A multi-parameter self-adaptive regulation and control directional enzymatic sleep-aiding peptide preparation system and method

Abstract

The present application relates to the technical field of preparation of directional enzymatic sleep-aiding peptide, and more particularly to a multi-parameter self-adaptive regulation and control system and method for preparation of directional enzymatic sleep-aiding peptide. Multidimensional process state data of the enzymatic hydrolysate are acquired in real time by a collection module, the hydrolysis rate, molecular weight and enzyme activity decay ratio are determined by an index calculation module, and a spectral quality index is constructed, a composite deviation index is constructed by a state determination module to determine the process state grade and is modulated by the spectral quality index, a strategy selection module selects an adjustment strategy and a spectral quality recovery strategy accordingly, a parameter adjustment module executes parameter adjustment, a verification module verifies the adjustment effect, and an evolution module updates the determination threshold and the spectral library maintenance strategy based on historical records, so that closed-loop control of enzymatic process detection, determination, adjustment, verification and evolution is realized.

Description

Technical Field

[0001] This invention relates to the field of targeted enzymatic hydrolysis for the preparation of sleep-aiding peptides, and more specifically to a multi-parameter adaptively regulated targeted enzymatic hydrolysis system and method for preparing sleep-aiding peptides. Background Technology

[0002] Targeted enzymatic hydrolysis of proteins is a major industrial method for preparing bioactive peptides. By selecting appropriate proteases and hydrolyzing peptide bonds in protein substrates under specific conditions, large protein molecules are degraded into bioactive peptide fragments within a target molecular weight range. Among these, sleep-aiding bioactive peptides prepared using casein as a substrate through sequential enzymatic hydrolysis with alkaline proteases and flavor proteases, among other complex enzymes, possess good intestinal absorption and the ability to cross the blood-brain barrier due to their moderate molecular weight, making them valuable for improving sleep.

[0003] Existing methods for monitoring protein enzymatic hydrolysis mainly include pH-stat and offline sampling analysis. pH-stat indirectly calculates the degree of hydrolysis by maintaining a constant pH in the reaction system and recording the amount of alkali consumed during hydrolysis. Essentially, it is a single-parameter indirect monitoring method, reflecting only the total number of hydrolyzed peptide bonds and failing to reflect the molecular weight distribution and yield of active peptides in the hydrolysis products. Because there is no linear correlation between the location of peptide bond breakage and pH changes, the same increase in degree of hydrolysis may correspond to drastically different molecular weight distributions in the pH-stat method. For example, in the early stages of hydrolysis, the initial breakage of large protein molecules occurs, leading to a rapid increase in the degree of hydrolysis but a slow decrease in the average molecular weight of the products. In the later stages, further degradation of small peptides causes the degree of hydrolysis to continue increasing, while the yield of active peptides decreases. Therefore, relying solely on pH-stat cannot determine whether the hydrolysis has entered an over-hydrolysis stage.

[0004] Offline sampling analysis involves periodically sampling from the enzymatic hydrolysis vessel, inactivating the enzyme, diluting the sample, and then determining the molecular weight distribution using high-performance liquid chromatography (HPLC) or gel permeation chromatography (GPC). The results are then fed back to adjust process parameters. This method suffers from a long detection lag, typically requiring 30 to 60 minutes from sampling to obtaining results. During this time, the enzymatic hydrolysis continues. By the time the results show the product molecular weight is below the target range, the actual enzymatic hydrolysis has already been severely overdone, resulting in an excessively high proportion of free amino acids and an irreversible reduction in the yield of active peptides. Furthermore, offline sampling analysis only provides data at discrete time points and cannot reflect the continuous dynamic changes in the enzymatic hydrolysis process. It lacks the ability to promptly detect process state drift caused by batch-to-batch differences in substrate proteins, batch fluctuations in enzyme activity, and changes in ambient temperature.

[0005] Near-infrared spectroscopy online monitoring technology acquires spectral data in real time by immersing a fiber optic probe in the enzymatic hydrolysate. Combined with chemometric models, it predicts physicochemical indicators such as the degree of hydrolysis and peptide concentration, thus addressing the lag problem of offline analysis to some extent. However, existing spectral monitoring-based control schemes only compare the predicted degree of hydrolysis or molecular weight with a preset threshold. When the predicted value reaches the threshold, enzymatic hydrolysis is terminated or fixed parameter adjustments are performed, representing a single-parameter threshold determination mode. In multi-enzyme sequential hydrolysis scenarios, the optimal pH and temperature conditions for different proteases vary significantly. The timing of enzyme switching depends not only on whether the degree of hydrolysis has reached the preset value but also on whether the current molecular weight distribution has entered the switching window, whether the residual activity of the preceding enzyme has decreased to a level requiring inactivation, and whether the expected benefit of the target product yield after switching outweighs the cost of the switching operation. These multiple factors exhibit complex nonlinear coupling relationships, and the single-parameter threshold determination mode cannot capture these coupling effects, resulting in a need to improve the accuracy of multi-enzyme switching timing judgment. Summary of the Invention

[0006] This invention provides a multi-parameter adaptive control system and method for preparing sleep-aiding peptides by targeted enzymatic hydrolysis, which overcomes the problems in the prior art where the enzymatic hydrolysis process relies on a single parameter threshold for determination, fails to couple the deviation characteristics of hydrolysis rate, molecular weight shift and enzyme activity decay into a unified determination quantity, and does not perceive the interference of the quality change of the spectral measurement data on the determination result, resulting in the lag or inaccuracy of process status determination.

[0007] To achieve the above objectives, the present invention provides a system for preparing sleep-aiding active peptides by targeted enzymatic hydrolysis, comprising:

[0008] The data acquisition module is used to collect multidimensional process status data of the enzymatic hydrolysate in real time.

[0009] The index calculation module is used to determine the real-time hydrolysis rate, weight-average molecular weight and enzyme activity decay ratio based on the multidimensional process state data, and to construct the spectral quality index based on the multidimensional process state data.

[0010] The state determination module is used to construct a composite deviation index based on the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, the ratio of the weight-average molecular weight to the target molecular weight, and the enzyme activity decay ratio; determine the process state level based on the comparison result of the composite deviation index and a preset threshold; and modulate the process state level based on the spectral quality index.

[0011] The strategy selection module is used to select the corresponding adjustment strategy and enzyme switching strategy based on the process state level, and to select the spectral quality recovery strategy based on the spectral quality index.

[0012] A parameter adjustment module is used to adjust the operating parameters of the enzymatic hydrolysis process based on the adjustment strategy.

[0013] The verification module is used to verify the adjustment effect based on the multidimensional process state data collected after adjustment, and to verify the spectral quality recovery effect based on the multidimensional process state data collected after the spectral quality recovery operation.

[0014] The evolution module is used to update the preset threshold based on historical adjustment records, and to update the normal spectral library and probe maintenance strategy of the acquisition module based on the statistical characteristics of spectral quality degradation events between multiple batches.

[0015] The multidimensional process status data includes spectral measurement data and single-point physicochemical parameters. The index calculation module determines the real-time hydrolysis rate and the weight-average molecular weight based on the pre-established mapping relationship between the spectral measurement data and the degree of hydrolysis and molecular weight, and determines the enzyme activity decay ratio based on the single-point physicochemical parameters.

[0016] The composite deviation index is constructed based on the rate-enzyme activity coupling index and the molecular weight-rate consistency index. The rate-enzyme activity coupling index is determined based on the hydrolysis rate deviation ratio and the enzyme activity decay ratio. The molecular weight-rate consistency index is determined based on the molecular weight offset ratio and the hydrolysis rate deviation ratio. The hydrolysis rate deviation ratio is the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, and the molecular weight offset ratio is the ratio of the weight-average molecular weight to the target molecular weight.

[0017] The process status level is determined to be a concern level based on the composite deviation index being greater than a first preset threshold, and the process status level is determined to be an abnormal level based on the composite deviation index being greater than a second preset threshold, wherein the second preset threshold is greater than the first preset threshold.

[0018] The spectral quality index is determined based on the signal-to-noise ratio characteristics, baseline drift characteristics, and abnormal spectral identification characteristics of the spectral components in the multidimensional process status data. Specifically, the signal-to-noise ratio characteristics are evaluated based on the noise level of the spectral components in a preset non-absorption region, the baseline drift characteristics are evaluated based on the baseline drift of the spectral components within a preset time period, and the abnormal spectral identification characteristics are evaluated based on the degree of deviation of the spectral components from the normal spectral library.

[0019] Based on the spectral quality index being lower than a first quality threshold, the process state level is shifted in a more conservative direction. Based on the spectral quality index being lower than a second quality threshold, the spectral-based determination results in the multidimensional process state data are masked, and the process state level is determined based on the determination results corresponding to the non-spectral components in the multidimensional process state data, wherein the second quality threshold is lower than the first quality threshold.

