Freeze-drying process control method and system based on intelligent algorithm and storage medium

Abstract

The invention relates to the technical field of freeze-drying process control, and discloses a freeze-drying process control method and system based on an intelligent algorithm and a storage medium. The method comprises the following steps: arranging temperature sensor arrays inside and outside a building material, and collecting temperature gradient and mass change rate data to form a feature set; based on the feature set, using a chaos butterfly effect algorithm to identify a nonlinear transition point, and predicting an ice crystal growth trend; dividing the material into an inner core area, a middle area and a surface layer area according to prediction, and implementing differential cooling control; and a control instruction is generated according to the area state, and the cold press unit, the vacuum pump set and the heating array are driven to execute cooperative control. The problems of uneven ice crystal distribution and material structure damage caused by extensive temperature control in a traditional building material freeze-drying process are solved. And meanwhile, the technical problems of energy waste and freeze-drying period prolonging caused by lack of identification capability for nonlinear transition points in the freeze-drying process in the prior art are effectively solved.

Description

Technical Field

[0001] This application relates to the field of freeze-drying process control technology, and in particular to a freeze-drying process control method, system and storage medium based on intelligent algorithms. Background Technology

[0002] Freeze-drying technology, as an important low-temperature drying method, is widely used in the processing of building materials, especially in the preparation of high-performance building materials such as ceramics and lightweight concrete, where it has unique advantages. Traditional freeze-drying techniques mainly rely on empirical parameter settings, controlling the freeze-drying process by pre-setting temperature curves, pressure curves, and time periods. In these traditional processes, operators typically select control parameters based on material type and batch size by consulting experience tables, then set a constant temperature gradient and chamber pressure value, relying on a simple single-loop PID controller to maintain the setpoint. These methods can achieve relatively stable results when processing samples with simple structures and small dimensions, but as building materials develop towards high performance and composite materials, the limitations of traditional control methods are becoming increasingly apparent.

[0003] However, traditional control methods lack real-time monitoring of the internal temperature field and moisture distribution of materials, resulting in the inability to optimize control parameters based on the actual internal state of the material. This leads to uneven ice crystal formation and sublimation processes, resulting in inconsistent product quality. Secondly, single temperature and pressure curves cannot meet the differentiated freeze-drying requirements of different regions of the material. Inconsistent freeze-drying degrees often occur between internal and surface areas, causing damage to the internal structure or excessive surface drying. Thirdly, traditional control systems lack the ability to identify nonlinear change points during freeze-drying. Key turning points such as the critical point of ice crystal formation, eutectic transition point, and main sublimation initiation point are often overlooked, missing optimal control opportunities, reducing energy efficiency, and prolonging the freeze-drying cycle. Furthermore, the level of coordinated control between the vacuum system, refrigeration system, and heating system is low. Each subsystem operates independently, lacking overall optimization, making it difficult to achieve precise control and efficient energy utilization.

[0004] Developing intelligent algorithms to identify nonlinear transition points during freeze-drying has become an important research direction for improving freeze-drying control accuracy. Simultaneously, achieving differentiated control over different regions of building materials and precise adjustment of vacuum system parameters are also key to improving freeze-drying efficiency and energy utilization. With the expanding application areas and increasing performance requirements of freeze-dried building materials, developing a method capable of real-time monitoring of the material's internal state and implementing intelligent control based on the monitoring data has become particularly important. Furthermore, current technologies lack research on applying chaotic dynamics theory to freeze-drying process prediction. While the chaotic butterfly effect algorithm has unique advantages in identifying nonlinear transition points in the system, its effective application in the freeze-drying process of building materials remains insufficiently addressed. Summary of the Invention

[0005] This application provides a method, system, and storage medium for controlling the freeze-drying process based on intelligent algorithms, which addresses the problems of uneven ice crystal distribution and material structural damage caused by crude temperature control in traditional freeze-drying processes for building materials. Simultaneously, it effectively overcomes the technical problems of energy waste and prolonged freeze-drying cycles caused by the lack of ability to identify nonlinear transition points in the freeze-drying process in existing technologies.

[0006] In a first aspect, this application provides a freeze-drying process control method based on intelligent algorithms. The method includes: arranging a multi-point temperature sensor array inside and on the surface of building materials to collect temperature gradient data and combine it with the mass change rate to form a freeze-drying process feature dataset; based on the freeze-drying process feature dataset, identifying nonlinear transition points in the freezing process of building materials using a chaotic butterfly effect algorithm to predict ice crystal growth trends; dividing the building materials inside the freeze-drying chamber into a core region, an intermediate region, and a surface region according to the ice crystal growth trends, and implementing differentiated cooling curve control for each region; generating multiple control commands according to the differentiated states of each region to drive the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control.

[0007] In a first embodiment of the first aspect, the arrangement of a multi-point temperature sensor array inside and on the surface of the building material to collect temperature gradient data and combine it with the mass change rate to form a freeze-drying process characteristic dataset includes: PT100 platinum resistance sensors and K-type thermocouples were pre-embedded in ceramic-based building materials and lightweight concrete specimens to collect temperature signals during the freeze-drying process. The mass change of building materials was recorded using a strain gauge electronic weighing system located under the freeze dryer rack, and the rate of water sublimation during the freeze drying process was calculated. The temperature signal is sampled in real time at a sampling frequency of 5Hz, and after processing by a preamplifier and signal conditioning circuit, power frequency interference and thermal noise are removed. The water sublimation rate data is digitized using an A / D converter and then compensated and corrected by combining the readings of the freeze-drying chamber pressure sensor. The spatial difference algorithm is used to calculate the temperature gradient and its trend between points inside the building material using the processed temperature signal data. The temperature gradient values ​​and the moisture sublimation rate data are matched by timestamps to form a freeze-drying process feature dataset that includes the temperature field evolution and moisture migration characteristics of the entire freeze-drying process.

[0008] In the second embodiment of the first aspect, the step of identifying nonlinear transition points in the freezing process of building materials and predicting ice crystal growth trends based on the freeze-drying process feature dataset and using a chaotic butterfly effect algorithm includes: Time-series sampling of temperature data in the freeze-drying process feature dataset is performed to construct the initial state matrix of the chaotic butterfly effect algorithm; Add small random perturbations to the initial state matrix to generate multiple sets of perturbation initial conditions, and calculate the evolution results according to the core iteration rules of the chaotic butterfly effect. Calculate the average Euclidean distance between multiple sets of evolution results, plot the distance change curve over time, and mark the corresponding time point as the chaotic critical point of the freeze-drying process of building materials when the distance growth rate exceeds the preset threshold. A time-delay coordinate transformation is performed on the freeze-drying process feature dataset. The temperature value at the current moment is compared with the temperature value after a certain delay to generate a two-dimensional phase diagram, from which the spiral structure and singularity of the building material freezing process are identified. Align the chaotic critical point and the singular point on the time axis, and filter out the moment points that simultaneously satisfy both characteristics to determine them as nonlinear transition points in the freezing process of building materials. Based on the temperature change rate and moisture migration rate data before and after the nonlinear transition point, the recursive prediction module of the chaotic butterfly effect algorithm is used to calculate the ice crystal growth path and distribution density in the future time period in parallel under multiple different initial conditions, forming a prediction result set of the ice crystal growth trend of building materials.

[0009] In the third embodiment of the first aspect, based on the temperature change rate and moisture migration rate data before and after the nonlinear transition point, the recursive prediction module of the chaotic butterfly effect algorithm calculates the ice crystal growth path and distribution density in parallel under multiple different initial conditions to form a prediction result set of the ice crystal growth trend of building materials, including: A fixed-length data segment of temperature change rate and moisture migration rate is extracted from before and after the nonlinear transition point to construct a state description vector of the freeze-drying process of building materials. Add a small perturbation that conforms to a normal distribution to the state description vector to generate multiple sets of initial condition vectors, which serve as the starting point for recursive prediction of the chaotic butterfly effect. The multiple sets of initial condition vectors are input into the six-dimensional state evolution device of the chaotic butterfly effect, and the temperature field and moisture field at different locations within the building material are obtained over time through step-by-step iterative calculation. Statistical analysis was performed on the change trajectory to calculate the probability distribution of temperature and moisture content at each spatial location point, and a thermal map of the probability of ice crystal formation during the freeze-drying process of building materials was constructed. Based on the aforementioned probability heat map, cluster analysis is performed on the internal regions of building materials to identify regions with similar ice crystal formation probabilities and classify them into ice crystal growth regions with similar freezing characteristics. Trend fitting is performed on the temperature-moisture trajectory of the ice crystal growth region to calculate the ice crystal growth rate and direction of each region, generating a set of predicted results for the ice crystal growth trend of building materials that includes spatial distribution and temporal evolution.

