Parameter adjustment method, device and equipment of low-temperature evaporation crystallization machine and storage medium

Abstract

The application relates to the technical field of parameter adjustment, and discloses a low-temperature evaporation crystallization machine parameter adjustment method, device, equipment and storage medium, the method comprises the following steps: collecting M groups of temperature sensor data and N groups of pressure sensor data of a low-temperature evaporation crystallization machine, and calculating a first uncertainty representation of temperature parameter measurement and a second uncertainty representation of pressure parameter measurement; data fusion is performed to obtain a temperature state optimal estimation value and a pressure state optimal estimation value; parallel prediction is performed through a thermodynamic equilibrium model, a crystallization kinetics model and a system response model to obtain a crystallization process prediction result; optimal process parameter target point and dynamic fault tolerance boundary value analysis are performed to generate a process parameter control strategy, the application can effectively cope with disturbances such as feed fluctuation and environmental change, and realizes real-time updating of the control strategy and accurate adjustment of temperature and pressure parameters.

Description

Technical Field

[0001] The present application relates to the field of parameter adjustment technology, and in particular to a method, device, equipment and storage medium for adjusting parameters of a low-temperature evaporation crystallizer. Background Art

[0002] The temperature and pressure measurements of low-temperature evaporation crystallizers are affected by various factors, such as equipment vibration, ambient temperature fluctuations, and sensor drift, resulting in large uncertainties in the measurement data. These uncertainties directly affect the stability of the crystallization process and product quality.

[0003] In the existing technology, the temperature and pressure control systems of low-temperature evaporation crystallizers mainly use single PID control or simple model predictive control methods. These methods are often based on single sensor data and simplified models, and cannot effectively deal with measurement uncertainty and multivariable coupling problems. When the system encounters external interference such as fluctuations in feed composition or changes in environmental conditions, the control system is prone to constraint violations. The temperature deviation can reach ±1.2°C and the pressure fluctuation can reach ±0.08MPa, far exceeding the accuracy requirements required for high-quality crystallization. In addition, the existing control methods lack an accurate description of the dynamic characteristics of the crystallization process. A single model is difficult to fully characterize the complex crystallization mechanism under low-temperature conditions, making it difficult for the control strategy to adapt to different working conditions. Summary of the Invention

[0004] The present application provides a method, device, equipment and storage medium for adjusting the parameters of a low-temperature evaporation crystallizer. The present application can effectively cope with interferences such as feed fluctuations and environmental changes, and realize real-time updating of control strategies and precise adjustment of temperature and pressure parameters.

[0005] In a first aspect, the present application provides a method for adjusting parameters of a low-temperature evaporation crystallizer, the method comprising:

[0006] Collecting M sets of temperature sensor data and N sets of pressure sensor data from a low-temperature evaporation crystallizer, and calculating a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement;

[0007] Performing data fusion on the M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and performing data fusion on the N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state;

[0008] Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model, a crystallization kinetics model, and a system response model for parallel prediction to obtain a crystallization process prediction result;

[0009] Based on the crystallization process prediction results, an optimal process parameter target point and a dynamic fault tolerance boundary value analysis are performed to generate a process parameter control strategy.

[0010] A second aspect of the present application provides a low-temperature evaporation crystallization machine parameter adjustment device, the low-temperature evaporation crystallization machine parameter adjustment device comprising:

[0011] An acquisition module is used to acquire M groups of temperature sensor data and N groups of pressure sensor data of the low-temperature evaporation crystallizer, and calculate a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement;

[0012] a data fusion module, configured to perform data fusion on the M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and to perform data fusion on the N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state;

[0013] A parallel prediction module is used to input the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model, a crystallization kinetics model and a system response model for parallel prediction to obtain a crystallization process prediction result;

[0014] A generation module is used to perform optimal process parameter target point and dynamic fault tolerance boundary value analysis based on the crystallization process prediction results to generate a process parameter control strategy.

[0015] The third aspect of the present application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned low-temperature evaporation crystallizer parameter adjustment method.

[0016] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for adjusting parameters of a low-temperature evaporation crystallizer.

[0017] Compared with the existing technology, the present application has the following beneficial effects: by collecting multiple sets of temperature and pressure sensor data and performing Gaussian distribution uncertainty analysis, the reliability of the measurement data is quantitatively evaluated, effectively solving the problem of inaccurate and unreliable measurement of a single sensor. Adaptive Kalman filtering is performed on multiple sets of sensor data based on uncertainty characterization, enabling the system to dynamically adjust the weights of different sensors, filter outliers and suppress noise interference, ensuring more reliable temperature and pressure state estimates. By combining the thermodynamic equilibrium model, crystallization kinetics model and system response model, the low-temperature evaporation crystallization process is comprehensively described through complementary theoretical foundations, overcoming the limitation that a single model cannot accurately describe complex crystallization processes. The weight coefficients of each model are dynamically adjusted based on the prediction error to achieve adaptive fusion of model prediction results, enabling the system to automatically select the most suitable model combination according to different working conditions, thereby improving prediction accuracy and adaptability. By analyzing the impact of external interference on the system, the optimal process parameter target point and dynamic fault tolerance boundary value are calculated, so that the control system has sufficient fault tolerance while ensuring control accuracy, and can effectively cope with interference such as feed fluctuations and environmental changes. The use of a soft-constraint optimization method with elastic variables improves the system's ability to handle constraint violations while ensuring accurate control of temperature and pressure parameters. This allows the control system to maintain stable operation under various interference conditions, avoiding the potential instability caused by traditional hard-constraint control methods when constraints are violated. Through continuous optimization and feedback adjustment within a rolling time domain, real-time updates of the control strategy and precise adjustment of temperature and pressure parameters are achieved, enabling the low-temperature evaporation crystallizer to maintain stable operation and high-quality crystallization results under dynamically changing operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] The structures, proportions, sizes, etc. depicted in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not intended to limit the conditions under which the present invention can be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportional relationships, or adjustments in size should still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and objectives that can be achieved by the present invention.

[0020] Figure 1 1 is a flow chart of a method for adjusting parameters of a low-temperature evaporation crystallizer provided by an embodiment of the present invention;

[0021] Figure 2 1 is a schematic block diagram of the structure of a parameter adjustment device for a low-temperature evaporation crystallizer provided by an embodiment of the present invention;

[0022] Figure 3 It is a schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0025] It should also be understood that the terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0026] It should be further understood that the term "and / or" used in this specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 In the embodiment of the present application, an embodiment of the method for adjusting the parameters of the low-temperature evaporation crystallizer includes:

[0027] Step 100: collecting M sets of temperature sensor data and N sets of pressure sensor data from a low-temperature evaporation crystallizer, and calculating a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement;

[0028] It is understandable that the execution subject of the present application can be a low-temperature evaporation crystallizer parameter adjustment device, or a terminal or a server, which is not limited here. The embodiment of the present application is described by taking the server as the execution subject as an example.

[0029] Specifically, M groups of temperature sensors and N groups of pressure sensors are deployed at key locations in the crystallizer system. Each group of sensors consists of different types of measuring elements. For example, temperature sensors use platinum resistance thermometers, thermocouples, and infrared thermometers, while pressure sensors employ piezoelectric, capacitive, and thin-film structures. During system operation, these sensors synchronously sample at a fixed frequency and transmit the collected temperature and pressure data in real time to the central processing module. Upon data arrival, the real-time measurement values ​​collected by the M groups of temperature sensors are statistically analyzed. By comparing the differences in measurement results between different sensors, the root mean square (RMS) deviation between the temperature measurement values ​​is calculated to quantify the measurement consistency between sensors under the same conditions. Historical measurement data is then compared with the current measurement data to analyze whether the current sensor data has any abnormal offsets, thereby determining the temperature measurement deviation value. Similarly, the data from the N groups of pressure sensors is processed using the same method, and the RMS deviation between the different sensors is calculated. This is then combined with historical pressure measurement data for trend comparison to determine the pressure measurement deviation value. A temperature uncertainty model conforming to a Gaussian distribution is established based on the temperature measurement deviation and the mean temperature measurement data from M groups of temperature sensors. This model characterizes the uncertainty distribution of the current temperature parameter based on the concentration and fluctuation range of all current temperature measurements. A credibility analysis is performed on each set of temperature measurements using statistical methods. Its reliability is determined based on its relative position in the measurement distribution and assigned a corresponding weight, forming a primary uncertainty representation of the temperature parameter measurement. A pressure uncertainty model is constructed based on the pressure measurement deviation and the average level of the current measurement values ​​from N groups of pressure sensors. The credibility of each set of pressure measurements is assessed. The confidence level is determined based on its distribution within the measurement distribution, and a secondary uncertainty representation of the pressure parameter measurement is generated.

