Freeze-dried particle intelligent production control system and method based on Internet of Things

Abstract

The invention provides an intelligent freeze-drying particle production control system and method based on the Internet of Things, and belongs to the technical field of freeze-drying particle production. The control system comprises a multi-source sensing unit and a data processing unit connected with the multi-source sensing unit; the multi-source sensing unit is arranged on a near infrared spectrum sensor, an optical fiber temperature sensor, a vacuum degree sensor and a visual camera of freeze-drying equipment and is used for acquiring the temperature gradient change rate, crystallinity data and vacuum fluctuation amplitude of a material; the data processing unit comprises a mapping module, a decision module and an optimization module; the mapping module converts collected data into a mass transfer resistance coefficient K. The conversion process comprises the following steps: step 1, selecting a weight factor combination of a temperature gradient change rate, a crystallinity logarithmic transformation value and a vacuum fluctuation amplitude according to a material formula type; according to the invention, real-time monitoring, dynamic optimization and quality tracing of the freeze-drying process are realized, and the production efficiency and the product consistency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of freeze-drying, and particularly relates to a freeze-dried particle intelligent production control system and method based on Internet of Things. BACKGROUND

[0002] Freeze-drying, also known as lyophilization, is a process for drying biological materials such as pharmaceuticals, proteins, enzymes and heat-sensitive materials. Freeze-drying provides drying of the target product by sublimation of ice crystals into water vapor, i.e. by direct conversion of the water content from a solid state to a gaseous state. Freeze-drying is usually performed under vacuum conditions, but also works at different atmospheric pressures.

[0003] The patent application with the publication number CN103917842A discloses a process line for producing freeze-dried particles under closed conditions, the process line comprising at least the following independent devices: a spraying chamber for generating droplets and for freezing the droplets to form particles, and a bulk freeze-dryer for freeze-drying the particles, wherein a transfer section is provided to transfer the product from the spraying chamber to the freeze-dryer, each of the devices and the transfer section being independently adapted for closed operation in order to produce particles under end-to-end closed conditions, and the spraying chamber being adapted to separate the droplets from any cooling circuit.

[0004] The freeze-drying control system in the prior art relies on fixed process curves, cannot respond to material state changes in real time, has control lag, relies on manual experience to set process parameters, cannot respond to material state changes in real time, causes product quality fluctuations, meanwhile, material characteristics differences (such as freeze-drying behaviors of proteins and polysaccharides) cause the traditional method to be difficult to be universal, and process parameter adjustment relies on trial and error, which is long in cycle and high in cost. SUMMARY

[0005] The present application aims to provide a freeze-dried particle intelligent production control system and method based on Internet of Things, realize real-time monitoring, dynamic optimization and quality tracing of the freeze-drying process, and improve production efficiency and product consistency.

[0006] To solve the above technical problems, the technical solution adopted by the present application is as follows:

[0007] A freeze-dried particle intelligent production control system based on Internet of Things, comprising a multi-source sensing unit and a data processing unit connected with the multi-source sensing unit.

[0008] The multi-source sensing unit is disposed on a near-infrared spectrum sensor, an optical fiber temperature sensor, a vacuum degree sensor and a visual camera of a freeze-drying device, and is used to collect material temperature gradient change rate, crystallinity data and vacuum fluctuation amplitude.

[0009] The data processing unit comprises a mapping module, a decision module and an optimization module; the mapping module converts the collected data into a mass transfer resistance coefficient K, and the conversion process comprises the following steps:

[0010] Step 1: according to the type of material formula, the weight factor combination of the temperature gradient change rate, the crystallinity logarithmic transformation value and the vacuum fluctuation amplitude is selected;

[0011] Step 2: the weighted temperature gradient change rate, the crystallinity logarithmic transformation value and the vacuum fluctuation amplitude are linearly superimposed to generate the K value;

[0012] The decision module is used for triggering the reinforcement learning algorithm to adjust the freeze-drying process curve when the real-time K value exceeds the dynamic threshold value; and the optimization module is used for optimizing the weight factor library by federated learning aggregation of historical K values, and driving digital twin simulation.