[0010] In the fourth embodiment of the first aspect, the step of dividing the building materials inside the freeze-drying cavity into a core region, an intermediate region, and a surface region according to the ice crystal growth trend, and implementing differentiated cooling curve control for each region, includes: Based on the ice crystal growth trend prediction result set, the internal temperature conduction characteristics of building materials are evaluated, and the thermal conductivity distribution map of building materials is determined. Based on the thermal conductivity distribution map and the actual distribution of temperature sensors, a hierarchical clustering method is used to divide the building materials into core area, intermediate area and surface area, generating a spatial area division scheme for building materials. Different cooling curves with different slopes are designed for the core region, the intermediate region and the surface region respectively. A slow cooling slope is used for the core region, a medium cooling slope is used for the intermediate region and a fast cooling slope is used for the surface region, forming a set of differentiated cooling curves for the three regions. The set of differentiated cooling curves in the three zones is converted into a sequence of temperature setpoints within the control cycle, and the target temperature value at each control moment is calculated by piecewise linear interpolation. The target temperature value is compared with the real-time measured temperature value to calculate the temperature control deviation of each area, and the cooling power distribution of each area is dynamically adjusted according to the magnitude of the deviation. Based on the dynamically adjusted cooling power distribution, the speed control parameters of the refrigeration compressor and the opening parameters of the refrigerant flow control valve are calculated to generate a multi-zone coordinated cooling control parameter set, and to execute differentiated cooling curve control for each zone of the building materials.

[0011] In the fifth embodiment of the first aspect, the step of generating multiple control commands according to the differentiated states of each region to drive the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control includes: Real-time temperature data of the core region, intermediate region and surface region are collected, and the deviation between the data and the target value of the cooling curve of each region is calculated to form a temperature control status evaluation matrix. Based on the temperature control status evaluation matrix, fuzzy rule reasoning is used to assign cooling capacity weights and heating power weights to each region, generating a regional energy allocation scheme. The regional energy distribution scheme is broken down into three sets of parameters: refrigeration compressor speed, electronic expansion valve opening, and evaporator fan speed, forming a zoned control command for the refrigeration compressor unit. Based on the moisture evaporation rate data of the core region, intermediate region and surface region, the required vacuum environment parameters are calculated, and coordinated control commands for vacuum pump speed and vacuum valve opening are generated. The temperature gradient directions of the core region, intermediate region and surface region are analyzed to determine the power distribution pattern of the heating plate array and generate power adjustment control commands for each heating channel. The control commands for the refrigeration compressor unit, vacuum pump unit, and heating plate array are synchronously transmitted to each execution device via an industrial bus, and the response status of the devices is monitored to complete multi-level coordinated control of the equipment at each stage of freeze-drying.

[0012] In the sixth embodiment of the first aspect, the step of calculating the required vacuum environment parameters based on the moisture evaporation rate data of the core region, intermediate region, and surface region, and generating coordinated control commands for the vacuum pump speed and vacuum valve opening, includes: The moisture evaporation rate data of the core region, intermediate region and surface region are combined into a moisture load matrix to calculate the total amount of water vapor per unit time. Based on the total water vapor volume and the freeze-drying chamber volume, the target chamber pressure value required to maintain the optimal sublimation rate is calculated. The target pressure value is compared with the current measured pressure value of the cavity, the pressure deviation and the rate of change of deviation are calculated, and a pressure control state vector is constructed. The pressure control state vector is input into the pressure-flow control calculation module to calculate the required vacuum system pumping rate and flow control parameters. Based on the pumping rate, the required rotational speed is calculated in reverse using the vacuum pump characteristic curve, generating a coordinated control command for the vacuum pump rotational speed and vacuum valve opening.

[0013] Secondly, this application provides a freeze-drying process control system based on intelligent algorithms, the freeze-drying process control system based on intelligent algorithms comprising: The data acquisition module is used to deploy a multi-point temperature sensor array on the interior and surface of building materials, collect temperature gradient data, and combine it with the mass change rate to form a feature dataset of the freeze-drying process. The identification module is used to identify the nonlinear transition points in the freezing process of building materials based on the freeze-drying process feature dataset and predict the ice crystal growth trend through the chaotic butterfly effect algorithm. The control module is used to divide the building materials inside the freeze-drying cavity into a core area, an intermediate area and a surface area according to the ice crystal growth trend, and to implement differentiated cooling curve control for each area. The generation module is used to generate multiple control commands according to the differentiated states of each region, and drive the refrigeration compressor unit, vacuum pump unit and heating plate array to perform multi-level coordinated control.

[0014] Thirdly, a freeze-drying process control device based on intelligent algorithms is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the freeze-drying process control device based on intelligent algorithms to execute the aforementioned freeze-drying process control method based on intelligent algorithms.

[0015] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described freeze-drying process control method based on intelligent algorithms.

[0016] The technical solution provided in this application achieves precise monitoring of the internal temperature gradient of building materials by arranging a multi-point temperature sensor array inside and on the surface of the materials. Combined with mass change rate data, this forms a comprehensive feature dataset reflecting the freeze-drying process, solving the problem of insufficient monitoring of the internal state of materials in traditional freeze-drying technology and providing a high-quality data foundation for intelligent control. Based on these feature datasets, this invention successfully identifies nonlinear transition points in the freezing process of building materials using a chaotic butterfly effect algorithm, accurately predicting ice crystal growth trends and significantly improving the predictability and scientific rigor of freeze-drying process control. The algorithm's features and the technical features of the temperature sensor array mutually support each other. The high-precision temperature gradient data provided by the sensors creates ideal input conditions for the chaotic butterfly effect algorithm, while the algorithm's in-depth mining of this data maximizes the value of the sensor network, forming a virtuous cycle. By dividing the building materials into core, intermediate, and surface regions and implementing differentiated cooling curve control for each region, this invention successfully solves the problem of internal stress and structural damage caused by mismatched freezing rates in different regions in traditional freeze-drying processes, significantly improving the uniformity and integrity of freeze-dried products. This regional division strategy is closely integrated with the ice crystal growth trend predicted by the chaotic butterfly effect algorithm. The nonlinear transition points accurately identified by the algorithm provide a scientific basis for regional division, while the implementation of differentiated control verifies and strengthens the accuracy of the algorithm's predictions. The two work synergistically to achieve a significant technical effect. Finally, this invention generates multiple control commands according to the differentiated states of each region, driving the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control. This achieves the overall optimized operation of each subsystem of the freeze-drying equipment, solving the problems of energy waste and insufficient control precision caused by the independent operation of each system in traditional control methods. This multi-level coordinated control mechanism forms a complete closed-loop control link with the aforementioned intelligent algorithm. The algorithm predicts system changes by processing sensor data, the control commands scientifically adjust equipment parameters based on the prediction results, and the equipment operating status is fed back to the sensor network to form new data inputs, enabling the entire control process to continuously improve itself. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of one embodiment of the freeze-drying process control method based on intelligent algorithms in this application. Figure 2 This is a schematic diagram of one embodiment of the freeze-drying process control system based on intelligent algorithms in this application. Detailed Implementation

[0019] This application provides a freeze-drying process control method, system, and storage medium based on intelligent algorithms. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the freeze-drying process control method based on intelligent algorithms in this application includes: Step S101: Arrange a multi-point temperature sensor array inside and on the surface of the building material, collect temperature gradient data and combine it with the mass change rate to form a feature dataset of the freeze-drying process. Step S102: Based on the freeze-drying process feature dataset, the nonlinear transition points in the freezing process of building materials are identified by the chaotic butterfly effect algorithm, and the ice crystal growth trend is predicted. Step S103: Based on the ice crystal growth trend, the building materials inside the freeze-drying cavity are divided into the core area, the middle area and the surface area, and differentiated cooling curve control is implemented for each area. Step S104: Generate multiple control commands according to the differentiated status of each region, and drive the refrigeration compressor unit, vacuum pump unit and heating plate array to perform multi-level coordinated control.

[0021] It is understood that the executing entity of this application can be a freeze-drying process control system based on intelligent algorithms, or it can be a terminal or a server; no specific limitation is made here. This application's embodiments use a server as an example for illustration.

[0022] Specifically, a multi-point temperature sensor array is deployed inside and on the surface of the building materials. These sensors include PT100 platinum resistance sensors embedded inside the materials and K-type thermocouples on the surface. These sensors are distributed at predetermined locations to ensure comprehensive capture of the temperature distribution inside and outside the materials. The acquired temperature signals are amplified by a preamplifier and then processed by a signal conditioning circuit to remove power frequency interference and thermal noise, achieving high-precision temperature data acquisition. Specifically, after acquiring multi-point temperature data, a spatial difference algorithm is used to calculate the temperature difference between any two adjacent measuring points, divided by the spatial distance, to obtain the temperature gradient value. Simultaneously, a strain gauge electronic weighing system under the freeze dryer shelf continuously records the mass change of the building materials. The mass change data over time is differentially calculated to obtain the moisture sublimation rate. The weighing system data is digitized by an A / D converter and then compensated and corrected using readings from the cavity pressure sensor to eliminate environmental interference factors. Finally, the temperature gradient value and moisture sublimation rate data are precisely aligned using timestamps to form a freeze-drying process characteristic dataset containing temperature field evolution and moisture migration characteristics.

