Peptide milk powder activity quality detection method based on wet process production
By deploying multiple sensors and constructing causal logic operation indicators during the wet production process, the problem of the disconnect between optical measurement deviation and bioactivity detection in peptide milk powder during wet production was solved, achieving high-precision activity quality detection and closed-loop control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHAANXI SHENGQUAN DAIRY TECH CO LTD
- Filing Date
- 2026-04-09
- Publication Date
- 2026-05-08
AI Technical Summary
Existing spectroscopic detection technologies cannot accurately reflect the biological activity of peptide milk powder in wet production. Furthermore, due to the distortion of microstructure, the optical scattering baseline drift and concentration measurement deviation are severe, making it impossible to achieve effective quality detection of peptide components.
By deploying multiple sensors in the wet production process to acquire fluid time series and reflectance spectral data, and combining singular value decomposition and partial least squares regression algorithms, a causal logic operation index is constructed, which is then deeply corrected to output the true active retention concentration that combines both concentration and activity attributes.
It enables precise removal of measurement biases caused by physical masking in complex industrial environments, ensuring the uniformity of the activity quality of products leaving the factory and providing rapid and accurate biological potency assessment.
Smart Images
Figure CN121994734A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of data processing technology, specifically relating to a method for detecting the activity and quality of peptide milk powder produced by wet processing. Background Technology
[0002] With the continuous development of the functional dairy industry, modified milk powder with added active small molecule peptides occupies an important position in the market. Modern dairy companies generally use wet processes to produce this type of high-quality milk powder. The complete wet process chain covers core steps such as liquid phase batching, concentrated milk heating, high-pressure pump delivery, and fluidized bed spray drying. After drying and shaping, in order to achieve rapid quality detection of peptide concentration and activity in the dry powder at the end of the production line, the industry usually uses near-infrared spectroscopy analysis technology for non-destructive testing. The concentration value is obtained by collecting spectral sequences and combining them with a partial least squares regression model.
[0003] The aforementioned conventional physicochemical testing methods suffer from optical scattering baseline drift defects caused by microstructural distortions in specific engineering scenarios of wet production. During the concentrated milk heating and high-pressure pumping stages of wet processing, small molecule peptides are prone to hydrophobic aggregation with whey proteins and other proteins due to the coupling effect of thermal stress and mechanical shear force. This change in physical state significantly increases the surface roughness of the final dried milk powder particles. When near-infrared light beams irradiate these rough particle surfaces, a strong Mie scattering effect is triggered, resulting in a severe nonlinear upward shift of the reflectance spectrum baseline across the entire wavelength range. Conventional smoothing filtering algorithms can only filter out high-frequency noise and cannot suppress the low-frequency baseline drift caused by particle morphology changes. This leads the regression model to misinterpret the enhanced scattering light path as absorption attenuation, resulting in serious omissions in concentration measurements and negative calculation biases.
[0004] Existing spectroscopic detection technologies suffer from a technical blind spot: a disconnect between physicochemical concentration parameters and biological activity characterization. The underlying detection principle of partial least squares regression models heavily relies on the chemical absorption peaks of peptide bonds within specific wavelength bands for concentration mapping. However, the biological potency of peptides depends on their complete three-dimensional conformation. The instantaneous high-temperature environment and high-pressure shear forces of the transport fluid in wet processes can cause irreversible unwinding of some peptide chains, resulting in the loss of actual biological activity, while the physical structure of the peptide bonds in the main chain remains intact. Because existing detection methods rely solely on the transient spectral characteristics of the end products, they fail to incorporate the physical parameter history of the material throughout its entire process lifecycle. This results in the final inversion output concentration value being contaminated with a large amount of inactive, inactive components, failing to objectively reflect the effective biological activity retention rate at product delivery. Summary of the Invention
[0005] This invention provides a method for detecting the activity and quality of peptide milk powder produced by wet processing, in order to solve the technical problems of optical measurement omission errors caused by the physical aggregation of micro-particles in the wet processing of peptide milk powder, and the deactivation of peptide spatial conformation induced by the comprehensive stress of complex processes, resulting in a disconnect between the apparent detection concentration and the true biological potency.
