An on-line monitoring system for the water content of a freeze-dried product

Abstract

The present application relates to the technical field of online monitoring, and more particularly to an online monitoring system for the moisture content of freeze-drying, which comprises an acquisition module, an adjustment module, an updating module, a correction module and an early warning module. The present application generates a pre-freezing index by fusing the temperature characteristics of the pre-freezing stage and the nuclear magnetic resonance relaxation characteristics, thereby adaptively adjusting the sublimation temperature rising rate. The present application online identifies the sublimation root cause type through the time sequence evolution law of the water vapor partial pressure, the cavity pressure and the temperature difference and other parameters in the sublimation stage, and differentiates the analysis duration and the upper limit of the plate temperature according to the diagnostic results. Meanwhile, the present application corrects the previous control parameters through the nuclear magnetic resonance relaxation characteristics and the point temperature difference in the analysis stage, and issues a risk warning based on the analysis index, effectively solving the problems of freeze-drying endpoint misjudgment and residual moisture in the fruit cavity affecting the quality of dried fruit caused by the formation of dense sugar shells, the blockage of mass transfer channels and the resulting heat insulation effect.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology, and in particular to an online monitoring system for the moisture content of freeze-dried food. Background Technology

[0002] With the continuous expansion of freeze-drying technology in the food processing field, high-sugar fruits have become an important processed product category due to their unique flavor and nutritional value. However, traditional freeze-drying processes struggle to effectively address the dense sugar shell formed when sugar migrates with moisture and solidifies on the surface. This sugar shell, on the one hand, blocks mass transfer channels, leading to systematic biases in traditional endpoint monitoring methods; on the other hand, its insulating effect prevents the complete removal of residual moisture. After the product leaves the warehouse, this residual moisture activates at room temperature, causing quality problems such as delayed collapse, stickiness, and mold growth, severely impacting batch yield and storage stability. Overcoming the sugar shell's shielding effect has become a pressing technical challenge in the freeze-drying of high-sugar fruits.

[0003] Chinese Patent Publication No. CN108519398A discloses a method for intelligent detection of moisture content and texture of high-sugar fruits using microwave-driven freeze-drying. The method includes: using a microwave-driven freeze-drying device to perform negative pressure low-frequency microwave-driven freeze-drying on the fruit; taking samples periodically during the process for low-field nuclear magnetic resonance (NMR) analysis to obtain the NMR response signal parameters of the samples, and determining the moisture content and hardness value of the samples; establishing a fruit texture characteristic database using the measured NMR response signal parameters, moisture content, and hardness value, and fitting the correspondence between the database and the degree of drying; then using this correspondence as a model to perform low-field NMR analysis on new samples to be tested, and substituting the detection results into the obtained correspondence to determine the degree of drying.

[0004] Therefore, the existing technology has the following problems: This method relies on staged offline sampling for low-field nuclear magnetic resonance analysis, which can easily lead to the detection results lagging behind the real-time changes in the drying process; This method uses a statistical regression model to predict quality by establishing nuclear magnetic parameters and moisture content and hardness values, which can easily lead to misjudgment of the freeze-drying endpoint due to the dense sugar shell formed on the surface of high-sugar fruits; This method relies on a single microwave-driven freeze-drying device and a fixed parameter combination, which can easily lead to insufficient generalization ability of the established model due to batch differences in materials or process fluctuations. Summary of the Invention

[0005] Therefore, the present invention provides an online monitoring system for freeze-dried moisture content, which overcomes the problems in the prior art caused by the formation of dense sugar shells, blockage of mass transfer channels and heat insulation effect, leading to misjudgment of freeze-drying endpoint and the impact of residual moisture in the fruit cavity on the quality of dried fruit through multi-parameter fusion and dynamic process control.

[0006] To achieve the above objectives, the present invention provides an online monitoring system for freeze-dried food moisture content, comprising:

[0007] The acquisition module is used to acquire the minimum temperature at a preset fruit cavity point at the end of the pre-freezing stage of the high-sugar fruit, the pre-freezing relaxation time, the water vapor partial pressure ratio in the freeze dryer cavity during the sublimation stage under a preset heating rate, the cavity pressure, the temperature difference between the high-sugar fruit and the shelf, the point temperature difference between two preset fruit pulp points during the analysis stage based on the preset analysis time and the preset upper limit of the plate temperature, and the analysis relaxation time at the end of the analysis stage.

[0008] An adjustment module is used to adjust the preset heating rate based on the threshold comparison result of the pre-freezing index, wherein the pre-freezing index is determined based on the minimum temperature, the pre-freezing relaxation time, and a preset temperature coefficient.

[0009] The update module is used to update the preset analysis time and the preset plate temperature limit according to the sublimation root cause type and the preset time step size, wherein the sublimation root cause type is determined based on the water vapor partial pressure ratio, the cavity pressure and the temperature difference value obtained after adjusting the preset heating rate.

[0010] The correction module is used to correct the preset temperature coefficient or the preset time step size according to the root cause type, wherein the root cause type is determined based on the analysis relaxation time and the point temperature difference obtained after updating the preset analysis time and the preset plate temperature limit.

[0011] The early warning module is used to issue an early warning based on the resolution index, wherein the resolution index is determined based on the resolution relaxation time, the point temperature difference, and the cavity pressure, which are re-acquired after correcting the preset temperature coefficient or the preset time step.

[0012] Furthermore, the adjustment module includes:

[0013] A temperature determination unit is used to determine the temperature risk level as a preset risk upper limit or a preset risk lower limit based on the comparison results between the minimum temperature and the preset eutectic point temperature and the preset temperature difference threshold, or to determine the temperature risk level based on the degree of deviation of the minimum temperature from the preset eutectic point temperature.

[0014] The relaxation determination unit is used to determine the relaxation risk level based on the proportion of free water signal, wherein the proportion of free water signal is determined based on the proportion of relaxation signal corresponding to free water during the pre-freezing relaxation time.

[0015] An adjustment unit is used to adjust the preset heating rate according to a preset inverse proportional model based on the threshold comparison result of the pre-freezing index, wherein the pre-freezing index is determined based on the weighted fusion result of the temperature risk degree, the preset temperature coefficient, and the relaxation risk degree.

[0016] Furthermore, the update module includes:

[0017] A boiling determination unit is used to determine whether the sublimation root cause type is internal boiling type based on the time-series changes of the temperature difference and the cavity pressure.

[0018] A sugar shell determination unit is used to determine whether the sublimation root type is a sugar shell germination type based on the relevant characteristics and changing trends of the temperature difference and the water vapor partial pressure ratio when the sublimation root type is not the internal boiling type.

[0019] The hysteresis determination unit is used to determine whether the sublimation root cause type is a heat transfer hysteresis type based on the correlation and fluctuation characteristics of the water vapor partial pressure ratio and the temperature difference when the sublimation root cause type is not the internal boiling type and the sugar shell germination type.

[0020] The update unit is used to update the preset parsing duration and the preset plate temperature limit according to the sublimation root cause type and the preset duration step size.

[0021] Furthermore, the boiling determination unit includes:

[0022] A pressure screening subunit is used to determine the moment of significant pressure based on the comparison between the cavity pressure and the average forward pressure and the average backward pressure, wherein the average forward pressure and the average backward pressure are determined based on the cavity pressure.

[0023] A temperature difference filtering subunit is used to determine the moment when the temperature difference is significant based on the comparison result between the temperature difference value and the mean forward temperature difference and the mean backward temperature difference, wherein the mean forward temperature difference and the mean backward temperature difference are determined based on the temperature difference value.

[0024] A boiling determination subunit is used to determine whether the sublimation root cause type is the internal boiling type based on the threshold comparison result of the average temperature and pressure interval, wherein the average temperature and pressure interval is determined based on the significant pressure moment and the significant temperature difference moment.

[0025] Furthermore, the sugar shell determination unit includes:

[0026] A related calculation subunit is used to determine the temperature-pressure correlation degree based on the correlation characteristics of the water vapor partial pressure ratio and the temperature difference.

[0027] A change calculation subunit is used to determine the partial pressure slope and temperature difference slope respectively based on the temporal changes of the water vapor partial pressure ratio and the temperature difference.

[0028] The sugar shell determination subunit is used to determine, based on the threshold comparison result of the temperature-pressure correlation, that the sublimation root type is the sugar shell germination type when both the partial pressure slope and the temperature difference slope are negative.

[0029] Furthermore, the hysteresis determination unit includes:

[0030] The change determination subunit is used to determine the partial pressure fluctuation value based on the standard deviation of the water vapor partial pressure ratio, and to determine the average temperature difference based on the average value of the temperature difference.