[0030] Step 200: performing data fusion on M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and performing data fusion on N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state;

[0031] Specifically, based on the temperature uncertainty model, the uncertainty distribution of temperature measurement is analyzed. A temperature state update equation is constructed to describe the dynamic process of temperature change over time. A temperature measurement equation is also constructed to express the mapping relationship between sensor data and the true temperature state. The state update equation characterizes the inherent evolutionary logic of temperature during system operation, while the measurement equation reflects how the observed data reflects the true system state. Together, these two equations form the temperature state-space model. Based on the pressure uncertainty representation, corresponding pressure state update equations and pressure measurement equations are established. The former describes how pressure changes over time or with operation, while the latter explains how the data measured by various sensors relate to the true pressure state. Together, these two equations form the pressure state-space model. A temperature fusion objective function is created based on the temperature state-space model, and a pressure fusion objective function is created based on the pressure state-space model. The temperature fusion objective function aims to minimize the deviation between sensor observations and the true temperature state, while taking into account inter-sensor consistency and measurement uncertainty, ensuring that the fusion process maintains accuracy while enhancing robustness. The pressure fusion objective function follows a similar approach, aiming to achieve high-precision reconstruction of the true pressure state by minimizing both estimation error and observation deviation. Based on the aforementioned objective function, an adaptive Kalman filter algorithm is introduced as the core computational tool for data fusion. The adaptive Kalman filter algorithm dynamically adjusts its internal covariance structure based on the uncertainty of the input data, thereby maintaining stable estimation accuracy even in scenarios with unstable noise or fluctuating sensor performance. When executing this algorithm, M sets of temperature sensor data and their uncertainty parameters are input into the temperature fusion objective function. Combined with the state-space model, a series of prediction and correction operations are performed to obtain the optimal estimate of the temperature state. Similarly, N sets of pressure sensor data and their uncertainty are input into the pressure fusion objective function. Through the corresponding adaptive Kalman filter process, the pressure state is continuously predicted and corrected, ultimately outputting the optimal estimate of the pressure state.

[0032] Step 300: Input the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into the thermodynamic equilibrium model, the crystallization kinetics model, and the system response model for parallel prediction to obtain a crystallization process prediction result;

[0033] It should be noted that the optimal estimated values ​​of the temperature state and the optimal estimated values ​​of the pressure state are input into the thermodynamic equilibrium model, the crystallization kinetics model, and the system response model. In the thermodynamic equilibrium model, the optimal estimated values ​​of temperature and pressure are used as key variables, and the equilibrium state of the crystallization system under the current working conditions is predicted by the internally constructed thermodynamic state equation, thereby outputting a series of thermodynamic equilibrium prediction values ​​such as solubility, saturation concentration, and phase equilibrium conditions, which characterize the phase boundary and energy state of the crystallization process under the current physical environment. At the same time, the same temperature and pressure state estimates are input into the crystallization kinetics model in combination with the material concentration parameters obtained in real time. The model comprehensively considers key factors such as the kinetic rate constant of crystal formation, the crystal face growth mechanism, and the change in crystal surface area, and then predicts the crystallization rate, crystal size change trend, and solid-liquid conversion efficiency under the current temperature, pressure, and material conditions, thereby forming a crystallization kinetics prediction value, which reflects the changing trajectory of crystal quality and yield over a period of time in the future. In the system response model, based on optimal estimates of temperature and pressure, the control system's response characteristics are analyzed, including the feedback behavior of the cooling, heating, and vacuum control systems to current parameter changes. The system's internal actuators' hysteresis characteristics, control time constants, and output stability are predicted to generate a predicted system response. Multi-model predictions are generated based on the predicted thermodynamic equilibrium, crystallization kinetics, and system response values. The errors in each model's historical prediction tasks are calculated in real time, and the historical predictions are compared with actual measurements to assess the accuracy of each model over different time periods and operating conditions. Based on these prediction errors, the weight coefficients corresponding to each model are dynamically adjusted using a predefined weight update rule. This ensures that models with high prediction accuracy receive a greater weight in the ensemble, while models with large prediction errors are automatically weakened. These adjusted model weights become the dynamic weight coefficients. The prediction results from multiple models are then integrated using the dynamic weight coefficients, integrating the prediction capabilities of each model across different dimensions. A weighted combination approach yields a crystallization prediction, including the trends in temperature, pressure, and crystallization rate over the future prediction time domain.

[0034] In this embodiment, the optimal estimated values ​​of the temperature and pressure states are standardized to eliminate the impact of dimensional differences and enhance the adaptability of the model. Through state normalization, the temperature and pressure states are converted into normalized temperature and pressure state values ​​within a uniform numerical range. Based on the normalized temperature and pressure state values, a pre-built historical control database is accessed and a set of historical control data most similar to the current state is extracted from the database. This data includes the refrigeration system input values, vacuum pump system input values, and heating system input values ​​at the corresponding time. These three types of input variables are combined to form a control input vector, which represents the typical control behavior of the crystallization system in the current state. Simultaneously, external disturbance factors present at the current moment are identified. Using mechanical vibration data collected by the equipment vibration sensor, ambient temperature data of the plant or equipment recorded by the ambient temperature sensor, and feed property change data provided by the feed composition detection module, a set of external disturbance vectors reflecting environmental and process disturbances is comprehensively constructed. This vector dynamically reflects the potential impact of non-control variables on system stability and serves as a necessary input to the response model, assisting the model in identifying the coupling relationship between control effects and disturbance effects. The normalized temperature state value, normalized pressure state value, control input vector, and external disturbance vector are input into the temperature response function of the system response model. This function, modeled based on historical system behavior, heat exchange characteristics, and device dynamics, calculates the temperature variation trend under the current control and disturbance combination and outputs a predicted temperature response value, representing the temperature variation due to the control input and external disturbance in the future time domain. Similarly, the same set of normalized state, control input, and disturbance variables are input into the pressure response function. This function, modeled based on factors such as the leakage characteristics of the system vacuum chamber, the influence of gas-liquid phase equilibrium, and the response delay of the pump group, combines historical operating patterns to output a predicted pressure response value, i.e., an estimate of the pressure variation under the control input and disturbance. Based on the predicted temperature and pressure response values ​​and the current initial state, a stepwise iteration is performed to calculate the system temperature and pressure values ​​for several future time points (e.g., k time points), restoring the temperature and pressure variation trend over the entire prediction time interval. The temperature and pressure prediction series for the k future time points are combined to generate the predicted system response value.

[0035] Step 400: Execute optimal process parameter target point and dynamic fault tolerance boundary value analysis based on the crystallization process prediction result to generate a process parameter control strategy.