[0006] This invention provides a method for detecting the activity and quality of peptide milk powder produced by wet processing, comprising the following steps: S1, acquire the fluid time series data of the production node and the original reflectance spectrum sequence of the spray drying node, and perform smoothing filtering operation on the original reflectance spectrum sequence to obtain the smoothed spectrum after standardization. S2, import the smoothed spectrum into the pre-configured partial least squares regression algorithm model, and calculate and output the apparent chemical concentration of peptide components in the current batch of milk powder; S3, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, constructs a computational index with causal logic to complete the deep correction calculation of apparent chemical concentration, and outputs the true activity retention concentration that has both concentration and activity attributes. S4, the industrial control host receives and records the actual active retention concentration value in real time, compares it with the preset quality outline indicator bottom line to determine the test result and execute closed-loop control of the equipment.
[0007] Its effects are as follows: By integrating fluid dynamic parameters and real-time spectral data throughout the entire production process, this invention constructs a quality monitoring system with deep correction logic. It can accurately remove measurement deviations caused by physical masking in complex industrial environments and eliminate ineffective components that have chemical components but have been deactivated. This achieves a leap from simple concentration measurement to dynamic evaluation of biological potency, ensuring a high degree of uniformity in the activity quality of the products leaving the factory.
[0008] Furthermore, the fluid time-series data of the production node and the original reflectance spectral sequence of the spray drying node are acquired. The original reflectance spectral sequence is then smoothed using a smoothing filter to obtain a standardized smooth spectrum, including: Time series of fluid temperature and fluid pressure were collected at the heater pipeline and the outlet valve of the high-pressure pump. The original reflectance spectrum sequence of peptide milk powder was collected in a continuous wavelength range at the powder outlet of the fluidized bed; The smoothing filter algorithm is called to perform first-order difference differentiation and polynomial smoothing operations on the original reflectance spectrum sequence to output a smoothed spectrum.
[0009] Its effect is that by precisely deploying sensors at core process nodes such as heater pipes, high-pressure pump outlets, and fluidized bed powder outlets, and combining them with a combined filtering algorithm of first-order difference and polynomial smoothing, compared with conventional single-point sampling or basic filtering methods, it can more effectively capture the thermodynamic and pressure fluctuation characteristics of materials during drastic physical changes, significantly improve the signal-to-noise ratio of the original signal, and lay a solid and realistic physical data foundation for the subsequent construction of a high-precision multi-dimensional activity inversion model.
[0010] Furthermore, the smoothed spectrum is imported into a pre-configured partial least squares regression algorithm model to calculate and output the apparent chemical concentration of peptide components in the current batch of milk powder, including: The high-dimensional spectral absorbance matrix is reduced in dimension by singular value decomposition, and the orthogonal latent variable features are extracted. Calculate the linear regression weight matrix between latent variable characteristics and peptide mass concentration; The apparent chemical concentration of peptides in the current milk powder is obtained by calculating the smoothed spectrum using a linear regression weight matrix.
[0011] Its effect is that by using singular value decomposition and linear regression weight matrix to reduce the dimensionality of high-dimensional spectral matrix, compared with ordinary full-band spectral fitting, it can extract latent variables that are highly correlated with the chemical characteristics of peptide chains more efficiently, greatly reducing redundant computational overhead, enabling the industrial control system to provide feedback on the apparent chemical concentration of the current batch in a very short time, providing a fast and accurate initial numerical reference for real-time monitoring of the production line.
[0012] Furthermore, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time-series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes, including: The absolute value of absorbance shift within the non-absorption reference frequency band is extracted to construct the spectral scattering drift factor, and the calculation formula is as follows:
[0013] In the formula, It is a dimensionless spectral scattering drift factor. For the first in the preset frequency band Each wavelength point, This represents the total number of discrete wavelength points contained within this frequency band. This is the measured standardized smooth spectral absorbance. The absorbance is the baseline for ideal non-agglomeration.
[0014] Its effect is that by extracting the absorbance shift of the non-absorption reference frequency band to construct the spectral scattering drift factor, compared with the traditional detection scheme that ignores the change in particle morphology, it can quantitatively characterize the decrease in optical transmittance caused by particle roughening and agglomeration in wet process, thus providing a scientific basis for subsequent elimination of low-frequency baseline drift caused by particle microstructure distortion.
