comparing the component characteristics with the component standard characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 is a flow chart of an intelligent milk powder production monitoring and control method according to an embodiment of the present application;

[0027] Figure 2 is a schematic diagram of a Raman spectrum according to an embodiment of the present application;

[0028] Figure 3 is a schematic diagram of a waveform diagram and a spectrum diagram of IMF5 according to an embodiment of the present application;

[0029] Figure 4 is a schematic diagram of a waveform diagram and a spectrum diagram of IMF3 according to an embodiment of the present application;

[0030] Figure 5 is a schematic diagram of a denoised Raman spectrum according to an embodiment of the present application;

[0031] Figure 6 It is a structural block diagram of an intelligent milk powder production monitoring and control system according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0033] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0034] According to the first aspect of the present application, the present application provides an intelligent milk powder production monitoring and control method. Figure 1 1 is a flow chart of an intelligent milk powder production monitoring and control method according to an embodiment of the present application. Figure 1As shown, the intelligent milk powder production monitoring and control method 100 includes steps S101 to S105, which are described in detail below.

[0035] S101, decomposing the Raman spectrum of the milk powder sample into multiple IMFs using a modal decomposition algorithm.

[0036] In one embodiment, during a milk powder production process, all control parameters during the production process are stored as a set of control parameter samples, and a milk powder sample after the production process is completed is obtained; and a Raman spectrum of the milk powder sample is obtained. The horizontal coordinate of the Raman spectrum is the Raman shift, the unit is the wave number, and the Raman shift can characterize the type of chemical bond or molecular structure. Different chemical bonds or molecular structures correspond to different Raman shifts; the vertical coordinate of the Raman spectrum is the Raman scattering intensity, which can reflect the number of corresponding chemical bonds or molecular structures in the milk powder sample. The greater the Raman scattering intensity, the greater the number of chemical bonds or molecular structures.

[0037] In the embodiment of the present application, the milk powder sample is goat milk powder, which includes a variety of nutrients such as isomaltooligosaccharide, polydextrose, lactoferrin, oligofructose and arachidonic acid, and the type and quantity of chemical bonds or molecular structures contained in each nutrient are different, which can be reflected in the Raman spectrum of the milk powder sample.

[0038] In actual scenarios, when collecting Raman spectra of milk powder samples, noise is inevitable. Figure 2 , is a schematic diagram of a Raman spectrum according to an embodiment of the present application.

[0039] In one embodiment, in order to ensure accurate milk powder quality monitoring results, it is necessary to remove the noise in the Raman spectrum; first, the Raman spectrum of the milk powder sample needs to be decomposed into multiple IMFs using a modal decomposition algorithm. The modal decomposition algorithm includes empirical mode decomposition, variational mode decomposition, ensemble empirical mode decomposition or complementary ensemble empirical mode decomposition; the IMF (Intrinsic Mode Functions) is the signal components of each layer obtained after the Raman spectrum is modally decomposed, and each layer of signal components represents the frequency components of different frequencies in the Raman spectrum.

[0040] Exemplarily, the modal decomposition algorithm uses empirical mode decomposition to decompose the Raman spectrum into multiple IMFs.

[0041] S102, dividing the multiple IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum, and performing denoising on the noise IMFs.

[0042] In one embodiment, each IMF is a frequency component of a different frequency decomposed from a Raman spectrum, some IMFs contain a large amount of noise information, and some IMFs contain a large amount of effective information of the Raman spectrum, therefore, the multiple IMFs are divided into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum. Specifically, dividing the multiple IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum includes: calculating the correlation between an IMF and the Raman spectrum, in response to the correlation being greater than a correlation threshold, marking the IMF as a non-noise IMF, and in response to the correlation being not greater than the correlation threshold, marking the IMF as a noise IMF.

[0043] The correlation threshold is 0.6, the IMF and the Raman spectrum can be regarded as two sequences, the correlation can be calculated by the absolute value of the Pearson correlation coefficient, or by other existing methods.