[0036] Specifically, the system analyzes the set of standard process parameters that the system follows. This set includes recommended ranges for key variables such as temperature, pressure, and crystallization rate. By analyzing historically representative parameter configurations under stable operating conditions, a set of representative reference values ​​is extracted as standard process reference points. Based on the predicted results of the crystallization process at the current moment, the system's feasible domain under disturbance constraints is analyzed, taking into account the expected temperature and pressure trends over a period of time. In combination with external disturbance information, the system's feasible region is analyzed. This region represents the acceptable temperature and pressure ranges for the system while maintaining stable operation. A disturbance sensitivity analysis is performed on this feasible range of temperature and pressure parameters. E sets of representative disturbance samples are constructed from the probability space of the disturbance variables. These samples cover common disturbance scenarios such as equipment vibration, ambient temperature fluctuations, and changes in feed composition. Simulations are performed for each set of disturbance samples. While maintaining the same control strategy, the degree of temperature and pressure deviation under these disturbance conditions is measured, forming a data set of temperature and pressure parameter deviations. By summarizing the parameter variation data under these E disturbance conditions, a parameter deviation distribution diagram is constructed, demonstrating the trend and deviation range of key system variables under different disturbance levels. On this basis, using a standard process reference point as the target and combining it with a parameter deviation distribution map, an optimization algorithm is employed to find a set of process parameter configurations that best balances control accuracy and system robustness under actual disturbance conditions. This is known as the optimal process parameter target point. This optimal target point is then compared with the deviation distribution map to evaluate the maximum acceptable deviations in temperature and pressure under various disturbance levels. The corresponding elastic variables are then determined using extreme value analysis. These elastic variables extend the traditional hard constraint boundaries to form dynamic tolerance boundaries with tolerances. These boundaries flexibly adapt to fluctuations in disturbance amplitude, providing a buffer during control strategy execution to avoid frequent constraint alarm triggering or actuator saturation. Using the optimal process parameter target point and the dynamic tolerance boundaries as input, a control objective analysis with a soft constraint mechanism is performed. A process parameter control strategy is generated by comprehensively considering multiple dimensions, including control error, energy consumption variation, actuator response speed, and system safety margin.

[0037] In this embodiment, a temperature-pressure state evolution model is constructed based on the predicted trend of future temperature, pressure and crystallization rate. The model describes the dynamic change process of temperature and pressure in the future time period, while taking into account factors such as system response delay, disturbance feedback and coupling between states, so as to improve the expression ability of the dynamic characteristics of the system. After completing the construction of the state evolution model, an objective function for control optimization is created based on the process parameter control strategy. The objective function aims to minimize the difference between the current system state and the optimal process parameter target point, while taking into account the stability and execution efficiency of the control action, so that the overall control behavior is both accurate and smooth. On the basis of establishing the objective function, constraints are introduced into the optimization problem. In order to ensure the feasibility and safety of the control sequence in actual operation, the optimization problem contains four types of constraints, namely state soft constraints, control input constraints, control increment constraints and non-negative constraints of elastic variables. State soft constraints limit temperature and pressure to the upper and lower limits of the dynamic fault tolerance boundary, but allow a certain degree of overshoot, and adjust the tolerance by introducing elastic variables. Control input constraints ensure that the cooling power, heating intensity, and vacuum pump speed are within the permissible operating range of the equipment. Control increment constraints limit the amplitude of changes between two consecutive control actions to prevent drastic adjustments that could cause instability. The non-negativity constraint of the elastic variables ensures that all soft constraint adjustments remain physically reasonable. By integrating these constraints into the optimization model, a multivariable, multi-constrained constrained optimization problem is formed. To solve this optimization problem, an efficient solver based on the interior point method is introduced. This method effectively handles complex constraint relationships while maintaining convergence stability. Using the interior point method, the optimization problem is iteratively calculated to obtain a robust set of temperature and pressure control sequences. This control sequence clearly indicates the control instructions to be taken by the cooling, vacuum pump, and heating systems over multiple future time steps, ensuring sufficient robustness of the control strategy under different disturbance scenarios. After obtaining the control sequence, only the first control action is actually executed. This strategy combines the idea of ​​rolling time domain, that is, re-updating the system status and prediction results in each sampling period, and repeating the entire optimization solution process based on the new input data, thereby continuously generating new control sequences covering the current time domain, making the entire control process real-time and adaptable. In the actual execution process, the control instructions obtained by the above calculation are converted into specific physical drive instructions through the control execution unit. Among them, the control signal of the refrigeration system includes the temperature setting value and the refrigeration power, the vacuum pump system receives the speed and valve opening adjustment signal, and the heating system receives the heating power and on-off control instructions. These instructions are parsed in the control execution unit and applied to the physical equipment, so that the temperature and pressure are closed-loop regulated according to the calculation results.

[0038] In the embodiment of the present application, by collecting multiple sets of temperature and pressure sensor data and performing Gaussian distribution uncertainty analysis, the reliability of the measurement data is quantitatively evaluated, which effectively solves the problem of inaccurate and unreliable measurement of a single sensor, and adaptively fuses the multiple sets of sensor data according to the uncertainty characterization, so that the system can dynamically adjust the weights of different sensors, filter outliers and suppress noise interference, and ensure more reliable temperature and pressure state estimation. The thermodynamic equilibrium model, crystallization kinetics model and system response model are used in combination to comprehensively describe the low-temperature evaporation crystallization process through complementary theoretical foundations, overcoming the limitation that a single model cannot accurately describe a complex crystallization process. The weight coefficients of each model are dynamically adjusted based on the prediction error to achieve adaptive fusion of the model prediction results, so that the system can automatically select the most suitable model combination according to different working conditions, thereby improving the prediction accuracy and adaptability. By analyzing the impact of external interference on the system, the optimal process parameter target point and dynamic fault tolerance boundary value are calculated, so that the control system has sufficient fault tolerance while ensuring control accuracy, and effectively responds to interference such as feed fluctuations and environmental changes. The use of a soft-constraint optimization method with elastic variables improves the system's ability to handle constraint violations while ensuring accurate control of temperature and pressure parameters. This allows the control system to maintain stable operation under various interference conditions, avoiding the potential instability caused by traditional hard-constraint control methods when constraints are violated. Through continuous optimization and feedback adjustment within a rolling time domain, real-time updates of the control strategy and precise adjustment of temperature and pressure parameters are achieved, enabling the low-temperature evaporation crystallizer to maintain stable operation and high-quality crystallization results under dynamically changing operating conditions.

[0039] In a specific embodiment, the process of executing step 100 may specifically include the following steps:

[0040] Collect M groups of temperature sensor data and N groups of pressure sensor data of the low-temperature evaporation crystallizer;

[0041] Calculate the root mean square value of the deviation between the measurement values ​​of different temperature sensors based on M groups of temperature sensor data and perform data comparison and analysis in combination with historical temperature measurement data to obtain the temperature measurement deviation value. At the same time, calculate the root mean square value of the deviation between the measurement values ​​of different pressure sensors based on N groups of pressure sensor data and perform data comparison and analysis in combination with historical pressure measurement data to obtain the pressure measurement deviation value.

[0042] A temperature uncertainty model conforming to the Gaussian distribution is established based on the temperature measurement deviation value and the temperature measurement mean of M groups of temperature sensor data. The credibility of each temperature measurement value is calculated using the probability density function to obtain the first uncertainty representation of the temperature parameter measurement.

[0043] A pressure uncertainty model conforming to the Gaussian distribution is established based on the pressure measurement deviation value and the pressure measurement mean of N groups of pressure sensor data. The credibility of each pressure measurement value is calculated through the probability density function to obtain the second uncertainty representation of the pressure parameter measurement.