[0015] Furthermore, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes. This also includes: The composite stress index is constructed by nonlinearly integrating the dynamic temperature variable and the dynamic pressure variable over a time scale. The calculation formula is as follows:
[0016] In the formula, The composite stress index, This represents the total number of discrete time steps. The actual fluid temperature. The critical temperature threshold for conformational unwinding. This refers to the absolute pressure of the fluid. The standard atmospheric pressure calibration constant is used to eliminate the dimension of pressure. This is the sampling time interval constant.
[0017] Its effect is that it innovatively integrates dynamic temperature and pressure variables nonlinearly to construct a composite stress index. Compared with the scheme that only monitors transient temperature, it can more objectively restore the history of heat and energy accumulation that the material has endured throughout the entire processing cycle, thereby accurately simulating the unwinding process of the three-dimensional spatial conformation of peptide chains under extreme stress. This effectively solves the technical blind spot of existing technologies that cannot identify the damage to biological activity through spectral features.
[0018] Furthermore, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes. This also includes: The true active retention concentration is constructed by coupling apparent chemical concentration, spectral scattering drift factor, and composite stress index. The calculation relationship is as follows:
[0019] In the formula, To retain the concentration of true activity, For apparent chemical concentration, is a dimensionless optical amplification compensation gain constant. It is a dimensionless molecular stress-sensitive loss constant. The maximum allowable stress constant of the system. It is a dimensionless spectral scattering drift factor. It is the composite stress index.
[0020] Furthermore, the industrial control host receives and records the actual active retention concentration value in real time, compares it with the preset quality outline target baseline to determine the test result, and executes closed-loop control of the equipment, including: When the measured values of the true active retention concentration in multiple consecutive sampling batches show a monotonically decreasing trend and approach the lower control limit, the feedback control program is triggered. The exhaust temperature is reduced by finely adjusting the opening of the exhaust damper of the spray drying tower by sending a command voltage to the field control equipment. The output frequency of the high-pressure pump inverter is simultaneously reduced to decrease the physical shear work done during fluid transport.
[0021] Furthermore, the absolute value of absorbance shift within the non-absorption reference frequency band is extracted to construct the spectral scattering drift factor. The shift measure is output by accumulating and averaging the absolute difference between the measured absorbance and the ideal absorbance, thereby characterizing the severity of the target polypeptide molecule being physically masked by the external rough structure.
[0022] Furthermore, in constructing the composite stress index by nonlinearly integrating dynamic temperature and dynamic pressure variables over a time scale, a maximum value discrimination function is introduced to filter fluid temperature data that has not reached the damage threshold. For fluid temperature data that exceeds the damage threshold, a positive temperature difference accumulation is generated. The absolute fluid pressure is nonlinearly mapped through the natural logarithm function to match the marginal diminishing law in fluid mechanics.
[0023] Furthermore, in constructing the true activity retention concentration by coupling apparent chemical concentration, spectral scattering drift factor and composite stress index, the apparent chemical concentration is optically masked and inversely compensated using the spectral scattering drift factor to restore the physical absolute total concentration, and biological activity is eliminated based on the natural exponential decay mapping term of the composite stress index to obtain the true activity retention concentration that reflects biological potency.
[0024] The beneficial effects are: This invention abandons the traditional framework of material composition analysis, which only involves filtering at the spectral mathematical morphology level. Instead, it introduces dynamic thermodynamic and hydrodynamic parameters of materials throughout their entire manufacturing lifecycle, deeply integrating them with end-stage transient optical detection data to achieve multimodal information. Based on underlying physicochemical mechanisms and biochemical protein denaturation mechanisms, the scheme specifically designs a scattering drift operator capable of accurately capturing the microscopic aggregation morphology of materials, as well as a comprehensive stress index capable of assessing the work done by high-pressure fluid damage. This achieves a dimensional leap from conventional static detection of purely chemical bonds to dynamic evaluation of biological potency.