[0044] In one embodiment, the non-noise IMF contains information of a specific frequency and is concentrated in the low-frequency region, and the energy value in the high-frequency region is 0; whereas in the noise IMF, both the low-frequency region and the high-frequency region have large energy values. For example, the fifth IMF is recorded as IMF5, and IMF5 is a non-noise IMF, see Figure 3 , is a schematic diagram of the waveform diagram and spectrum diagram of IMF5 according to an embodiment of the present application. It can be seen that the frequency of IMF5 is mainly concentrated in the low-frequency area; the third IMF is recorded as IMF3, and IMF3 is a noise IMF. Figure 4 , is a schematic diagram of a waveform diagram and a spectrum diagram of IMF3 according to an embodiment of the present application. It can be seen that IMF3 has a large energy value in both the low-frequency region and the high-frequency region.

[0045] However, not all information in the high-frequency region is noise. Therefore, in order to ensure the accuracy of the milk powder quality monitoring results, it is necessary to remove the noise in the noise IMF and retain the effective information. Specifically, the denoising process of the noise IMF includes: obtaining the spectrum of the noise IMF, and arranging the frequencies in descending order of energy value to obtain the frequency sequence of the noise IMF; calculating the cutoff energy of each frequency in the frequency sequence, and the cutoff energy satisfies the relationship:

[0046] ,in, is the frequency sequence The energy value of the frequency, is the sum of the energy values ​​of each frequency in the frequency sequence, is the frequency sequence In response to the cutoff energy of a frequency being greater than an energy threshold, the frequency and the frequency before the frequency in the frequency sequence are taken as target frequencies, the energy value of the target frequency is retained and the energy values ​​of other frequencies except the target frequency are set to 0, so as to obtain the noise IMF after denoising; wherein the energy threshold is positively correlated with the correlation of the noise IMF.

[0047] The smaller the correlation between the noise IMF and the Raman spectrum, the more noise is contained in the noise IMF. Therefore, in order to effectively remove the noise in the noise IMF, the energy threshold is set to a smaller value, which can remove more noise and ensure the denoising effect. Specifically, the energy threshold of the noise IMF satisfies the relationship: ,in is the energy reference value, is the correlation between the noise IMF and the Raman spectrum.

[0048] In this way, the denoising process of the noise IMF is completed, and the noise information in the noise IMF is removed.

[0049] S103, adding the non-noise IMF and the noise IMF after denoising to obtain a denoised Raman spectrum.

[0050] In one embodiment, all noise IMFs after denoising and all non-noise IMFs are added together to obtain a denoised Raman spectrum. The denoised Raman spectrum removes noise information in the Raman spectrum, removes key point information such as extreme points and inflection points brought by the noise information in the Raman spectrum, and ensures the accuracy of the quality monitoring results of the milk powder sample. Figure 5 , is a schematic diagram of a denoised Raman spectrum according to an embodiment of the present application.

[0051] S104, constructing a component feature of the milk powder sample, wherein the component feature includes a neighborhood feature of a key point in the denoised Raman spectrum, and the key point includes an extreme point and an inflection point.

[0052] In one embodiment, a position point in a denoised Raman spectrum has a coordinate position including a horizontal coordinate and a vertical coordinate, wherein the horizontal coordinate corresponds to the type of chemical bonds or molecular structures, and the vertical coordinate corresponds to the number of the chemical bonds or molecular structures; the extreme points and inflection points in the denoised Raman spectrum can reflect the component content of the chemical bonds or molecular structures, and therefore, the extreme points and inflection points are recorded as key points.

[0053] Specifically, the method for obtaining the neighborhood features of key points includes: for a key point, calculating the distances between other key points and the key point, and selecting K other key points in order of distance from small to large as neighborhood key points; respectively forming key point pairs with the coordinate positions of the key point and each neighborhood key point, and obtaining the hash value of each key point pair; using the hash value of each key point pair as a characterization feature, and using the horizontal coordinate of the key point and the characterization feature as the neighborhood features of the key point.

[0054] For example, K is 3, which is a key point. Corresponding to 3 neighborhood key points , and , then a total of 3 sets of key point pairs can be obtained, which are recorded as , and ;Will The hash value of is recorded as Hash1. Similarly, the hash values ​​Hash2 and Hash3 can be obtained. The key point is The neighborhood characteristics are ;in, is the horizontal coordinate of the key point, To characterize the features.

[0055] The component characteristics of the milk powder sample can be obtained by combining the neighborhood features of all key points in the denoised Raman spectrum.