[0031] The hysteresis determination subunit is used to determine whether the sublimation root cause type is the heat transfer hysteresis type based on the threshold comparison results of the partial pressure fluctuation value, the average temperature difference value, and the temperature-pressure correlation.

[0032] Furthermore, the update unit includes:

[0033] The boiling update subunit is used to update the preset analysis time and the preset plate temperature upper limit according to the preset boiling time model and the preset boiling plate temperature model when the sublimation root cause type is the internal boiling type.

[0034] The sugar shell updating subunit is used to update the preset parsing duration and the preset plate temperature limit according to the preset sugar shell duration model and the preset sugar shell plate temperature model when the sublimation root cause type is the sugar shell germination type.

[0035] The hysteresis update subunit is used to update the preset analysis time according to the preset hysteresis time model when the sublimation root cause type is the heat transfer hysteresis type.

[0036] Furthermore, the correction module includes:

[0037] The root cause determination unit is used to determine whether the root cause type is sugar shell pseudo-endpoint type or over-drying and coking type based on the threshold comparison results of the analysis free water ratio and the average temperature difference at the point. The analysis free water ratio is determined based on the proportion of relaxation time greater than the preset relaxation threshold in the analysis relaxation time, and the average temperature difference at the point is determined based on the average value of the temperature difference at the point within the preset monitoring time.

[0038] The correction unit is used to correct the preset temperature coefficient or the preset time step according to the root cause type.

[0039] Furthermore, the correction unit includes:

[0040] The weight correction subunit is used to correct the preset temperature coefficient based on the percentage of free water in the analysis and a preset ratio threshold when the root cause type is the sugar shell pseudo-terminal type.

[0041] The step size correction subunit is used to correct the preset time step size based on the relative deviation between the average temperature difference at the point and the preset point threshold when the root cause type is the over-drying coking type.

[0042] Furthermore, the early warning module includes:

[0043] An index determination unit is used to determine the analytical index based on the percentage of free water, the average temperature difference at the points, and the geometric mean characteristics of the cavity pressure.

[0044] An early warning unit is used to issue an early warning when the analytical index is greater than or equal to a preset index threshold.

[0045] Compared with existing technologies, the advantages of this invention lie in the fact that it generates a pre-freezing index by fusing temperature characteristics and nuclear magnetic resonance relaxation characteristics during the pre-freezing stage. Based on this index, the sublimation heating rate is adaptively adjusted, reducing the risk of sugar migration induced by incomplete pre-freezing from the source of the process. By analyzing the temporal evolution of parameters such as water vapor partial pressure, cavity pressure, and temperature difference during the sublimation stage, three abnormal modes—internal boiling, sugar shell germination, and heat transfer hysteresis—are identified online. The analysis time and plate temperature upper limit are updated differentially based on the diagnostic results, enabling early identification and process compensation of the sugar shell formation process. Simultaneously, the drying effect during the analysis stage is inversely evaluated using nuclear magnetic resonance relaxation characteristics and point temperature differences. Based on this, preceding control parameters are corrected, and risk warnings are issued based on the analysis index. This effectively solves the problems of misjudgment of the freeze-drying endpoint and the impact of residual moisture in the fruit cavity on dried fruit quality caused by dense sugar shell formation, blockage of mass transfer channels, and insulation effects.

[0046] Furthermore, by employing a segmented threshold quantification strategy to determine the temperature risk level, the saturation of risk indicators under extreme conditions is ensured, and control oscillations caused by minor temperature fluctuations are prevented. Since free water is the main carrier of sugar migration, its residual proportion directly reflects the severity of unfrozen water in the pre-freezing stage. By using the minimum value of the free water signal proportion and 1 as the relaxation risk level, the accuracy of pre-freezing effect judgment is improved. In addition, the calculation of the comprehensive pre-freezing index avoids misjudgments caused by a single indicator, and a preset inverse proportional model is used to nonlinearly adjust the sublimation heating rate, so that the heating rate decreases inversely with the increase of the pre-freezing index. This ensures high efficiency under low risk conditions and significantly slows down the heating rate under high risk conditions. Simultaneously, the inverse proportional function has smooth decay characteristics, avoiding the impact of step adjustments on process stability.

[0047] Furthermore, by observing the rapid sublimation of ice crystals within the material during internal boiling, which generates a large amount of water vapor, causing a sudden increase in cavity pressure, and simultaneously releasing latent heat of phase change leading to a sharp rise in material temperature and a rapid decrease in the shelf temperature difference, this anomaly can be identified by detecting the close correlation between the timing of these pressure and temperature abrupt changes. Furthermore, by observing the gradual blockage of mass transfer channels by the dense surface layer during the initial formation of the sugar shell, leading to a continuous decrease in the water vapor partial pressure ratio, and the significant synchronous negative correlation between the material temperature and the shelf temperature difference caused by the insulating effect of the sugar shell, early signals of sugar shell germination can be effectively captured by observing changes in the water vapor partial pressure ratio and temperature difference. The heat transfer lag anomaly occurs when the heat required for the sublimation of ice within the material cannot be transferred from the shelf to the sublimation interface in a timely manner, while the mass transfer channels remain unobstructed and water vapor escape is unimpeded. During this anomaly, the water vapor partial pressure ratio does not remain relatively stable. Simultaneously, due to the persistently insufficient heat input, the material temperature lags behind the shelf temperature for an extended period, resulting in a high temperature difference. Furthermore, because this anomaly is concentrated in the heat transfer process rather than the mass transfer process, there is no significant correlation between the water vapor partial pressure ratio and the temperature difference. Therefore, the occurrence of the heat transfer lag anomaly can be determined by monitoring the statistical and correlation characteristics of the average water vapor partial pressure ratio and temperature difference. In addition, for the internal boiling type, a compensation model that includes forward pressure difference and temperature difference is adopted to extend the analysis time and reduce the upper limit of plate temperature. By increasing the adjustment range, boiling damage is suppressed and sufficient recovery time is given. For the sugar shell germination type, a composite model that integrates temperature and pressure correlation, partial pressure slope and temperature difference slope is adopted to adjust the analysis time and the upper limit of plate temperature. The gentle temperature and pressure synergy softens or destroys the sugar shell in the germination stage. For the heat transfer lag type, the analysis time is extended only according to the ratio of the average temperature difference to the preset average threshold, while keeping the upper limit of plate temperature unchanged. Heat penetration is achieved by exchanging time for heat, realizing a closed-loop linkage of diagnosis and control.

[0048] Furthermore, the sugar-shell pseudo-endpoint type is characterized by a dense sugar shell forming on the surface of high-sugar fruits, causing surface temperature and other indicators to show that drying is complete, but a high proportion of free water remains inside the fruit cavity. This is identified when the proportion of free water exceeds a preset threshold, thus identifying the risk of substandard drying due to sugar shell camouflage. The over-drying and charring type occurs when, despite the basic removal of free water, uneven heat transfer or localized overheating leads to significant temperature differences between different parts of the same fruit, resulting in over-drying or even charring in some areas. Therefore, even when the proportion of free water is below a preset threshold, the over-drying and charring type is identified when the average temperature difference between two preset fruit pulp points exceeds a preset point threshold. This point temperature difference captures information on drying uniformity, avoiding the neglect of quality degradation due to reliance solely on moisture content. In addition, by identifying different root cause types and making specific corrective actions, an adaptive closed-loop improvement is achieved, optimizing preceding parameters from the drying results.

[0049] Furthermore, by analyzing the abnormal features exposed during the analysis phase, the preceding control parameters are optimized in reverse. When it is determined to be a sugar shell pseudo-endpoint type, it indicates that a high proportion of free water remains inside the fruit cavity. This reflects that the weight of temperature risk in the pre-freezing index calculation is too high, causing the system to over-rely on the degree of temperature compliance and ignore the risk of internal free water residue reflected by the relaxation risk. At this time, reducing the preset temperature coefficient makes subsequent batches more sensitive to the relaxation risk. In the correction formula, B / B0 represents the degree of exceedance of the ratio of free water in the analysis relative to the preset ratio threshold. The more serious the exceedance, the more significant the misjudgment of internal free water in the pre-freezing stage, and the greater the temperature coefficient that needs to be reduced. The preset weight correction coefficient is used to control the sensitivity of reduction and ensure that the correction magnitude is proportional to the degree of abnormality. When the condition is determined to be over-drying and charring, it indicates that although the free water content has reached the standard after the analysis, the temperature difference between the two preset pulp points is too large. This reflects uneven heat transfer or local overheating during the drying process. It is also found that the analysis time adjustment step size is too large, causing the system to overcompensate when updating the analysis time, resulting in over-drying in some areas. By shortening the preset time step size, the adjustment range of subsequent batches when updating the analysis time according to the sublimation root cause type can be more refined. In the correction formula, (D-D0) / D0 represents the relative deviation of the average temperature difference at the point relative to the preset point threshold. The larger the deviation, the more severe the over-drying, and the larger the step size reduction ratio needs to be. The preset attenuation coefficient is used to control the intensity of the reduction, ensuring that the step size reduction is proportional to the degree of temperature difference exceeding the standard. The subtraction structure based on 1 ensures that the corrected step size is always positive and will not change abruptly.

