Intelligent measurement and control robot system and method for concentration process

Abstract

The application discloses a kind of concentrated process intelligent measurement and control robot system and method, belong to industrial process control field.For the technical problem that existing concentration process is difficult to adapt to nonlinear time-varying characteristics, cannot realize multi-objective dynamic optimization, the initial model based on heat mass transfer mechanism is constructed, the digital twin model of concentration process is obtained by step parameter self-tuning to temperature calculation module and heat transfer calculation module;Based on the model, end point concentration, concentration time and total energy consumption are used as optimization target to carry out multi-objective Pareto optimization, and the optimal process parameter setting value is determined and issued for execution;In batch operation, closed-loop control is carried out through real-time feedback deviation, so that actual parameters track optimal trajectory;Cross-batch cumulative data triggers model parameter re-identification and iterative update.The application also provides a robot system for realizing the above method.The control precision and self-adaptive ability of the concentration process are significantly improved, and the collaborative optimization of quality, efficiency and energy consumption is realized.

Description

Technical Field

[0001] This invention belongs to the field of industrial process control and intelligent manufacturing technology, specifically a concentration process intelligent measurement and control robot system and method. Background Technology

[0002] Concentration is a key unit operation widely used in modern industrial production, including chemical, pharmaceutical, food, and environmental protection industries. Its core objective is to remove solvents (usually water) from materials through methods such as heating and depressurization, thereby increasing the concentration of the target product to meet the requirements of subsequent processing, storage, or transportation. For example, in the pharmaceutical industry, extracts of traditional Chinese medicine are concentrated to achieve a specified ointment density; in wastewater treatment, evaporation and concentration reduce the volume of waste liquid and recover valuable salts.

[0003] Currently, the main technical bottlenecks in the measurement and control methods of the concentration process in industrial production are as follows.

[0004] First, there is a significant lag in process awareness and quality control. Currently, many concentration processes still rely on manual, timed sampling, with samples sent to laboratories for offline analysis to obtain key quality parameters such as concentration and composition. This "sampling-sending-testing" model results in a significant time lag (typically tens of minutes to several hours) between the test results and the actual production status, preventing operators from timely and accurately grasping the instantaneous state of the concentration process. This lag not only makes it difficult to achieve refined process control but also leads to a severe delay in responding to abnormal operating conditions (such as feed fluctuations and equipment performance degradation). Furthermore, the manual sampling and offline analysis processes themselves are prone to introducing operational errors, further reducing the reliability of the data.

[0005] Secondly, it is difficult to adapt to the complex characteristics of the process in terms of process parameter control. Concentration processes, especially vacuum evaporation concentration, are typical complex thermal processes characterized by nonlinearity, large time-varying variations, and multivariate coupling. Their dynamic characteristics change in real time with variations in factors such as material viscosity, concentration, heating steam pressure, and vacuum level. Traditional control strategies, such as manual adjustment based on experience or PID control with fixed parameters, are difficult to effectively handle these complex characteristics. Specifically: (1) Poor adaptability: When the batch of materials changes, the equipment status (such as heat exchanger scaling) drifts, or there is external disturbance, the preset control parameters cannot adapt automatically, resulting in a decrease in control effect; (2) Difficulty in achieving multi-objective dynamic optimization: Actual production often requires seeking a balance between multiple conflicting objectives, such as pursuing high production efficiency (shortening batch time), reducing energy consumption, and ensuring product quality (stable final concentration). Traditional methods can usually only achieve fixed-value control of a single parameter (such as temperature and pressure), and cannot dynamically adjust the operating strategy according to real-time operating conditions to dynamically achieve the optimal balance between efficiency, energy consumption, and quality.