[0056] In other embodiments, the method for obtaining the neighborhood feature of the key point includes: taking the area within a preset distance on both sides of the key point as the neighborhood range, taking the de-noised Raman spectrum within the neighborhood range as the characterization feature, and taking the characterization feature and the horizontal coordinate of the key point as the neighborhood feature of the key point. Wherein, the preset distance is 2.

[0057] In this way, accurate composition characteristics of the milk powder sample are obtained, which are used to reflect the number and type of chemical bonds or molecular structures of the milk powder sample, and further reflect the influencing components in the milk powder sample.

[0058] S105, comparing the ingredient characteristics with the ingredient standard characteristics to obtain the quality monitoring result of the milk powder sample.

[0059] In one embodiment, the component standard feature is the component feature of the Raman spectrum when the milk powder is of qualified quality, and the component standard feature includes neighborhood features of multiple standard key points; by comparing the matching degree between the component feature and the component standard feature, the quality monitoring result of the milk powder sample can be obtained; the greater the matching degree between the component feature and the component standard feature, the better the quality of the milk powder sample.

[0060] Specifically, the component standard feature includes neighborhood features of multiple standard key points, and the component feature is compared with the component standard feature to obtain the quality monitoring result of the milk powder sample, including: calculating the similarity between the key point in the component feature and each standard key point, if the maximum similarity is greater than the similarity threshold, taking the key point and the standard key point corresponding to the maximum similarity as a matching pair, and taking the maximum similarity as the matching degree of the matching pair; obtaining multiple matching pairs, and calculating the quality monitoring value, the quality monitoring value for:

[0061] ;in, is the number of matching pairs, is the number of key points in the component features, For the In response to the quality monitoring value being greater than the monitoring threshold, the quality monitoring result is qualified, otherwise, the quality monitoring result is unqualified.

[0062] The similarity threshold is 0.8, and the monitoring threshold is 0.8. It should be noted that the number of matching pairs is less than or equal to the number of key points in the component features.

[0063] Wherein, the similarity satisfies the relationship:

[0064] ;in and Key points The horizontal axis and characterization characteristics, and Standard key points The horizontal axis and characterization characteristics, Key points in component characteristics Key points with standards The similarity.

[0065] In one embodiment, in response to the quality monitoring result being unqualified, an abnormal reminder is issued, and all control parameters in the production process are checked to eliminate the cause of the unqualified quality monitoring result to ensure that the quality of the milk powder produced next time meets the requirements.

[0066] In other optional embodiments, Fourier transform infrared spectroscopy can be used instead of Raman spectroscopy to realize intelligent monitoring and control of milk powder production, wherein Fourier transform infrared spectroscopy analyzes molecular vibrations by measuring the absorption of infrared light by molecules, and then determines the molecular structure in the milk powder sample.

[0067] The technical principle and implementation details of the intelligent milk powder production monitoring and control method of the present application are introduced above through specific embodiments. First, the Raman spectrum of the milk powder sample is collected. In order to avoid the influence of key points (extreme points and inflection points) caused by noise information in the Raman spectrum on the quality monitoring results, the Raman spectrum is decomposed into multiple IMFs, and the multiple IMFs are divided into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum. The non-noise IMF contains the effective information of the Raman spectrum, and the noise IMF contains both the noise information and the effective information of the Raman spectrum. Therefore, it is necessary to denoise the noise IMF, and add the non-noise IMF and the noise IMF after denoising to obtain the denoised Raman spectrum. The extreme points and inflection points in the denoised Raman spectrum can accurately reflect the content information of each chemical bond or molecular structure in the milk powder sample; further, the key points in the denoised Raman spectrum are used to construct the component characteristics of the milk powder sample, and accurate quality monitoring results are obtained by comparing the component characteristics with the component standard characteristics.

[0068] According to the second aspect of the present application, the present application also provides an intelligent milk powder production monitoring and control system. Figure 6 : is a structural block diagram of an intelligent milk powder production monitoring and control system according to an embodiment of the present application. Figure 6 As shown, the system 50 includes a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent milk powder production monitoring and control method according to the first aspect of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art, so they are not described here.