[0050] Furthermore, by analyzing the ratio of the free water percentage to its threshold, the degree to which residual free water in the cavity exceeds the safety threshold after parameter correction is reflected, which is a quantitative characterization of the risk of false endpoints in the sugar shell. The ratio of the average temperature difference at each point to its threshold represents the degree of deviation of drying uniformity from acceptable deviations, used to capture risks such as over-drying and coking due to uneven heat transfer. The ratio of the cavity pressure to its threshold reflects the stability of the system's vacuum condition relative to the reference pressure, supplementing the impact of equipment operating status on drying effect. Weighted fusion allows the analytical index to integrate multi-dimensional key characteristics. By comparing the analytical index with the preset index threshold, batches with remaining quality risks can be ultimately intercepted after the system has completed all adaptive adjustments, achieving quality control. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the online monitoring system for freeze-dried moisture content in this embodiment;

[0052] Figure 2 This is a logic diagram of the adjustment unit in this embodiment determining the preset heating rate.

[0053] Figure 3This is the logic diagram for determining whether the sublimation root cause type is internal boiling in the boiling determination subunit of this embodiment.

[0054] Figure 4 This is a logic diagram for the early warning unit in this embodiment to determine when to issue an early warning. Detailed Implementation

[0055] 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.

[0056] 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.

[0057] Please see Figure 1 As shown, this is a schematic diagram of an online monitoring system for freeze-dried food moisture content according to this embodiment. This embodiment provides an online monitoring system for freeze-dried food moisture content, including:

[0058] The acquisition module is used to acquire the minimum temperature at a preset fruit cavity point at the end of the pre-freezing stage of the high-sugar fruit, the pre-freezing relaxation time, the water vapor partial pressure ratio in the freeze dryer cavity during the sublimation stage under a preset heating rate, the cavity pressure, the temperature difference between the high-sugar fruit and the shelf, the point temperature difference between two preset fruit pulp points during the analysis stage based on the preset analysis time and the preset upper limit of the plate temperature, and the analysis relaxation time at the end of the analysis stage.

[0059] An adjustment module, connected to the acquisition module, is used to adjust the preset heating rate based on the threshold comparison result of the pre-freezing index, wherein the pre-freezing index is determined based on the minimum temperature, the pre-freezing relaxation time, and a preset temperature coefficient.

[0060] An update module, which is connected to the acquisition module and the adjustment module respectively, is used to update the preset analysis time and the preset plate temperature limit according to the sublimation root cause type and the preset time step size. The sublimation root cause type is determined based on the water vapor partial pressure ratio, the cavity pressure and the temperature difference value re-acquired after adjusting the preset heating rate.

[0061] A correction module, which is connected to the acquisition module, the adjustment module and the update module respectively, is used to correct the preset temperature coefficient or the preset time step size according to the root cause analysis type, wherein the root cause analysis type is determined based on the analysis relaxation time and the point temperature difference obtained again after updating the preset analysis time and the preset plate temperature upper limit.

[0062] An early warning module, which is connected to the acquisition module and the correction module respectively, is used to issue an early warning based on the analytical index, wherein the analytical index is determined based on the analytical relaxation time, the point temperature difference and the cavity pressure after correcting the preset temperature coefficient or the preset time step size.

[0063] In this embodiment, high-sugar fruits refer to fruit varieties with a sugar content greater than 10%, which are prone to forming a dense sugar shell due to sugar migration and concentration during freeze-drying. An online monitoring system for freeze-drying moisture content is applied to the freeze-drying production process of high-sugar fruits such as strawberries, mangoes, and figs, which are prone to forming a dense sugar bloom on the surface during the later stages of drying. During the freeze-drying of high-sugar fruits, improper temperature control in the pre-freezing stage can lead to unfrozen liquid sugar water remaining in the fruit cavity. This sugar migrates to the surface with the moisture during the sublimation stage, and the dense sugar shell formed by its concentration and solidification under vacuum blocks the water vapor escape channels, causing traditional pressure rise tests to misjudge the drying endpoint. Furthermore, it creates an insulating effect, resulting in a falsely high surface temperature while moisture remains deep within the fruit cavity. This residual moisture, reactivated at room temperature, can easily cause delayed collapse, stickiness, and mold growth in the product within hours.

[0064] In this embodiment, the lowest temperature in the acquisition module is the minimum temperature measured at the end of the pre-freezing stage by inserting a thermocouple at a preset fruit cavity point of the high-sugar fruit. This directly reflects whether the pre-freezing has reached below the eutectic point. It is obtained by continuously recording the thermocouple temperature data throughout the pre-freezing stage and taking the lowest value within 5 minutes before the end of the pre-freezing stage. The pre-freezing relaxation time is the distribution characteristic of the transverse relaxation time of water molecules at the preset fruit cavity point at the end of the pre-freezing stage. This can be determined by spatially encoding the preset fruit cavity point using low-field nuclear magnetic resonance (NMR) technology, collecting the NMR signal of this area, and performing inversion calculations. The water vapor partial pressure ratio is the proportion of water vapor partial pressure in the freeze dryer cavity during the sublimation stage. It is measured synchronously on the same gas by a thermal conductivity vacuum gauge and a capacitor diaphragm vacuum gauge installed on the wall of the drying cavity. The real-time reading of the thermal conductivity vacuum gauge is divided by the capacitance. The ratio obtained from the real-time reading of the diaphragm vacuum gauge is used to determine the pressure. The chamber pressure is the absolute pressure value inside the freeze dryer chamber during the sublimation stage, obtained by real-time pressure data measured by a diaphragm vacuum gauge installed on the wall of the drying chamber. The temperature difference is the difference between the average temperature of the preset fruit cavity points at the same moment during the sublimation stage and the temperature of the freeze dryer shelf, determined by thermocouples inserted near the preset fruit cavity points and temperature sensors of the freeze dryer shelf. The resolution relaxation time is the distribution characteristic of the transverse relaxation time of water molecules obtained by spatial encoding measurement of the same preset fruit cavity area as in the pre-freezing stage, and its acquisition method is the same as that of the pre-freezing relaxation time. The point temperature difference is the difference between the measured temperatures of two preset fruit pulp points on the same fruit during the resolution stage, determined by temperature data from two thermocouples that are simultaneously collected at the preset fruit pulp points.

[0065] In this embodiment, the preset heating rate, preset resolution time, preset plate temperature upper limit, and preset eutectic point temperature are set using a vision recognition system embedded in the equipment, which is deployed at the feeding end of the freeze-drying chamber. At the software level, a YOLOv8 deep learning network model is used. This model has been trained on an expanded dataset containing over 50,000 images simulating actual production environments, achieving a recognition accuracy of 98.7% for different types of fruits. After the material is placed into the chamber, a camera captures multispectral images of the material in real time. The microcomputer runs the model to complete three recognition tasks: category identification, attitude estimation, and loading distribution perception, and sends the recognition results to the lower-level controller. The controller uses the recognition results as the primary key to query the built-in process knowledge base. This knowledge base pre-stores the optimal process parameter sets for each type of material during the freeze-drying process in the pre-freezing, sublimation, and resolution stages, calibrated through experiments. It can load preset heating rates, preset resolution times, preset plate temperature upper limits, and preset eutectic point temperatures for each material as the initial benchmark for the corresponding stage of operation.