[0006] The aforementioned problems are particularly pronounced in large-scale, continuous industrial production scenarios, and have become a bottleneck restricting the automation and intelligentization of concentration processes. Therefore, there is an urgent need for a new method that can achieve real-time, accurate sensing, intelligent decision-making, and closed-loop optimization control of the concentration process to overcome the shortcomings of existing technologies. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide an intelligent measurement and control robot system and method for the concentration process, which aims to solve the technical problems that the existing concentration process parameter control is difficult to adapt to complex nonlinear time-varying characteristics and cannot achieve multi-objective dynamic optimization; by constructing an online correctable digital twin model and combining multi-objective optimization and closed-loop feedback control, the invention achieves accurate perception, intelligent decision-making and coordinated optimization of energy efficiency, quality and efficiency in the concentration process.

[0008] To achieve the above objectives, the present invention provides the following technical solution: This invention first proposes an intelligent measurement and control method for the concentration process, comprising the following steps: Step 1: Constructing a digital twin model of the concentration process: The physical structure parameters and material property parameters of the concentration equipment are obtained, and an initial mechanism model based on the heat and mass transfer mechanism is constructed. Historical operating data is collected, and the parameters of the temperature calculation module in the initial mechanism model are calibrated so that the predicted values ​​of steam temperature and liquid temperature are close to the measured values. Based on the calibrated temperature parameters, the parameters of the heat transfer calculation module in the initial mechanism model are identified so that the predicted values ​​of evaporation rate, concentration time, and total energy consumption are close to the measured values, thus obtaining a digital twin model of the concentration process. Step Two: Multi-objective Optimization and Decision Making Based on the digital twin model, with the optimization objectives of achieving the target concentration, minimizing the concentration time, and minimizing the total energy consumption, and combined with the process parameter constraints, a multi-objective optimization algorithm is used to generate the optimal combination of Pareto process parameters, and the optimal process parameter settings are determined based on the current production status. Step 3: Closed-loop control: During batch operation, the actuator of the concentration equipment is controlled according to the optimal process parameter set value; process data is collected in real time and compared with the predicted value of the digital twin model; the steam flow rate, vacuum degree and feed flow rate are adjusted in real time according to the deviation obtained from the comparison, so that the actual process parameters track the trajectory of the optimal process parameters. Step 4: Online Update When the accumulated sample size reaches a set threshold or the model prediction deviation exceeds the limit, the parameters of the digital twin model are re-identified, and the parameters of the temperature calculation module and the heat transfer calculation module are iteratively updated.

[0009] Furthermore, in step one, the parameter calibration method for the temperature calculation module is as follows: Establish the mapping relationship between steam flow rate and steam pressure, and the mapping relationship between steam pressure and steam temperature; The boiling point of pure water is calculated based on the vacuum degree, and the boiling point rise is corrected based on the real-time concentration to obtain the predicted value of the drug solution temperature. We construct the weighted sum of squares of the predicted deviations of steam temperature and liquid temperature, respectively, and minimize the sum of the two as the objective function for temperature calibration to obtain the calibration values ​​of the temperature mapping parameter and the boiling point elevation parameter.

[0010] Furthermore, in step one, the parameter identification method for the heat transfer calculation module is as follows: Along the time axis of the concentration process, the real-time concentration is calculated based on the current mass, and then the current heat transfer temperature difference and net heat transfer power are calculated. The current instantaneous evaporation rate is calculated based on the corrected latent heat of vaporization, and the current batch mass is updated. Repeat the above steps until the current quality reaches the target endpoint quality, and obtain the predicted enrichment time and total energy consumption. We construct the weighted sum of squares of the prediction errors of instantaneous evaporation rate, concentration time, and total energy consumption, respectively, and use the sum of these three as the objective function for heat transfer identification to minimize the solution, thereby obtaining the identification values ​​of heat transfer coefficient, heat loss ratio, and latent heat of vaporization correction coefficient.

[0011] Furthermore, in step two, a hierarchical search strategy is adopted for multi-objective optimization, as follows: The first layer uses coarse-grained search to generate candidate parameter points within the process parameter range, calls the digital twin model for evaluation, and eliminates candidate points that do not meet the constraints. The second layer performs a fine-grained search centered on the optimal region obtained from the coarse search, scores candidate points based on the comprehensive evaluation function, and selects the parameter combination with the best score as the recommended operation point.