[0069] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in this application may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0070] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. An intelligent milk powder production monitoring and control method, characterized in that: The monitoring and control method comprises: The Raman spectrum of the milk powder sample was decomposed into multiple IMFs using the modal decomposition algorithm; Dividing the multiple IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum, and performing denoising on the noise IMFs; The non-noise IMF and the noise IMF after denoising are added to obtain the denoised Raman spectrum; Constructing a component feature of the milk powder sample, wherein the component feature includes a neighborhood feature of a key point in the denoised Raman spectrum, wherein the key point includes an extreme point and an inflection point; Comparing the ingredient characteristics with the ingredient standard characteristics to obtain the quality monitoring result of the milk powder sample; The denoising process of the noise IMF comprises: Obtain a frequency spectrum of the noise IMF, and arrange the frequencies in descending order of energy value to obtain a frequency sequence of the noise IMF; The cutoff energy of each frequency in the frequency sequence is calculated, and the cutoff energy satisfies the relationship: ,in, is the frequency sequence The energy value of the frequency, is the sum of the energy values ​​of each frequency in the frequency sequence, is the frequency sequence In response to a frequency cutoff energy being greater than an energy threshold, the frequency and the frequency before the frequency in the frequency sequence are taken as target frequencies, the energy value of the target frequency is retained and the energy values ​​of other frequencies other than the target frequency are set to 0, to obtain a noise IMF after denoising; Wherein, the energy threshold is positively correlated with the noise IMF; Energy threshold of noise IMF Satisfies the relationship: ,in is the energy reference value, is the correlation between the noise IMF and the Raman spectrum; The component standard feature includes neighborhood features of multiple standard key points. The component feature is compared with the component standard feature to obtain the quality monitoring result of the milk powder sample, including: Calculate the similarity between the key point in the component feature and each standard key point. If the maximum similarity is greater than the similarity threshold, take the key point and the standard key point corresponding to the maximum similarity as a matching pair, and take the maximum similarity as the matching degree of the matching pair; Obtain multiple matching pairs and calculate quality monitoring values, wherein the quality monitoring values for: ;in, is the number of matching pairs, is the number of key points in the component features, For the In response to the quality monitoring value being greater than the monitoring threshold, the quality monitoring result is qualified, otherwise, the quality monitoring result is unqualified; The similarity satisfies the relationship: ;in and Key points The horizontal axis and characterization characteristics, and Standard key points The horizontal axis and characterization characteristics, Key points in composition characteristics Key points with standards The similarity.

2. An intelligent milk powder production monitoring and control method according to claim 1, characterized in that: The modal decomposition algorithm includes empirical mode decomposition, variational mode decomposition, ensemble empirical mode decomposition or complementary ensemble empirical mode decomposition.

3. The intelligent milk powder production monitoring and control method according to claim 1, characterized in that: Dividing the plurality of IMFs into noise IMFs and non-noise IMFs according to the correlation between each IMF and the Raman spectrum includes: A correlation between an IMF and the Raman spectrum is calculated, and in response to the correlation being greater than a correlation threshold, the IMF is marked as a non-noise IMF, and in response to the correlation being not greater than the correlation threshold, the IMF is marked as a noise IMF.

4. The intelligent milk powder production monitoring and control method according to claim 1, characterized in that: The method for acquiring the neighborhood features of the key points includes: For a key point, calculate the distance between other key points and the key point, and select K other key points in order of distance from small to large as neighborhood key points; The coordinate positions of the key point and each neighboring key point are respectively combined into key point pairs, and a hash value of each key point pair is obtained; The hash value of each key point pair is used as a characterization feature, and the horizontal coordinate of the key point and the characterization feature are used as neighborhood features of the key point.

5. The intelligent milk powder production monitoring and control method according to claim 1, characterized in that: The method for acquiring the neighborhood features of the key points includes: The area within a preset distance on both sides of the key point is taken as the neighborhood range, the denoised Raman spectrum within the neighborhood range is taken as the characterization feature, and the characterization feature and the horizontal coordinate of the key point are taken as the neighborhood feature of the key point.

6. An intelligent milk powder production monitoring and control system, characterized in that: It comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent milk powder production monitoring and control method according to any one of claims 1 to 5 is implemented.

Images