[0013] Further, the control system further comprises an ice crystal morphology identification module, the ice crystal morphology identification module captures the ice crystal fractal dimension through a high-resolution visual camera, and generates a morphology correction coefficient, and the mapping module introduces a morphology correction term when calculating the mass transfer resistance coefficient.

[0014] Further, the weight factor combination method is as follows: the temperature gradient change rate in the material accounts for 40-70%, the crystallinity accounts for 20-50%, and the vacuum fluctuation accounts for 5-20%, and the dynamic threshold value is automatically adjusted in the range of 0.8-1.2 according to the freeze-drying stage.

[0015] Further, the mass transfer resistance coefficient K and the freeze-drying rate present strong negative correlation, the correlation coefficient is greater than or equal to 0.95, the calculation precision error is 3-5%, and the edge processing delay is 40-50 ms.

[0016] Further, the decision module reinforcement learning temperature control strategy takes the approximation of the target K value as the core optimization direction, and simultaneously introduces an energy consumption penalty mechanism to adjust the temperature curve.

[0017] Further, the data processing unit further comprises a phase state adaptive module, the phase state adaptive module identifies the glass state transition region based on the spatial distribution relationship between the crystallinity and the temperature gradient, and when the glass state region accounts for more than 15%, the weight proportion of the crystallinity data in the K value calculation is automatically increased.

[0018] The application also discloses a freeze-dried particle intelligent production control method based on an Internet of Things.

[0019] S101: real-time acquisition of material temperature gradient change rate, crystallinity and vacuum fluctuation amplitude original data;

[0020] S102: According to the current material formula type matching preset weight factor, the three-element data is weighted and superimposed to generate the mass transfer resistance coefficient K;

[0021] S103: Comparing the real-time K value with the stage dynamic threshold value, if the K value continuously exceeds the standard, the heating rate is dynamically inhibited or the vacuum degree is compensated through reinforcement learning, and if the K value continuously deviates, the energy consumption optimization mode is started;

[0022] S104: Based on the historical K value data set federated learning optimization weight factor configuration rule.

[0023] The control logic of S103 is as follows: when the K value exceeds the upper limit of the threshold value for 3 minutes continuously, the cooling amplitude is linearly enhanced according to the exceeding proportion; when the K value is lower than the lower limit of the threshold value for 10 minutes, the vacuum compensation program is started, and the compensation amount is proportional to the reciprocal of the K value.

[0024] Further, the present application also includes S105, a method for establishing K value and quality association rules, statistical batch freeze-drying endpoint K value fluctuation variance, when the variance value is greater than 0.05, the batch is automatically determined to be unqualified, and the remaining service life of the vacuum pump is predicted according to the K value fluctuation frequency characteristics.

[0025] Further, in the pre-freezing stage, the K value reference is constructed, and the weighted average value of the crystallinity is calculated as K0 reference under the condition of-38℃ ～ -40℃ state, and setting the deviation of the real-time K value from the K0 reference as 15-25% as the process abnormality alarm threshold value.

[0026] Compared with the prior art, the present application has the following beneficial effects:

[0027] The present application realizes dynamic adjustment of freeze-drying process parameters through multi-source data fusion and intelligent algorithm, reduces quality fluctuation, and effectively improves product consistency. The present application unifies multi-source data dimensions through K value, controls response delay, introduces energy consumption penalty through reinforcement learning, and comprehensively reduces energy consumption. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and other related drawings can also be obtained by those skilled in the art without creative labor.

[0029] Figure 1 The overall principle block diagram of the control method described in the present application. DETAILED DESCRIPTION

[0030] In the following, certain example embodiments are simply described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the embodiments of the present application. Therefore, the drawings and description are to be regarded as being illustrative in nature rather than restrictive. The following description is made in connection with the accompanying drawings and description of the preferred embodiments. Figure 1 The embodiments of the present application are described in detail.