[0012] Furthermore, in step two, a multi-objective Bayesian optimization method is used for optimization. The method is as follows: a surrogate model of the digital twin model is constructed based on the initial sample, and new candidate process parameter points are selected using the hypervolume lifting criterion to gradually approach the optimal solution region; after determining the optimal region, a local fine search method is further used to refine and optimize the region.

[0013] Furthermore, in step two, the evaluation function is composed of a weighted sum of the endpoint quality deviation function, the concentration time function, and the energy consumption function; The weight coefficients of the evaluation function are determined by dynamically fusing and normalizing the preset initial weights of the production mode with the CRITIC weights and entropy weights calculated based on real-time operating data, so that the weight coefficients can adapt to the priority changes of different production scenarios.

[0014] Furthermore, in step two, the process parameter constraints include: steam flow rate constraints, vacuum degree constraints, liquid temperature constraints, heat transfer temperature difference constraints, single batch running time constraints, and final concentration constraints.

[0015] Furthermore, step three also includes online incremental correction of the heat transfer coefficient in the digital twin model: the current heat transfer coefficient is calculated back based on the real-time steam flow rate and the real-time heat transfer temperature difference, and the calculated heat transfer coefficient is written back to the digital twin model after amplitude limiting to update the model state.

[0016] This invention also proposes an intelligent measurement and control robot system for the concentration process, used to implement the intelligent measurement and control method for the concentration process as described above, comprising: The sensing module, configured on the concentration equipment, is used to collect data on material temperature, tank vacuum, material density, feed flow rate, steam flow rate, and valve opening in real time during the concentration process. The data processing module includes a data preprocessing unit, a digital twin model unit, and a database unit. The data preprocessing unit is used to clean, denoise, and time-stamp align the raw data collected by the sensing module. The digital twin model unit has a built-in digital twin model of the condensation process, which is used to update the model state in real time based on the preprocessed data to achieve dynamic simulation, state prediction, and energy consumption calculation of the condensation process. The database unit is used to store the raw sensing data, preprocessed data, model parameters, optimization results, and historical production data. The decision control module includes a multi-objective optimization unit and an intelligent decision unit; the multi-objective optimization unit is used to solve for the Pareto optimal process parameter combination based on the digital twin model and the set optimization objectives; the intelligent decision unit is used to determine the optimal process parameter set value from the Pareto optimal solution set according to the current production status and generate control commands. The feedback execution module is used to receive the control commands and drive the actuator to adjust the process parameters; and to compare the real-time collected measured data with the predicted values ​​of the digital twin model, and to make real-time feedback adjustments based on the comparison deviation. The cross-batch learning module is used to accumulate running data across batches. When the accumulated sample size reaches a set threshold or the model prediction deviation exceeds a set range, it triggers the parameter re-identification of the digital twin model and iteratively updates the parameters of the temperature calculation module and the heat transfer calculation module.

[0017] Furthermore, it also includes a human-computer interaction module, which comprises: The visual monitoring interface provides a graphical display of real-time parameter trend curves in the concentration process, simulation results of digital twin models, equipment operating status, and the achievement of optimization goals. It also supports operators in querying, setting, and manually intervening in key parameters. An alarm unit is used to issue a warning message when the real-time collected data deviates from the normal prediction range of the digital twin model.