[0066] In this embodiment, the visual recognition system consists of a microcomputer and an industrial-grade CMOS camera, deployed at the feeding end of the freeze-drying chamber to achieve real-time perception of material type, shape, and loading status. The deep learning model used in this system adopts the Yolov8 architecture, and its training dataset comes from two sources: first, the publicly available Fruits-360 benchmark dataset, providing basic fruit type samples; second, 50,000 self-collected images simulating actual production environments, covering complex conditions such as water vapor dispersion, lens fogging, chamber metal reflection, different lighting angles, material occlusion, and multi-pose placement, ensuring the model's strong generalization ability to real freeze-drying scenarios. The dataset is divided into training, validation, and test sets in a 7:1.5:1.5 ratio. After a series of optimization training processes, including dynamic adjustment of the learning rate, expansion of the feature extraction layer, and re-clustering based on anchor frame size, an average recognition accuracy of 98.7% was achieved on the test set, demonstrating strong robustness to interference factors such as fogging and reflection. During material identification, the visual recognition system takes as input multispectral images of the material captured in real time by a camera and outputs as the material's category label, geometric morphology features, and pallet density level. When the operator pushes the freeze-drying pallet loaded with material to the designated workstation and triggers the start command, the microcomputer calls the deployed Yolov8 model to perform frame-by-frame inference on the image stream, completing target detection and classification. The inference results are parsed by the post-processing module to generate a unique category code for the current batch of material, such as STRAWBERRY_01, average particle size range, and stacking looseness index. This structured data is then sent to the lower-level controller via a serial communication protocol. The controller uses the category code as the primary key and the morphology and density parameters as secondary keys to perform a composite condition search in the built-in process knowledge base, accurately matching and loading the optimal process operating parameters for that condition as preset benchmark values ​​for each stage of the control loop.

[0067] In this embodiment, the built-in process knowledge base is a structured collection of freeze-drying process parameters. Its construction is based on multi-source data fusion and experimental calibration, used to achieve accurate mapping from material characteristics to process parameters throughout the entire process. During construction, the glass transition temperature and eutectic point temperature of each material are determined using differential scanning calorimetry. Disintegration temperature thresholds and collapse critical values ​​are determined through freeze-drying microscopic observation. The freezing rate, heating power, and resolution temperature are systematically calibrated under different loading thicknesses and tray densities based on single-factor experiments. Furthermore, sensor time-series data of qualified products from historical production batches are extracted, and cluster analysis and association rule mining are used to inversely summarize the optimal parameter ranges under different operating conditions. Simultaneously, for some materials, basic physical property parameters are supplemented through authoritative literature databases and publicly available freeze-drying process datasets, and recommended process curves are generated based on heat and mass transfer numerical simulations. After cleaning, normalization, and outlier removal, the aforementioned data is imported into a relational database or key-value storage engine. A multi-dimensional index is established using the material category code as the primary key and particle size range and padding looseness as secondary keys, forming a complete process record that includes preset eutectic point temperature, preset heating rate for the sublimation stage, preset resolution time for the resolution stage, and preset upper limit of plate temperature. During the recognition process, the input to the process knowledge base is the material category code and morphological feature vector output by the visual recognition system, and the output is a set of preset process parameters that match the input. The lower-level controller receives standardized data frames from the microcomputer, parses out the material category code and auxiliary key information, and performs a multi-dimensional index search in the process knowledge base using this combination key. If a complete match is found, the corresponding preset eutectic point temperature, preset heating rate for the sublimation stage, preset analysis time for the analysis stage, and preset upper limit of plate temperature are directly loaded. If no exact match is found, the nearest neighbor interpolation algorithm is executed to select material parameters with similar physical properties and perform linear correction based on thickness differences to generate the above three preset parameters. If the material category code does not exist in the current knowledge base, the system calls the default safe process template and prompts the operator to enter the rapid calibration process. This process guides the operator to complete three sets of stepped experiments. The experimental results are automatically fitted by a Bayesian optimization algorithm to generate a complete process curve for the new material and stored in the knowledge base. All loaded preset parameters are written to the controller memory in the form of floating-point register values, which serve as the initial given values ​​for the preset eutectic point temperature, preset heating rate for the sublimation stage, preset analysis time for the analysis stage, and preset upper limit of plate temperature, respectively.

[0068] In this embodiment, the preset fruit cavity point is a characteristic location used to insert a temperature sensor to monitor the internal temperature of the fruit cavity during the pre-freezing stage. Depending on the fruit's geometry and thermal conductivity, it is typically set at the center of the fruit's largest cross-section or at the deepest point of the flesh. In this embodiment, it is set 10 mm from the midpoint of the line connecting the stem and calyx towards the fruit cavity, ensuring that the measured temperature is the lowest temperature in the fruit cavity area during the pre-freezing stage. The preset flesh point is a characteristic point used to simultaneously monitor the temperature at two different locations during the analysis stage. Depending on the fruit's size and the temperature distribution gradient during the drying process, it is typically set with one point near the peel and one near the core. In this embodiment, it is set at a surface flesh point 5 mm from the peel and a deep flesh point 5 mm from the core, allowing the temperature difference between the two points to characterize the drying uniformity. The preset temperature coefficient... This is a dimensionless coefficient used to adjust the weight of temperature risk in the pre-freezing index calculation. It depends on the weight of the influence of temperature risk and relaxation risk on the final pre-freezing effect. It can be determined through regression analysis of historical batch data and is usually set between 0.3 and 0.7. In this embodiment, it is set to 0.5, which can balance the contribution of temperature compliance and free water residue in the weighted fusion of pre-freezing index, making the adjustment of heating rate more in line with actual risk. The preset time step is the benchmark increment when adjusting the analysis stage time. It depends on the equipment control accuracy, material drying rate and process stability requirements. It is calibrated according to the minimum adjustable time unit of the equipment and process experience and is usually set between 10 minutes and 30 minutes. In this embodiment, it is set to 15 minutes, which can adjust the analysis time with an appropriate step size after each root cause type diagnosis.

[0069] By fusing temperature characteristics and NMR relaxation features during the pre-freezing stage to generate a pre-freezing index, the sublimation heating rate is adaptively adjusted to reduce the risk of sugar migration induced by incomplete pre-freezing from the source of the process. By analyzing the temporal evolution of parameters such as water vapor partial pressure, cavity pressure, and temperature difference during the sublimation stage, three abnormal modes—internal boiling, sugar shell germination, and heat transfer hysteresis—are identified online. Based on the diagnostic results, the analysis time and plate temperature upper limit are updated differentially to achieve early identification and process compensation of the sugar shell formation process. Simultaneously, the drying effect during the analysis stage is inversely evaluated using NMR relaxation features and point temperature differences. Based on this, preceding control parameters are corrected, and risk warnings are issued based on the analysis index. This effectively solves the problems of misjudgment of the freeze-drying endpoint and the impact of residual moisture in the fruit cavity on dried fruit quality caused by dense sugar shell formation, blockage of mass transfer channels, and insulation effects.

[0070] Please see Figure 2 As shown, this is a logic diagram for the adjustment unit to determine the preset heating rate in this embodiment. In this embodiment, the adjustment module includes:

[0071] A temperature determination unit is used to determine the temperature risk level as a preset risk upper limit when the minimum temperature is greater than a preset eutectic point temperature, and to determine the temperature risk level as a preset risk lower limit when the difference between the preset eutectic point temperature and a preset temperature difference threshold is greater than the minimum temperature. Furthermore, when the difference between the preset eutectic point temperature and the preset temperature difference threshold is less than the minimum temperature, and the minimum temperature is less than the preset eutectic point temperature, the unit determines the temperature risk level based on the degree of deviation of the minimum temperature from the preset eutectic point temperature, wherein W... F =(T0-T1) / T S Among them, W F This refers to the temperature risk level. T0 is the preset eutectic point temperature, T1 is the minimum temperature, and T... S It is a preset temperature difference threshold;

[0072] The relaxation determination unit is used to determine the relaxation risk level based on the proportion of pre-frozen free water, where R C =min(1,C Z ), where R C It is the degree of relaxation risk, C Z It is the proportion of pre-frozen free water, wherein the proportion of pre-frozen free water is determined based on the proportion of relaxation time in the pre-frozen relaxation time that is greater than a preset relaxation threshold;

[0073] An adjustment unit, connected to both the temperature determination unit and the relaxation determination unit, is used to adjust the preset heating rate according to a preset inverse proportional model when the pre-freezing index is greater than a preset pre-freezing threshold. The preset inverse proportional model is V'=V / (1+α×R). Y ), where V' is the adjusted preset heating rate, V is the original preset heating rate, α is the preset rate adjustment coefficient, and R Y It is the pre-freezing index, which is determined based on the weighted fusion result of the temperature risk degree, the preset temperature coefficient, and the relaxation risk degree.