[0031] Embodiment one: the embodiment discloses a freeze-dried particle intelligent production control system based on Internet of Things, comprising a multi-source sensing unit and a data processing unit connected with the multi-source sensing unit;

[0032] Among them, the multi-source sensing unit is deployed in the near-infrared spectrum sensor, optical fiber temperature sensor, vacuum degree sensor and visual camera of the freeze-drying equipment, which is used to collect material temperature gradient change rate, crystallinity data and vacuum fluctuation amplitude;

[0033] The data processing unit includes a mapping module, a decision module and an optimization module; the mapping module converts the collected data into a mass transfer resistance coefficient K, and the conversion process includes the following steps:

[0034] Step 1: according to the type of material formula, the weight factor combination of temperature gradient change rate, crystallinity logarithmic transformation value and vacuum fluctuation amplitude is selected;

[0035] Step 2: linearly superimpose the weighted temperature gradient change rate, crystallinity logarithmic transformation value and vacuum fluctuation amplitude to generate K value;

[0036] The decision module is used to trigger the reinforcement learning algorithm to adjust the freeze-drying process curve when the real-time K value exceeds the dynamic threshold; the optimization module is used to optimize the weight factor library by federated learning aggregation of historical K value, and drive digital twin simulation.

[0037] In this embodiment, the multi-source sensing unit transmits data to the mapping module, the decision module outputs control instructions to the freeze-drying equipment according to the K value, and the optimization module updates the weight factor library through federated learning.

[0038] Among them, the weight factor combination method is as follows: the weight of temperature gradient change rate in the material accounts for 40-70%, the crystallinity accounts for 20-50%, and the vacuum fluctuation accounts for 5-20%, and the dynamic threshold is automatically adjusted in the interval of 0.8-1.2 according to the freeze-drying stage.

[0039] In this embodiment, the weight of temperature gradient change rate in the material accounts for 60%, the crystallinity accounts for 30%, and the vacuum fluctuation accounts for 10%.

[0040] Among them, the mass transfer resistance coefficient K and the freeze-drying rate show strong negative correlation, the correlation coefficient is greater than or equal to 0.95, the calculation accuracy error is 3-5%, and the edge processing delay is 40-50ms.

[0041] In this embodiment, the calculation accuracy error is 4%, and the edge processing delay is 45 ms.

[0042] Among them, the decision module strengthens the learning temperature control strategy to approximate the target K value as the core optimization direction, and introduces an energy consumption penalty mechanism to adjust the temperature curve.

[0043] Further, the data processing unit further comprises a phase state adaptive module, which identifies a glass state transition region based on the spatial distribution relationship between crystallinity and temperature gradient, and automatically increases the weight proportion of crystallinity data in K value calculation when the glass state region proportion exceeds 15%.

[0044] The specific steps are as follows:

[0045] S101: Real-time acquisition of material temperature gradient change rate, crystallinity and vacuum fluctuation amplitude original data;

[0046] S102: According to the current material formula type, match the preset weight factor, and add the three element data to generate the mass transfer resistance coefficient K;

[0047] S103: Compare the real-time K value with the stage dynamic threshold value, if the K value is continuously over-standard, suppress the heating rate or compensate the vacuum degree through reinforcement learning, if the K value is continuously low, start the energy consumption optimization mode;

[0048] S104: Based on the historical K value data set federated learning optimization weight factor configuration rule.

[0049] Among them, the control logic of S103 is as follows: when the K value exceeds the upper limit of the threshold value for 3 minutes continuously, the cooling amplitude is linearly enhanced according to the exceeding proportion; when the K value is lower than the lower limit of the threshold value for 10 minutes, the vacuum compensation program is started, and the compensation amount is proportional to the reciprocal of the K value.

[0050] Further, the present embodiment further comprises S105, a method for establishing K value and quality association rules, statistics of batch freeze-drying endpoint K value fluctuation variance, when the variance value is greater than 0.05, automatically determine that the batch is unqualified, and predict the remaining service life of the vacuum according to the K value fluctuation frequency characteristics.

[0051] Among them, in the pre-freezing stage, the K value reference is constructed, and the weighted average value of the crystallinity is calculated as K0 reference at-38℃ ～ -40℃ state, and set the real-time K value deviation from K0 reference 15-25% as the process abnormal alarm threshold.