[0044] Specifically, data from M sets of temperature sensors and N sets of pressure sensors are collected at key locations in the crystallizer. M and N represent the number of temperature and pressure sensor groups deployed, respectively. Each group is composed of multiple sensor types to create redundancy and enhance interference resistance. Temperature sensors include platinum resistance thermometers, thermocouples, and infrared thermometers, while pressure sensors include piezoelectric, capacitive, and thin-film sensors. All sensors are sampled at a uniform time step, typically at a frequency above 10 Hz, to ensure the system captures rapidly changing physical state characteristics. After collecting real-time data from the M sets of temperature sensors, the measured values ​​from each group are compared horizontally. By comparing the measurement results of each sensor group at the same time point and calculating the root mean square deviation between the values, a preliminary assessment of the current temperature measurement dispersion—that is, the consistency of the responses of each temperature sensor under the same physical environment—is made. Furthermore, to enhance objectivity and robustness, the current temperature measurement results are compared with historical temperature data under the same operating conditions over the same sampling period to assess whether the current measurement data exhibits abnormal drift, significant deviation, or the possibility of sensor failure. After completing the above comparison, the lateral RMS deviations are integrated with the longitudinal historical trend deviations to generate a temperature measurement deviation value. Simultaneously, the N sets of pressure sensor data are processed using the same process. The RMS deviations of the measurements from each set of pressure sensors at the same time point are calculated to determine their response consistency. Combined with historical pressure data under similar temperature and feed conditions, longitudinal trend analysis is performed to output the pressure measurement deviation value. Based on the temperature measurement deviations and the current mean measurement values ​​of the M sets of temperature sensors, a temperature uncertainty model conforming to the characteristics of a Gaussian distribution is constructed. As a statistical modeling form with excellent mathematical properties and practical applicability, the Gaussian distribution effectively describes the concentration and fluctuation range of measurement errors. In this model, the mean temperature measurement value represents the center of the distribution, while the temperature measurement deviation value serves as the basis for variance estimation, characterizing the dispersion of the measurement data. Once this distribution is established, the confidence level of each set of temperature sensor measurements is calculated using the probability density function of the Gaussian distribution. Each temperature data point is substituted into the probability density function, and its confidence level within the overall distribution is determined by its distance from the mean and the density of the distribution. Sensor data with higher confidence levels, whose measured values ​​are closer to the statistically true values, should be given higher weight in subsequent multi-sensor fusion and state estimation. Conversely, data with lower confidence levels will be suppressed during data fusion to avoid misleading the system's estimation results. This step provides a first uncertainty representation for the temperature parameter measurement. A Gaussian distribution uncertainty model for the pressure parameter is constructed based on the pressure measurement deviation and the mean of the current N sets of pressure sensor measurements.In this model, the mean pressure value determines the center of the distribution, while the pressure measurement deviation determines the distribution's morphological width, reflecting the statistical characteristics of the data from different pressure sensors. The model uses the probability density function to perform a credibility analysis on each set of pressure measurements. By calculating the position of each pressure value on the probability density curve, its credibility relative to the population is determined. Based on this, each set of pressure sensors is assigned a weight parameter representing its credibility, resulting in a second uncertainty representation of the pressure parameter measurements.

[0045] In a specific embodiment, the process of executing step 200 may specifically include the following steps:

[0046] Establishing a temperature state update equation and a temperature measurement equation based on the first uncertainty representation, and generating a temperature state space model according to the temperature state update equation and the temperature measurement equation;

[0047] Establishing a pressure state update equation and a pressure measurement equation based on the second uncertainty representation, and generating a pressure state space model according to the pressure state update equation and the pressure measurement equation;

[0048] Creating a temperature fusion objective function based on the temperature state space model, and creating a pressure fusion objective function based on the pressure state space model;

[0049] An adaptive Kalman filter operation is performed based on the temperature fusion objective function to obtain an optimal estimate of the temperature state, and an adaptive Kalman filter operation is performed based on the pressure fusion objective function to obtain an optimal estimate of the pressure state.

[0050] Specifically, based on the first uncertainty representation, a temperature state update equation and a temperature measurement equation are established, thereby constructing a temperature state-space model. This state update equation describes the inherent laws of temperature evolution over time, comprehensively considering the impact of factors such as system heat capacity, heat transfer efficiency, control input, and external disturbances on temperature trends. Process noise is also introduced to represent the cumulative disturbance of unpredictable factors. The measurement equation, on the other hand, describes the sensor's observation of the current system's true temperature state, reflecting the indirect relationship between the temperature state and the measured values ​​of multiple sensors. In this process, the confidence interval and bias level defined in the temperature uncertainty model are converted into statistical characteristics of measurement noise, and the estimation error during the state evolution process is introduced to provide a probabilistic basis for estimation algorithms such as Kalman filtering. In this way, when constructing the temperature state-space model, the dynamic response characteristics of the temperature control system are preserved, while the statistical characteristics of measurement errors are organically integrated into the model framework, making the model more adaptable to real-world situations. Similarly, for pressure data, based on the second uncertainty representation, the corresponding pressure state update equation and pressure measurement equation are constructed to generate a pressure state-space model. When constructing the pressure state update equation, the transition relationship between the gas and liquid phases within the crystallization apparatus, the impact of vacuum pump regulation on chamber pressure, and the long-term effect of equipment sealing on pressure fluctuations are considered. These physical and process factors jointly determine the pressure evolution pattern in continuous time. In the construction of the pressure measurement equation, factors such as the characteristic differences, response speed, and system noise of different pressure sensor types are incorporated into the model. Furthermore, a weighting mechanism and an observation error control term are introduced into the measurement equation, drawing on the quantitative analysis of the reliability of each sensor in the second uncertainty characterization. An objective function for data fusion is created based on the structural characteristics and parameter information of each model. The core of the temperature fusion objective function is to minimize the weighted deviation between the predicted temperature state and the sensor measurement value. The uncertainty levels of different sensors are dynamically weighted to increase the influence of high-confidence measurements on the estimation result. A state estimation error suppression term is introduced into the objective function to improve estimation stability and avoid short-term oscillations. In this function design, state transition error and observation error are respectively set as the two main components of the function cost. Dynamic adjustment of the weighting factors achieves dual optimization of model prediction performance and observation consistency. Similarly, the pressure fusion objective function is constructed based on the dynamic and observation equations in the pressure state-space model. The credibility distribution of the pressure uncertainty model is incorporated into the weight design to ensure minimal overall estimation bias and stable system response. In these two fusion objective functions, the smoothness of state evolution, consistency between observations, and preservation of historical states are simultaneously considered, ensuring the final estimation results are highly robust and operational.An adaptive Kalman filter algorithm is introduced as the core state estimation tool in determining the fusion objective function, providing optimal estimates for both temperature and pressure states. An adaptive mechanism is incorporated during execution to dynamically adjust the proportion of state covariance to measurement covariance in the Kalman gain, updating the correction margin between the predicted state and the actual measurement in real time. For temperature estimation, the adaptive Kalman filter uses real-time variance information provided by the first uncertainty representation to correct the weight of the prediction error, ensuring that highly reliable temperature sensor data dominates the state update. When sensor data fluctuates significantly within a short period of time or deviates significantly from the overall measurement trend, the filter automatically reduces its influence in the update process, preventing abnormal data from affecting the stability of the system's temperature estimation. For pressure state estimation, the filter dynamically integrates pressure sensor reliability information with historical measurement consistency trends to automatically adjust the prediction error gain coefficient, enhancing the guiding role of stable data. It also rapidly corrects prediction deviations when necessary, ultimately outputting an optimal pressure estimate that is closer to the actual system state. The entire estimation process utilizes a rolling time series update mechanism, recalculating the state prediction, covariance matrix, Kalman gain, and correction term within each sampling period to ensure that the system maintains state estimation accuracy and response sensitivity over a continuous time axis. Through the independent operation and coordinated output of two adaptive Kalman filters for temperature and pressure, the optimal estimated values ​​of the temperature state and the optimal estimated values ​​of the pressure state reflecting the current operating status of the crystallization device are obtained.

[0051] In a specific embodiment, the process of executing step 300 may specifically include the following steps:

[0052] Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into the thermodynamic equilibrium model for thermodynamic prediction to obtain a thermodynamic equilibrium prediction value;

[0053] Inputting the optimal estimated value of the temperature state, the optimal estimated value of the pressure state and the material concentration into the crystallization kinetics model to perform crystallization prediction and obtain a crystallization kinetics prediction value;

[0054] Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into the system response model to perform system response prediction and obtain the system response prediction value;

[0055] Generate multi-model prediction results based on thermodynamic equilibrium prediction values, crystallization kinetics prediction values, and system response prediction values;

[0056] Calculate the prediction error based on the prediction results of multiple models, and dynamically adjust the model weight coefficient according to the prediction error to obtain the dynamic weight coefficient;

[0057] The prediction results of multiple models are integrated and calculated according to the dynamic weight coefficient to obtain the crystallization process prediction results, which include the changing trends of temperature, pressure and crystallization rate in the future prediction time domain.