[0074] The preset temperature difference threshold is a temperature range used to define the risk level when the pre-freezing temperature deviates from the eutectic point temperature. By collecting data on the difference between the minimum pre-freezing temperature and the eutectic point temperature for multiple batches, the probability distribution of anomalies such as false endpoints in the sugar shell and over-drying / caramelization in subsequent analysis stages is statistically analyzed within each difference range. The inflection point where the anomaly probability begins to rise significantly is selected as the benchmark value, and corrections are made based on the equipment temperature control accuracy to determine the preset temperature difference threshold. Based on this, the preset temperature difference threshold is typically set between 3℃ and 8℃; in this embodiment, it is set to 5℃. This normalizes the difference between the pre-freezing temperature and the eutectic point temperature to a range of 0 to 1 when linearly interpolating the temperature risk component, thereby quantitatively characterizing the degree of compliance of the pre-freezing temperature. The preset relaxation threshold is used for... To distinguish the relaxation time boundary between free water and bound water, low-field nuclear magnetic resonance (NMR) scanning was performed on multiple fruit cavity samples with different water contents to obtain the relaxation time spectra of each sample. Simultaneously, the free water content of each sample was determined by weighing or drying. Then, the relaxation time spectrum of each sample was divided into multiple consecutive time windows according to the relaxation time from shortest to longest. The signal intensity within each time window was calculated, and a scatter plot was plotted with free water content as the x-axis and the signal intensity of each time window as the y-axis to observe the correlation between signal intensity and free water content in different time windows. When the signal intensity of a certain time window showed a clear synchronous increasing trend with the increase of free water content, the signal intensity before that time window showed a correlation with the free water content. When there is no significant correlation, the initial relaxation time of this time window is the critical point at which the free water signal begins to appear. The statistical average of this critical point in multiple samples is selected as the preset relaxation threshold. Based on this, the preset relaxation threshold is usually set between 80ms and 120ms. In this embodiment, it is set to 100ms, which can effectively extract the free water relaxation signal from the pre-freezing relaxation time spectrum. The preset pre-freezing threshold is a risk threshold used to determine whether the pre-freezing risk needs to trigger an adjustment of the heating rate. By collecting the pre-freezing index calculation values ​​and the corresponding long component proportions (i.e., the proportion of relaxation times greater than the preset relaxation threshold in the pre-freezing relaxation time) of the most recent 15 to 30 historical batches, the pre-freezing index is divided into multiple intervals at 0.1 intervals, and the long component in each interval is statistically analyzed. For batches accounting for more than 5%, the pre-freezing index value at which this proportion begins to rise significantly is selected as the preset pre-freezing threshold. Based on this, the preset pre-freezing threshold is usually set between 0.2 and 0.4. In this embodiment, it is set to 0.3, which enables automatic adjustment of the sublimation heating rate when the pre-freezing index exceeds this threshold. The preset rate adjustment coefficient is a parameter used to control the sensitivity of the pre-freezing index to the adjustment of the heating rate. By obtaining the performance parameters of the freeze dryer shelf temperature control system, the minimum heating rate value at which the equipment can operate stably is determined. The ratio of this minimum value to the reference heating rate is calculated, and then the ratio is corrected according to the degree of matching between the target rate to be achieved when the pre-freezing index is 1 and the target rate. Based on this, the preset rate adjustment coefficient is usually set between 0.5 and 1.The value is between 5 and 0.8 in this embodiment. This allows the heating rate to be reduced to 20% of the baseline rate when the pre-freezing index is at its maximum, and the heating rate to decrease slowly when the pre-freezing index is low, achieving a smooth trade-off between risk and efficiency.

[0075] In this embodiment, the preset risk upper limit is set to 1. When the minimum temperature is greater than the preset eutectic point temperature, it means that the pre-freezing temperature has not reached below the eutectic point, and the free water in the material may not have completely crystallized. At this time, the temperature risk level reaches the highest level and is directly assigned a value of 1 to significantly indicate that there is a serious temperature risk in the pre-freezing process. The preset risk lower limit is set to 0. When the minimum temperature is lower than the difference between the preset eutectic point temperature and the preset temperature difference threshold, it indicates that the pre-freezing temperature has been sufficiently lower than the eutectic point temperature, and the deviation is within the safe range. The temperature risk level is the lowest and is assigned a value of 0, representing that the temperature risk is negligible. The method can define the risk level of different temperature ranges in the calculation of temperature risk. In the process of determining the pre-freezing index by weighted summation, the weight corresponding to the relaxation risk is the contribution ratio of the relaxation risk to the pre-freezing index. By performing logistic regression analysis on the degree of compliance of pre-freezing temperature in historical batches and the probability of subsequent abnormalities such as false endpoints of sugar shells and over-drying and caramelization, the contribution coefficient of temperature risk to abnormal risk is calculated and normalized. It is usually set between 0.2 and 0.8. In this embodiment, it is set to 0.5, which enables the pre-freezing index to integrate macroscopic temperature characteristics and microscopic moisture characteristics.

[0076] By employing a segmented threshold quantification strategy to determine the temperature risk level, the saturation of risk indicators under extreme operating conditions is ensured, and control oscillations caused by minor temperature fluctuations are prevented. Since free water is the main carrier of sugar migration, its residual proportion directly reflects the severity of unfrozen water in the pre-freezing stage. By using the minimum value of the free water signal proportion and 1 as the relaxation risk level, the accuracy of pre-freezing effect judgment is improved. Furthermore, the calculation of the comprehensive pre-freezing index avoids misjudgments caused by a single indicator, and a preset inverse proportional model is used to nonlinearly adjust the sublimation heating rate, causing the heating rate to decrease inversely with the increase of the pre-freezing index. This ensures high efficiency under low risk conditions and significantly slows down the heating rate under high risk conditions. Simultaneously, the inverse proportional function has smooth decay characteristics, avoiding the impact of step adjustments on process stability.

[0077] Specifically, the update module includes:

[0078] A boiling determination unit is used to determine whether the sublimation root cause type is internal boiling based on the time-series changes in the temperature difference and the cavity pressure.

[0079] A sugar shell determination unit, which is connected to the boiling determination unit, is used to determine whether the sublimation root cause type is sugar shell germination type based on the relevant characteristics and changing trends of the temperature difference and the water vapor partial pressure ratio when the sublimation root cause type is not the internal boiling type.

[0080] The hysteresis determination unit is connected to the boiling determination unit and the sugar shell determination unit respectively, and is used to determine whether the sublimation root cause type is a heat transfer hysteresis type based on the correlation characteristics and fluctuation characteristics of the water vapor partial pressure ratio and the temperature difference when the sublimation root cause type is not the internal boiling type and the sugar shell germination type.

[0081] An update unit, which is connected to the boiling determination unit, the sugar shell determination unit, and the hysteresis determination unit respectively, is used to update the preset resolution time and the preset plate temperature upper limit according to the sublimation root cause type and the preset time step size.

[0082] Please see Figure 3 As shown, this is the logic diagram for determining whether the sublimation root cause type is internal boiling in the boiling determination subunit of this embodiment. In this embodiment, the boiling determination unit includes:

[0083] A pressure screening subunit is used to determine the moment when the cavity pressure is simultaneously greater than the average forward pressure and the average backward pressure as a significant pressure moment, wherein the average forward pressure and the average backward pressure are determined based on the average value of the cavity pressure within a preset neighborhood window centered on each moment within a preset time period.

[0084] A temperature difference filtering subunit is used to determine the moment when the temperature difference value is simultaneously greater than the forward temperature difference average value and the backward temperature difference average value as a significant temperature difference moment, wherein the forward temperature difference average value and the backward temperature difference average value are determined based on the average value of the temperature difference value within a preset neighborhood window centered on each moment within a preset time period.

[0085] A boiling determination subunit, which is connected to the pressure screening subunit and the temperature difference screening subunit respectively, is used to determine that the sublimation root cause type is the internal boiling type when the average temperature and pressure interval is less than a preset temperature and pressure threshold. The average temperature and pressure interval is determined based on the average value of the interval between the pressure significant moment and the adjacent temperature difference significant moment.

[0086] Specifically, the sugar shell determination unit includes:

[0087] The relevant calculation subunit is used to calculate the Pearson correlation coefficient between the water vapor partial pressure ratio and the temperature difference within the preset time period, so as to obtain the temperature-pressure correlation.

[0088] The change calculation subunit is used to calculate the linear fitting slope of the water vapor partial pressure ratio and the temperature difference within the preset time period, so as to obtain the partial pressure slope and the temperature difference slope.

[0089] The sugar shell determination subunit is connected to the correlation calculation subunit and the change calculation subunit respectively, and is used to determine that the sublimation root cause type is the sugar shell germination type when the temperature-pressure correlation is greater than a preset correlation threshold and the partial pressure slope and the temperature difference slope are both negative.

[0090] Specifically, the hysteresis determination unit includes:

[0091] The change determination subunit is used to calculate the standard deviation of the water vapor partial pressure ratio within the preset determined time period to obtain the partial pressure fluctuation value, and to calculate the average value of the temperature difference within the preset determined time period to obtain the average temperature difference value.