[0052] In practical applications, for protein material freeze-drying control, a single monoclonal antibody preparation, freeze-drying machine is equipped with near-infrared spectrum sensor, wavelength range 900-1700nm, optical fiber temperature sensor, accuracy ±0.1℃, vacuum degree sensor, resolution 0.01mbar and visual camera, frame rate 30fps.

[0053] Weight factor: temperature gradient change rate 60%, crystallinity 30%, vacuum fluctuation 10%.

[0054] Dynamic threshold: pre-freezing stage 0.8, sublimation stage 1.0, desorption stage 1.2.

[0055] K0 reference: pre-freezing stage crystallinity weighted average value at-38℃ as reference, real-time K value deviation 15% trigger alarm.

[0056] The specific control process is as follows:

[0057] S101: sensor real-time data acquisition, temperature gradient change rate through optical fiber sensor array calculation, crystallinity by near-infrared spectrum data through PLS model prediction, vacuum fluctuation amplitude by vacuum degree sensor high frequency sampling(10Hz)obtained.

[0058] S102: mapping module generates K value by weight superposition, calculation accuracy error control in 5%, edge processing delay 45ms.

[0059] S103: sublimation stage K value continuous 3 minutes more than 1.05, decision module starts reinforcement learning algorithm, heating plate temperature from 25℃ to 20℃, and compensates vacuum degree to 0.2mbar, original set 0.3mbar, energy consumption penalty mechanism synchronous reduce temperature regulation rate, avoid excessive energy consumption, when glassy region proportion reduced to 10% below, restore original weight factor.

[0060] S104: optimization module through federal learning aggregation historical K value, find crystallinity weight can fine-tune to 35% in sublimation later period, further improve freeze-drying rate, correlation coefficient≥0.95.

[0061] In this embodiment, real-time detection temperature gradient change rate 0.15℃ / s, crystallinity 0.85, vacuum fluctuation amplitude 12Pa, vacuum fluctuation amplitude need to be divided by reference value 10Pa for dimensionless.

[0062] Calculate K value:

[0063]

[0064] Wherein, The temperature gradient change rate is the rate of change of the temperature gradient, log(A) is the crystallinity, dimensionless, 0-1, and AP is the normalized vacuum fluctuation amplitude, AP = PI / PO, PI is the original fluctuation amplitude, and PO is the reference value, specifically 10 Pa.

[0065] w1 is the temperature gradient change rate weight, w2 is the crystallinity weight, and w3 is the vacuum fluctuation weight.

[0066] K = 0.15 * 0.6 + log(0.85) * 0.3 + 12 * 0.1 = 0.09 + (-0.07) + 1.2 = 1.22;

[0067] log(0.85) is rounded to two decimal places;

[0068] At this time, the K value is greater than the threshold value 1.2, triggering the reinforcement learning cooling:

[0069] AT = -2 ℃ * (1.22-1.0) = -0.44 ℃;

[0070] The vacuum fluctuation amplitude needs to be divided by the reference value (for example, the vacuum fluctuation amplitude / 10 Pa) to ensure that the K value is dimensionless; after adjustment, the K value returns to 1.05, and tests show that the freeze-drying cycle is shortened by 18%, the product moisture content is reduced from 2.5% to 1.8%, the active retention rate is increased from 92% to 95%, and the energy consumption is reduced by 15%.

[0071] Example Two: This example is basically the same as Example One, except that in this example, the freeze-drying control of polysaccharide materials is realized, specifically a plant polysaccharide extract, and the freeze-drying machine configuration is the same as in Example One.

[0072] In this example, the weight factors are: temperature gradient change rate 30%, crystallinity 50%, and vacuum fluctuation 20%.

[0073] Phase self-adaptation: the glass transition region identification threshold is set to 15%, and after triggering, the crystallinity weight is temporarily increased to 60%.