[0020] The spectral quality recovery strategy includes: performing an enzymatic solution foam removal operation based on the spectral quality index being in a first degradation range; performing an automatic probe cleaning operation based on the spectral quality index being in a second degradation range; and outputting an alarm signal and increasing the offline sampling frequency based on the spectral quality index being in a third degradation range. The first degradation range is higher than the second degradation range, and the second degradation range is higher than the third degradation range.

[0021] The strategy selection module determines to switch from the first enzymatic hydrolysis stage to the second enzymatic hydrolysis stage based on the multidimensional process state data indicating that the degree of hydrolysis reaches a preset switching threshold, the rate of decrease of the weight-average molecular weight slows down to a preset slowdown threshold, and the enzyme activity decay ratio exceeds a preset decay threshold.

[0022] The verification module recalculates the composite deviation index within a preset verification period after the parameter adjustment takes effect. If the recalculated result is greater than the second preset threshold, the strategy selection module is triggered to select the adjustment strategy corresponding to the anomaly level again. If the recalculated result is still greater than the second preset threshold after two consecutive adjustments, an alarm signal is output. After the spectral quality recovery operation is executed, the verification module redetermines the spectral quality index based on the multidimensional process status data. If the spectral quality index recovers to above the preset recovery threshold, the contribution of the spectral-based judgment component to the process status level is restored.

[0023] A method for preparing sleep-aiding active peptides by targeted enzymatic hydrolysis, using the above-mentioned preparation system, includes the following steps:

[0024] Real-time acquisition of multidimensional process status data of enzymatic hydrolysate;

[0025] The real-time hydrolysis rate, weight-average molecular weight, and enzyme activity decay ratio are determined based on the multidimensional process state data, and a spectral quality index is constructed based on the multidimensional process state data.

[0026] A composite deviation index is constructed based on the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, the ratio of the weight-average molecular weight to the target molecular weight, and the enzyme activity decay ratio. The process state level is determined based on the comparison result of the composite deviation index and the preset threshold, and the process state level is modulated based on the spectral quality index.

[0027] Based on the process state level, select the corresponding adjustment strategy and enzyme switching strategy, and based on the spectral quality index, select the spectral quality recovery strategy;

[0028] The operating parameters of the enzymatic hydrolysis process are adjusted based on the aforementioned adjustment strategy;

[0029] The adjustment effect is verified based on the multidimensional process state data collected after the adjustment, and the spectral quality recovery effect is verified based on the multidimensional process state data collected after the spectral quality recovery operation.

[0030] The preset threshold is updated based on historical adjustment records, and the normal spectral library and probe maintenance strategy are updated based on the statistical characteristics of spectral quality degradation events between multiple batches.

[0031] The beneficial effects of the multi-parameter adaptively regulated targeted enzymatic hydrolysis preparation system and method for sleep-aiding peptides of the present invention are as follows:

[0032] By collecting multidimensional process state data of the enzymatic hydrolysate and extracting deviation features in three dimensions—hydrolysis rate, molecular weight distribution, and enzyme activity decay—a composite deviation index is constructed to perform multi-level judgments on the process state and select targeted parameter adjustment strategies accordingly. After adjustment, the adjustment effect is verified in real time, and based on the verification results, decisions are made to maintain, make secondary adjustments, or issue alarms. The detection, judgment, adjustment, verification, and adaptive evolution of the enzymatic hydrolysis process are integrated into a closed-loop control system, which solves the problem that the existing single-parameter threshold judgment mode cannot capture the nonlinear coupling relationship between hydrolysis rate, molecular weight deviation, and enzyme activity decay. In the multi-enzyme sequential enzymatic hydrolysis scenario, adaptive regulation of different substrate batches and enzyme activity fluctuations is achieved, improving the consistency of product molecular weight distribution and active peptide yield between different batches.

[0033] Furthermore, by assessing the quality of each original spectrum from three dimensions—signal-to-noise ratio, baseline drift, and abnormal spectrum identification—and constructing a spectral quality index, when the spectral quality index decreases, the process status judgment based on spectral characteristics is modulated, shifted, or masked according to the degree of degradation. Simultaneously, a graded recovery operation is performed, including foam removal, automatic probe cleaning, or alarm-induced degradation operation. After recovery, the spectral quality is verified in real time. The root cause trend of degradation events is statistically analyzed across multiple batches to trigger preventive probe maintenance and batch-to-batch adaptive updates of the normal spectral library. This prevents misjudgment and misadjustment based on distorted data when foam or protein adsorption film adheres to the probe window, causing spectral distortion, and restores the probe in the early stages of contamination to avoid repeated degradation due to the same root cause.

[0034] Furthermore, by simultaneously introducing molecular weight window conditions, enzyme activity conditions, and target product yield comparison conditions based on the current molecular weight distribution into the multi-enzyme switching determination, the selection of enzyme switching timing is no longer dependent on fixed enzymatic hydrolysis time or a single degree of hydrolysis value. Under different substrate protein batches and enzyme activity differences, the optimal switching point is adaptively selected, avoiding insufficient utilization of alkaline protease due to premature switching and excessive product degradation due to late switching. At the same time, the determination weight of molecular weight shift is reduced within the preset grace period in the early stage of enzymatic hydrolysis to adapt to the normal behavior of the initial nonlinear stage of enzymatic hydrolysis.

[0035] Furthermore, the batch-to-batch threshold adaptive module statistically updates the judgment threshold based on historical adjustment records, causing the threshold to converge towards a lower judgment value as batches accumulate. This gradually improves the system's sensitivity to process state shifts. The batch-to-batch adaptive correction of the enzymatic hydrolysis kinetic model eliminates the impact of substrate protein batch differences on judgment accuracy. The offline sampling verification module periodically calibrates the online prediction model and feeds back offline high-precision molecular weight distribution data to the prediction model for correction, thereby preventing systematic deviations in the online prediction model due to substrate changes or spectral drift. Attached Figure Description

[0036] Figure 1 A block diagram of a multi-parameter adaptively regulated targeted enzymatic hydrolysis system for preparing sleep-aiding peptides is shown.

[0037] Figure 2 A state judgment logic diagram of a multi-parameter adaptively regulated directional enzymatic hydrolysis method for preparing sleep-aiding peptides is shown. Detailed Implementation

[0038] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0039] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0040] This embodiment provides a multi-parameter adaptively regulated targeted enzymatic hydrolysis system for preparing sleep-aiding peptides, referencing... Figure 1 The enzymatic hydrolysis vessel is equipped with a jacketed temperature control and stirring device. The enzymatic hydrolysis vessel is equipped with an online pH electrode, temperature electrode and conductivity electrode, as well as an immersion near-infrared fiber optic probe and automatic alkali, acid and enzyme replenishment pipelines.

[0041] The data acquisition module is connected to the index calculation module to transmit multi-dimensional process status data to the index calculation module.

[0042] The index calculation module is connected to the acquisition module and the state determination module to transmit real-time hydrolysis rate, weight-average molecular weight, enzyme activity decay ratio and spectral quality index to the state determination module.

[0043] The status determination module connects to the index calculation module and the strategy selection module to transmit the process status level to the strategy selection module. The status determination module also connects to the evolution module to transmit the deviation characteristic data of each dimension after determination to the evolution module for threshold statistical analysis, and to receive updated preset thresholds from the evolution module.

[0044] The strategy selection module is connected to the state determination module and the parameter adjustment module to transmit the adjustment strategy and enzyme switching strategy to the parameter adjustment module.

[0045] The parameter adjustment module, connected to the strategy selection module, adjusts the operating parameters of the enzymatic hydrolyzer based on the adjustment strategy. The parameter adjustment module also connects to the validation module to transmit the operation records of parameter adjustments to the validation module for validation triggering.

[0046] The verification module connects to the parameter adjustment module to trigger re-data acquisition after adjustment. The verification module also connects to the strategy selection module to trigger a new adjustment strategy selection if the adjustment effect does not meet preset conditions. The verification module also connects to the evolution module to transmit the verification results of the adjustment effect to the evolution module.

[0047] The evolution module connects to the status determination module, verification module, and acquisition module. It receives deviation data from each batch from the status determination module, receives adjustment effect verification results from the verification module, updates preset thresholds statistically based on historical adjustment records, and sends the updated preset thresholds to the status determination module. Furthermore, it updates the normal spectral library and probe maintenance strategy of the acquisition module based on the statistical characteristics of spectral quality degradation events across multiple batches, and then sends the updated information to the acquisition module.

[0048] Specifically, the acquisition module includes a near-infrared spectroscopy acquisition module and a conventional parameter acquisition module. The near-infrared spectroscopy acquisition module acquires spectral measurement data of the enzymatic hydrolysate online through an immersion near-infrared fiber optic probe, while the conventional parameter acquisition module acquires single-point physicochemical parameters in real time through pH, ​​temperature, and conductivity electrodes. Together, they constitute the multi-dimensional process state data source upon which all subsequent judgments and adjustments depend.