[0058] Specifically, the optimal estimates of the temperature and pressure states are input into a thermodynamic equilibrium model for predictive analysis at the thermodynamic level. Based on the correlation between phase diagram theory, solubility curves, and physical thermodynamic parameters, the thermodynamic equilibrium model analyzes the thermodynamic behavior of the material at the current temperature and pressure state, including whether saturation has been reached, whether crystallization is possible, the saturation level of the remaining solute in the solution, and whether energy conservation is maintained in the system. This model, embedded with an equation of state or empirical regression equation, rapidly determines the direction, rate, and critical temperature and pressure conditions of the system's approach to equilibrium based on the input temperature and pressure state. It then outputs a predicted thermodynamic equilibrium value, providing fundamental data for determining the presence of thermodynamic potential energy driving crystallization and the location of the crystallization zone. The optimal estimates of the temperature and pressure states, along with the material concentration collected in real time, are then input into a crystallization kinetics model for predictive calculations focused on microscopic processes. The crystallization kinetics model is based on mass transfer theory, nucleation rate equations, and crystal growth rates, and is adjusted based on the type of crystallizing agent and crystal structure parameters. Its core objective is to estimate the number of crystals formed per unit time, the rate of mass change, and the potential evolution of the particle size distribution based on the current state. In practical applications, this model, implemented using ordinary differential equations or neural networks, accurately simulates characteristics such as changes in nucleation rate caused by temperature fluctuations and phase boundary changes caused by pressure fluctuations. The model outputs dynamic predictions such as crystallization rate, average crystal size, and saturation trend, reflecting the crystallization process and product quality trends under the current state over a period of time. Furthermore, to more accurately understand the feedback effect of control inputs on the system's internal state, optimal estimates of the temperature and pressure states are introduced as input variables into the system response model to predict and analyze device-level response characteristics. The system response model, centered on the response inertia of physical devices, controller adjustment speed, and the coupling between actuators, simulates the physical response changes that may occur in devices such as refrigeration units, heating systems, and vacuum pump systems under specific temperature and pressure conditions. This model analyzes hysteresis, oscillation, or nonlinear behavior caused by changes in a control variable and adaptively adjusts parameters based on historical operating conditions to more realistically reproduce the system response dynamics. The system response prediction output by the model describes the actuator's operating trends, adjustment capacity boundaries, and the temperature and pressure change paths after the control inputs, under the current estimated state conditions. Multi-model predictions are generated based on predicted values ​​for thermodynamic equilibrium, crystallization kinetics, and system response. The predictive capabilities of each model at different time scales and output variables are uniformly mapped, yielding multi-source predictions for the three-dimensional parameters of temperature, pressure, and crystallization rate along the time axis. To evaluate the effectiveness of this multi-model integrated prediction, a comparison mechanism with historical real-world data is introduced. The deviation between each prediction model's output and the actual measured value is evaluated, and the corresponding prediction error is calculated.Prediction error assessment is not limited to the difference calculation at the final output. Instead, a sliding window strategy is introduced at multiple time points to statistically analyze the mean square, absolute deviation, or residual trend of the prediction error to form a quantitative evaluation indicator for the model's credibility. Based on this prediction error, a dynamic weight adjustment algorithm is used to update the weight coefficient of each model in the ensemble prediction. The weight adjustment strategy takes into account short-term prediction accuracy, long-term trend consistency, and model robustness. Models that maintain high accuracy over multiple time windows are given higher ensemble weights, while models with frequent deviation fluctuations or sensitivity to external interference are downgraded. This weight adjustment mechanism is adaptive and automatically redistributes model importance as operating conditions change, forming a set of dynamic weight coefficients. Based on a weighted fusion mechanism, the prediction results of multiple models are integrated and calculated to generate a crystallization process prediction result. This prediction result covers multiple future prediction moments and provides the changing trends of temperature, pressure, and crystallization rate in the continuous time domain.

[0059] In a specific embodiment, the step of inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into the system response model to perform system response prediction, and the process of obtaining the system response prediction value may specifically include the following steps:

[0060] Performing state normalization processing on the optimal estimated value of the temperature state and the optimal estimated value of the pressure state to obtain a normalized temperature state value and a normalized pressure state value;

[0061] extracting a refrigeration system input value, a vacuum pump system input value, and a heating system input value of a current control cycle from a historical control database according to the normalized temperature state value and the normalized pressure state value, and combining the refrigeration system input value, the vacuum pump system input value, and the heating system input value into a control input vector;

[0062] Calculate the external interference vector based on equipment vibration sensor data, ambient temperature sensor data and feed composition detection data;

[0063] The normalized temperature state value, the normalized pressure state value, the control input vector and the external disturbance vector are input into the temperature response function of the system response model to calculate the temperature variation and obtain the temperature response prediction value;

[0064] The normalized temperature state value, the normalized pressure state value, the control input vector and the external disturbance vector are input into the pressure response function of the system response model to calculate the pressure variation and obtain the pressure response prediction value;

[0065] The system temperature value and the system pressure value at the next k moments are calculated based on the temperature response prediction value and the pressure response prediction value, and the system response prediction value is generated based on the system temperature value and the system pressure value.

[0066] Specifically, the optimal estimated temperature and pressure states are normalized, unifying physical quantities of different dimensions and distributions into the same numerical range. This results in normalized temperature and pressure state values, facilitating subsequent gradient optimization and numerical fitting in various functional structures, such as neural networks and regression functions, while avoiding slow convergence and reduced accuracy due to dimensional inconsistencies. Based on these normalized temperature and pressure state values, a pre-built historical control database is accessed. This database records the correspondence between system states and control inputs over a large number of historical operating cycles, including temperature setpoints and power control commands for the cooling system, speed and valve opening control signals for the vacuum pump system, and power setpoints and on / off commands for the heating system under various temperature and pressure combinations. The current normalized state value is matched to the most similar or optimally correlated set of historical data in the database, and the corresponding control input parameters are extracted. The cooling, vacuum pump, and heating system input values ​​are then combined to construct a multidimensional control input vector. Furthermore, information about environmental disturbances that affect the accuracy of the response prediction at the current moment is obtained. To capture these disturbances, three dimensions of external sensor input are introduced: vibration sensors installed on the equipment, temperature sensors deployed within the plant environment, and a detection module for real-time monitoring of feed material composition changes. Data from the equipment vibration sensors reflects the stability of the system's mechanical structure or actuators during operation. Abnormal vibration indicates equipment performance degradation or load changes. Data from the ambient temperature sensor captures changes in the thermal environment within the crystallizer, directly impacting the load on the cooling and heating systems. Data on feed composition changes reflects the impact of raw material fluctuations on the thermodynamic and kinetic conditions of the crystallization process. These data from various sensors are consolidated, dimensionally transformed, and combined in a unified format to form a set of well-structured external disturbance vectors. The normalized temperature state value, normalized pressure state value, control input vector, and external disturbance vector are input into the temperature response function of the system response model to calculate the temperature change. This temperature response function is constructed based on the system's physical characteristics, actuator dynamic behavior, and historical response curves. It combines current control instructions and the intensity of the external disturbance to calculate the temperature change within the predicted time domain. The model structure uses an analytical model based on first-order or second-order response differential equations, using a data-driven recurrent neural network or recursive network structure. The specific structure is optimized and adjusted according to the actual system characteristics. This function predicts the temperature offset caused by the current control action and interference input, and combines it with the initial temperature state to calculate the temperature response prediction value. This prediction value represents the possible evolution trend of the system temperature in the short term future without changing the control strategy. In parallel with the temperature response process, the same input combination is passed into the pressure response function in the system response model. This function focuses on simulating the combined impact of the vacuum pump system, the system sealing structure, and the phase change behavior on the pressure change.Because the pressure response is more susceptible to sudden disturbances and equipment fluctuations, the function's design requires robustness and real-time correction capabilities. The pressure response prediction calculated by the model represents the predicted trend of pressure changes in the future time domain under the influence of the current control input and external disturbances, and is corrected based on the initial pressure state. Based on the predicted temperature and pressure response values, the prediction stage for k future time steps begins. At each prediction time step, the current state and the predicted change are superimposed, and the temperature and pressure prediction trajectories are recursively generated until a continuous prediction of the system's temperature and pressure states over k prediction steps is completed. This sequential prediction depicts the system's operating trend under the current control strategy and known disturbance conditions, and is used to predict whether the system is about to enter a critical range or whether there is a risk of temperature and pressure exceeding the limit. The temperature and pressure prediction values ​​for all k future time steps are then consolidated and combined to generate the system response prediction.