[0092] The hysteresis determination subunit is used to determine that the sublimation root cause type is the heat transfer hysteresis type when the partial pressure fluctuation value is less than a preset fluctuation threshold, the average temperature difference is greater than a preset average threshold, and the temperature-pressure correlation is less than a preset correlation threshold.

[0093] Specifically, the update unit includes:

[0094] The boiling update subunit is used to update the preset resolution time and the preset plate temperature upper limit according to the preset boiling time model and the preset boiling plate temperature model, respectively, when the sublimation root cause type is the internal boiling type. The preset boiling time model is T... J =T J +t 0B ×(1+C P / P0+C T / T0), where T J ' is the preset parsing duration updated based on the preset boiling duration model, T J This is the preset parsing time before the update, t 0B It is the preset duration step size, C P It is the forward pressure difference, P0 is the mean forward pressure at the moment when the pressure is significant, and C is the forward pressure difference. T It is the forward temperature difference value, T0 is the preset plate temperature step size, and the preset boiling plate temperature model is T. w =T w +t 0w ×(1+C 0w / P0+C T / T0), where T w ' is the upper limit of the preset plate temperature based on the updated preset boiling plate temperature model, T w This is the previous preset upper limit of board temperature, t 0wIt is a preset plate temperature step size, wherein the forward pressure difference is determined based on the difference between the cavity pressure at the moment of significant pressure and the average forward pressure, and the forward temperature difference is determined based on the difference between the temperature difference at the moment of significant temperature difference and the average forward temperature difference.

[0095] The sugar shell update subunit is used to update the preset resolution duration and the preset plate temperature upper limit according to a preset sugar shell duration model and a preset sugar shell plate temperature model when the sublimation root cause type is the sugar shell germination type. The preset sugar shell duration model is T. J ''=T J +t 0B ×[1+X WC ×(L Y / Y A +L T / T A )] , where T J '' is the preset parsing duration updated based on the preset sugar shell duration model, X WC Temperature-pressure correlation, L Y It is the slope of the partial pressure, Y A It is the average value of the water vapor partial pressure ratio over a predetermined time period, L T It is the slope of the temperature difference, T A It is the average value of the temperature difference within a preset time period, and the preset sugar shell plate temperature model is T. w ''=T w +t 0w ×[1+X WC ×(L Y / Y A +L T / T A )], where T w '' is the upper limit of the preset plate temperature based on the preset sugar shell plate temperature model after updating;

[0096] The hysteresis update subunit is used to update the preset analysis time according to a preset hysteresis time model when the sublimation root cause type is the heat transfer hysteresis type, wherein the preset hysteresis time model is T. J '''=T J +A / A0×t 0B Where A is the average temperature difference and A0 is the preset average threshold.

[0097] The preset duration is the length of the time window used to calculate statistical features. It is achieved by collecting data on the duration from the occurrence to the recovery of each anomaly, plotting a probability distribution, and selecting the duration covering more than 80% of the anomaly events as a benchmark. This is then corrected by combining the sensor sampling period and system response time. Based on this, it is typically set between 10 and 30 minutes; in this embodiment, it is set to 20 minutes to ensure sufficiently stable statistical features are extracted within the window. The preset neighborhood window is the neighborhood time width used to calculate local averages. It is achieved by collecting pressure and temperature difference data from multiple batches during stable operation phases, calculating their random fluctuation amplitude and autocorrelation time, and testing the accuracy and delay time of abrupt change detection under different window widths using simulation or historical data. A window width that effectively suppresses noise without significantly delaying detection is selected. Based on this, it is typically set between 1 and 5 minutes; in this embodiment, it is set to 3 minutes to smooth high-frequency noise while preserving the abrupt change characteristics of pressure and temperature differences. The preset temperature and pressure threshold is the critical time interval value used to determine internal boiling anomalies. It is achieved by collecting historical batch data confirmed as internal boiling, extracting the significant pressure moments and adjacent significant temperature difference moments for each case. The time difference at specific moments is used to plot a histogram of the time difference distribution and calculate the cumulative probability. The time difference corresponding to a cumulative probability of 90% is selected as the critical value, and fine-tuned in conjunction with the equipment response delay. Based on this, it is usually set between 1 minute and 3 minutes. In this embodiment, it is set to 2 minutes, which can cover most internal boiling events. The preset correlation threshold is a critical value used to determine the strength of the correlation between the water vapor partial pressure ratio and the temperature difference. By extracting the temperature and pressure correlation values ​​of the normal sublimation stage and the confirmed sugar shell germination stage from historical data, the probability density curves of the two sets of data are plotted, and the two curves are selected. The threshold is set at the intersection of the lines or the value that minimizes the classification error rate. Based on this, it is usually set between 0.6 and 0.8. In this embodiment, it is set to 0.7, which can effectively distinguish between sugar shell germination and normal operating conditions. The preset fluctuation threshold is a critical value used to determine the stability of the water vapor partial pressure ratio. By collecting the standard deviation of the water vapor partial pressure ratio in the historical batches of heat transfer lag and the standard deviation under normal operating conditions, a box plot or cumulative distribution of the two sets of data is drawn, and a critical value that can separate heat transfer lag cases from most normal cases is selected. Based on this, it is usually set between 0.05 and 0.15. In this embodiment, it is set to 0.1. It can accurately identify the characteristic of no significant fluctuation in partial pressure ratio when heat transfer lags. The preset average threshold is a critical value used to determine whether the temperature difference is too large. By collecting the distribution of the average temperature difference in batches with heat transfer lag, and considering both the maximum allowable temperature difference of the material and the heating capacity of the equipment, a value is selected that indicates heat transfer lag without exceeding the material's tolerance limit. Based on this, it is usually set between 5℃ and 15℃; in this embodiment, it is set to 10℃, which can effectively identify continuous large temperature differences caused by insufficient heat transfer. The preset plate temperature step size is the reference increment for adjusting the upper limit of the plate temperature. The minimum adjustable step value for plate temperature control is obtained through the equipment manual or actual measurement. Combined with the influence of different step sizes on the drying effect in process experiments, a step size that can be effectively adjusted without causing overshoot is selected. Based on this, it is usually set between 1℃ and 3℃; in this embodiment, it is set to 2℃, which can adjust the plate temperature with a moderate range.

[0098] In this embodiment, the internal boiling type refers to the material's internal temperature exceeding the boiling point or melting point under the pressure, causing ice crystals to melt instantly or boil violently, generating a large amount of steam that damages the material structure. The sugar shell germination type refers to the formation of a dense thin shell on the surface of high-sugar fruits due to the concentration of sugar as it migrates with water, blocking the mass transfer channels and producing a heat insulation effect. The heat transfer lag type refers to insufficient heat transfer efficiency, resulting in the material temperature being much lower than the shelf temperature, but the mass transfer channels remain unobstructed. This is characterized by a stable water vapor partial pressure ratio without fluctuations, a continuously large temperature difference, and a weak temperature-pressure correlation.

[0099] The internal boiling type occurs when the ice crystals inside the material instantly sublimate, generating a large amount of water vapor, causing a sudden increase in cavity pressure. Simultaneously, the release of latent heat of phase change causes a rapid rise in material temperature, which coincides with a sharp decrease in the temperature difference between the material and the shelf. This anomaly can be identified by detecting the close correlation between the timing of these pressure and temperature changes. The initial formation of the sugar shell is based on the gradual blockage of mass transfer channels by the dense surface layer, leading to a continuous decrease in the water vapor partial pressure ratio. Simultaneously, the insulating effect of the sugar shell causes a significant synchronous negative correlation between material temperature and shelf temperature difference. This allows for the effective capture of early signals of the sugar shell germination type by observing changes in the water vapor partial pressure ratio and temperature difference. The delayed heat transfer type occurs when the heat required for the sublimation of the ice layer inside the material cannot be transferred from the shelf to the sublimation interface in time, but the mass transfer channels remain unobstructed and water vapor escape is not hindered. During this type, the water vapor partial pressure ratio does not remain relatively stable. Furthermore, due to the continuous insufficient heat input, the material temperature lags behind the shelf temperature for a long time, resulting in a high temperature difference. Since this occurs primarily in the heat transfer stage rather than the mass transfer stage, there is no significant correlation between the water vapor partial pressure ratio and the temperature difference. Based on this, the occurrence of heat transfer lag anomalies can be determined by monitoring the statistical and correlation characteristics of the water vapor partial pressure ratio and the average temperature difference. Furthermore, for internal boiling anomalies, a compensation model incorporating forward pressure and temperature differences is used to simultaneously extend the analysis time and lower the upper limit of the plate temperature. This increases the adjustment range to suppress boiling damage and allow sufficient recovery time. For sugar shell germination anomalies, a composite model integrating temperature-pressure correlation, partial pressure slope, and temperature difference slope is used to simultaneously adjust the analysis time and the upper limit of the plate temperature. This uses a gentle temperature-pressure synergy to soften or destroy the sugar shell during the germination stage. For heat transfer lag anomalies, the analysis time is extended only based on the ratio of the average temperature difference to a preset average threshold, while keeping the upper limit of the plate temperature unchanged. This allows for heat penetration through time-exchange, achieving a closed-loop linkage between diagnosis and control.