[0049] The index calculation module includes a spectral quality monitoring module, a spectral feature extraction module, an online hydrolysis degree prediction module, a molecular weight distribution prediction module, and an enzyme activity decay estimation module. The spectral quality monitoring module performs real-time quality assessment on each raw spectrum from three dimensions: signal-to-noise ratio, baseline drift, and abnormal spectrum identification, and constructs a spectral quality index. The spectral feature extraction module extracts features from the raw spectral vectors, including characteristic wavenumber absorbance values ​​and principal component scores. The online hydrolysis degree prediction module uses a built-in Si-PLS prediction model to calculate the predicted hydrolysis degree and hydrolysis rate in real time based on spectral features. The molecular weight distribution prediction module uses a built-in artificial neural network prediction model to calculate the predicted weight-average molecular weight and molecular weight distribution width index in real time based on spectral features. The enzyme activity decay estimation module uses a sliding time window to extract the hydrolysis rate sequence and compares it with a baseline hydrolysis rate to determine the enzyme activity decay ratio. The outputs of these five sub-modules together constitute the four data streams—real-time hydrolysis rate, weight-average molecular weight, enzyme activity decay ratio, and spectral quality index—transmitted by the index calculation module to the state determination module.

[0050] The state determination module includes an enzymatic hydrolysis process state determination module and an enzyme digestion characteristic offset identification module. The enzymatic hydrolysis process state determination module constructs a composite deviation index based on real-time hydrolysis rate deviation ratio, molecular weight deviation ratio, and enzyme activity decay ratio, and performs multi-level process state determination. Simultaneously, it receives a spectral quality index to modulate the determination level. The enzyme digestion characteristic offset identification module identifies whether the enzyme selectivity of the current enzyme preparation has shifted based on a distribution width index. The process state level and enzyme digestion offset information output by both modules are jointly transmitted to the strategy selection module.

[0051] The strategy selection module includes a multi-parameter strategy selection module and a multi-enzyme switching determination module. The multi-parameter strategy selection module selects corresponding parameter adjustment strategies and spectral quality recovery strategies based on the process state level. The multi-enzyme switching determination module comprehensively determines the optimal timing for switching from alkaline protease to flavor protease based on three dimensions: molecular weight window conditions, enzyme activity conditions, and product yield prediction conditions. The adjustment strategies and enzyme switching strategies are output to the parameter adjustment module.

[0052] The parameter adjustment module includes a parameter adjustment execution module and an enzymatic hydrolysis termination execution module. The parameter adjustment execution module performs specific pH, temperature, enzyme supplementation, and stirring speed adjustments via the alkali and acid addition pipelines, enzyme replenishment pipelines, jacket temperature control system, and stirring motor of the enzymatic hydrolysis tank. The enzymatic hydrolysis termination execution module performs alkaline protease inactivation or final inactivation operations when switching or termination conditions are met. The parameter adjustment operation records are transmitted to the verification module.

[0053] The verification module includes an immediate verification module for adjustment effects and an offline sampling verification module. The immediate verification module recalculates multidimensional evaluation values ​​to verify the adjustment effect after parameter adjustments take effect. If preset conditions are not met, the strategy selection module is triggered to reselect an adjustment strategy. The offline sampling verification module triggers offline sampling according to the sampling frequency and compares offline measured values ​​with online predicted values. When the deviation exceeds the limit, the prediction model is updated or a prompt to check the probe status is triggered. The verification results of the adjustment effects are transmitted to the evolution module.

[0054] The evolution module includes a batch-to-batch threshold adaptive evolution module and an enzymatic hydrolysis kinetic model evolution module. The batch-to-batch threshold adaptive evolution module statistically updates the preset thresholds of the state determination module based on historical adjustment records; the enzymatic hydrolysis kinetic model evolution module performs batch-to-batch adaptive corrections to the expected hydrolysis rate. The evolution module also feeds back the statistical characteristics of multiple batch-to-batch spectral quality degradation events to the acquisition module to update the normal spectral library and probe maintenance strategy.

[0055] Specifically, a substrate solution is prepared by mixing casein and water at a mass ratio of 1:8 to 1:12. The substrate concentration ranges from 8% to 12% w / v, and in this embodiment, it is set to 10%. This value is chosen to balance product concentration and the fluidity of the enzymatic hydrolysate. Excessively high substrate concentration leads to excessive viscosity of the hydrolysate, affecting mass transfer and spectral acquisition, while excessively low substrate concentration reduces the production capacity per batch. The substrate solution is added to the enzymatic hydrolysis tank, and the temperature is raised to the preset temperature for the alkaline protease hydrolysis stage. The preset temperature for the alkaline protease hydrolysis stage ranges from 50°C to 60°C, and in this embodiment, it is set to 55°C. This value is chosen because the alkaline protease 2.4L exhibits stable catalytic activity within this temperature range. The pH is adjusted to the preset pH for the alkaline protease hydrolysis stage. The preset pH for the alkaline protease hydrolysis stage ranges from 7.5 to 8.5, and in this embodiment, it is set to 8.0. This value is chosen based on the optimal pH range for Alcalase 2.4L. The alkaline protease preparation is added, with the enzyme-to-enzyme ratio ranging from 1% to 3% w / w. In this embodiment, it is set to 2%, a value chosen to control enzyme preparation costs while ensuring the hydrolysis rate. The stirring speed ranges from 100 to 200 rpm, with 150 rpm set in this embodiment. This value is chosen to ensure uniform mass transfer while avoiding the generation of excessive foam that could affect near-infrared spectral acquisition.

[0056] During the flavor protease hydrolysis stage, when the multi-enzyme switching determination module determines that the switching conditions are met, the temperature is raised to 85°C to 95°C and maintained for 10 to 15 minutes to inactivate the alkaline protease. The temperature is then lowered to the preset temperature for the flavor protease hydrolysis stage, which ranges from 45°C to 55°C; in this embodiment, it is set to 50°C, based on the optimal temperature range for the flavor protease. The pH is adjusted to the preset pH for the flavor protease hydrolysis stage, which ranges from 6.5 to 7.5; in this embodiment, it is set to 7.0. The flavor protease enzyme preparation is added, with an enzyme-to-base ratio ranging from 0.5% to 2% w / w; in this embodiment, it is set to 1%. The flavor protease hydrolysis stage continues until the hydrolysis termination execution module determines that the termination conditions are met.

[0057] The near-infrared spectroscopy acquisition module acquires the transmission or diffuse reflectance spectra of the enzymatic hydrolysate online using an immersion near-infrared fiber optic probe, with a spectral wavenumber range of 4000 cm⁻¹. -1 Up to 12500cm -1 In this embodiment, it is set to 4000cm. -1 Up to 10000cm -1 The acquisition interval ranges from 15 to 60 seconds, and is set to 30 seconds in this embodiment. This value is based on the fact that, on the time scale of enzymatic hydrolysis kinetics, an acquisition interval of 15 to 60 seconds can capture continuous changes in the degree of hydrolysis and molecular weight without missing key inflection points. Each acquired spectral data is preprocessed, including baseline correction, standard normal variable transformation, and Savitzky-Golay smoothing. The preprocessed spectral data is output as a raw spectral vector to the spectral feature extraction module.

[0058] The spectral quality monitoring module receives each raw spectral vector output by the near-infrared spectral acquisition module, performs real-time evaluation of spectral quality from three dimensions: signal-to-noise ratio, baseline drift, and abnormal spectral identification, and constructs the spectral quality index Q based on the evaluation results of the three dimensions.

[0059] The wavenumber region with no absorption is selected based on the enzyme hydrolysate system; in this example, 8000 cm⁻¹ is selected. -1 Up to 8500cm -1 In the region where no significant absorption peaks were observed in any component of the enzymatic hydrolysate, signal fluctuations could be attributed to instrument noise and scattering interference. The baseline noise standard deviation of the current spectrum was compared with a preset signal-to-noise ratio (SNR) warning threshold. The SNR warning threshold was determined as follows: 100 spectra were continuously collected under conditions of an aqueous phase system and normal stirring in the enzymatic hydrolysate tank. The mean and standard deviation of the baseline noise standard deviation were calculated, and the SNR warning threshold was calculated by adding three times the standard deviation to the mean.

[0060] The drift monitoring duration ranges from 5 to 10 minutes, and is set to 5 minutes in this embodiment. This value is based on the fact that normal random fluctuations and systematic baseline drift can be distinguished within a 5-minute time window. If the unidirectional shift of the average absorbance value within the drift monitoring duration exceeds a preset drift threshold, it is determined to be an abnormal baseline drift. The drift threshold ranges from 0.002 to 0.005 AU, and is set to 0.003 AU in this embodiment. This value is based on the fact that, under normal operating conditions, the baseline fluctuation in the non-absorption region of the enzymatic hydrolysate spectrum does not exceed 0.002 AU, and 0.003 AU covers 99% of normal operating conditions while leaving a reasonable tolerance margin.