[0067] In a specific embodiment, the process of executing step 400 may specifically include the following steps:

[0068] Analyze the standard process parameter set of low-temperature evaporation crystallizer to obtain the standard process reference point;

[0069] According to the crystallization process prediction results, the feasible domain of the system state constraints under the current external interference conditions is analyzed to obtain the feasible range of temperature and pressure parameters;

[0070] Perform interference sensitivity analysis on the feasible range of temperature and pressure parameters, generate E groups of interference samples, and calculate the temperature and pressure parameter offset under each group of interference samples to obtain the parameter deviation distribution diagram;

[0071] Perform optimization based on the parameter deviation distribution map and the standard process reference point to obtain the optimal process parameter target point;

[0072] Based on the optimal process parameter target point and parameter deviation distribution diagram, the maximum allowable deviation of temperature and pressure under various interference levels is calculated, and the elastic variables are determined based on extreme value analysis to obtain the dynamic fault tolerance boundary value;

[0073] According to the optimal process parameter target point and dynamic fault tolerance boundary value, the control target analysis with soft constraints is performed to obtain the process parameter control strategy.

[0074] Specifically, historical operating data of a low-temperature evaporation crystallizer under ideal operating conditions without external interference was categorized and analyzed to develop a set of standard process parameters. This set includes key parameters such as temperature, pressure, crystallization rate, concentration stability, and crystal size uniformity. The range of these values ​​represents the optimal state of the equipment under designed conditions. Through clustering and statistical analysis of the standard data, a set of parameter configurations that best represent optimal operating behavior was extracted, known as the standard process reference points. Based on the crystallization process prediction results, the feasible domain of system state constraints under current external interference conditions was analyzed. Combining the predicted future trends of temperature, pressure, and crystallization rate with currently collected external disturbance information (such as feed fluctuations, ambient temperature changes, and equipment vibration levels), a dynamic state constraint mapping model was constructed. This model analyzes the range of temperature and pressure values ​​that can be achieved under a given control capability and disturbance context, identifies the physical boundaries allowed during system operation, and derives the feasible range of temperature and pressure parameters—the set of temperature and pressure combinations that allows the system to operate safely and stably without violating its constraints under the current interference conditions and control capability. A disturbance sensitivity analysis is performed on the feasible range of temperature and pressure parameters. By constructing a disturbance sample space, the system generates E sets of representative external disturbance samples based on the statistical characteristics of the disturbance variables. Each set of samples corresponds to a possible disturbance combination scenario, such as high feed concentration fluctuations, high ambient temperature, and high mechanical vibration. Each set of disturbance samples is input into the established system state prediction framework. The temperature and pressure responses under each disturbance are simulated, and the temperature and pressure parameter offsets for each set of disturbance samples are calculated. These offsets form discrete results of the system response under multiple disturbance conditions. This data is aggregated to construct a parameter deviation distribution map, showing the range and frequency characteristics of possible temperature and pressure variations under multiple disturbance scenarios. Using this parameter deviation distribution map and the standard process reference point as optimization inputs, a process parameter target optimization model is constructed. While satisfying the current feasible range and future disturbance impacts, the target temperature and pressure are kept as close as possible to the standard process reference point, while maintaining sufficient fault tolerance and adjustment margin. Using a multi-objective optimization algorithm, a cost function is constructed to balance multiple indicators such as the degree of parameter deviation, system stability, and control energy consumption. By solving this optimization problem, the optimal process parameter target point that is most suitable for the crystallization process under the current interference scenario is obtained. On the basis of obtaining the optimal process parameter target point, combined with the parameter deviation distribution diagram, the maximum deviation value that can be tolerated by temperature and pressure under various interference levels is analyzed. This analysis calculates the fluctuation limit of the statistical parameters under each type of interference situation and calculates the elastic variable based on the extreme value theory, that is, the temperature and pressure are allowed to exceed the acceptable range of the target point under short-term or local disturbances. The elastic variable is used to soften the original hard process constraints. By introducing a certain controllable elastic buffer zone, the system can gain greater adjustment space and operation flexibility under high disturbance conditions.The elastic variables obtained through extreme value analysis will be used as dynamic fault-tolerant boundary values ​​to give the system a time-varying parameter allowable fluctuation range, so as to cope with complex disturbance changes without triggering system alarms or shutdowns. The optimal process parameter target point and the dynamic fault-tolerant boundary value are input together into the control objective function with soft constraints to conduct a comprehensive analysis of the entire control system objectives. This objective function covers the system's tracking accuracy requirements for temperature and pressure, and uniformly models the minimization of energy consumption of control inputs, the smoothness of parameter adjustment, and the feasibility of actuator responses. At the same time, a constraint penalty term for the elastic variable is introduced to prevent the system from deviating from the target point from expanding indefinitely. By optimizing and solving this objective function, a set of process parameter control strategies is output.

[0075] In a specific embodiment, the method for adjusting the parameters of the low-temperature evaporation crystallizer further includes the following steps:

[0076] Based on the crystallization process prediction results, a temperature-pressure state evolution model is constructed, and the objective function of the optimization problem is created according to the process parameter control strategy;

[0077] Adding state soft constraints, control input constraints, control increment constraints, and elastic variable non-negative constraints to the objective function of the optimization problem to obtain a constrained optimization problem;

[0078] The constrained optimization problem is solved using the interior point method to obtain a robust temperature and pressure control sequence, which includes control instructions for the refrigeration system, vacuum pump system, and heating system.

[0079] The first control action of the robust temperature and pressure control sequence is executed, and at the next sampling moment, the optimization process is repeated based on the new measurement state to obtain the temperature and pressure control sequence in the rolling time domain;

[0080] The control execution unit converts the control instructions in the temperature and pressure control sequence in the rolling time domain into physical drive signals. The physical drive signals include the temperature setting value and cooling power of the refrigeration system, the speed and valve opening of the vacuum pump system, and the power and on-off instructions of the heating system, thereby realizing closed-loop control of the temperature and pressure in the low-temperature evaporation crystallizer.