[0023] Based on the aforementioned freeze-drying process feature dataset, the chaotic butterfly effect algorithm is applied to identify nonlinear transition points in the freezing process of building materials and predict ice crystal growth trends. The algorithm first samples temperature data over time to construct an initial state matrix, then adds small random perturbations to generate multiple sets of perturbation initial conditions. Evolution results are calculated for these initial conditions according to the core iterative rules of the chaotic butterfly effect, and the average Euclidean distance between multiple sets of evolution results is calculated and a distance-time curve is plotted. When the distance growth rate exceeds a preset threshold, the corresponding time point is marked as a chaotic critical point. Simultaneously, a time-delay coordinate transformation is performed on the freeze-drying process feature dataset to generate a two-dimensional phase diagram, from which spiral structures and singularities are identified. The chaotic critical points and singularities are aligned on the time axis, and time points that simultaneously satisfy both characteristics are selected as nonlinear transition points in the freezing process of building materials. Based on the temperature change rate and moisture migration rate data before and after these nonlinear transition points, the recursive prediction module of the chaotic butterfly effect algorithm calculates the ice crystal growth path and distribution density in parallel under multiple different initial conditions, forming a prediction result set for the ice crystal growth trend of building materials.

[0024] Based on the ice crystal growth trend, the building materials inside the freeze-drying chamber are divided into a core region, a middle region, and a surface region, and differentiated cooling curve control is implemented for each region. Specifically, firstly, based on the ice crystal growth trend prediction result set, the internal temperature conduction characteristics of the building materials are evaluated to determine the thermal conductivity distribution map. Then, based on the thermal conductivity distribution map and the actual distribution of temperature sensors, a hierarchical clustering method is used to divide the building materials into a core region, a middle region, and a surface region, generating a spatial region division scheme. Different cooling curves with different slopes are designed for these three regions: a slow cooling slope for the core region, a medium cooling slope for the middle region, and a fast cooling slope for the surface region, forming a set of differentiated cooling curves for the three regions. These cooling curves are converted into a temperature setpoint sequence within the control cycle, and the target temperature value at each control moment is calculated through piecewise linear interpolation. By comparing the target temperature value with the real-time measured temperature value, the temperature control deviation of each region is calculated, and the cooling power distribution of each region is dynamically adjusted according to the magnitude of the deviation. Based on dynamically adjusted cooling power allocation, the speed control parameters of the refrigeration compressor and the opening parameters of the refrigerant flow control valve are calculated to generate a set of multi-zone coordinated cooling control parameters.

[0025] Multiple control commands are generated based on the differentiated states of each region to drive the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control. First, real-time temperature data for the core, intermediate, and surface regions is collected, and the deviation is calculated against the target values ​​of the cooling curves for each region to form a temperature control status evaluation matrix. Based on this matrix, fuzzy rule reasoning is used to assign refrigeration capacity weights and heating power weights to each region, generating a regional energy allocation scheme. This scheme is broken down into three sets of parameters: refrigeration compressor speed, electronic expansion valve opening, and evaporator fan speed, forming zoned control commands for the refrigeration compressor unit. Based on the moisture evaporation rate data of the core, intermediate, and surface regions, the required vacuum environment parameters are calculated, generating coordinated control commands for vacuum pump speed and vacuum valve opening. The temperature gradient directions of the core, intermediate, and surface regions are analyzed to determine the power distribution pattern of the heating plate array, generating power adjustment control commands for each heating channel. Finally, the control commands for the refrigeration compressor unit, vacuum pump unit, and heating plate array are synchronously transmitted to each executing device via an industrial bus, and the device response status is monitored, completing the multi-level coordinated control of the equipment at each stage of freeze-drying.

[0026] In one specific embodiment, the process of performing step S101 may specifically include the following steps: PT100 platinum resistance sensors and K-type thermocouples were pre-embedded in ceramic-based building materials and lightweight concrete specimens to collect temperature signals during the freeze-drying process. The mass change of building materials was recorded using a strain gauge electronic weighing system located under the freeze dryer rack, and the rate of water sublimation during the freeze drying process was calculated. The temperature signal is sampled in real time at a sampling frequency of 5Hz. After processing by a preamplifier and signal conditioning circuit, power frequency interference and thermal noise are removed. The moisture sublimation rate data is digitized using an A / D converter and then compensated and corrected by combining the readings of the freeze-drying chamber pressure sensor. The spatial difference algorithm is used to calculate the temperature gradient and its trend between points inside the building material using the processed temperature signal data. By matching temperature gradient values ​​and moisture sublimation rate data with timestamps, a freeze-drying process feature dataset containing the temperature field evolution and moisture migration characteristics of the entire freeze-drying process is formed.

[0027] Specifically, temperature sensors were embedded in ceramic-based building materials and lightweight concrete specimens, specifically PT100 platinum resistance thermometers and K-type thermocouples. The PT100 platinum resistance thermometers offer high accuracy (0.1℃) and a temperature range of -200℃ to 850℃, making them particularly suitable for monitoring temperature changes near the freezing point within building materials. The K-type thermocouples have a temperature range of -270℃ to 1372℃, a fast response time, and are suitable for monitoring rapid temperature changes on material surfaces. The sensors were arranged radially, with equal spacing from the material center outwards to form a grid pattern, ensuring the capture of temperature changes at different depths. The density was determined based on the thermal conductivity of the building materials, with materials with lower thermal conductivity requiring denser placement. Mass change monitoring utilized a strain gauge electronic weighing system located beneath the freeze dryer shelves. This system consisted of a weighing sensor, an amplifier circuit, and a data acquisition unit. The weighing sensor was a strain gauge structure with a measurement range of 0-50 kg and a resolution of 0.1 g, accurately recording the mass loss of building materials due to moisture sublimation during freeze-drying. The freeze dryer rack is directly fixed to the weighing system, and the mass change of the building materials is directly converted into an electrical signal output. The moisture sublimation rate is calculated using the differential method, that is, the mass change value in each time interval is divided by the time interval to obtain the amount of moisture sublimation per unit time. During temperature signal acquisition, a sampling frequency of 5Hz is used for real-time data acquisition. This frequency is sufficient to capture the temperature change characteristics during the freeze-drying process of building materials without generating too much redundant data. The raw temperature signal is first amplified by a preamplifier, increasing the millivolt-level sensor signal to the volt level, with the gain typically set to 100-1000 times. The amplified signal is then processed by a signal conditioning circuit, mainly including a low-pass filter, a notch filter, and mean smoothing. The low-pass filter cutoff frequency is set to 2Hz to filter out high-frequency noise; the notch filter center frequency is set to 50Hz to specifically filter out power frequency interference; the mean smoothing algorithm uses a sliding window method with a window width of 5 data points to reduce the impact of random fluctuations. After these processing steps, the temperature signal noise is significantly reduced, and the signal-to-noise ratio is improved by approximately 20dB. Moisture sublimation rate data processing involves analog-to-digital conversion and pressure compensation. The analog signal output by the strain gauge electronic weighing system is digitized by a 16-bit A / D converter, with a sampling rate of 5Hz synchronized with the temperature signal. Oversampling technology is employed during the A / D conversion, meaning the internal sampling frequency is four times the external sampling frequency. Then, averaging downsampling further improves the digitization accuracy. The digitized water sublimation rate data also needs to be compensated and corrected using the readings from the freeze-drying chamber pressure sensor, as the chamber pressure affects the water sublimation rate. The compensation method involves establishing a mapping table between the water sublimation rate and the chamber pressure. When the chamber pressure changes, the water sublimation rate is corrected according to this mapping relationship. A typical compensation range is when the pressure varies within the 10-100 Pa range, with a water sublimation rate compensation coefficient between 0.9 and 1.1.

[0028] The temperature gradient calculation uses a spatial difference algorithm to construct a three-dimensional temperature field by calculating the temperature values ​​of each temperature measuring point within the building material according to their spatial position. Then, the temperature difference between adjacent points is calculated and divided by the actual physical distance to obtain the temperature gradient vector.

[0029] Spatial difference algorithms include two implementation methods: central difference and forward difference. For internal measuring points, the central difference method is used, with the calculation formula T'(x) = (T(x+Δx) - T(x-Δx)) / (2Δx); for surface measuring points, the forward difference method is used, with the calculation formula T'(x) = (T(x+Δx) - T(x)) / Δx. Here, T'(x) represents the temperature gradient at location x (unit: ℃ / cm), T(x) represents the temperature value at location x (unit: ℃), T(x+Δx) represents the temperature value at location x+Δx (unit: ℃), T(x-Δx) represents the temperature value at location x-Δx (unit: ℃), and Δx represents the actual physical distance between adjacent measuring points (unit: cm). The temperature gradient calculated using this method includes not only magnitude but also direction information, reflecting the heat transfer path within building materials. In the data integration stage, timestamp matching technology is used to correlate the temperature gradient values ​​with the moisture sublimation rate data. Each data point contains precise time information. By comparing timestamps, temperature gradient data and moisture sublimation rate data at the same moment are paired to form a multidimensional feature vector. When the sampling times of the two data sources are not completely consistent, linear interpolation is used to generate estimates for the corresponding moments. The final freeze-drying process feature dataset is a multidimensional time series database containing complete information on temperature field evolution and moisture migration.