[0061] The normal spectral library is constructed from a predetermined number of spectra obtained through continuous operation under normal enzymatic hydrolysis conditions. The predetermined number ranges from 200 to 500 spectra; in this embodiment, it is set to 300. If the Mahalanobis distance of a spectrum exceeds a predetermined Mahalanobis distance threshold, it is determined to be an abnormal spectrum. The Mahalanobis distance threshold is determined based on the 99th quantile of the chi-square distribution of the Mahalanobis distances of all spectra in the normal spectral library. This value is based on the assumption that, under the normality assumption, the 99th quantile would misclassify only about 1% of normal spectra as abnormal.

[0062] In some embodiments, the spectral quality index Q=1 indicates that the spectral quality is normal and the spectral feature extraction module is using the spectral vector normally. When only the signal-to-noise ratio dimension triggers an anomaly, Q=0.7, indicating mild scattering interference, possibly caused by enzymatic hydrolysate foam adhering to the probe window. The spectral data is still usable but has some noise. When only the baseline drift dimension triggers an anomaly, Q=0.5, indicating that a protein adsorption film may have formed on the probe window, the spectral baseline is shifting, and the prediction accuracy is decreasing. When only the abnormal spectral dimension triggers an anomaly, and the number of consecutive triggers does not exceed the preset consecutive number threshold, Q=0.5, indicating an occasional spectral anomaly. When any two dimensions trigger an anomaly simultaneously, or when the number of consecutive triggers in the abnormal spectral dimension exceeds the preset consecutive number threshold, Q=0.2, indicating a severe deterioration in spectral quality and that the spectral data is unreliable. The preset consecutive number threshold ranges from 3 to 5; in this embodiment, it is set to 3. This value is based on the premise that 1 to 2 occasional abnormal spectra are tolerated, but 3 consecutive abnormal spectra indicate persistent interference. The spectral quality index Q is output to the enzymatic hydrolysis process status determination module.

[0063] The spectral feature extraction module extracts features from the received raw spectral vector. The extracted features include absorbance values ​​at preset feature wavenumbers and the scores of the first few principal components in the spectral principal component analysis. The preset feature wavenumbers are selected based on the chemical composition of the enzymatic hydrolysate, including 4800 cm⁻¹, which is related to the NH stretching vibration of peptide bonds. -1 Up to 4900cm -1 Region, 5150cm related to free carboxyl OH -1 Up to 5250cm-1 Region, 5800 cm related to aromatic amino acid side chains -1 Up to 6000cm -1 The number of principal components ranges from 5 to 15; in this embodiment, it is set to 8. This value is based on the fact that the cumulative variance contribution rate of the first 8 principal components usually exceeds 95%. The spectral feature extraction module outputs the feature absorbance vector and principal component score vector to the online hydrolysis degree prediction module and the molecular weight distribution prediction module.

[0064] The online hydrolysis degree prediction module incorporates a Si-PLS-based hydrolysis degree prediction model. This model is established as follows: Under offline conditions, near-infrared spectroscopy and corresponding pH-stat hydrolysis degree measurements are performed on multiple batches of casein protease hydrolysis processes with different substrate concentrations and enzyme-to-substrate ratios. The Si-PLS model is trained using spectral features as input and the measured hydrolysis degree as output. Before use, the model undergoes a root mean square error (RMSE) assessment on a validation set, with the RMS error not exceeding 0.8% of the hydrolysis degree. During online operation, the online hydrolysis degree prediction module receives the feature vector output from the spectral feature extraction module, calculates the predicted hydrolysis degree value at the current moment in real time, and calculates the real-time hydrolysis rate based on the ratio of the difference between the predicted hydrolysis degree values ​​at two consecutive acquisition moments to the acquisition interval. The online hydrolysis degree prediction module outputs the real-time hydrolysis degree, real-time hydrolysis rate, and expected hydrolysis rate to the enzymatic hydrolysis process status determination module. The expected hydrolysis rate is provided by the enzymatic hydrolysis kinetic model evolution module. The initial expected hydrolysis rate is preset based on empirical kinetic parameters of substrate protein type and enzyme preparation type. In this embodiment, for the enzymatic hydrolysis system of casein and Alcalase 2.4L, the initial expected hydrolysis rate is 0.15% to 0.25% DH / min within 30 minutes after the start of enzymatic hydrolysis, 0.05% to 0.15% DH / min within 30 to 90 minutes, and less than 0.05% DH / min after 90 minutes.

[0065] The molecular weight distribution prediction module incorporates a molecular weight distribution prediction model based on an artificial neural network. This model takes the feature vector output from the spectral feature extraction module as input and outputs the predicted weight-average molecular weight, number-average molecular weight, and molecular weight distribution width index of the enzymatic hydrolysis products. The prediction model is trained using molecular weight data measured offline by GPC, and the root mean square error of the prediction on the validation set is no greater than 500 Da. The molecular weight distribution prediction module outputs the predicted weight-average molecular weight, number-average molecular weight, and distribution width index to the enzymatic hydrolysis process status determination module and the enzyme digestion characteristic offset identification module.

[0066] The conventional parameter acquisition module collects real-time data on the pH, temperature, and conductivity of the enzymatic hydrolysate using pH, temperature, and conductivity electrodes, with the acquisition frequency synchronized with that of the near-infrared spectroscopy acquisition module. This conventional parameter acquisition module then outputs these parameters to the enzymatic hydrolysis process status determination module and the parameter adjustment execution module.

[0067] During enzymatic hydrolysis, the enzyme activity decay estimation module extracts the hydrolysis rate sequence over a recent period using a preset sliding time window. The time span of the hydrolysis rate sequence ranges from 5 to 15 minutes; in this embodiment, it is set to 10 minutes. This value is based on the fact that a 10-minute time window contains 20 sampling points, which is sufficient to smooth out random fluctuations while retaining the trend information of enzyme activity changes. The hydrolysis rate sequence within the sliding time window is compared with the baseline hydrolysis rate in the initial stage, and the enzyme activity decay ratio is defined as the ratio of the average hydrolysis rate within the current time window to the baseline hydrolysis rate. The baseline hydrolysis rate is the average hydrolysis rate from the 10th to the 30th minute after the start of enzymatic hydrolysis, during which substrate is sufficient and enzyme inactivation has not occurred significantly. The enzyme activity decay ratio is output to the enzymatic hydrolysis process status determination module.

[0068] The enzymatic hydrolysis process state determination module receives the following data from the detection layer: real-time degree of hydrolysis, real-time hydrolysis rate, expected hydrolysis rate, predicted weight-average molecular weight, enzyme activity decay ratio, pH, temperature, and conductivity. Three raw determination features are extracted from the above data:

[0069] The normalized hydrolysis rate deviation ratio, defined as the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, is denoted as α. α greater than 1 indicates that the current hydrolysis rate is faster than expected, while α less than 1 indicates that the current hydrolysis rate is slower than expected.

[0070] The molecular weight offset ratio, defined as the ratio of the predicted weight-average molecular weight to the median target molecular weight, is denoted as β. In this embodiment, the median target molecular weight is 2000 Da. A β greater than 1 indicates that the current product molecular weight is higher than the target, and a β less than 1 indicates that the current product molecular weight is lower than the target.

[0071] The enzyme activity attenuation ratio, denoted as γ, is provided by the enzyme activity attenuation estimation module.

[0072] Based on the three original judgment features mentioned above, two intermediate composite indices are constructed.

[0073] The fluctuation offset coupling index, denoted as φ, is defined as the product of the absolute values ​​of α deviating from 1 and γ deviating from 1, i.e., φ = |α-1| × |γ-1|. The physical meaning of φ lies in capturing the severity of the nonlinear coupling between hydrolysis rate shift and enzyme activity decline. Specifically, a decrease in α alone may be caused by either substrate depletion or enzyme activity decline, while a decrease in γ alone may be caused by measurement noise or local temperature fluctuations. When both deviate simultaneously, i.e., φ increases significantly, it indicates that a systematic decrease in enzyme activity is substantially affecting the hydrolysis rate, rather than a temporary or accidental factor. Therefore, φ provides information that a single indicator cannot provide for distinguishing between "enzyme inactivation-related rate decline" and "substrate-limited rate decline."