[0025] The feature extraction logic and computational formulas designed in this solution are all based on a highly mature standard mathematical function system, resulting in extremely low computational resource overhead. It eliminates the need to deploy massive computing clusters and prepare huge amounts of labeled samples, thus avoiding the risk of singular value collapse in black-box computation. This lightweight and causally deterministic logic architecture can be directly compiled and burned into low-power programmable controllers commonly used in industrial settings. While ensuring the rigor of stringent process decisions, it achieves fully automated detection and real-time feedback closed-loop control at extremely low modification costs, demonstrating excellent adaptability and economic benefits for field applications. Attached Figure Description
[0026] Figure 1 This is a flowchart of the method for detecting the activity and quality of peptide milk powder produced by wet process according to the present invention.
[0027] Figure 2 This invention provides a scatter plot comparing the deviation distribution between predicted concentration and actual concentration.
[0028] Figure 3 This invention provides a line graph comparing the stability of various methods for monitoring continuous production batches. Detailed Implementation
[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0030] An embodiment of the method for detecting the activity and quality of peptide milk powder based on wet production provided by this invention: like Figure 1 As shown, the method for detecting the activity and quality of peptide milk powder produced by wet processing includes the following steps: S1: Obtain the fluid time series data of the production node and the original reflectance spectrum sequence of the spray drying node, and perform smoothing filtering operation on the original reflectance spectrum sequence to obtain the smoothed spectrum after standardization.
[0031] In a modern wet-process milk powder production line, a multi-node fusion data acquisition and reference time synchronization mechanism is established. High-frequency industrial-grade contact temperature sensors and high-precision diaphragm pressure transmitters are fixedly deployed at key operational nodes: the main pipeline of the concentrated milk heater, the fluid outlet valve of the high-pressure homogenizing pump, and the fluidized bed spray drying tower. Relying on a field-controlled programmable logic controller (PLC) based on a preset fixed sampling frequency, high-density time-series data is continuously acquired within a single processing batch cycle. The acquired data sources accurately include fluid temperature time series characterizing the thermodynamic state inside the pipeline, and fluid pressure time series characterizing the shear strength of fluid transport. All acquisition methods rely on industrial sensors that directly contact the material, ensuring data authenticity and zero latency.
[0032] An industrial online near-infrared spectrometer is vertically suspended above the drying outlet at the bottom of the fluidized bed. The instrument probe is aimed at the stably flowing finished powder layer, continuously acquiring the raw reflectance spectrum sequence of the peptide milk powder covering a specific wavelength range. The system internally calls a standard smoothing filtering algorithm to perform noise reduction preprocessing on the raw reflectance spectrum sequence. Based on the intensity of electromagnetic interference at the site, a matching sliding window width constant is set, and the polynomial fitting order constant is determined. This drives the microprocessor to perform first-order difference differentiation and polynomial smoothing numerical calculations on the raw spectral sequence, thereby filtering out high-frequency spikes caused by external ambient light interference and optimizing the feature recognition of the absorbance curve. Finally, a standardized smoothed spectral data sequence that meets the requirements of subsequent dimensionality reduction calculations is output.
[0033] By deploying multi-point sensor hardware to acquire fluid time-series data throughout the production process and smoothing and filtering the raw reflectance spectrum sequence collected at the end, the influence of environmental stray signals can be effectively eliminated and the historical parameters of material production can be locked in, providing a solid and high signal-to-noise ratio raw physical data foundation for establishing a high-precision multi-dimensional inversion model.
[0034] S2 imports the smoothed spectrum into the pre-configured partial least squares regression algorithm model and calculates and outputs the apparent chemical concentration of peptide components in the current batch of milk powder.
[0035] The field control system imports the standardized, smoothed spectral matrix output from the pre-filtering stage into the pre-configured partial least squares regression algorithm model's computational core via a high-speed industrial Ethernet bus. This regression algorithm model underwent thorough offline calibration before production deployment, acquiring multiple sets of standard spectral matrices of pure peptide solutions with defined concentration gradients. The algorithm utilizes singular value decomposition (SVD) to perform principal component compression dimensionality reduction on the high-dimensional spectral absorbance matrix containing numerous discrete wavelength nodes, removing redundant information and extracting multiple orthogonal latent variable feature components. Subsequently, the least squares method is used to calculate the linear regression weight mapping matrix between these latent variable feature vectors and the true mass concentration of the target peptide.