[0030] In one specific embodiment, the process of performing step S102 may specifically include the following steps: Time-series sampling of temperature data from the freeze-drying process feature dataset was performed to construct the initial state matrix for the chaotic butterfly effect algorithm; A small random perturbation is added to the initial state matrix to generate multiple sets of perturbation initial conditions, and the evolution results are calculated according to the core iteration rules of the chaotic butterfly effect. Calculate the average Euclidean distance between multiple sets of evolution results, plot the distance change curve over time, and mark the corresponding time point as the chaotic critical point of the freeze-drying process of building materials when the distance growth rate exceeds the preset threshold. A time-delay coordinate transformation is performed on the feature dataset of the freeze-drying process. The temperature value at the current moment is compared with the temperature value after a certain delay to generate a two-dimensional phase diagram, from which the spiral structure and singularity of building materials during the freezing process can be identified. Align the chaotic critical point and the singular point on the time axis, and select the moment point that simultaneously satisfies both characteristics to determine the nonlinear transition point in the freezing process of building materials. Based on the temperature change rate and moisture migration rate data before and after the nonlinear transition point, the recursive prediction module of the chaotic butterfly effect algorithm is used to calculate the ice crystal growth path and distribution density in the future under multiple different initial conditions, forming a prediction result set of the ice crystal growth trend of building materials.

[0031] Specifically, the application of the chaotic butterfly effect algorithm in the freeze-drying process control of building materials begins with time-series sampling of temperature data from the freeze-drying process feature dataset. First, temperature data sequences are extracted from the freeze-drying process feature dataset. A sliding window sampling method is used to truncate continuous temperature data, with a window width typically chosen to be 512 data points and a sliding step size of 64 data points, ensuring sufficient overlap between adjacent windows to capture the continuous variation characteristics of the data. The sampled temperature data sequences are normalized, mapping their value range to the [-1,1] interval to eliminate dimensional influences, and then reorganized into an initial state matrix. The initial state matrix is ​​a six-dimensional matrix, with each dimension corresponding to the temperature field, pressure field, moisture concentration field, energy flow field, mass transfer field, and microstructure field of the building material, respectively. Each row of the matrix represents the complete state at a given time point, and each column represents the change sequence of a state variable at different time points. Small random perturbations are added to the initial state matrix to generate multiple sets of perturbation initial conditions. The perturbation is added using a controlled stochastic method. Based on the original initial state matrix, a random perturbation following a Gaussian distribution is added to each element, with the perturbation amplitude controlled within 0.1% of the original value. Typically, 20 different sets of perturbation initial conditions are generated, with perturbation values ​​produced by a random number generator, and each set of perturbations is independent of the others. These perturbation initial conditions are then evolved according to the core iterative rules of the chaotic butterfly effect. The core iterative rules are based on the improved Lorentz equations, extending the original three-dimensional equations into a six-dimensional system of equations to describe the coupled evolution of various physical fields during the freeze-drying of building materials. The iterative calculation uses the fourth-order Runge-Kutta numerical integration method, with a time step set to 0.01 seconds. Typically, 500 time steps are calculated backward to obtain the complete evolution trajectory of the system under each set of perturbation initial conditions.

[0032] After obtaining multiple sets of evolution results, the average Euclidean distance between these trajectories is calculated to quantify the differences in system state caused by initial small perturbations. The specific calculation process is as follows: at each time step, the Euclidean distance between all trajectory pairs is calculated, and the average value is taken as the average Euclidean distance at that time point. The Euclidean distance calculation considers the weights of each dimension in the six-dimensional space, with the temperature field and moisture concentration field having higher weights, reflecting their dominant role in the freeze-drying process. The calculated average Euclidean distance is plotted as a curve over time, and the rate of increase in distance is calculated. When the rate of increase exceeds a preset threshold (usually set to 50% / min), the corresponding time point is marked as a chaotic critical point. These chaotic critical points typically correspond to phase transition points in the freeze-drying process of building materials, such as the initiation point of ice crystal formation, the eutectic transformation point, or the initiation point of primary sublimation.

[0033] Simultaneously, a time-delay coordinate transformation is performed on the freeze-drying process feature dataset. The temperature time series T(t) is paired with its value T(t+τ) after a delay of a certain time τ to form a coordinate pair [T(t), T(t+τ)]. This is used to plot a point set on a two-dimensional plane, forming a two-dimensional phase diagram. The choice of the time delay τ is crucial and is usually determined by the first local minimum of the mutual information function. For the freeze-drying process of building materials, a typical value is between 8 and 15 seconds. In the obtained two-dimensional phase diagram, spiral structures and singularities are identified. A spiral structure represents the trajectory of a point set rotating around a central point, while a singularity is the location where the trajectory abruptly changes, typically manifested as a bifurcation point, saddle point, or attractor in the phase diagram. These special structures correspond to critical transition moments in the freeze-drying process. The chaotic critical points identified in the previous steps are compared and aligned with the singularities in the phase diagram on the time axis. Time points that simultaneously satisfy both characteristics—that is, time points that are both chaotic critical points and singularities in the phase diagram—are identified as nonlinear transition points in the freezing process of building materials. This dual verification mechanism significantly improves the accuracy and robustness of transition point identification, reducing false positives and false negatives. These nonlinear transition points precisely mark the key stages of change in the freeze-drying process of building materials, providing a time reference for subsequent precise control.

[0034] Based on the identified temperature change rate and moisture migration rate data before and after the nonlinear transition point, ice crystal growth is predicted using a recursive prediction module based on the chaotic butterfly effect algorithm. The recursive prediction module employs a recursive multi-starting-point prediction strategy, extracting temperature change rate and moisture migration rate features from the state data before and after the nonlinear transition point to construct a local dynamic model. Then, under multiple different initial conditions, this dynamic model is used to calculate the system evolution in parallel over future time periods. In the specific calculation process, an adaptive time-step numerical integration method is used, automatically reducing the step size when the system state changes drastically to ensure calculation accuracy. The predicted multiple evolution trajectories are statistically processed to form a probability distribution of ice crystal growth paths and distribution densities in different regions of the building material, ultimately generating a prediction result set of ice crystal growth trends.

[0035] In one specific embodiment, the process of performing parallel calculations of ice crystal growth paths and distribution densities in future time periods under multiple different initial conditions can specifically include the following steps: A state description vector of the freeze-drying process of building materials is constructed by extracting fixed-length data segments of temperature change rate and moisture migration rate before and after the nonlinear transition point. Add a small perturbation that conforms to a normal distribution to the state description vector to generate multiple sets of initial condition vectors, which serve as the starting point for recursive prediction of the chaotic butterfly effect. Multiple sets of initial condition vectors are input into the six-dimensional state evolution generator of the chaotic butterfly effect. Through step-by-step iterative calculation, the trajectories of temperature and moisture fields at different locations within the building material over time are obtained. Statistical analysis of the change trajectory was performed to calculate the probability distribution of temperature and moisture content at each spatial location point, and a thermal map of the probability of ice crystal formation during the freeze-drying process of building materials was constructed. Based on the probability heat map, cluster analysis is performed on the internal regions of building materials to identify regions with similar ice crystal formation probabilities and classify them into ice crystal growth regions with similar freezing characteristics. Trend fitting is performed on the temperature-moisture trajectory of the ice crystal growth region to calculate the ice crystal growth rate and direction of each region, generating a set of prediction results for the ice crystal growth trend of building materials that includes spatial distribution and temporal evolution.

[0036] Specifically, data is extracted from the area surrounding identified nonlinear transition points. Nonlinear transition points refer to key state change points during the freeze-drying process of building materials, such as the initiation point of ice crystal formation, the eutectic transition point, or the initiation point of primary sublimation. Fixed-length data segments of temperature change rate and moisture migration rate are extracted before and after the nonlinear transition point. Specifically, 100 sampling points are taken forward and 100 points backward from the nonlinear transition point, forming a data segment with a total length of 201 points. The temperature change rate data is calculated by dividing the temperature difference between adjacent time points by the time interval, reflecting the rate of temperature change over time. The moisture migration rate data is calculated by dividing the mass change rate by the initial moisture content of the building material, reflecting the rate at which moisture transforms into gas. After time alignment and normalization, these extracted data segments are combined to form a vector describing the state of the freeze-drying process of the building material, called the state description vector. This vector is a high-dimensional vector containing the temperature change rate sequence, the moisture migration rate sequence, and their spatial distribution information, comprehensively describing the dynamic behavior characteristics of the building material before and after the nonlinear transition point. Adding a small perturbation conforming to a normal distribution to the constructed state description vector is a crucial step in the recursive prediction of the chaotic butterfly effect. The perturbation is added using the Monte Carlo method, adding a random number conforming to a normal distribution N(0, σ²) to each element of the state description vector, where σ is the perturbation strength parameter, typically set to 0.5%-2% of the original value. The choice of perturbation strength needs to balance two aspects: small enough to preserve the basic characteristics of the state description vector, and large enough to produce meaningful predictive differences. Typically, 20-50 sets of perturbed state description vectors are generated. Each set contains the basic features of the original data but differs in microscopic details. These vectors serve as the starting points for the recursive prediction of the chaotic butterfly effect, i.e., the initial condition vectors.