[0074] The molecular weight-rate consistency index, denoted as ψ, is defined as the product of the deviation of β from 1 and the deviation of α from 1, i.e., ψ = (β-1) × (α-1). The sign and value of ψ reflect the consistency between the direction of change in product molecular weight and the direction of change in hydrolysis rate. When ψ is greater than zero, the deviation of β and α are in the same direction, indicating that the change in molecular weight and the change in hydrolysis rate are consistent in direction. For example, a decrease in hydrolysis rate and a cessation of the decrease in molecular weight are both positive, or an increase in hydrolysis rate and a rapid decrease in molecular weight are both negative, which is normal enzymatic kinetic behavior. When ψ is less than zero, the deviation of β and α are opposite, indicating a disconnect between the change in molecular weight and the change in hydrolysis rate. For example, an abnormal increase in hydrolysis rate but no corresponding decrease in molecular weight may indicate that non-specific enzymatic cleavage or protease autolysis has produced small molecule interfering substances that affect the spectral signal without actually accelerating enzymatic cleavage. Similarly, a decrease in hydrolysis rate but an abnormally rapid decrease in molecular weight may indicate an alteration in enzymatic selectivity.

[0075] A multidimensional evaluation value E is constructed based on two intermediate composite indices. E is defined as the weighted sum of the absolute values ​​of φ and ψ, i.e., E = w1 × φ + w2 × |ψ|. Here, w1 and w2 are weighting coefficients, satisfying w1 + w2 = 1. In this embodiment, w1 is set to 0.5 and w2 to 0.5. This value is based on the fact that during normal enzymatic hydrolysis, the magnitudes of φ and |ψ| are similar, and equal weighting can balance their contributions. No specific limitations are made here. Under normal process conditions, α and β are close to 1, γ is close to 1, φ is close to zero, |ψ| is close to zero, and E is close to zero, indicating that there are no significant deviations in any dimension.

[0076] When the spectral quality index Q output by the spectral quality monitoring module is less than 1, it indicates a decrease in the reliability of the spectral data. In this case, the values ​​of α and β derived from the spectrum may be biased. The enzymatic hydrolysis process status determination module modulates the determination method of the multidimensional evaluation value E based on the value of Q. Specifically, if Q = 0.7, it indicates only a slight decrease in the signal-to-noise ratio. At this time, the prediction accuracy of the online hydrolysis degree prediction module and the molecular weight distribution prediction module is still within an acceptable range, and the determination result of E remains unchanged. However, the abnormal signal-to-noise ratio information is also output to the multi-parameter strategy selection module to trigger the spectral quality recovery operation. If Q = 0.5, it indicates a clear abnormality in the spectral quality. The values ​​of α and β may have non-negligible biases. The process status determination level based on E is shifted one level to a more conservative approach: the original normal level is treated as a concern level, and the original concern level is treated as an abnormal level. If Q=0.2, it indicates a severe deterioration in spectral quality, rendering the spectral data unreliable. In this case, α and β-based spectral judgments are disabled, and the process status is determined solely based on γ and conventional parameters such as pH, temperature, and conductivity. Simultaneously, a signal indicating a severe deterioration in spectral quality is output to the multi-parameter strategy selection module. The shifting of the judgment level is automatically canceled when the spectral quality recovers to Q=1. In the Q=0.2 state, if γ is still within the normal range and conventional parameters do not deviate, the process status is considered normal. However, the system proactively increases the offline sampling frequency to twice the normal frequency to supplement online monitoring with offline data.

[0077] The enzymatic hydrolysis process status determination module compares the multidimensional evaluation value E with a preset first threshold and a second threshold. The first threshold ranges from 0.06 to 0.12, and is set to 0.08 in this embodiment. This value is based on the fact that in the enzymatic hydrolysis of 30 batches of normal casein Alcalase, the E value of more than 99% of the sampling points does not exceed 0.08. The second threshold ranges from 0.15 to 0.25, and is set to 0.20 in this embodiment. This value is based on the fact that in the enzymatic hydrolysis process where 5 batches experienced abnormalities, including enzyme inactivation, substrate protein deterioration, and temperature control failure, the E value of the abnormal events all exceeded 0.20. Based on the comparison results, the current process status is determined as normal, concerning, or abnormal.

[0078] If E is not greater than the first threshold, it is judged as normal. Under the normal level, the current enzymatic hydrolysis parameters are maintained unchanged, and online monitoring continues.

[0079] If E is greater than the first threshold and not greater than the second threshold, it is determined to be at the attention level. Under the attention level, the dominant deviation feature among the three original features is further analyzed. The dominant deviation feature is determined as follows: calculate the absolute value of the deviation of α, β, and γ from their respective baseline values ​​of 1, i.e., |α-1|, |β-1|, |γ-1|, and take the feature corresponding to the largest absolute value of the deviation as the dominant deviation feature. The dominant deviation feature and the current enzymatic hydrolysis stage information are output to the multi-parameter strategy selection module.

[0080] If E is greater than the second threshold, it is determined to be an anomaly. An anomaly indicates that multiple dimensions deviate significantly at the same time. Further subdivision of the dominant deviation feature is no longer performed; instead, multi-parameter collaborative adjustment is initiated. The anomaly level information is output to the multi-parameter strategy selection module.

[0081] At the attention level, in addition to the ratio of E value to threshold, the enzymatic hydrolysis process status determination module also independently judges whether the three original characteristics exceed their respective single-characteristic warning thresholds. The single-characteristic warning threshold for α ranges from 0.4 to 2.5; in this embodiment, it is set to 0.5 and 2.0. This value is based on the fact that during 60 batches of normal enzymatic hydrolysis, the probability of α being in the range of 0.5 to 2.0 exceeds 98%. The upper limit of the single-characteristic warning threshold for β ranges from 1.2 to 1.4; in this embodiment, it is set to 1.25. This value is based on the fact that when β exceeds 1.25, i.e., the weight-average molecular weight exceeds 2500 Da, it indicates that the degree of enzymatic hydrolysis is significantly insufficient. The lower limit of the single-characteristic warning threshold for γ ranges from 0.5 to 0.7; in this embodiment, it is set to 0.6. This value is based on the fact that when enzyme activity is lower than 60% of the initial value, the decrease in hydrolysis rate will lead to a significant extension of the enzymatic hydrolysis time. When a certain original feature exceeds its single feature warning threshold but E has not yet exceeded the first threshold, it is determined to be a single feature warning level. At this time, the corresponding warning feature is also output to the multi-parameter strategy selection module for targeted adjustment, so as to intervene in the independent offset before the coupling deterioration is formed.

[0082] The enzyme digestion characteristic shift identification module receives the number-average molecular weight prediction value and distribution width index output by the molecular weight distribution prediction module. The distribution width index is defined as the ratio of weight-average molecular weight to number-average molecular weight. The larger the distribution width index, the wider the molecular weight distribution of the enzymatic hydrolysis product, i.e., the worse the product uniformity. The enzyme digestion characteristic shift identification module compares the current distribution width index with a preset distribution width warning threshold. The distribution width warning threshold ranges from 1.5 to 2.5, and is set to 2.0 in this embodiment. This value is based on the fact that, under the condition that the target product molecular weight is 1000 Da to 3000 Da, a distribution width index exceeding 2.0 indicates that the product contains a large number of short peptides or free amino acids below 500 Da, indicating over-digestion, or more than 5000 Da of long peptides, indicating insufficient digestion, resulting in decreased product uniformity. If the distribution width index exceeds the distribution width warning threshold, it indicates that the enzyme digestion selectivity of the current enzyme preparation has shifted, and the enzyme digestion characteristic shift identification module outputs this information to the multi-parameter strategy selection module and the offline sampling verification module.

[0083] In some embodiments, if the process status is at the normal level, no adjustments are made. If the process status is at the concern level, a strategy is selected based on the type of dominant deviation characteristic. If the dominant deviation characteristic is α, and the hydrolysis rate deviates from the dominant one, a branch is further executed according to the direction of α. If α is less than 1, the rate is slow, and γ is not greater than 0.6, indicating that enzyme activity decay is the main reason for the rate decrease, an enzyme supplementation strategy is selected, adding the current stage enzyme preparation, with the addition amount being 10% to 30% of the initial enzyme dosage, set to 20% in this embodiment. If α is less than 1 but γ is greater than 0.6, indicating that the rate decrease may be caused by a decrease in substrate concentration rather than enzyme activity decay, in this case, no enzyme supplementation is performed, but the stirring speed is appropriately increased to enhance mass transfer, with the stirring speed increased by 10% to 20% of the current value. If α is greater than 2.0, the rate is fast, indicating that enzymatic hydrolysis is accelerating away from the expected trajectory, a temperature reduction strategy is selected, lowering the enzymatic hydrolysis temperature by 2°C to 5°C from the current value, lowered by 3°C in this embodiment. If the dominant deviation characteristic is β, and the molecular weight shift is dominant, a branch is further executed according to the direction of β. If β is greater than 1, the molecular weight is too large, indicating insufficient enzymatic hydrolysis. In this case, extend the hydrolysis time and check if the pH and temperature deviate from the enzyme's optimal range. If they do, restore them to the optimal pH and temperature. If β is less than 1, the molecular weight is too small, indicating over-hydrolysis. Execute different branches based on the current hydrolysis stage: If currently in the alkaline protease hydrolysis stage, immediately trigger the multi-enzyme switching judgment module to assess whether to switch to the flavor protease hydrolysis stage prematurely; if currently in the flavor protease hydrolysis stage and β remains below the target lower limit, select a termination strategy and trigger the hydrolysis termination execution module. If the dominant deviation characteristic is γ, indicating dominant enzyme activity decline, and the deviations of α and β are still within the normal range, it indicates that enzyme activity is decreasing but has not yet significantly affected the degree of hydrolysis and molecular weight of the product. In this case, supplement with enzyme or adjust the pH to the enzyme's optimal value to delay further decrease in enzyme activity.