[0081] Specifically, based on the prediction results of the crystallization process, that is, the trend information of temperature, pressure and crystallization rate in the future prediction time domain, combined with the historical state evolution path and the system structure response characteristics, a temperature and pressure state evolution model including control input, state transition, disturbance response and feedback delay is established. The core task of this model is to describe the logical law of the transition of the current state to the next moment under the influence of control input and external disturbance in any control cycle. Therefore, explicit state update terms, external disturbance terms and control input terms are introduced as driving variables to ensure that the model has strong adaptability under disturbance dominance, parameter fluctuations and nonlinear influences. The objective function of the optimization problem is created according to the process parameter control strategy. The objective function includes multiple weight adjustment terms to balance the priorities between different objectives. For example, one part of the objective function is used to minimize the deviation between the predicted state and the optimal process parameter target, thereby ensuring that the process output approaches the optimal operating conditions. Another part is used to control the smoothness of the input increment to prevent drastic fluctuations in the controller output that could cause system oscillations. In addition, a soft constraint penalty term is introduced to constrain the temperature and pressure states to operate within the dynamic fault tolerance boundary, while allowing violations of hard constraints within a certain range to enhance the system's flexible adjustment capabilities under sudden disturbances. All of these objective function components together form the basic structure for controller optimization. To ensure the engineering feasibility of the objective function, multiple types of constraints are added to the optimization problem. State soft constraints set adjustable upper and lower limits for temperature and pressure, allowing them to temporarily exceed standard limits under the control of certain elastic variables but preventing them from deviating indefinitely. Control input constraints explicitly limit the output range of actuators, such as cooling power, heating intensity, and vacuum pump speed, ensuring that control instructions do not exceed physical execution capabilities. Control increment constraints limit the amplitude of control variable changes between adjacent control cycles to prevent equipment load shocks caused by excessively rapid adjustments. Elastic variable non-negativity constraints require that all allowed state extension boundaries must be positive and have a boundary adjustment mechanism to prevent the system from losing control in the face of large fluctuations. Integrating these constraints into the optimization framework creates a complex constrained optimization problem with soft and hard constraints, multiple objective terms, and disturbance variables. To efficiently solve this constrained optimization problem, an interior point method is introduced as a solution algorithm for the optimal control sequence. The interior point method is an iterative solution method suitable for high-dimensional, strongly constrained optimization problems. It exhibits excellent convergence and computational stability, and can accurately satisfy multiple constraints while effectively converging on the optimization objective. The optimization problem is solved within each sampling period using an interior point method, resulting in an optimal temperature and pressure control sequence covering the future rolling forecast horizon. This control sequence clearly indicates how the output parameters of the refrigeration, vacuum pump, and heating systems should be adjusted over the next several control cycles to maintain the temperature and pressure near the process target while adhering to the constraint boundaries and avoiding hard violations.This control sequence employs a rolling optimization approach, using only the first control action in the current sampling cycle as the actual control output. The remaining actions are recalculated in the next sampling cycle. Each time the system enters the next sampling point, it reassesses the current state based on the new estimated temperature and pressure state and disturbance monitoring data, updates the state evolution model, and repeatedly solves the entire optimization problem. This forms a continuously iterative rolling optimization closed loop, resulting in a rolling time-domain temperature and pressure control sequence. The control instructions in this sequence are transmitted to the control execution unit, completing the conversion from digital signals to physical actions. As the end of the system control chain, the control execution unit splits the received control instructions and maps them to specific actuators. The cooling system receives the temperature setpoint and cooling power adjustment parameters to adjust heat exchange efficiency and refrigerant flow; the vacuum pump system receives the speed setpoint and valve opening control signals to adjust the system's internal pressure and pumping rate; and the heating system adjusts the heating output of the electric heating device based on the received power control variable and on / off signal, achieving dynamic control of the heating rate. The entire instruction is converted into electrical, hydraulic or mechanical actions at the execution layer through drive signals, ensuring that the system is physically adjusted according to the control strategy, ultimately achieving high-precision, predictable and robust closed-loop control of temperature and pressure in the low-temperature evaporation crystallizer.

[0082] The above describes the method for adjusting the parameters of the low-temperature evaporation crystallizer in the embodiment of the present application. The following describes the device 10 for adjusting the parameters of the low-temperature evaporation crystallizer in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the low-temperature evaporation crystallizer parameter adjustment device 10 includes:

[0083] An acquisition module 11 is configured to acquire M sets of temperature sensor data and N sets of pressure sensor data from a low-temperature evaporation crystallizer, and calculate a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement;

[0084] a data fusion module 12 configured to perform data fusion on M sets of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and to perform data fusion on N sets of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state;

[0085] A parallel prediction module 13 is used to input the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into the thermodynamic equilibrium model, the crystallization kinetics model and the system response model for parallel prediction to obtain a crystallization process prediction result;

[0086] The generation module 14 is used to perform optimal process parameter target point and dynamic fault tolerance boundary value analysis based on the crystallization process prediction results to generate a process parameter control strategy.

[0087] Through the collaborative efforts of these components, the system collects data from multiple temperature and pressure sensors and performs Gaussian uncertainty analysis to quantitatively assess the reliability of the measurement data, effectively addressing the inaccuracy and unreliability of single sensor measurements. Adaptive Kalman filtering is then applied to the data based on uncertainty characterization, enabling the system to dynamically adjust the weights of different sensors, filter outliers, and suppress noise interference, ensuring more reliable temperature and pressure state estimates. By combining thermodynamic equilibrium models, crystallization kinetic models, and system response models, the system comprehensively describes the low-temperature evaporation crystallization process based on complementary theoretical foundations, overcoming the limitation of single models in accurately describing complex crystallization processes. By dynamically adjusting the weight coefficients of each model based on prediction error, the system achieves adaptive fusion of model prediction results, automatically selecting the most appropriate model combination for different operating conditions, improving prediction accuracy and adaptability. By analyzing the impact of external disturbances on the system and calculating optimal process parameter targets and dynamic fault tolerance boundaries, the control system maintains control accuracy while maintaining sufficient fault tolerance to effectively cope with disturbances such as feed fluctuations and environmental changes. The use of a soft-constraint optimization method with elastic variables improves the system's ability to handle constraint violations while ensuring accurate control of temperature and pressure parameters. This allows the control system to maintain stable operation under various interference conditions, avoiding the potential instability caused by traditional hard-constraint control methods when constraints are violated. Through continuous optimization and feedback adjustment within a rolling time domain, real-time updates of the control strategy and precise adjustment of temperature and pressure parameters are achieved, enabling the low-temperature evaporation crystallizer to maintain stable operation and high-quality crystallization results under dynamically changing operating conditions.

[0088] See also Figure 3 , Figure 3 This is a schematic block diagram of the structure of an electronic device 300 provided in an embodiment of the present application. The electronic device 300 includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected via a device bus 303, wherein the memory 302 may include a non-volatile storage medium and an internal memory.

[0089] The non-volatile storage medium can store a computer program. The computer program includes program instructions. When the program instructions are executed by the processor 301, the processor 301 can execute any of the above-mentioned low-temperature evaporation crystallizer parameter adjustment methods.

[0090] The processor 301 is used to provide computing and control capabilities to support the operation of the entire electronic device 300 .

[0091] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned low-temperature evaporation crystallizer parameter adjustment methods.

[0092] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the electronic device 300 involved in the solution of the present application. The specific electronic device 300 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0093] It should be understood that the processor 301 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0094] It should be noted that, those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the electronic device 300 described above can refer to the corresponding process of the aforementioned low-temperature evaporation crystallizer parameter adjustment method, and will not be repeated here.

[0095] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, the one or more processors implement the low-temperature evaporation crystallizer parameter adjustment method provided in the embodiment of the present application.

[0096] The computer-readable storage medium may be an internal storage unit of the electronic device 300 in the aforementioned embodiment, such as a hard disk or memory of the electronic device 300. The computer-readable storage medium may also be an external storage device of the electronic device 300, such as a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, etc., equipped with the electronic device 300.

[0097] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0098] If the integrated unit is implemented in the form of 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 application, 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. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0099] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for adjusting parameters of a low-temperature evaporation crystallizer, characterized in that: include: Collecting M sets of temperature sensor data and N sets of pressure sensor data from a low-temperature evaporation crystallizer, and calculating a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement; Performing data fusion on the M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and performing data fusion on the N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state; Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model, a crystallization kinetics model, and a system response model for parallel prediction to obtain a crystallization process prediction result; Based on the crystallization process prediction results, the optimal process parameter target point and dynamic fault tolerance boundary value analysis are performed to generate a process parameter control strategy; specifically, the following steps are performed: analyzing the standard process parameter set of the low-temperature evaporation crystallizer to obtain a standard process reference point; analyzing the feasible domain of the system state constraints under the current external interference conditions according to the crystallization process prediction results to obtain a feasible interval of temperature and pressure parameters; performing interference sensitivity analysis on the feasible interval of temperature and pressure parameters to generate E groups of interference samples, and calculating the temperature and pressure parameter offsets under each group of interference samples to obtain a parameter deviation distribution diagram; performing optimization solution based on the parameter deviation distribution diagram and the standard process reference point to obtain an optimal process parameter target point; calculating the maximum allowable deviations of temperature and pressure under each interference level according to the optimal process parameter target point and the parameter deviation distribution diagram, and determining the elastic variables based on extreme value analysis to obtain a dynamic fault tolerance boundary value; performing control target analysis with soft constraints according to the optimal process parameter target point and the dynamic fault tolerance boundary value to obtain a process parameter control strategy.