[0037] The six-dimensional state evolver is the core computational unit of the chaotic butterfly effect algorithm. It contains six coupled nonlinear state equations, which describe the dynamic evolution of the temperature field, pressure field, moisture concentration field, energy flow field, mass transfer field, and microstructure field within the building material. The evolution calculation employs an adaptive step-size numerical integration method, automatically reducing the step size when the system state changes drastically to ensure computational accuracy. A typical initial step size is set to 0.5 seconds, with a minimum step size of 0.01 seconds. For each set of initial condition vectors, at least 300 time steps of iterative calculation are performed to track the evolution of the system state over time. The calculation results include complete time-varying trajectories of the temperature and moisture fields at each measuring point within the building material. These trajectories reflect the different possible evolutionary paths of the system under small initial differences.

[0038] Statistical analysis first calculates the mean and standard deviation of temperature and moisture content changes over time at each spatial grid point, and then constructs a probability distribution model based on these statistics. For the temperature field, a threshold method is used to determine the conditions for ice crystal formation: when the temperature at a point is below the freezing point and the moisture content is above a critical value, the conditions for ice crystal formation are considered met. By statistically analyzing the probability of each point in multiple sets of trajectories meeting the ice crystal formation conditions, a probability heat map of ice crystal formation during the freeze-drying process of building materials is constructed. The heat map is represented by a three-dimensional grid, with the color intensity of each grid point indicating the probability of ice crystal formation: red represents high-probability areas, and blue represents low-probability areas. This intuitive visualization method showcases the spatial distribution characteristics of ice crystal formation within building materials.

[0039] Based on the generated heatmap of ice crystal formation probability, cluster analysis was performed on the internal regions of the building materials. The K-means clustering algorithm was used to group spatial points with similar ice crystal formation probabilities into one cluster. The clustering process first determined the number of clusters K, typically set to 3-5, representing different ice crystal growth regions; then, K initial cluster centers were randomly selected; next, the distance from each spatial point to each cluster center was calculated, and the point was assigned to the nearest cluster; finally, the cluster centers were updated, and the iteration was repeated until convergence. Euclidean distance was used as the distance metric, with the weight set to the reciprocal of the probability value, making the clustering of high-probability regions more refined. The clustering results divided the building materials into several ice crystal growth regions with similar freezing characteristics, such as the core region, intermediate region, and surface region. The ice crystal growth behavior within each region is similar, facilitating the implementation of differentiated control strategies.

[0040] For the defined ice crystal growth regions, the temporal evolution characteristics of their temperature-moisture trajectories are further analyzed. Specifically, trend fitting is performed on the time series of temperature and moisture content within each region, using multinomial regression or exponential function fitting to obtain key parameters such as the temperature decrease rate and moisture reduction rate. Based on the fitting results, the ice crystal growth rate and direction for each region are calculated. The ice crystal growth rate is jointly determined by the temperature decrease rate and moisture reduction rate, typically expressed as the increase in ice crystal volume per unit time; the ice crystal growth direction is determined by the temperature gradient direction, with ice crystals preferentially growing along the temperature gradient direction. Integrating this information, a set of predicted results for the growth trend of ice crystals in building materials, encompassing both spatial distribution and temporal evolution, is formed, providing a decision-making basis for subsequent refined control.

[0041] In one specific embodiment, the process of executing step S103 may specifically include the following steps: Based on the prediction results set of ice crystal growth trends, the internal temperature conduction characteristics of building materials are evaluated, and the thermal conductivity distribution map of building materials is determined. Based on the thermal conductivity distribution map and the actual placement of temperature sensors, a hierarchical clustering method is used to divide building materials into core area, intermediate area and surface area, generating a spatial area division scheme for building materials. Different cooling curves with different slopes were designed for the core region, the intermediate region, and the surface region. A slow cooling slope was used for the core region, a medium cooling slope was used for the intermediate region, and a fast cooling slope was used for the surface region, forming a set of differentiated cooling curves for the three regions. The set of differentiated cooling curves in the three zones is converted into a sequence of temperature setpoints within the control cycle, and the target temperature value at each control moment is calculated by piecewise linear interpolation. The target temperature value is compared with the real-time measured temperature value to calculate the temperature control deviation of each area, and the cooling power distribution of each area is dynamically adjusted according to the magnitude of the deviation. Based on dynamically adjusted cooling power allocation, the speed control parameters of the refrigeration compressor and the opening parameters of the refrigerant flow control valve are calculated to generate a set of multi-zone coordinated cooling control parameters, and to execute differentiated cooling curve control for each zone of the building materials.

[0042] Specifically, temperature field data from the obtained ice crystal growth trend prediction results are used to infer the thermal conductivity distribution of building materials through time series analysis of the temperature field. The specific calculation method is based on Fourier's law of heat conduction. Temperature distribution data of building materials at multiple time points during the initial freeze-drying stage are collected, and the temperature gradient and heat flux density between adjacent measuring points are calculated. Then, the local thermal conductivity is obtained by dividing the heat flux density by the temperature gradient. Due to the heterogeneity of the internal structure of building materials, the thermal conductivity varies at different locations. Therefore, it is necessary to perform gridded calculations throughout the entire material space to form a complete thermal conductivity distribution map. The thermal conductivity distribution map is a three-dimensional data matrix, where each matrix element represents the thermal conductivity value at the corresponding spatial location, in units of watts per meter of Kelvin. Regions with high thermal conductivity have a fast heat transfer rate and a corresponding fast cooling rate; regions with low thermal conductivity have a slow heat transfer rate and a corresponding slow cooling rate. This difference in heat transfer characteristics is the basis for designing differentiated cooling curves.

[0043] Based on the thermal conductivity distribution map and the actual placement of temperature sensors, a hierarchical clustering method was used to divide the building materials into a core region, an intermediate region, and a surface region. Hierarchical clustering is a bottom-up clustering algorithm that first treats each measurement point as an independent cluster, and then gradually merges the most similar clusters until a predetermined number of clusters is reached. The similarity measure is the Euclidean distance of thermal conductivity, which is the square of the difference between the thermal conductivity values ​​of two points. The clustering process starts from the actual placement of temperature sensors, considering the thermal conductivity value of each sensor location and the thermal conductivity distribution of the surrounding space. Through iterative calculation, regions with similar thermal conductivity are continuously merged, ultimately forming three main clusters: a core region with low thermal conductivity, an intermediate region with medium thermal conductivity, and a surface region with high thermal conductivity. This partitioning method considers the natural distribution of the internal thermal conductivity characteristics of the building materials, avoiding the irrationality that may be caused by simple geometric partitioning. The partitioning result is represented by a three-dimensional mesh, where each cell is labeled with its region type, forming a spatial region partitioning scheme for the building materials.

[0044] Designing cooling curves with different slopes for the defined core, intermediate, and surface regions is crucial for precisely controlling the freeze-drying process. The cooling curve design is based on the heat transfer characteristics and ice crystal growth kinetics of each region. The core region, with its low thermal conductivity and slow heat transfer, uses a slow cooling slope, typically set at 0.2–0.5 °C / min. The intermediate region, with its moderate thermal conductivity, uses a medium cooling slope, typically set at 0.5–1.0 °C / min. The surface region, with its high thermal conductivity and fast heat transfer, uses a rapid cooling slope, typically set at 1.0–2.0 °C / min. The cooling curve for each region includes multiple stages: a pre-cooling stage (from room temperature to near 0 °C), a supercooling stage (down to -5 to -10 °C, promoting ice nucleus formation), and a cryogenic stage (down to the final target temperature of -30 to -50 °C). The cooling slopes also differ between stages: the pre-cooling stage has a steeper slope, the supercooling stage has a moderate slope, and the cryogenic stage has a shallower slope. These designs combine to form a set of differentiated cooling curves for three zones, with each curve representing a series of time-temperature pairs.

[0045] Converting the set of differentiated cooling curves in three zones into a sequence of temperature setpoints within the control cycle is the foundation of actual control execution. The control cycle is typically set to 10–60 seconds, depending on the response speed and computational power of the control system. The conversion process employs a piecewise linear interpolation algorithm. For each time point ti in the control cycle, two adjacent time points tj and tj+1 are found on the cooling curve, and then the target temperature value at that moment is calculated through linear interpolation. The interpolation formula is: T(ti) = T(tj) + (T(tj+1) - T(tj))×(ti - tj) / (tj+1 - tj), where T(ti) is the target temperature value at time ti, and T(tj) and T(tj+1) are the temperature values ​​at times tj and tj+1 on the cooling curve, respectively. In this way, the continuous cooling curve is converted into a discrete sequence of temperature setpoints, adapting to the requirements of the digital control system.