[0074] The control process is as follows:

[0075] During the pre-freezing stage, the K0 reference is established at -40℃, and an alarm is issued when the real-time K value deviates by 20%, the pre-freezing time is extended from 2 hours to 2.5 hours to avoid collapse caused by large ice crystals;

[0076] During the sublimation stage, the phase self-adaptation module identifies that the glassy region accounts for 18% through the crystallinity and temperature gradient spatial distribution, automatically increases the crystallinity weight to 65%, and reduces the K value from 1.1 to 0.98 to avoid incomplete drying due to high mass transfer resistance.

[0077] The K value is calculated = crystallinity contribution 0.92 x 0.65 + temperature gradient contribution 0.10 x 0.3 + vacuum fluctuation contribution 0.8 x 0.05 = 0.628;

[0078] K value < 0.8, start vacuum compensation: ΔP = 0.5 / K = 0.8 Pa;

[0079] In step S105, the batch end K value fluctuation variance is 0.03 (qualified threshold is 0.05), and it is determined to be qualified; at the same time, according to the K value fluctuation frequency (3 times / hour), the remaining life of the vacuum pump is predicted to be 200 hours, and maintenance is arranged in advance.

[0080] After testing, the reconstitution time of the freeze-dried product is shortened from 120 seconds to 80 seconds, the clarity of the solution after reconstitution is improved, and the unqualified rate of the batch is reduced from 1.2% to 0.3%.

[0081] Example three: This embodiment is mainly used to realize the cooperative optimization of multiple devices. In this embodiment, five freeze-dryers are used to realize the cooperative optimization of process parameters of multiple devices through federated learning, and the specific process is as follows:

[0082] The historical K value, weight factor and process parameters of each device are uploaded to the central server, and the data is aggregated after desensitization processing; the FedAWO algorithm is used to optimize the weight factor library, the differential privacy technology is used to aggregate the edge device data, and the central server only updates the weight factor model; the thermal distribution heterogeneity of different devices is identified, and personalized weight configuration is generated, and the temperature gradient weight of the edge device is increased by 5%; the digital twin of the freeze-dryer is constructed, the freeze-drying curve under different weight configurations is simulated, and the optimal parameter combination is predicted. The pressure of the equipment in the sublimation stage is adjusted from 0.25 mbar to 0.22 mbar.

[0083] After testing, the average freeze-drying cycle of multiple devices is shortened by 12%, the energy consumption standard deviation is reduced from 7.3% to 1.8%, and the product quality consistency is significantly improved.

[0084] The specific data comparison is as follows:

[0085] Index Conventional system This embodiment Batch pass rate 88.5％ 99.1％ Average energy consumption 1.2 MW·h 0.94 MW·h Abnormal response speed 150 ms 45 ms

[0086] Example four: This embodiment is further optimized on the basis of example one. In this embodiment, when the mass transfer resistance coefficient K is calculated by the mapping module, the unsteady-state mass transfer equation is introduced:

[0087]

[0088] Wherein, A is crystallinity, and a is inversely calculated from material porosity, and is calculated by the absorbance change rate of near-infrared spectrum in 1200-1400 nm band; β is a phase change correction factor, β = 1 when there is no phase change, and when the glass state area ratio is > 15%, Ea R is the gas constant, and T is the average temperature of the material.

[0089] Due to the high viscosity of liposomes (> 50 cP) leading to freeze-drying collapse, the surface fluctuation relaxation time τ is captured by a visual camera to calculate the viscosity, η = k·ΔT / τ (k is the calibration coefficient), k is obtained by calibrating the known viscosity standard in the freeze dryer; when η > 50 cP, the crystallinity weight is automatically increased to 70%.

[0090] The control process is as follows:

[0091] The pre-freezing stage detects η = 65 cP, and the crystallinity weight is increased from 50% to 70%;

[0092] K value calculation: K = 0.12 × 0.2 + log(0.75) × 0.7 + 15 × 0.1 = 1.18;

[0093] The decision module is because K value > threshold 1.1;

[0094] Start cooling ΔT = -0.5℃ × (1.18-1.1) = -0.4℃.

[0095] After testing, the encapsulation rate is increased from 85% to 93%, and the collapse rate is reduced from 8% to 0.5%.