[0084] If the process status is abnormal, increase the stirring speed to 120% to 150% of the current value to eliminate any potential local concentration or temperature gradients; adjust the pH to the optimal pH center value of the enzyme preparation at the current stage; adjust the temperature to the optimal temperature center value of the enzyme preparation at the current stage; if the adjusted E value does not fall below the second threshold within the preset validation time, replenish the enzyme preparation at the current stage. The above adjustments should be performed in the following order of priority: pH adjustment, temperature adjustment, stirring adjustment, and enzyme replenishment. After each adjustment, wait one validation data collection cycle before evaluating whether to continue with the next adjustment.

[0085] If the enzymatic hydrolysis process status determination module outputs single-feature warning level information, then targeted adjustments will be made based on the type of warning feature. For example, if α exceeds the single-feature warning upper limit, the temperature or stirring will be reduced; if β exceeds the single-feature warning upper limit, the enzymatic hydrolysis time will be extended or the pH deviation will be checked; if γ is below the single-feature warning lower limit, enzyme will be supplemented or the temperature deviation will be checked.

[0086] If the spectral quality monitoring module outputs a spectral quality degradation signal, the multi-parameter strategy selection module executes a spectral quality recovery strategy based on the value of Q. When Q=0.7, light scattering may be caused by enzymatic hydrolysate foam adhering to the probe window. A foam removal operation is performed: pause stirring for 5 to 10 seconds to allow the foam on the liquid surface to dissipate naturally, then resume stirring; or briefly increase the stirring speed to 150% of the current value for 5 seconds to flush the probe window with liquid flow. After this operation, wait two acquisition cycles for the spectral quality monitoring module to reassess Q. When Q=0.5, this may be caused by the formation of a protein adsorption film on the probe window or continuous foam interference. A probe cleaning operation is performed: use an automatic telescopic mechanism to remove the probe from the enzymatic hydrolysate, rinse the probe window with deionized water, and then re-immerse it. After the cleaning operation, wait three acquisition cycles for Q to be reassessed. When Q=0.2, the spectral quality recovery strategy primarily uses alarm and degradation operation: output audible and visual alarm signals to prompt the operator to check the probe status, and automatically increase the offline sampling frequency. The alarm status is manually deactivated after the operator confirms that the probe status has recovered. The above-mentioned spectral quality recovery operation is independent of E-based process status adjustment. That is, during the spectral quality degradation, spectral quality is restored first, and then process status adjustment is performed.

[0087] The multi-enzyme switching determination module is responsible for determining the optimal timing for switching from alkaline protease to flavor protease. The timing of the switch is not solely based on the degree of hydrolysis reaching a preset value, but rather considers three conditions simultaneously. The first condition is the molecular weight condition: the predicted weight-average molecular weight has entered a preset switching window. The switching window ranges from 2500 Da to 4000 Da; in this embodiment, it is set to 3000 Da. This means that the molecular weight condition is met when the predicted weight-average molecular weight first enters a range below 3000 Da. This value is based on the fact that when the molecular weight of the alkaline protease's digestion product is around 3000 Da, continuing to use alkaline protease has limited effect on further reducing the molecular weight, while the exonuclease activity of the flavor protease is more efficient at further trimming the peptide chain ends within this molecular weight range. The second condition is the enzyme activity condition: α is not greater than 1.0 and not less than 0.3, meaning the hydrolysis rate of the alkaline protease has slowed down but has not yet been completely inactivated. At this point, the loss of product yield due to inactivation is within an acceptable range. If α is still greater than 1.0 and the molecular weight condition is met, it indicates that the enzyme is still working efficiently and the product molecular weight has entered the window. At this time, there is no rush to switch; continue to use alkaline protease for further hydrolysis to reduce the energy consumption cost of enzyme preparation and inactivation operations. If α is lower than 0.3 and the molecular weight condition is not met, it indicates that the enzyme has been severely inactivated and the product molecular weight is still too large. At this time, switching is also triggered, but after switching, the hydrolysis time in the flavor protease hydrolysis stage needs to be appropriately extended to compensate for the insufficient hydrolysis in the alkaline protease hydrolysis stage. The third condition is the product yield prediction condition. Based on the current molecular weight distribution output by the molecular weight distribution prediction module, the difference in target product yield between the current switching and delayed switching is estimated. If the target product yield of the current switching is higher than the predicted yield of the delayed switching, the product yield prediction condition is met. When the molecular weight condition, enzyme activity condition, and product yield prediction condition are all met, the multi-enzyme switching determination module outputs a switching signal to the parameter adjustment execution module and the hydrolysis termination execution module, triggering the start of the alkaline protease inactivation and flavor protease hydrolysis stages.

[0088] The parameter adjustment execution module receives outputs from the multi-parameter strategy selection module and the multi-enzyme switching determination module. Through the automatic alkali / acid addition pipeline, enzyme replenishment pipeline, jacket temperature control system, and stirring motor of the enzymatic hydrolysis tank, it executes specific pH adjustment, temperature adjustment, enzyme replenishment, and stirring speed adjustment operations. Before performing any adjustment, the parameter adjustment execution module records a snapshot of the current process parameters, including the pH, temperature, stirring speed, predicted degree of hydrolysis, predicted molecular weight, and E value at the time of adjustment trigger. These snapshots are stored as historical adjustment records.

[0089] Regarding the triggering of the enzymatic hydrolysis termination execution module: First, when the multi-enzyme switching determination module outputs a switching signal, the enzymatic hydrolysis termination execution module performs an alkaline protease inactivation operation: the hydrolysate is heated to 90°C and maintained for 10 minutes, and after inactivation, it is cooled to the preset temperature for the flavor protease hydrolysis stage. Second, when the multi-parameter strategy selection module determines that early termination is necessary, for example, if β is below the target lower limit in the flavor protease hydrolysis stage and continued hydrolysis would lead to excessive product degradation, the enzymatic hydrolysis termination execution module performs a final inactivation operation: the hydrolysate is heated to 90°C to 95°C and maintained for 15 minutes to completely inactivate all residual enzyme activity. Third, in the flavor protease hydrolysis stage, when β first enters the preset target window and the real-time hydrolysis rate is below the preset low rate threshold for a preset number of consecutive sampling cycles, it indicates that the molecular weight of the target product is basically within the target range of 1000 Da to 3000 Da and the residual enzyme activity no longer has the ability to continue efficient hydrolysis, thus triggering the final inactivation operation normally. The preset target window β is in the range of 0.5 to 1.5, the low rate threshold is in the range of 0.01% to 0.03% DH / min, and is set to 0.02% DH / min in this embodiment. The preset number of continuous acquisition cycles is in the range of 3 to 8, and is set to 5 in this embodiment.

[0090] The instant verification module verifies the adjustment effect within a preset verification period after each parameter adjustment takes effect. The preset verification period ranges from 3 to 10 near-infrared spectral acquisition cycles; in this embodiment, it is set to 5 acquisition cycles, or 2.5 minutes. This value is based on the assumption that the enzymatic hydrolysis rate stabilizes to the new level within approximately 2 to 3 minutes after pH and temperature adjustments. The verification method involves recalculating the multidimensional evaluation value for each acquisition cycle within the verification period and averaging it as E'. If E' is not greater than the first threshold, the adjustment is effective, and the adjusted parameters continue to be maintained. If E' is still greater than the first threshold, the strategy selection corresponding to the concern level or abnormality level is executed again based on the comparison result of E' with the first and second thresholds. If E' is still greater than the second threshold after two consecutive adjustments, it indicates that the process state deviation cannot be recovered through automatic adjustment, which may be caused by problems that cannot be solved by parameter adjustment, such as substrate protein abnormalities, enzyme inactivation, or sensor malfunctions. The control system outputs an alarm signal.

[0091] After the spectral quality restoration operation is performed, the instant verification module simultaneously verifies the restoration of Q. If Q recovers from 0.7 to 1, or from 0.5 to 1 or 0.7 within the preset verification time after the restoration operation, it is considered a valid restoration, restoring the normal judgment logic and judgment level based on E. If Q recovers from 0.2 to above 0.7, it indicates that the spectral quality has been basically restored, the unreliable state of the spectral data is resolved, and the contributions of α and β to E are restored. If Q remains at 0.2 after two consecutive restoration operations, it indicates that the probe may have suffered severe contamination or hardware failure that cannot be recovered by automatic cleaning. The control system outputs an alarm signal and suspends the spectral-based automatic adjustment function, switching to a semi-automatic control mode driven by offline sampling. Only basic process status monitoring is performed based on γ and conventional parameters, and the operator manually decides on the adjustment plan based on the offline detection results.