[0036] The system microprocessor calls the aforementioned fixed linear regression weight matrix to perform matrix multiplication mapping operations on the currently continuously acquired batch smoothed spectra. After the operation is completed, it outputs the apparent chemical concentration of peptide components in the tested milk powder. Since the model is built based on an ideal pure solution state, its output apparent chemical concentration is limited by the inherent defects of the algorithm. On the one hand, it fails to cover the optical transmittance reduction error caused by particle physical roughening and agglomeration due to the wet high-temperature spray granulation process; on the other hand, it only relies on the peptide bond characteristic frequency band for inversion, and does not exclude the proportion of substances that have been damaged by thermomechanical forces and lost their biological activity. Therefore, this initial calculated concentration needs to be used as an intermediate state parameter and passed to the next level correction link.
[0037] By importing the preprocessed smoothed spectrum into the partial least squares regression algorithm model to extract orthogonal latent variables and complete linear mapping, the basic chemical concentration of the target chemical substance can be rapidly calculated at the system level, thus constructing a necessary initial numerical reference platform for overcoming interference from complex processing scenarios and carrying out in-depth concentration correction.
[0038] S3, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, constructs a computational index with causal logic to complete the deep correction calculation of apparent chemical concentration, and outputs the true activity retention concentration that has both concentration and activity attributes.
[0039] The system constructs a spectral scattering drift factor to characterize the rise in the spectral baseline, addressing optical scattering errors caused by microstructural distortion. It is known that the 1100nm to 1200nm frequency band in the near-infrared range is a non-absorption reference band for water and common whey components. Coarse particle aggregation can excite a strong Mie scattering effect, causing an abnormal rise in the background baseline in this band. The calculation formula is set as follows:
[0040] In the formula, It is a dimensionless spectral scattering drift factor. For the first in the preset frequency band Each wavelength point, This represents the total number of discrete wavelength points contained within this frequency band. This is the measured standardized smooth spectral absorbance. The absorbance is the baseline for ideal non-agglomeration.
[0041] The formula includes the total number of discrete wavelength points within the frequency band, and calculates the absolute difference between the measured smooth spectral absorbance and the non-agglomerated reference absorbance obtained in the laboratory for high-purity homogeneous substrate calibration, and then averages the results. This factor extracts a dimensionless optical background shift to characterize the severity of microscopic physical masking phenomena.
[0042] The system simultaneously extracts the composite stress index for bioactive damage; given the extreme vulnerability of the three-dimensional hydrogen bond structure of peptides to heat accumulation and fluid shear, the calculation formula is set as follows:
[0043] In the formula, The composite stress index, This represents the total number of discrete time steps. The actual fluid temperature. The critical temperature threshold for conformational unwinding. This refers to the absolute pressure of the fluid. The standard atmospheric pressure calibration constant is used to eliminate the dimension of pressure. This is the sampling time interval constant.
[0044] The effective temperature difference data exceeding the unwinding critical temperature threshold is extracted from the formula and multiplied and coupled with the natural logarithmic value of the fluid absolute pressure after dimensionless calibration. The pressure term uses logarithmic operation to match the physical consensus that the high-pressure shear failure efficiency tends to diminish marginally. The temperature-pressure product is integrated and accumulated over a constant time interval along the discrete time step, and the output is the total stress work parameter in degrees Celsius multiplied by seconds.
[0045] The system performs cross-decoupling operations on the aforementioned physical-optical and biochemical-thermodynamic characteristics, outputting the final characterization result of the true activity retention concentration. The calculation relationship is set as follows:
[0046] In the formula, To retain the concentration of true activity, For apparent chemical concentration, This is the optical amplification compensation gain constant. It is the molecular stress-sensitive loss constant. This is the maximum allowable stress constant for the system.
[0047] The dimensions of each parameter were verified as follows: the apparent chemical concentration is in grams per 100 grams; the scattering drift factor and the optical amplification compensation gain constant are both dimensionless constants; the composite stress exponent and the maximum allowable stress constant of the system are both in degrees Celsius multiplied by seconds, and after division, the exponent term is internally converted into a dimensionless pure number. The final calculated true active retention concentration maintains a reasonable dimension of grams per 100 grams.