2. The low-temperature evaporation crystallizer parameter adjustment method according to claim 1, characterized in that: The method of collecting M groups of temperature sensor data and N groups of pressure sensor data of a low-temperature evaporation crystallizer and calculating a first uncertainty representation of temperature parameter measurement and a second uncertainty representation of pressure parameter measurement includes: Collect M groups of temperature sensor data and N groups of pressure sensor data of the low-temperature evaporation crystallizer; Calculating the root mean square value of the deviation between the measurement values ​​of different temperature sensors based on the M sets of temperature sensor data and performing data comparison and analysis in combination with historical temperature measurement data to obtain a temperature measurement deviation value; and calculating the root mean square value of the deviation between the measurement values ​​of different pressure sensors based on the N sets of pressure sensor data and performing data comparison and analysis in combination with historical pressure measurement data to obtain a pressure measurement deviation value; Establishing a temperature uncertainty model that conforms to a Gaussian distribution based on the temperature measurement deviation value and the temperature measurement mean of the M groups of temperature sensor data, and calculating the credibility of each temperature measurement value through a probability density function to obtain a first uncertainty representation of the temperature parameter measurement; A pressure uncertainty model conforming to Gaussian distribution is established based on the pressure measurement deviation value and the pressure measurement mean of the N groups of pressure sensor data, and the credibility of each pressure measurement value is calculated through a probability density function to obtain a second uncertainty representation of the pressure parameter measurement.

3. The method for adjusting parameters of a low-temperature evaporation crystallizer according to claim 1, wherein: The performing data fusion on the M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and performing data fusion on the N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state, includes: Establishing a temperature state update equation and a temperature measurement equation based on the first uncertainty representation, and generating a temperature state space model according to the temperature state update equation and the temperature measurement equation; Establishing a pressure state update equation and a pressure measurement equation based on the second uncertainty representation, and generating a pressure state space model according to the pressure state update equation and the pressure measurement equation; Creating a temperature fusion objective function according to the temperature state space model, and creating a pressure fusion objective function according to the pressure state space model; An adaptive Kalman filter operation is performed based on the temperature fusion objective function to obtain an optimal estimate of the temperature state, and an adaptive Kalman filter operation is performed based on the pressure fusion objective function to obtain an optimal estimate of the pressure state.

4. The method for adjusting parameters of a low-temperature evaporation crystallizer according to claim 1, wherein: The step of inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model, a crystallization kinetics model, and a system response model for parallel prediction to obtain a crystallization process prediction result includes: Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model for thermodynamic prediction to obtain a thermodynamic equilibrium prediction value; Inputting the optimal estimated value of the temperature state, the optimal estimated value of the pressure state, and the material concentration into a crystallization kinetics model to perform crystallization prediction to obtain a crystallization kinetics prediction value; Inputting the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a system response model to perform system response prediction to obtain a system response prediction value; generating a multi-model prediction result according to the thermodynamic equilibrium prediction value, the crystallization kinetics prediction value, and the system response prediction value; Calculating a prediction error based on the multi-model prediction results, and dynamically adjusting a model weight coefficient according to the prediction error to obtain a dynamic weight coefficient; The multi-model prediction results are integrated and calculated according to the dynamic weight coefficient to obtain the crystallization process prediction result, which includes the change trends of temperature, pressure and crystallization rate in the future prediction time domain.

5. The method for adjusting parameters of a low-temperature evaporation crystallizer according to claim 4, wherein: Inputting the temperature state optimal estimate value and the pressure state optimal estimate value into a system response model to perform system response prediction to obtain a system response prediction value includes: Performing state normalization processing on the optimal estimated value of the temperature state and the optimal estimated value of the pressure state to obtain a normalized temperature state value and a normalized pressure state value; extracting a refrigeration system input value, a vacuum pump system input value, and a heating system input value of a current control cycle from a historical control database according to the normalized temperature state value and the normalized pressure state value, and combining the refrigeration system input value, the vacuum pump system input value, and the heating system input value into a control input vector; Calculate the external interference vector based on equipment vibration sensor data, ambient temperature sensor data and feed composition detection data; Inputting the normalized temperature state value, the normalized pressure state value, the control input vector, and the external disturbance vector into a temperature response function of a system response model to calculate a temperature variation, thereby obtaining a temperature response prediction value; Inputting the normalized temperature state value, the normalized pressure state value, the control input vector, and the external disturbance vector into the pressure response function of the system response model to calculate the pressure variation and obtain a pressure response prediction value; The system temperature value and the system pressure value at k moments in the future are calculated according to the temperature response prediction value and the pressure response prediction value, and the system response prediction value is generated according to the system temperature value and the system pressure value.

6. The method for adjusting parameters of a low-temperature evaporation crystallizer according to claim 1, wherein: The low-temperature evaporation crystallizer parameter adjustment method further includes: Constructing an evolution model of the temperature-pressure state based on the crystallization process prediction results, and creating an objective function of the optimization problem according to the process parameter control strategy; Adding state soft constraints, control input constraints, control increment constraints, and elastic variable non-negative constraints to the objective function of the optimization problem to obtain a constrained optimization problem; The constrained optimization problem is solved using an interior point method to obtain a robust temperature and pressure control sequence, wherein the robust temperature and pressure control sequence includes control instructions for a refrigeration system, a vacuum pump system, and a heating system; Executing the first control action of the robust temperature and pressure control sequence, and repeating the optimization process based on the new measurement state at the next sampling moment to obtain a temperature and pressure control sequence in a rolling time domain; The control execution unit converts the control instructions in the temperature and pressure control sequence in the rolling time domain into physical drive signals. The physical drive signals include the temperature setting value and cooling power of the refrigeration system, the speed and valve opening of the vacuum pump system, and the power and on-off instructions of the heating system, thereby realizing closed-loop control of the temperature and pressure in the low-temperature evaporation crystallizer.

7. A low-temperature evaporation crystallizer parameter adjustment device, characterized in that: Used to perform the low-temperature evaporation crystallization machine parameter adjustment method according to any one of claims 1 to 6, the low-temperature evaporation crystallization machine parameter adjustment device comprises: An acquisition module is used to acquire M groups of temperature sensor data and N groups of pressure sensor data of the low-temperature evaporation crystallizer, and calculate a first uncertainty representation of the temperature parameter measurement and a second uncertainty representation of the pressure parameter measurement; a data fusion module, configured to perform data fusion on the M groups of temperature sensor data according to the first uncertainty representation to obtain an optimal estimate of the temperature state, and to perform data fusion on the N groups of pressure sensor data according to the second uncertainty representation to obtain an optimal estimate of the pressure state; A parallel prediction module is used to input the optimal estimated value of the temperature state and the optimal estimated value of the pressure state into a thermodynamic equilibrium model, a crystallization kinetics model and a system response model for parallel prediction to obtain a crystallization process prediction result; A generation module is used to perform optimal process parameter target point and dynamic fault tolerance boundary value analysis based on the crystallization process prediction results to generate a process parameter control strategy.

8. An electronic device, characterized in that: The electronic device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the electronic device to execute the low-temperature evaporation crystallizer parameter adjustment method according to any one of claims 1 to 6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the low-temperature evaporation crystallizer parameter adjustment method according to any one of claims 1 to 6 is implemented.

Images