[0046] Comparing the target temperature value with the real-time measured temperature value to calculate the temperature control deviation is the core step of closed-loop control. The temperature control deviation is defined as the difference between the measured temperature and the target temperature: e(t) = Tmeas(t) - Tset(t), where e(t) is the temperature deviation at time t, Tmeas(t) is the measured temperature, and Tset(t) is the target temperature. Based on the magnitude and trend of the deviation, the cooling power distribution in each region is dynamically adjusted. The adjustment method employs a fuzzy control strategy, using the temperature deviation e and the rate of change de / dt as input variables, and the cooling power adjustment ΔP as the output variable. A series of fuzzy rules are used to achieve mapping. Typical fuzzy rules include: if the deviation is large and increasing, the cooling power is increased significantly; if the deviation is small and stable, the cooling power is fine-tuned; if the deviation is negative and decreasing, the cooling power is reduced, etc. After fuzzy inference and defuzzification, the specific cooling power adjustment is obtained, thereby achieving fine control of the cooling process.

[0047] Based on dynamically adjusted cooling power distribution, the speed control parameters of the refrigeration compressor and the opening parameters of the refrigerant flow control valve are calculated. This step converts the abstract cooling power demand into specific equipment control commands. The refrigeration compressor speed control parameter mainly refers to the compressor's operating frequency or speed, usually using frequency converter control with a frequency range of 20~60Hz. The compressor frequency and cooling capacity have an approximately linear relationship; the higher the frequency, the greater the cooling capacity. The refrigerant flow control valve opening parameters control the distribution of refrigerant flow to each cooling zone, with an opening range typically from 0~100%. The valve opening in different zones is set according to the proportion of cooling power required in that zone, while also considering the thermal interaction between zones. All control parameters are combined to form a multi-zone coordinated cooling control parameter set, which is transmitted to the execution equipment through a communication interface to achieve differentiated cooling curve control for each zone of the building material.

[0048] In one specific embodiment, the process of executing step S104 may specifically include the following steps: Real-time temperature data of the core region, intermediate region and surface region are collected, and the deviation between the data and the target value of the cooling curve of each region is calculated to form a temperature control status evaluation matrix. Based on the temperature control status evaluation matrix, fuzzy rule reasoning is used to assign cooling capacity weights and heating power weights to each region, generating a regional energy allocation scheme. The regional energy distribution scheme is broken down into three sets of parameters: refrigeration compressor speed, electronic expansion valve opening, and evaporator fan speed, forming zoned control commands for the refrigeration compressor unit. Based on the moisture evaporation rate data of the core region, intermediate region and surface region, the required vacuum environment parameters are calculated, and coordinated control commands for vacuum pump speed and vacuum valve opening are generated. The temperature gradient directions of the core region, intermediate region and surface region are analyzed to determine the power distribution pattern of the heating plate array and generate power adjustment control commands for each heating channel. The industrial bus synchronously transmits control commands for the refrigeration compressor unit, vacuum pump unit, and heating plate array to each actuator, monitors the equipment response status, and completes multi-level coordinated control of the equipment at each stage of freeze drying.

[0049] Specifically, PT100 platinum resistance sensors and K-type thermocouples pre-embedded in each region are used to continuously acquire temperature data at a sampling frequency of 5Hz. The data is processed by a preamplifier and signal conditioning circuit to form a real-time temperature data stream for each region. Simultaneously, the target temperature value for each region at the current moment is extracted from the cooling curve database. The real-time temperature is compared with the target temperature to calculate the temperature deviation. The temperature deviation is calculated by subtracting the target temperature from the measured temperature. A positive value indicates that the actual temperature is higher than the target value, requiring enhanced cooling; a negative value indicates that the actual temperature is lower than the target value, requiring reduced cooling or enhanced heating. The rate of change of temperature deviation is also calculated for each region, which is the difference in temperature deviation between two adjacent sampling moments divided by the time interval. The temperature deviation and the rate of change of deviation for each region are combined to form a temperature control status evaluation matrix, which is an m×n two-dimensional matrix. Here, m is the number of regions (usually 3, corresponding to the core region, intermediate region, and surface region), and n is the number of state parameters (usually 2, corresponding to temperature deviation and the rate of change of deviation). The temperature control status evaluation matrix comprehensively reflects the real-time status of temperature control in each region during the freeze-drying process and is an important basis for subsequent control decisions. Based on the formed temperature control state evaluation matrix, fuzzy rule reasoning is used to assign cooling capacity weights and heating power weights to each region. Fuzzy rule reasoning is a method of reasoning based on human experience and knowledge, suitable for handling imprecise and uncertain systems. First, the two input variables, temperature deviation and deviation change rate, are fuzzified and divided into five fuzzy sets: negative large (NB), negative small (NS), zero (ZO), positive small (PS), and positive large (PB). Triangular or trapezoidal membership functions are used to describe membership relationships. Then, a fuzzy rule library is set, containing 25 rules such as "If the temperature deviation is positive large and the deviation change rate is positive large, then the cooling capacity weight is large and the heating power weight is zero," covering all possible input combinations. Fuzzy reasoning is performed using the MAX-MIN synthesis method, matching the input fuzzy sets with the antecedents of the rules to obtain the rule activations. The contribution of each rule to the output is then calculated. Finally, the fuzzy output is defuzzified using the centroid method to obtain the specific cooling capacity weights and heating power weights. These weight values ​​combine to form a regional energy allocation scheme, guiding subsequent calculations of equipment control parameters.

[0050] Breaking down the regional energy allocation scheme into specific equipment control parameters is a key step in achieving precise control. First, the total cooling capacity requirement is converted into the refrigeration compressor speed. The refrigeration compressor speed and cooling capacity have an approximately linear relationship. By consulting the compressor characteristic curve or using performance equations, the corresponding speed value for the required cooling capacity is determined, typically ranging from 600-3000 rpm. Then, based on the cooling capacity weight of each region, the refrigerant flow rate allocated to each region is calculated and converted into the electronic expansion valve opening parameter. The relationship between the electronic expansion valve opening and refrigerant flow rate usually follows a square root relationship, meaning the flow rate is proportional to the square root of the opening. Finally, the evaporator fan speed is determined. The fan speed directly affects the cooling capacity transfer efficiency and is usually adjusted linearly according to the required cooling capacity transfer rate. These three sets of parameter values ​​constitute the zone control commands for the refrigeration compressor unit, achieving precise control over the generation and distribution of cooling capacity.

[0051] The process of calculating the required vacuum environment parameters based on moisture evaporation rate data from the core, intermediate, and surface regions involves a freeze-drying kinetic model. Moisture evaporation rate data, obtained from a mass monitoring system, reflects the amount of moisture converted to gaseous state per unit time in each region. According to basic freeze-drying theory, the moisture sublimation rate is closely related to the chamber pressure and material surface temperature, and is approximately proportional to the difference between the saturated vapor pressure and the actual chamber pressure. Through inverse calculation, the optimal chamber pressure value required at the current moisture sublimation rate is determined. This pressure value maintains an appropriate moisture mass transfer driving force, preventing both excessively high pressure leading to a slow sublimation rate and excessively low pressure causing frost formation on the material surface. The determined target chamber pressure value is typically in the range of 10-100 Pa. Then, based on the difference between the current actual chamber pressure and the target pressure, the required pumping rate is calculated and converted into a vacuum pump speed parameter. Simultaneously, based on the spatial distribution characteristics of the moisture evaporation rate, the optimal opening degree of the vacuum valve is calculated to adjust the pumping intensity in different regions. These parameters combine to form a coordinated control command for the vacuum pump speed and vacuum valve opening. Analyzing the temperature gradient directions in the core, intermediate, and surface regions is crucial for determining the power distribution pattern of the heating plate array, which is key to accurate energy replenishment. The temperature gradient direction reflects the natural direction of heat transfer, pointing from high-temperature regions to low-temperature regions. By analyzing the temperature distribution data within each region, a temperature gradient vector field is calculated, describing the direction and magnitude of temperature change at each spatial point. The temperature gradient vector is calculated using the central difference method; for a point (x, y, z) in three-dimensional space, its temperature gradient vector consists of the partial derivatives in the x, y, and z directions. Based on the calculated temperature gradient vector field, the direction of natural heat flow is determined, and then a heating energy replenishment scheme with the opposite direction is designed to balance the temperature distribution in each region. Specifically, the heating plate array is divided into several independently controlled heating channels, each responsible for heating a specific region. Based on the temperature state and temperature gradient direction of each region, the power required by each heating channel is calculated, forming the power distribution pattern of the heating plate array, which is then converted into power adjustment control commands for each heating channel. The industrial bus synchronously transmits control commands for the refrigeration compressor unit, vacuum pump unit, and heating plate array to each actuator, enabling multi-level collaborative control. The industrial bus employs industrial communication protocols such as Profibus, Modbus, or EtherCAT, providing highly reliable and real-time data transmission. Control commands are packaged according to a predetermined format, including fields such as device address, function code, data area, and checksum, ensuring correct transmission. Upon receiving the control commands, each actuator converts them into actual physical operations through its internal drive circuitry, such as a frequency converter adjusting motor speed or an electric valve actuator adjusting valve opening. Simultaneously, status feedback information from each device is returned to the control unit via the same industrial bus, forming a closed-loop control system that monitors device response status in real time, ensuring correct execution of control commands.This multi-device collaborative control method ensures that the three major systems of refrigeration, vacuum and heating in each stage of freeze-drying can work in a coordinated manner and accurately realize the preset control strategy.