[0100] Specifically, the correction module includes:

[0101] The root cause determination unit is used to determine the root cause type as sugar shell pseudo-endpoint when the proportion of free water in the analysis is greater than or equal to the preset proportion threshold, and to determine the root cause type as over-drying and coking when the proportion of free water in the analysis is less than the preset proportion threshold and the average temperature difference at the point is greater than the preset point threshold. The proportion of free water in the analysis is determined based on the proportion of relaxation time greater than the preset relaxation threshold in the analysis relaxation time, and the average temperature difference at the point is determined based on the average value of the temperature difference at the point within a preset monitoring period.

[0102] A correction unit, connected to the root cause determination unit, is used to correct the preset temperature coefficient or the preset time step according to the analyzed root cause type.

[0103] The preset ratio threshold is a critical ratio used to determine whether the residual free water exceeds the standard during the analysis stage. By obtaining the free water percentage data at the end of the analysis and combining it with the subsequent stability test results of the corresponding batch, a distribution curve of free water percentage versus quality pass rate is plotted. The inflection point where the pass rate begins to decline significantly is selected as the benchmark value, and then normalization correction is performed through cluster analysis of different fruit varieties. Based on this, it is usually set between 10% and 30%, and in this embodiment, it is set to 20%, which can effectively distinguish between sugar shell pseudo-endpoint abnormalities and truly completely dried batches. The preset point threshold is a critical value used to determine whether the temperature difference between two fruit pulp points is too large during the analysis stage. By combining the sensory evaluation and microstructure observation of the corresponding batch, the correlation between temperature difference and over-drying charring phenomenon is statistically analyzed, and the point where the risk of over-drying charring increases significantly is selected. The corresponding temperature difference is used as a benchmark, and heat transfer simulation is normalized and corrected according to the differences in fruit size and thermal properties. Based on this, it is usually set between 3℃ and 8℃. In this embodiment, it is set to 5℃, which can accurately identify the tendency of over-drying and charring caused by uneven heat transfer or local overheating. The preset monitoring time is the length of the time window used to calculate the average temperature difference at the point. By determining the shortest time required for the temperature difference at the point to stably reflect the drying state, and combining the sensor sampling frequency and system response time, and by analyzing the time series autocorrelation of the temperature difference at the point in historical batches, the time length for the autocorrelation function to decay to below 0.5 is selected as the benchmark. Based on this, it is usually set between 10 minutes and 30 minutes. In this embodiment, it is set to 20 minutes, which can ensure that the calculation of the average temperature difference at the point has sufficient statistical stability.

[0104] The sugar-shell pseudo-endpoint type occurs when the dense sugar shell on the surface of high-sugar fruits indicates that drying is complete, but a high proportion of free water remains inside the fruit cavity. This is identified when the proportion of free water exceeds a preset threshold, thus identifying the risk of incomplete drying due to sugar shell camouflage. The over-drying and charring type occurs when, despite the removal of most free water, uneven heat transfer or localized overheating leads to significant temperature differences between different parts of the same fruit, resulting in over-drying or even charring in some areas. Therefore, even when the proportion of free water is below a preset threshold, this is identified when the average temperature difference between two preset fruit pulp points exceeds a preset point threshold. This point temperature difference captures information on drying uniformity, avoiding the neglect of quality degradation due to reliance solely on moisture content. Furthermore, by identifying different root cause types and making specific corrective actions, an adaptive closed-loop improvement is achieved, optimizing preceding parameters from the drying results.

[0105] Specifically, the correction unit includes:

[0106] The weight correction subunit is used to reduce the preset temperature coefficient according to the ratio of free water in the analysis and the preset ratio threshold when the root cause type is the sugar shell pseudo endpoint type. Wherein, β'=β-b×B / B0, where β' is the preset temperature coefficient after correction, β is the preset temperature coefficient before correction, b is the preset weight correction coefficient, B is the ratio of free water in the analysis, and B0 is the preset ratio threshold.

[0107] The step size correction subunit is used to correct the preset time step size based on the relative deviation between the average temperature difference at the point and the preset point threshold when the root cause type is the over-drying coking type. Here, S'=S×[1-η×(D-D0) / D0], where S' is the corrected preset time step size, S is the original preset time step size, η is the preset attenuation coefficient, D is the average temperature difference at the point, and D0 is the preset point threshold.

[0108] The preset weight correction coefficient is a proportional factor used to control the correction magnitude of the free water ratio in relation to the preset temperature coefficient. It is obtained by collecting the free water ratio from the sugar shell pseudo-endpoint type in multiple historical batches and comparing it with the optimal temperature coefficient correction determined through offline optimization. Linear regression analysis is used to fit the proportional relationship between the two, and the slope value that most significantly improves the batch pass rate after correction is selected as the benchmark. Normalization correction is then performed based on system response time and anti-interference capability. Therefore, it is typically set between 0.1 and 0.5; in this embodiment, it is set to 0.3, which allows for a reduction in the preset temperature coefficient. The degree is proportional to the proportion of residual free water; the preset attenuation coefficient is a proportional factor used to control the reduction of the preset time step by the relative deviation of the average temperature difference at the point. By analyzing the average temperature difference at the point of over-dry coking in historical batches and the appropriate time step reduction determined by simulation optimization, the correspondence between the relative deviation and the reduction magnitude is calculated. The proportional control optimization method is used to determine the proportional coefficient that makes the control effect optimal and does not cause step oscillation. Based on this, it is usually set between 0.2 and 0.8. In this embodiment, it is set to 0.5, which can appropriately reduce the time step according to the temperature difference deviation ratio.

[0109] By analyzing the abnormal features exposed during the analysis phase, the preceding control parameters are optimized in reverse. When the result is determined to be a sugar shell pseudo-endpoint type, it indicates that a high proportion of free water remains inside the fruit cavity. This reflects that the weight of temperature risk in the pre-freezing index calculation is too high, causing the system to over-rely on the degree of temperature compliance and ignore the risk of internal free water residue reflected by the relaxation risk. At this time, reducing the preset temperature coefficient makes subsequent batches more sensitive to the relaxation risk. In the correction formula, B / B0 represents the degree of exceedance of the ratio of free water in the analysis relative to the preset ratio threshold. The more severe the exceedance, the more significant the misjudgment of internal free water in the pre-freezing stage, and the greater the temperature coefficient that needs to be reduced. The preset weight correction coefficient is used to control the sensitivity of the reduction and ensure that the correction magnitude is proportional to the degree of abnormality. When the condition is determined to be over-drying and charring, it indicates that although the free water content has reached the standard after the analysis, the temperature difference between the two preset pulp points is too large. This reflects uneven heat transfer or local overheating during the drying process. It is also found that the analysis time adjustment step size is too large, causing the system to overcompensate when updating the analysis time, resulting in over-drying in some areas. By shortening the preset time step size, the adjustment range of subsequent batches when updating the analysis time according to the sublimation root cause type can be more refined. In the correction formula, (D-D0) / D0 represents the relative deviation of the average temperature difference at the point relative to the preset point threshold. The larger the deviation, the more severe the over-drying, and the larger the step size reduction ratio needs to be. The preset attenuation coefficient is used to control the intensity of the reduction, ensuring that the step size reduction is proportional to the degree of temperature difference exceeding the standard. The subtraction structure based on 1 ensures that the corrected step size is always positive and will not change abruptly.

[0110] Please see Figure 4 As shown, this is a logic diagram for the early warning unit to issue an early warning in this embodiment. In this embodiment, the early warning module includes:

[0111] An index determination unit is used to calculate the geometric average of the ratio of the proportion of free water to the preset ratio threshold, the ratio of the average temperature difference at the point to the preset point threshold, and the ratio of the cavity pressure to the preset pressure threshold, so as to obtain the analysis index.

[0112] An early warning unit, connected to the index determination unit, is used to issue an early warning when the analytical index is greater than or equal to a preset index threshold.