[0096] Further, due to the abnormal increase of mass transfer resistance at -50℃; add a tunneling effect term in K value calculation:

[0097] K new = K + δ·e -d / λ ;

[0098] Where δ is the barrier height, δ < 0.1, d is the dry layer thickness, h is the Planck constant, m is the water molecule mass, ΔE is the energy barrier.

[0099] Real-time calculation d = 0.2mm (near-infrared inversion), ΔE = 0.15eV;

[0100] Correction amount δ·e -d / λ = 2.5, original K = 1.3, modified K new = 1.05; The decision module is because the K value returns to the threshold, and the vacuum compensation is cancelled. For mRNA vaccine, the mRNA vaccine activity retention rate is increased from 80% to 95%.

[0101] It should be noted that in actual application, linear superposition is used for conventional materials, and non-steady-state equation is used for materials with viscosity η > 50 cP.

[0102] Example five: this embodiment is further optimized on the basis of example one, the existing freeze-drying system cannot monitor the microstructure of ice crystals in real time, leading to the formation of needle-shaped ice crystals in the pre-freezing stage of monoclonal antibody drugs, the fractal dimension D f <1.6, causing protein denaturation and turbidity after reconstitution, with a product rejection rate of up to 15%, the fractal dimension D f Grid counting method.

[0103] In this embodiment, a high-resolution visual camera is added, with a resolution of 5μm and a frame rate of 60fps. An ice crystal morphology recognition module is also set up, which captures the fractal dimension D f of the ice crystals through the high-resolution visual camera and generates a morphology correction coefficient γ. The mapping module introduces a morphology correction term K new1 when calculating the mass transfer resistance coefficient.

[0104] K new1 = K + γ · (1 / D f );

[0105]

[0106] The forced temperature rise rate is limited to ≤0.1℃ / min.

[0107] The specific process is as follows:

[0108] Step 1: Data collection:

[0109] The real-time value of the optical fiber temperature sensor is 0.15℃ / s, the crystallinity is 0.88, and the fractal dimension D f of the ice crystals is 1.55.

[0110] Step 2: K value calculation and correction;

[0111] Basic K value calculation, weights: temperature 60%, crystallinity 30%, vacuum 10%;

[0112] K = 0.15 × 0.6 + kog(0.88) × 0.3 + (8Pa / 10Pa) × 0.1 = 0.09 - 0.017 + 0.08 = 0.153;

[0113] Morphology correction, D f = 1.55, γ = 0.2;

[0114] K new1 = 0.153 + 0.2 × (1 / 1.55) = 0.153 + 0.129 = 0.282.

[0115] Step 3:

[0116] Df = 1.55 <1.6 is detected, triggering the temperature rise rate limit;

[0117] The original temperature curve is raised from -40°C to -20°C at 0.3°C / min; the adjusted temperature rate is reduced to 0.1°C / min, while the ice crystal reconstruction program is started at the same time, and nitrogen gas is injected to change the crystallization path. The historical data of polymerization shows that when Df < 1.6, the crystallinity weight is increased to 40% to reduce the risk of denaturation.

[0118] After actual testing, the test effects of the embodiment scheme and the conventional scheme are as follows:

[0119] Index Conventional scheme This embodiment Promotion range Product unqualified rate 15％ 0.5％ ↓96.7％ Protein activity retention rate 85％ 98％ ↑15.3％ Reconstitution clarity (NTU) 12.5 1.2 ↓90.4％ Freeze-drying cycle 48h 45h ↓6.3％