[0052] In one specific embodiment, the process of generating coordinated control commands for the vacuum pump speed and vacuum valve opening may specifically include the following steps: The moisture evaporation rate data of the core region, intermediate region and surface region are combined into a moisture load matrix to calculate the total amount of water vapor per unit time. Based on the total amount of water vapor and the volume of the freeze-drying chamber, the target pressure value of the chamber required to maintain the optimal sublimation rate is calculated. The target pressure value is compared with the current measured pressure value in the cavity, the pressure deviation and the rate of change of deviation are calculated, and a pressure control state vector is constructed. Input the pressure control state vector into the pressure-flow control calculation module to calculate the required vacuum system pumping rate and flow control parameters. Based on the pumping rate, the required rotational speed is calculated in reverse using the vacuum pump characteristic curve, generating a coordinated control command for the vacuum pump speed and vacuum valve opening.

[0053] Specifically, the moisture evaporation rate data of the core, intermediate, and surface regions are combined into a moisture load matrix, starting with the mass change data collected from the strain gauge electronic weighing system in each region. By performing differential calculations on the mass data at continuous time points, the moisture reduction in each region is obtained. This is then divided by the time interval to obtain the moisture evaporation rate value, expressed in grams per minute (g / min). The moisture load matrix is ​​a three-dimensional data structure containing spatial coordinates and corresponding moisture evaporation rate values. When constructing the moisture load matrix, the discrete measurement point data must first be extended to the entire material space using spatial interpolation methods. Commonly used interpolation methods include Kriging interpolation or inverse distance weighting to ensure a continuous moisture evaporation rate distribution field. After the moisture load matrix is ​​constructed, all elements in the matrix are integrated, i.e., the moisture evaporation rate of each spatial unit is multiplied by its volume and summed to obtain the total amount of water vapor released by the entire building material per unit time, expressed in grams per minute (g / min). This total water vapor directly reflects the gas load that needs to be removed by the vacuum system during the freeze-drying process and is the fundamental data for subsequent calculations of vacuum environment parameters. Based on the calculated total water vapor volume and freeze-drying chamber volume, the target chamber pressure required to maintain the optimal sublimation rate is calculated. This calculation is based on thermodynamic principles and a freeze-drying kinetic model. During freeze-drying, the water sublimation rate is directly proportional to the difference between the saturated vapor pressure at the ice surface and the actual chamber pressure; this difference is the driving force for sublimation. Maintaining appropriate chamber pressure is crucial for freeze-drying efficiency: too low a pressure will result in an excessively low ice surface temperature, slowing down the sublimation rate; too high a pressure will reduce the pressure difference, also decreasing the sublimation rate. The calculation of the optimal pressure value needs to consider the saturated vapor pressure at the ice surface temperature and an optimization coefficient, which typically ranges from 0.3 to 0.5. The ice surface temperature is obtained from temperature sensor data, while the saturated vapor pressure is obtained by looking up the temperature in a table or by calculation. The optimization coefficient considers factors such as the total water vapor volume, sublimation area, and vacuum system performance. A higher value means maintaining a higher chamber pressure, which is beneficial for improving heat transfer efficiency; a lower value means maintaining a lower chamber pressure, which is beneficial for improving water mass transfer efficiency. In this way, the calculated target pressure value of the cavity can maximize the freeze-drying efficiency while ensuring the quality of freeze-drying.

[0054] Comparing the target pressure value with the current measured pressure value in the cavity, calculating the pressure deviation and the rate of change of deviation, and constructing the pressure control state vector are fundamental to achieving precise control. The measured pressure value in the cavity is obtained through a Pirani vacuum gauge or a capacitive pressure sensor, with a measurement accuracy typically within ±1% of the reading. The pressure deviation is calculated by subtracting the target pressure value from the measured pressure value. A positive deviation indicates that the actual pressure is higher than the target value, requiring an increase in the pumping rate; a negative deviation indicates that the actual pressure is lower than the target value, requiring a decrease in the pumping rate. The rate of change of pressure deviation is calculated by subtracting the pressure deviation from the previous time step from the current pressure deviation, and then dividing by the time interval. The pressure deviation and the rate of change of deviation together form the pressure control state vector, which comprehensively reflects the current control state of the vacuum system. Inputting the pressure control state vector into the pressure-flow control calculation module, the process of calculating the required vacuum system pumping rate and flow control parameters involves the conversion relationships between multiple physical quantities. The pressure-flow control calculation module is a calculation unit based on a combination of physical and empirical models, with the pressure control state vector as input and the pumping rate and flow control parameters as output. The calculation process first determines the required adjustment to the pumping rate based on the magnitude and trend of the pressure deviation. The pumping rate is measured in cubic meters per second (m³ / s), representing the volume of gas the vacuum system can remove per unit time. The pumping rate calculation considers three factors: total water vapor volume, target pressure, and cavity volume. The basic principle is that the gas load divided by the target pressure equals the required pumping rate. After determining the pumping rate, the flow control parameters are further calculated, including the opening settings of the main vacuum valve and the zone vacuum valves. The main vacuum valve controls the total pumping volume, while the zone vacuum valves control the pumping distribution ratio in each zone. The relationship between valve opening and flow rate typically follows a specific flow characteristic curve, such as an equal percentage characteristic or a linear characteristic, requiring the selection of an appropriate conversion relationship based on the actual valve type.

[0055] The final execution step involves calculating the required rotational speed based on the calculated pumping rate using the vacuum pump characteristic curve, and then generating coordinated control commands for the vacuum pump speed and vacuum valve openings. The vacuum pump characteristic curve is a graph or mathematical expression describing the relationship between pump speed, pumping rate, and ultimate pressure; it is a crucial technical parameter provided by the pump manufacturer. Characteristic curves are typically represented by comparing pumping rates to pumping pressures at different speeds. The reverse calculation process involves finding or calculating the required pumping speed from the characteristic curve, given the desired pumping rate and operating pressure. Commonly used vacuum pump types include Roots pumps, screw pumps, and dry scroll pumps. Different types of pumps have different characteristic curves, requiring the selection of an appropriate calculation method for each specific pump type. The obtained rotational speed is usually achieved by controlling the operating frequency of the vacuum pump motor using a frequency converter. Simultaneously, combined with the previously calculated flow control parameters, a coordinated control command set including the vacuum pump speed and the openings of each vacuum valve is generated. These commands are sent to the corresponding actuators via an industrial control network, achieving precise control of the vacuum environment.

[0056] The above describes the freeze-drying process control method based on intelligent algorithms in the embodiments of this application. The following describes the freeze-drying process control system based on intelligent algorithms in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the freeze-drying process control system based on intelligent algorithms in this application includes: The acquisition module 201 is used to arrange a multi-point temperature sensor array inside and on the surface of building materials, collect temperature gradient data and combine it with the mass change rate to form a feature dataset of the freeze-drying process. The identification module 202 is used to identify the nonlinear transition points in the freezing process of building materials based on the freeze-drying process feature dataset and to predict the ice crystal growth trend through the chaotic butterfly effect algorithm. The control module 203 is used to divide the building materials in the freeze-drying cavity into a core area, an intermediate area and a surface area according to the ice crystal growth trend, and to implement differentiated cooling curve control for each area. The generation module 204 is used to generate multiple control commands according to the differentiated states of each region, and drive the refrigeration compressor unit, vacuum pump unit and heating plate array to perform multi-level coordinated control.

[0057] above Figure 2 The freeze-drying process control system based on intelligent algorithms in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The freeze-drying process control equipment based on intelligent algorithms in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0058] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the freeze-drying process control method based on intelligent algorithms.

[0059] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0060] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a freeze-drying process control device based on intelligent algorithms (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A freeze-drying process control method based on intelligent algorithms, characterized in that, The method includes: A multi-point temperature sensor array is arranged inside and on the surface of building materials to collect temperature gradient data and combine it with the mass change rate to form a characteristic dataset of the freeze-drying process; Based on the freeze-drying process feature dataset, the nonlinear transition points in the freezing process of building materials are identified by the chaotic butterfly effect algorithm, and the ice crystal growth trend is predicted. Based on the ice crystal growth trend, the building materials inside the freeze-drying cavity are divided into a core area, an intermediate area, and a surface area, and differentiated cooling curve control is implemented for each area. Based on the differentiated states of each region, multiple control commands are generated to drive the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control.