[0092] The offline sampling verification module triggers offline sampling at a preset sampling frequency, ranging from once every 30 to 60 minutes. In this embodiment, it is set to once every 45 minutes. This frequency is chosen to provide sufficient verification data points without significantly increasing the cost of offline detection. After enzyme inactivation and dilution, the offline samples are sent to an HPLC or GPC system to determine their molecular weight distribution. The results are compared with the online prediction values ​​from the molecular weight distribution prediction module. If the deviation between the offline measured value and the online predicted value exceeds a preset deviation threshold twice consecutively, ranging from 15% to 25% (20% in this embodiment), it indicates that the online prediction model may be biased due to batch changes in the substrate protein or drift in the spectral characteristics of the enzyme digest. This triggers an online adaptive update of the prediction model or prompts the operator to check the status of the spectral probe. Furthermore, if the enzyme digestion characteristic shift identification module outputs a signal indicating that the distribution width exceeds the limit, the offline sampling verification module will additionally analyze the free amino acid content and peptide distribution in the next offline sampling to confirm the root cause of the enzyme digestion selectivity shift.

[0093] The batch-to-batch threshold adaptive evolution module adaptively updates the first and second thresholds after accumulating a preset number of historical adjustment records. The preset number ranges from 20 to 50 records; in this embodiment, it is set to 30 records. This value is based on the fact that 30 valid adjustment records can provide a statistically significant threshold distribution estimate. The update method is as follows: records where the adjusted value E' recovers to below the first threshold are selected from the historical adjustment records. The pre-adjustment E values ​​of these records are extracted, sorted in ascending order, and the E value of the k-th record is taken as the updated first threshold, where k = 0.90 × M, and M is the total number of valid records. This uses the 90th percentile, ensuring that the updated first threshold converges towards a lower judgment value while guaranteeing that over 95% of normal data points are not misjudged. The updated second threshold is the 95th percentile of the pre-adjustment E value. After multiple batches of evolution, since only successfully adjusted events are included in the statistics, the distribution of pre-adjustment E values ​​gradually shifts towards lower values, the judgment threshold decreases accordingly, and the system's sensitivity to process state deviations gradually increases. Understandably, when there are significant changes to the substrate protein source or enzyme batch, the accumulated historical adjustment records can be partially or completely reset to readjust to the current characteristics of the enzymatic hydrolysis system.

[0094] The enzymatic hydrolysis kinetics model evolution module performs batch-to-batch adaptive corrections to the expected hydrolysis rate. After each complete enzymatic hydrolysis batch, the measured degree of hydrolysis-time curve of that batch is compared with the expected hydrolysis rate curve of the previous batch, and the average deviation between the measured hydrolysis rate and the expected hydrolysis rate is calculated for each time period. If the average deviation of two consecutive batches is in the same direction and the magnitude exceeds a preset deviation threshold (ranging from 10% to 20%, set to 15% in this embodiment), the expected hydrolysis rate curve is shifted or scaled towards the measured curve, and the updated expected hydrolysis rate is applied to the enzymatic hydrolysis process determination of the next batch. Thus, the enzymatic hydrolysis kinetics model evolution module automatically adapts the expected hydrolysis rate to the differences in substrate protein batches. For example, the difference in the ratio of α-casein to β-casein in casein in different batches will affect the initial hydrolysis rate of Alcalase. The above adaptive correction can eliminate the impact of this difference on the accuracy of determination.

[0095] The spectral quality monitoring module statistically analyzes events where the spectral quality index Q falls below 1 during each batch of enzymatic hydrolysis, recording the stage of each Q degradation event, the root cause dimension, and the effectiveness of the recovery operation. The root cause dimension refers to the dimension that first triggers the anomaly in the degradation event, such as signal-to-noise ratio, baseline drift, or abnormal spectrum. If the number of Q degradation events caused by the same root cause dimension continuously increases within a preset number of batches, it indicates that the probe is aging or accumulating contamination, and the system outputs a probe maintenance prompt. The preset number of batches ranges from 5 to 10; in this embodiment, it is set to 5. This value is based on the premise that under normal operating conditions, the spectral quality of 5 consecutive batches should remain stable. If the same root cause event increases within 5 batches, random factors are excluded. In addition, the normal spectral library is updated after each complete batch: all spectra with Q=1 and normal Mahalanobis distance in this batch are included in the normal spectral library, replacing the earliest equivalent spectrum in the library, so that the normal spectral library tracks the slow inter-batch drift of the spectral characteristics of the enzymatic hydrolysate. The updating of the normal spectral library allows the calculation benchmark for the Mahalanobis distance threshold to evolve accordingly, avoiding the misjudgment of normal spectral fluctuations caused by changes in substrate protein batches or fine-tuning of enzyme formulations as abnormal spectra.

[0096] Specifically, the following determination strategy in this embodiment has been verified in multiple batches of casein hydrolysis.

[0097] In the initial stage of enzymatic hydrolysis, both the hydrolysis rate and molecular weight deviate significantly from their baseline values. The real-time hydrolysis rate can reach 2 to 3 times the expected hydrolysis rate, and the predicted weight-average molecular weight is 4 to 5 times the median target molecular weight. Simultaneously, the enzyme activity decay ratio is close to 1. At this point, relying solely on the E value for anomaly detection will lead to misjudgment. Based on this situation, an initial grace period mechanism is added to the enzymatic hydrolysis process state determination module: within a preset grace period after the start of enzymatic hydrolysis, the weight of the single-feature determination of β is reduced. The preset grace period ranges from 30 to 60 minutes; in this embodiment, it is set to 45 minutes. This value is based on the fact that the initial nonlinear phase of casein Alcalase hydrolysis typically ends within 30 to 50 minutes, after which the hydrolysis rate tends to stabilize. Within the grace period, the value of |β-1| is multiplied by a grace period coefficient k, where k ranges from 0.1 to 0.3; in this embodiment, it is set to 0.2. If E exceeds the second threshold within the grace period, further check whether α exceeds its single-feature warning upper limit. If α exceeds the threshold but is within the grace period, no abnormality level judgment response is triggered; only the deviation is recorded in the log. That is, in the initial stage of normal caseinase hydrolysis, a higher hydrolysis rate and a higher initial molecular weight are expected behaviors, and neither the abnormality level nor the single-feature warning level within the grace period triggers adjustment.

[0098] Under normal process conditions during the mid-stage of enzymatic hydrolysis outside the grace period, for example, in one sampling cycle, α is between 0.80 and 0.90, β is between 1.80 and 2.50, γ is between 0.90 and 0.95, and E is between the first and second thresholds. The dominant deviation characteristic is usually β. Since a higher molecular weight during the mid-stage of enzymatic hydrolysis is normal, and the pH and temperature are within the optimal range for alkaline proteases at this time, no adjustment is triggered. Only the trend of β changes is tracked. If β does not show a decreasing trend in the next judgment, the pH and temperature deviations are further checked.

[0099] In cases where enzyme activity declines during the later stages of enzymatic hydrolysis, for example, if α drops to between 0.30 and 0.40, β to between 1.50 and 2.00, and γ to between 0.40 and 0.50 within a single sampling cycle, and E exceeds the second threshold, it is considered an abnormal level. Simultaneously, α and γ are both below the single-feature warning lower limit. Multi-parameter coordinated adjustment is initiated: First, check if pH and temperature deviate from their respective optimal ranges. If not, increase the stirring speed to 120% of the current value. After 5 sampling cycles, if E is still greater than the second threshold, replenish the current stage of enzyme preparation, adding 20% ​​of the initial enzyme dosage. After the above adjustments, α recovers to above 0.70, γ recovers to above 0.70, and E drops below the first threshold.

[0100] Regarding multi-enzyme switching determination, for example, in a certain acquisition cycle, if the predicted weight-average molecular weight has entered the preset switching window, with α between 0.70 and 0.90, γ between 0.65 and 0.75, and E not exceeding the first threshold, it is considered normal. At this time, the multi-enzyme switching determination module checks three conditions: the weight-average molecular weight is below the upper limit of the switching window, satisfying the molecular weight condition; α is between 0.3 and 1.0, satisfying the enzyme activity condition; and the difference between the predicted switching and delayed switching target product yields based on the current molecular weight distribution is compared, satisfying the product yield prediction condition. When all three conditions are met, enzyme switching is triggered: alkaline protease inactivation is performed followed by cooling to switch to flavor protease hydrolysis conditions, and flavor protease enzyme preparation is added to start flavor protease hydrolysis.