[0048] By constructing computational indices based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, deep correction calculations are completed. This can completely eliminate physical omissions caused by particle morphology at the end of the data and accurately remove failures caused by thermal damage, ultimately outputting high-fidelity material concentration values that closely match the in vivo potency.
[0049] S4, the industrial control host receives and records the actual active retention concentration value in real time, compares it with the preset quality outline indicator bottom line to determine the test result and execute closed-loop control of the equipment.
[0050] The industrial control host motherboard deployed on-site continuously receives the actual active retention concentration (AUC) values from the computing core based on a single measurement via its communication interface. The system's storage array establishes a dynamic traceability ledger, comparing it in real-time with the baseline indicators of the sports nutrition powder product formulation quality guidelines pre-set by the enterprise before production. When the monitoring algorithm detects that the AUC measurements over multiple consecutive sampling periods show a monotonically decreasing trend and are approaching the set lower limit warning line, the main program immediately triggers the underlying automated feedback control command. The control cabinet output module sends microvolt-level adjustment command voltages to the on-site actuators, fine-tuning the mechanical opening of the induced draft and exhaust dampers of the spray drying tower to reduce the exhaust temperature. Simultaneously, the control bus sends a frequency reduction command to the high-pressure fluid pump drive inverter to reduce the work done by the fluid machinery.
[0051] The actual monitoring data shown in the attached diagram further objectively demonstrates the optimization effect of this solution.
[0052] See Figure 2 The scatter plot represents the actual active ingredient concentration of the finished product on the horizontal axis and the concentration output by the system on the vertical axis. A standard baseline running through the entire diagonal interval is retained in the plot. Redundant explanatory boxes have been removed from the chart to highlight the distribution pattern. Dots representing predictions from conventional methods are scattered around the baseline, exhibiting divergent and large-scale offset characteristics, confirming the inaccuracy of existing technologies under coarse particle conditions. The matrix of squares representing the detection results of this scheme is highly convergent and evenly distributed on both sides of the baseline, exhibiting extremely high visual accuracy.
[0053] See Figure 3The linear trend chart uses the horizontal axis to track the batch sequence of the continuously advancing production line. The solid line used to indicate the true baseline concentration maintains a smooth progression, while existing technologies predict trajectory curves that oscillate violently and frequently exhibit abnormally high plateaus. The trajectory curve predicted by the method of this invention maintains a highly stable state throughout and closely follows the baseline solid line. The aforementioned detailed data clearly demonstrates the stability and accuracy of the solution's control link.
[0054] By receiving and recording the actual active retention concentration value in real time through the industrial control host, performing comparison and judgment, and associating with the underlying control terminal, the lagging control situation of relying on post-event sampling and scrapping has been changed, and the stability and high uniformity of the product's biological activity have been firmly locked in the full-speed mass production state.
[0055] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.
Claims
1. A method for detecting the activity and quality of peptide milk powder produced by wet processing, characterized in that, Includes the following steps: S1, acquire the fluid time series data of the production node and the original reflectance spectrum sequence of the spray drying node, and perform smoothing filtering operation on the original reflectance spectrum sequence to obtain the smoothed spectrum after standardization. S2, import the smoothed spectrum into the pre-configured partial least squares regression algorithm model, and calculate and output the apparent chemical concentration of peptide components in the current batch of milk powder; S3, based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, constructs a computational index with causal logic to complete the deep correction calculation of apparent chemical concentration, and outputs the true activity retention concentration that has both concentration and activity attributes. S4, the industrial control host receives and records the actual active retention concentration value in real time, compares it with the preset quality outline indicator bottom line to determine the test result and execute closed-loop control of the equipment.
2. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 1, characterized in that, Obtain fluid time-series data from the production nodes and raw reflectance spectral sequences from the spray drying nodes. Perform smoothing filtering on the raw reflectance spectral sequences to obtain standardized smoothed spectra, including: Time series of fluid temperature and fluid pressure were collected at the heater pipeline and the outlet valve of the high-pressure pump. The original reflectance spectrum sequence of peptide milk powder was collected in a continuous wavelength range at the powder outlet of the fluidized bed; The smoothing filter algorithm is called to perform first-order difference differentiation and polynomial smoothing operations on the original reflectance spectrum sequence to output a smoothed spectrum.
3. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 1, characterized in that, The smoothed spectrum is imported into a pre-configured partial least squares regression algorithm model to calculate and output the apparent chemical concentration of peptides in the current batch of milk powder, including: The high-dimensional spectral absorbance matrix is reduced in dimension by singular value decomposition, and the orthogonal latent variable features are extracted. Calculate the linear regression weight matrix between latent variable characteristics and peptide mass concentration; The apparent chemical concentration of peptides in the current milk powder is obtained by calculating the smoothed spectrum using a linear regression weight matrix.
4. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 1, characterized in that, Based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes, including: The absolute value of absorbance shift within the non-absorption reference frequency band is extracted to construct the spectral scattering drift factor, and the calculation formula is as follows: In the formula, It is a dimensionless spectral scattering drift factor. For the first in the preset frequency band Each wavelength point, This represents the total number of discrete wavelength points contained within this frequency band. This is the measured standardized smooth spectral absorbance. The absorbance is the baseline for ideal non-agglomeration.
5. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 4, characterized in that, Based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes. This also includes: The composite stress index is constructed by nonlinearly integrating the dynamic temperature variable and the dynamic pressure variable over a time scale. The calculation formula is as follows: In the formula, The composite stress index, This represents the total number of discrete time steps. The actual fluid temperature. The critical temperature threshold for conformational unwinding. This refers to the absolute pressure of the fluid. The standard atmospheric pressure calibration constant is used to eliminate the dimension of pressure. This is the sampling time interval constant.
6. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 5, characterized in that, Based on the background characteristics of smoothed spectra and the hydrodynamic characteristics of fluid time series data, a computational index with causal logic is constructed to perform deep correction calculations on apparent chemical concentrations, outputting the true activity retention concentration that combines both concentration and activity attributes. This also includes: The true active retention concentration is constructed by coupling apparent chemical concentration, spectral scattering drift factor, and composite stress index. The calculation relationship is as follows: In the formula, To retain the concentration of true activity, For apparent chemical concentration, is a dimensionless optical amplification compensation gain constant. It is a dimensionless molecular stress-sensitive loss constant. The maximum allowable stress constant of the system. It is a dimensionless spectral scattering drift factor. It is the composite stress index.
7. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 1, characterized in that, The industrial control host receives and records the actual active retention concentration value in real time, compares it with the preset quality outline target minimum to determine the test result, and executes closed-loop control of the equipment, including: When the measured values of the true active retention concentration in multiple consecutive sampling batches show a monotonically decreasing trend and approach the lower control limit, the feedback control program is triggered. The exhaust temperature is reduced by finely adjusting the opening of the exhaust damper of the spray drying tower by sending a command voltage to the field control equipment. The output frequency of the high-pressure pump inverter is simultaneously reduced to decrease the physical shear work done during fluid transport.
8. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 4, characterized in that, The absolute value of absorbance shift within the non-absorption reference frequency band is extracted to construct the spectral scattering drift factor. The shift measure is output by accumulating and averaging the absolute difference between the measured absorbance and the ideal absorbance, thereby characterizing the severity of the target peptide molecule being physically masked by the external rough structure.
9. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 5, characterized in that, In constructing a composite stress index, dynamic temperature and dynamic pressure variables are nonlinearly integrated over a time scale. A maximum value discrimination function is introduced to filter fluid temperature data that has not reached the damage threshold. For fluid temperature data that exceeds the damage threshold, a positive temperature difference accumulation is generated. The absolute fluid pressure is nonlinearly mapped through the natural logarithm function to match the marginal diminishing law in fluid mechanics.
10. The method for detecting the activity and quality of peptide milk powder produced by wet processing according to claim 6, characterized in that, In constructing the true activity retention concentration by coupling apparent chemical concentration, spectral scattering drift factor and composite stress index, the apparent chemical concentration is optically masked and inversely compensated using the spectral scattering drift factor to restore the physical absolute total concentration, and biological activity is eliminated based on the natural exponential decay mapping term of the composite stress index to obtain the true activity retention concentration that reflects biological potency.