2. The freeze-drying process control method based on intelligent algorithms according to claim 1, characterized in that, The method involves arranging a multi-point temperature sensor array inside and on the surface of the building materials to collect temperature gradient data and combine it with the mass change rate to form a feature dataset of the freeze-drying process, including: PT100 platinum resistance sensors and K-type thermocouples were pre-embedded in ceramic-based building materials and lightweight concrete specimens to collect temperature signals during the freeze-drying process. The mass change of building materials was recorded using a strain gauge electronic weighing system located under the freeze dryer rack, and the rate of water sublimation during the freeze drying process was calculated. The temperature signal is sampled in real time at a sampling frequency of 5Hz, and after processing by a preamplifier and signal conditioning circuit, power frequency interference and thermal noise are removed. The water sublimation rate data is digitized using an A / D converter and then compensated and corrected by combining the readings of the freeze-drying chamber pressure sensor. The spatial difference algorithm is used to calculate the temperature gradient and its trend between points inside the building material using the processed temperature signal data. The temperature gradient values ​​and the moisture sublimation rate data are matched by timestamps to form a freeze-drying process feature dataset that includes the temperature field evolution and moisture migration characteristics of the entire freeze-drying process.

3. The freeze-drying process control method based on intelligent algorithms according to claim 1, characterized in that, The step of identifying nonlinear transition points in the freezing process of building materials and predicting ice crystal growth trends based on the freeze-drying process feature dataset includes: Time-series sampling of temperature data in the freeze-drying process feature dataset is performed to construct the initial state matrix of the chaotic butterfly effect algorithm; Add small random perturbations to the initial state matrix to generate multiple sets of perturbation initial conditions, and calculate the evolution results according to the core iteration rules of the chaotic butterfly effect. Calculate the average Euclidean distance between multiple sets of evolution results, plot the distance change curve over time, and mark the corresponding time point as the chaotic critical point of the freeze-drying process of building materials when the distance growth rate exceeds the preset threshold. A time-delay coordinate transformation is performed on the freeze-drying process feature dataset. The temperature value at the current moment is compared with the temperature value after a certain delay to generate a two-dimensional phase diagram, from which the spiral structure and singularity of the building material freezing process are identified. Align the chaotic critical point and the singular point on the time axis, and filter out the moment points that simultaneously satisfy both characteristics to determine them as nonlinear transition points in the freezing process of building materials. Based on the temperature change rate and moisture migration rate data before and after the nonlinear transition point, the recursive prediction module of the chaotic butterfly effect algorithm is used to calculate the ice crystal growth path and distribution density in the future time period under multiple different initial conditions, forming a prediction result set of the ice crystal growth trend of building materials.

4. The freeze-drying process control method based on intelligent algorithms according to claim 3, characterized in that, Based on the temperature change rate and moisture migration rate data before and after the nonlinear transition point, the recursive prediction module of the chaotic butterfly effect algorithm calculates the ice crystal growth path and distribution density in future time periods under multiple different initial conditions, forming a prediction result set of the ice crystal growth trend of building materials, including: A fixed-length data segment of temperature change rate and moisture migration rate is extracted from before and after the nonlinear transition point to construct a state description vector of the freeze-drying process of building materials. Add a small perturbation that conforms to a normal distribution to the state description vector to generate multiple sets of initial condition vectors, which serve as the starting point for recursive prediction of the chaotic butterfly effect. The multiple sets of initial condition vectors are input into the six-dimensional state evolution device of the chaotic butterfly effect, and the temperature field and moisture field at different locations within the building material are obtained over time through step-by-step iterative calculation. Statistical analysis was performed on the change trajectory to calculate the probability distribution of temperature and moisture content at each spatial location point, and a thermal map of the probability of ice crystal formation during the freeze-drying process of building materials was constructed. Based on the aforementioned probability heat map, cluster analysis is performed on the internal regions of building materials to identify regions with similar ice crystal formation probabilities and classify them into ice crystal growth regions with similar freezing characteristics. A trend fitting is performed on the temperature-moisture trajectory of the ice crystal growth region to calculate the ice crystal growth rate and direction of each region, generating a set of predicted results for the ice crystal growth trend of building materials that includes spatial distribution and temporal evolution.

5. The freeze-drying process control method based on intelligent algorithms according to claim 1, characterized in that, Based on the ice crystal growth trend, the building materials inside the freeze-drying cavity are divided into a core region, a middle region, and a surface region. Differentiated cooling curve control is implemented for each region, including: Based on the ice crystal growth trend prediction result set, the internal temperature conduction characteristics of building materials are evaluated, and the thermal conductivity distribution map of building materials is determined. Based on the thermal conductivity distribution map and the actual distribution of temperature sensors, a hierarchical clustering method is used to divide the building materials into core area, middle area and surface area, generating a spatial area division scheme for building materials. Different cooling curves with different slopes are designed for the core region, the intermediate region and the surface region. A slow cooling slope is used for the core region, a medium cooling slope is used for the intermediate region and a fast cooling slope is used for the surface region, forming a set of differentiated cooling curves for the three regions. The set of differentiated cooling curves in the three zones is converted into a sequence of temperature setpoints within the control cycle, and the target temperature value at each control moment is calculated by piecewise linear interpolation. The target temperature value is compared with the real-time measured temperature value to calculate the temperature control deviation of each area, and the cooling power distribution of each area is dynamically adjusted according to the magnitude of the deviation. Based on the dynamically adjusted cooling power allocation, the speed control parameters of the refrigeration compressor and the opening parameters of the refrigerant flow control valve are calculated to generate a set of multi-zone coordinated cooling control parameters, and to execute differentiated cooling curve control for each zone of the building materials.

6. The freeze-drying process control method based on intelligent algorithms according to claim 1, characterized in that, The process of generating multiple control commands based on the differentiated states of each region to drive the refrigeration compressor unit, vacuum pump unit, and heating plate array to perform multi-level coordinated control includes: Real-time temperature data of the core region, intermediate region and surface region are collected, and the deviation between the data and the target value of the cooling curve of each region is calculated to form a temperature control status evaluation matrix. Based on the temperature control status evaluation matrix, fuzzy rule reasoning is used to assign cooling capacity weights and heating power weights to each region, generating a regional energy allocation scheme. The regional energy distribution scheme is broken down into three sets of parameters: refrigeration compressor speed, electronic expansion valve opening, and evaporator fan speed, forming a zoned control command for the refrigeration compressor unit. Based on the moisture evaporation rate data of the core region, intermediate region and surface region, the required vacuum environment parameters are calculated, and coordinated control commands for vacuum pump speed and vacuum valve opening are generated. The temperature gradient directions of the core region, intermediate region and surface region are analyzed to determine the power distribution pattern of the heating plate array and generate power adjustment control commands for each heating channel. The control commands for the refrigeration compressor unit, vacuum pump unit, and heating plate array are synchronously transmitted to each execution device via an industrial bus, and the response status of the devices is monitored to complete multi-level coordinated control of the equipment at each stage of freeze-drying.

7. The freeze-drying process control method based on intelligent algorithms according to claim 6, characterized in that, The calculation of required vacuum environment parameters based on the moisture evaporation rate data of the core region, intermediate region, and surface region, and the generation of coordinated control commands for vacuum pump speed and vacuum valve opening, including: The moisture evaporation rate data of the core region, intermediate region and surface region are combined into a moisture load matrix to calculate the total amount of water vapor per unit time. Based on the total water vapor volume and the freeze-drying chamber volume, the target chamber pressure value required to maintain the optimal sublimation rate is calculated. The target pressure value is compared with the current measured pressure value of the cavity, the pressure deviation and the rate of change of deviation are calculated, and a pressure control state vector is constructed. The pressure control state vector is input into the pressure-flow control calculation module to calculate the required vacuum system pumping rate and flow control parameters. Based on the pumping rate, the required rotational speed is calculated in reverse using the vacuum pump characteristic curve, generating a coordinated control command for the vacuum pump rotational speed and vacuum valve opening.

8. A freeze-drying process control system based on intelligent algorithms, characterized in that, For implementing the freeze-drying process control method based on intelligent algorithms as described in any one of claims 1-7, the freeze-drying process control system based on intelligent algorithms comprises: The data acquisition module is used to deploy a multi-point temperature sensor array on the interior and surface of building materials, collect temperature gradient data, and combine it with the mass change rate to form a feature dataset of the freeze-drying process. The identification module is used to identify the nonlinear transition points in the freezing process of building materials based on the freeze-drying process feature dataset and predict the ice crystal growth trend through the chaotic butterfly effect algorithm. The control module is used to divide the building materials inside the freeze-drying cavity into a core area, an intermediate area and a surface area according to the ice crystal growth trend, and to implement differentiated cooling curve control for each area. The generation module is used to generate multiple control commands according to the differentiated states of each region, and drive the refrigeration compressor unit, vacuum pump unit and heating plate array to perform multi-level coordinated control.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, it causes the processor to execute the freeze-drying process control method based on intelligent algorithms as described in any one of claims 1 to 7.

Images