[0101] Regarding the normal termination of the flavor protease hydrolysis stage, for example, if β first enters the target window of 0.5 to 1.5, α drops to approximately 0.50, and γ drops to approximately 0.50 in a sampling cycle, E is determined to be of concern level, with α being the dominant deviation characteristic. If the real-time hydrolysis rate is below 0.02% DH / min for five consecutive sampling cycles, it indicates that the target product molecular weight has been met and the residual enzyme activity does not have the ability to continue efficient hydrolysis, thus triggering final inactivation normally. Conversely, in cases where the molecular weight of the product in the flavor protease hydrolysis stage is too low, leading to premature termination, β is already far below the lower limit of the target window. Although E is still below the second threshold, the multi-parameter strategy selection module directly selects the termination strategy based on the deviation direction of β and the current stage information of flavor protease hydrolysis, triggering the final inactivation operation to prevent the active peptide from being further degraded into free amino acids.

[0102] Regarding spectral quality degradation and recovery, for example, during the middle stage of an alkaline protease digestion, protein foam generated by stirring gradually adhered to the probe window. The spectral quality monitoring module detected that the baseline noise standard deviation of four consecutive spectra consistently exceeded the signal-to-noise ratio warning threshold, and the average absorbance in the non-absorbance wavenumber region increased unidirectionally by 0.004 AU within 5 minutes, causing Q to drop from 1 to 0.5. At this point, the digestion process status determination module shifted the determination level based on E by one level. Simultaneously, the multi-parameter strategy selection module received a Q=0.5 signal, determining that the degradation was likely caused by protein adsorption at the probe window, triggering automatic probe cleaning: the telescopic mechanism lifted the probe out of the liquid surface, the window was rinsed with deionized water, and then it was re-immersed. After three acquisition cycles, Q recovered to 1. After the determination level recovered, normal E value determination continued, and no abnormalities occurred in subsequent digestion processes. If Q further drops to 0.2 in the above situation and fails to recover after two automatic cleanings, the control system outputs an alarm signal to prompt the operator to check whether the probe has mechanical damage or irreversible contamination. At the same time, it automatically increases the offline sampling frequency to once every 15 minutes and suspends the spectrum-based automatic adjustment function, maintaining basic process monitoring only based on γ and conventional parameters.

[0103] It is understood that the specific parameter values, coefficient values ​​in the calculation formulas, and threshold values ​​in this embodiment are illustrative examples provided to facilitate understanding of the invention, and are not intended to limit the invention. In practical applications, the above parameters can be adjusted according to the specific substrate protein type, enzyme preparation type and batch, target product molecular weight range, and production scale.

Claims

1. A system for preparing sleep-aiding active peptides by targeted enzymatic hydrolysis, characterized in that, include: The data acquisition module is used to collect multidimensional process status data of the enzymatic hydrolysate in real time. The index calculation module is used to determine the real-time hydrolysis rate, weight-average molecular weight and enzyme activity decay ratio based on the multidimensional process state data, and to construct the spectral quality index based on the multidimensional process state data. The state determination module is used to construct a deviation index based on the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, the ratio of the weight-average molecular weight to the target molecular weight, and the enzyme activity decay ratio; determine the process state level based on the comparison result of the composite deviation index and the preset threshold; and modulate the process state level based on the spectral quality index. The strategy selection module is used to select the corresponding adjustment strategy and enzyme switching strategy based on the process state level, and to select the spectral quality recovery strategy based on the spectral quality index. A parameter adjustment module, which is used to adjust the operating parameters of the enzymatic hydrolysis process based on the adjustment strategy; The verification module is used to verify the adjustment effect based on the multidimensional process state data collected after adjustment, and to verify the spectral quality recovery effect based on the multidimensional process state data collected after the spectral quality recovery operation. The evolution module is used to update the preset threshold based on historical adjustment records, and to update the normal spectral library and probe maintenance strategy of the acquisition module based on the statistical characteristics of spectral quality degradation events between multiple batches.

2. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 1, characterized in that, The multidimensional process status data includes spectral measurement data and single-point physicochemical parameters. The index calculation module determines the real-time hydrolysis rate and the weight-average molecular weight based on the pre-established mapping relationship between the spectral measurement data and the degree of hydrolysis and molecular weight, and determines the enzyme activity decay ratio based on the single-point physicochemical parameters.

3. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 1, characterized in that, The composite deviation index is constructed based on the rate-enzyme activity coupling index and the molecular weight-rate consistency index. The rate-enzyme activity coupling index is determined based on the hydrolysis rate deviation ratio and the enzyme activity decay ratio. The molecular weight-rate consistency index is determined based on the molecular weight offset ratio and the hydrolysis rate deviation ratio. The hydrolysis rate deviation ratio is the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, and the molecular weight offset ratio is the ratio of the weight-average molecular weight to the target molecular weight.

4. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 3, characterized in that, The process status level is determined to be a concern level based on the composite deviation index being greater than a first preset threshold, and the process status level is determined to be an abnormal level based on the composite deviation index being greater than a second preset threshold, wherein the second preset threshold is greater than the first preset threshold.

5. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 1, characterized in that, The spectral quality index is determined based on the signal-to-noise ratio characteristics, baseline drift characteristics, and abnormal spectral identification characteristics of the spectral components in the multidimensional process status data. Specifically, the signal-to-noise ratio characteristics are evaluated based on the noise level of the spectral components in a preset non-absorption region, the baseline drift characteristics are evaluated based on the baseline drift of the spectral components within a preset time period, and the abnormal spectral identification characteristics are evaluated based on the degree of deviation of the spectral components from the normal spectral library.

6. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 5, characterized in that, Based on the spectral quality index being lower than a first quality threshold, the process state level is shifted in a more conservative direction. Based on the spectral quality index being lower than a second quality threshold, the spectral-based determination results in the multidimensional process state data are masked, and the process state level is determined based on the determination results corresponding to the non-spectral components in the multidimensional process state data, wherein the second quality threshold is lower than the first quality threshold.

7. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 6, characterized in that, The spectral quality recovery strategy includes: performing an enzymatic solution foam removal operation based on the spectral quality index being in a first degradation range; performing an automatic probe cleaning operation based on the spectral quality index being in a second degradation range; and outputting an alarm signal and increasing the offline sampling frequency based on the spectral quality index being in a third degradation range. The first degradation range is higher than the second degradation range, and the second degradation range is higher than the third degradation range.

8. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 1, characterized in that, The strategy selection module determines to switch from the first enzymatic hydrolysis stage to the second enzymatic hydrolysis stage based on the multidimensional process state data indicating that the degree of hydrolysis reaches a preset switching threshold, the rate of decrease of the weight-average molecular weight slows down to a preset slowdown threshold, and the enzyme activity decay ratio exceeds a preset decay threshold.

9. The targeted enzymatic hydrolysis sleep-aiding active peptide preparation system according to claim 4, characterized in that, The verification module recalculates the composite deviation index within a preset verification time after the parameter adjustment takes effect. If the recalculated result is greater than the second preset threshold, the strategy selection module is triggered to select the adjustment strategy corresponding to the abnormality level again. If the recalculated result is still greater than the second preset threshold after two consecutive adjustments, an alarm signal is output. After the spectral quality recovery operation is performed, the verification module re-determines the spectral quality index based on the multidimensional process status data. If the spectral quality index recovers to above the preset recovery threshold, the contribution of the spectral-based judgment component to the process status level is restored.

10. A method for preparing sleep-aiding active peptides by targeted enzymatic hydrolysis, characterized in that, The system for preparing sleep-aiding active peptides by targeted enzymatic hydrolysis according to any one of claims 1 to 9 includes the following steps: Real-time acquisition of multidimensional process status data of enzymatic hydrolysate; The real-time hydrolysis rate, weight-average molecular weight, and enzyme activity decay ratio are determined based on the multidimensional process state data, and a spectral quality index is constructed based on the multidimensional process state data. A composite deviation index is constructed based on the ratio of the real-time hydrolysis rate to the expected hydrolysis rate, the ratio of the weight-average molecular weight to the target molecular weight, and the enzyme activity decay ratio. The process state level is determined based on the comparison result of the composite deviation index and the preset threshold, and the process state level is modulated based on the spectral quality index. Based on the process state level, select the corresponding adjustment strategy and enzyme switching strategy, and based on the spectral quality index, select the spectral quality recovery strategy; The operating parameters of the enzymatic hydrolysis process are adjusted based on the aforementioned adjustment strategy; The adjustment effect is verified based on the multidimensional process state data collected after the adjustment, and the spectral quality recovery effect is verified based on the multidimensional process state data collected after the spectral quality recovery operation. The preset threshold is updated based on historical adjustment records, and the normal spectral library and probe maintenance strategy are updated based on the statistical characteristics of spectral quality degradation events between multiple batches.

Images