[0113] The preset index threshold is a critical value used to determine whether there is a quality risk in the current batch. By collecting quality index data such as moisture content, rehydration ratio, and sensory evaluation of the finished product after multiple batches of freeze-drying, the batches are divided into qualified and unqualified groups. The resolution index of each group is calculated, and the cumulative probability distribution curve of the two resolution indices is plotted. The resolution index value that makes the identification rate of unqualified batches reach more than 90% and the false alarm rate of qualified batches less than 10% is selected as the benchmark. Then, it is corrected in combination with the long-term stability requirements of the equipment and the tolerance differences of different material varieties. Based on this, it is usually set between 0.6 and 0.9. In this embodiment, it is set to 0.8, which can trigger an early warning in time when the quality risk accumulates to a critical level.

[0114] In this embodiment, the warning is triggered automatically after the freeze-drying process is completed. The system compares the analytical index with a preset index threshold. When the analytical index reaches or exceeds the threshold, the alarm is triggered to indicate that the current batch has quality problems such as excessive residual moisture in the fruit cavity or uneven drying, and that non-conforming products need to be preventively intercepted.

[0115] The ratio of the free water percentage to its threshold reflects the extent to which residual free water in the cavity exceeds the safety threshold after parameter correction, providing a quantitative characterization of the risk of false endpoints in the sugar shell drying process. The ratio of the average temperature difference at different points to its threshold represents the deviation of drying uniformity from acceptable deviations, used to detect risks such as over-drying and coking due to uneven heat transfer. The ratio of the cavity pressure to its threshold reflects the stability of the system's vacuum condition relative to the reference pressure, supplementing the influence of equipment operating status on drying performance. Weighted fusion allows the analytical index to integrate multi-dimensional key characteristics. By comparing the analytical index with preset index thresholds, batches with remaining quality risks can be ultimately intercepted after all adaptive adjustments are completed, achieving quality control.

[0116] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An online monitoring system for the moisture content of freeze-dried food, characterized in that, include: The acquisition module is used to acquire the minimum temperature at a preset fruit cavity point at the end of the pre-freezing stage of the high-sugar fruit, the pre-freezing relaxation time, the water vapor partial pressure ratio in the freeze dryer cavity during the sublimation stage under a preset heating rate, the cavity pressure, the temperature difference between the high-sugar fruit and the shelf, the point temperature difference between two preset fruit pulp points during the analysis stage based on the preset analysis time and the preset upper limit of the plate temperature, and the analysis relaxation time at the end of the analysis stage. An adjustment module is used to adjust the preset heating rate based on the threshold comparison result of the pre-freezing index, wherein the pre-freezing index is determined based on the minimum temperature, the pre-freezing relaxation time, and a preset temperature coefficient. The update module is used to update the preset analysis time and the preset plate temperature limit according to the sublimation root cause type and the preset time step size, wherein the sublimation root cause type is determined based on the water vapor partial pressure ratio, the cavity pressure and the temperature difference value obtained after adjusting the preset heating rate. The correction module is used to correct the preset temperature coefficient or the preset time step size according to the root cause type, wherein the root cause type is determined based on the analysis relaxation time and the point temperature difference obtained after updating the preset analysis time and the preset plate temperature limit. The early warning module is used to issue an early warning based on the resolution index, wherein the resolution index is determined based on the resolution relaxation time, the point temperature difference, and the cavity pressure after correcting the preset temperature coefficient or the preset time step size. The adjustment module includes: A temperature determination unit is used to determine the temperature risk level as a preset risk upper limit or a preset risk lower limit based on the comparison results between the minimum temperature and the preset eutectic point temperature and the preset temperature difference threshold, or to determine the temperature risk level based on the degree of deviation of the minimum temperature from the preset eutectic point temperature. The relaxation determination unit is used to determine the relaxation risk level based on the proportion of free water signal, wherein the proportion of free water signal is determined based on the proportion of relaxation signal corresponding to free water during the pre-freezing relaxation time. An adjustment unit is used to adjust the preset heating rate according to a preset inverse proportional model based on the threshold comparison result of the pre-freezing index, wherein the pre-freezing index is determined based on the weighted fusion result of the temperature risk degree, the preset temperature coefficient, and the relaxation risk degree. The update module includes: A boiling determination unit is used to determine whether the sublimation root cause type is internal boiling type based on the time-series changes of the temperature difference and the cavity pressure. A sugar shell determination unit is used to determine whether the sublimation root type is a sugar shell germination type based on the relevant characteristics and changing trends of the temperature difference and the water vapor partial pressure ratio when the sublimation root type is not the internal boiling type. The hysteresis determination unit is used to determine whether the sublimation root cause type is a heat transfer hysteresis type based on the correlation and fluctuation characteristics of the water vapor partial pressure ratio and the temperature difference when the sublimation root cause type is not the internal boiling type and the sugar shell germination type. The update unit is used to update the preset parsing duration and the preset plate temperature limit according to the sublimation root cause type and the preset duration step size; The correction module includes: The root cause determination unit is used to determine whether the root cause type is sugar shell pseudo-endpoint type or over-drying and coking type based on the threshold comparison results of the analysis free water ratio and the average temperature difference at the point. The analysis free water ratio is determined based on the proportion of relaxation time greater than a preset relaxation threshold in the analysis relaxation time, and the average temperature difference at the point is determined based on the average value of the temperature difference at the point within a preset monitoring period. The correction unit is used to correct the preset temperature coefficient or the preset time step according to the root cause type. The early warning module includes: An index determination unit is used to determine the analytical index based on the percentage of free water, the average temperature difference at the points, and the geometric mean characteristics of the cavity pressure. An early warning unit is used to issue an early warning when the analytical index is greater than or equal to a preset index threshold.

2. The online monitoring system for freeze-dried moisture content according to claim 1, characterized in that, The boiling determination unit includes: A pressure screening subunit is used to determine the moment of significant pressure based on the comparison between the cavity pressure and the average forward pressure and the average backward pressure, wherein the average forward pressure and the average backward pressure are determined based on the cavity pressure. A temperature difference filtering subunit is used to determine the moment when the temperature difference is significant based on the comparison result between the temperature difference value and the mean forward temperature difference and the mean backward temperature difference, wherein the mean forward temperature difference and the mean backward temperature difference are determined based on the temperature difference value. A boiling determination subunit is used to determine whether the sublimation root cause type is the internal boiling type based on the threshold comparison result of the average temperature and pressure interval, wherein the average temperature and pressure interval is determined based on the significant pressure moment and the significant temperature difference moment.

3. The online monitoring system for freeze-dried moisture content according to claim 2, characterized in that, The sugar shell determination unit includes: A related calculation subunit is used to determine the temperature-pressure correlation degree based on the correlation characteristics of the water vapor partial pressure ratio and the temperature difference. A change calculation subunit is used to determine the partial pressure slope and temperature difference slope respectively based on the temporal changes of the water vapor partial pressure ratio and the temperature difference. The sugar shell determination subunit is used to determine, based on the threshold comparison result of the temperature-pressure correlation, that the sublimation root type is the sugar shell germination type when both the partial pressure slope and the temperature difference slope are negative.

4. The online monitoring system for freeze-dried moisture content according to claim 3, characterized in that, The hysteresis determination unit includes: The change determination subunit is used to determine the partial pressure fluctuation value based on the standard deviation of the water vapor partial pressure ratio, and to determine the average temperature difference based on the average value of the temperature difference. The hysteresis determination subunit is used to determine whether the sublimation root cause type is the heat transfer hysteresis type based on the threshold comparison results of the partial pressure fluctuation value, the average temperature difference value, and the temperature-pressure correlation.

5. The online monitoring system for freeze-dried moisture content according to claim 4, characterized in that, The update unit includes: The boiling update subunit is used to update the preset analysis time and the preset plate temperature upper limit according to the preset boiling time model and the preset boiling plate temperature model when the sublimation root cause type is the internal boiling type. The sugar shell updating subunit is used to update the preset parsing duration and the preset plate temperature limit according to the preset sugar shell duration model and the preset sugar shell plate temperature model when the sublimation root cause type is the sugar shell germination type. The hysteresis update subunit is used to update the preset analysis time according to the preset hysteresis time model when the sublimation root cause type is the heat transfer hysteresis type.

6. The online monitoring system for freeze-dried moisture content according to claim 5, characterized in that, The correction unit includes: The weight correction subunit is used to correct the preset temperature coefficient based on the percentage of free water in the analysis and a preset ratio threshold when the root cause type is the sugar shell pseudo-terminal type. The step size correction subunit is used to correct the preset time step size based on the relative deviation between the average temperature difference at the point and the preset point threshold when the root cause type is the over-drying coking type.

Images