[0120] Although the preferred embodiments of the present application have been described, those skilled in the art who understand the basic inventive concept can make further changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0121] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. It should be noted that any modifications, equivalent replacements and improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An Internet of Things based intelligent production control system for lyophilized granules, characterized by: The system comprises a multi-source sensing unit and a data processing unit connected with the multi-source sensing unit. The multi-source sensing unit is arranged on a near-infrared spectrum sensor, an optical fiber temperature sensor, a vacuum degree sensor and a visual camera of the freeze-drying equipment, and is used for collecting material temperature gradient change rate, crystallinity data and vacuum fluctuation amplitude. The data processing unit comprises a mapping module, a decision module and an optimization module; the mapping module converts the collected data into a mass transfer resistance coefficient K, and the conversion process comprises the following steps: Step 1: selecting a combination of weight factors of the temperature gradient change rate, the crystallinity logarithmic transformed value and the vacuum fluctuation amplitude according to the type of the material formula; Step 2: linearly superimposing the weighted temperature gradient change rate, the crystallinity logarithmic transformed value and the vacuum fluctuation amplitude to generate the K value; The decision module is used to trigger a reinforcement learning algorithm to adjust the freeze-drying process curve when the real-time K value exceeds a dynamic threshold; and the optimization module is used to optimize a weight factor library through federated learning of historical K value data sets, and to drive digital twin simulation. 2.The intelligent production control system for lyophilized granules based on Internet of Things according to claim 1, characterized in that: The weight factor combination method is as follows: the weight of the temperature gradient change rate in the material accounts for 40-70%, the weight of the crystallinity accounts for 20-50%, and the weight of the vacuum fluctuation accounts for 5-20%; and the dynamic threshold is automatically adjusted in the range of 0.8-1.2 according to the freeze-drying stage. 3.The intelligent production control system for lyophilized granules based on Internet of Things according to claim 1, characterized in that: The mass transfer resistance coefficient K is strongly negatively correlated with the freeze-drying rate. 4.The intelligent production control system for lyophilized granules based on Internet of Things according to claim 1, characterized in that: The decision module reinforcement learning temperature control strategy takes the approximation of the target K value as the core optimization direction, and introduces an energy consumption penalty mechanism to adjust the temperature curve. 5.The intelligent production control system for lyophilized granules based on Internet of Things according to claim 1, characterized in that: The data processing unit further comprises a phase state adaptive module, which identifies a glass state transition region based on the spatial distribution relationship between the crystallinity and the temperature gradient, and automatically increases the weight proportion of the crystallinity data in the K value calculation when the proportion of the glass state region exceeds 15%. 6.The intelligent production control system for lyophilized granules based on Internet of Things according to claim 1, characterized in that: The control system further comprises an ice crystal morphology identification module, which captures the ice crystal fractal dimension through a high-resolution visual camera, generates a morphology correction coefficient, and introduces the morphology correction term when the mapping module calculates the mass transfer resistance coefficient. 7.A method for intelligent production control of freeze-dried particles based on Internet of Things, characterized in that, The system comprises the freeze-dried particle intelligent production control system based on the Internet of Things according to any one of claims 1-6, and the specific steps are as follows: S101: real-time acquisition of material temperature gradient change rate, crystallinity and vacuum fluctuation amplitude original data; S102: matching the preset weight factor according to the current material formula type, and weighting and superimposing the three-element data to generate the mass transfer resistance coefficient K; S103: comparing the real-time K value with the stage dynamic threshold, and if the K value continuously exceeds the threshold, dynamically inhibiting the temperature rise rate or compensating the vacuum degree through reinforcement learning, or if the K value continuously deviates, starting the energy consumption optimization mode; S104: federated learning optimization of the weight factor configuration rule based on the historical K value data set. 8.The method according to claim 7, wherein, The control logic of S103 is as follows: when the K value continuously exceeds the upper limit of the threshold for 3 minutes, the temperature decrease amplitude is linearly increased by the exceeding proportion; and when the K value is lower than the lower limit of the threshold for 10 minutes, a vacuum compensation program is started, and the compensation amount is proportional to the reciprocal of the K value. 9.The method of claim 8, wherein the method further comprises: Also includes S105, the establishment of K value and quality correlation rule method, statistical batch freeze end K value fluctuation variance, variance value > 0.05 when automatically determine batch unqualified, according to the K value fluctuation frequency characteristics to predict the remaining life of vacuum. 10.The method of claim 7, wherein the method is characterized by: In the pre-freezing phase to build K value benchmark, at-38 ℃ ~-40 ℃ state to calculate the weighted average value of crystallinity as K0 benchmark, set the real-time K value deviation K0 benchmark 15-25% for process abnormal alarm threshold.

Images