[0018] The beneficial effects of this invention are as follows: The intelligent measurement and control method for the concentration process of the present invention achieves the following technical effects: (1) The accuracy and adaptability of the model are significantly improved: The present invention adopts a step-by-step decoupling strategy that combines "temperature calibration" and "heat transfer identification" to overcome the shortcomings of traditional mechanism models that rely on fixed empirical parameters and are difficult to adapt to batch differences of materials. By constructing weighted objective functions of temperature prediction deviation and heat transfer prediction deviation respectively for parameter self-tuning, the digital twin model can accurately characterize the nonlinear time-varying characteristics of the concentration process, providing a high-fidelity virtual mapping basis for subsequent optimization decisions. (2) Achieving synergistic optimization of quality, efficiency and energy consumption: Based on a high-fidelity digital twin model, this invention takes achieving the target concentration, minimizing the concentration time and minimizing the total energy consumption as multiple objectives. It adopts the Pareto optimization strategy to dynamically generate the optimal combination of process parameters, which effectively solves the contradiction in traditional control where "ensuring quality sacrifices efficiency and improving efficiency increases energy consumption", and achieves a dynamic balance among the three. (3) Possesses closed-loop adaptive anti-interference capability: By comparing the deviation between the measured data and the model prediction in real time, the steam flow rate, vacuum degree and feed flow rate are dynamically adjusted to ensure that the actual process parameters always track the optimal trajectory; when disturbances such as raw material fluctuations and equipment aging occur, the system can automatically trigger parameter re-identification and model update to form a robust control closed loop of "perception-decision-execution-evolution".

[0019] In summary, this invention, from model building to intelligent decision-making to closed-loop control and autonomous evolution, has interconnected and synergistic steps that significantly improve the control accuracy, adaptability, and overall operational efficiency of the concentration process. Attached Figure Description

[0020] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration: Figure 1 This is a flowchart of the intelligent measurement and control method for the concentration process of the present invention; Figure 2 This is a framework diagram of the intelligent measurement and control robot system for the concentration process of this invention. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0022] This invention proposes an intelligent monitoring and control robot system and method for the concentration process, aiming to solve the technical problems of existing concentration processes where process parameter control is difficult to adapt to complex process characteristics and cannot achieve multi-objective dynamic optimization. This improves the automation level, production efficiency, and product quality stability of the concentration process, while reducing energy consumption. Specifically, the intelligent monitoring and control robot system deeply integrates perception, cognition, and decision-making capabilities. By constructing a digital twin model of the concentration process, it achieves real-time, accurate perception and intelligent control. Specifically, the system first uses multiple types of sensors (such as temperature sensors, pressure sensors, online concentration detectors, and flow sensors) deployed at key locations in the concentration equipment to collect real-time process data, including material temperature, heating steam pressure, material flow rate, and real-time concentration, as well as equipment operating status parameters, forming a comprehensive process perception data chain. Then, based on the collected massive amounts of data, a dynamic prediction model for the concentration process is constructed using machine learning algorithms. This model can accurately predict the concentration effect and energy consumption indicators under different combinations of process parameters and continuously optimizes the model's accuracy through online learning to adapt to the time-varying characteristics and complex disturbances of the concentration process. Based on this, the decision-making system of the concentration cognitive robot combines production objectives (such as set concentration, efficiency requirements, energy consumption limits, etc.) and uses multi-objective optimization algorithms (such as NSGA-II, MOPSO, etc.) to dynamically optimize key process parameters (such as steam flow rate, vacuum degree, etc.) to generate the best control strategy. Finally, the optimized process parameters are precisely applied to the concentration equipment through the actuator to achieve closed-loop control. Through the above mechanism, the concentration cognitive robot of this invention can effectively overcome the limitations of traditional concentration process detection and control methods, realizing the transformation of the concentration process from experience-driven to data-driven, intelligent decision-making.

[0023] Specifically, such as Figure 1 As shown, the intelligent measurement and control method for the concentration process in this embodiment includes the following steps.

[0024] Step 1: Construct a digital twin model of the concentration process. Obtain the physical structure parameters and material property parameters of the concentration equipment, and construct an initial mechanism model based on the heat and mass transfer mechanism. Collect historical operating data and calibrate the parameters of the temperature calculation module in the initial mechanism model to make the predicted steam temperature and liquid temperature values ​​approximate the measured values. Based on the calibrated temperature parameters, identify the parameters of the heat transfer calculation module in the initial mechanism model to make the predicted evaporation rate, concentration time, and total energy consumption values ​​approximate the measured values, thus obtaining a digital twin model of the concentration process.

[0025] In this embodiment, the steps for constructing a digital twin model of the condensation process are as follows.

[0026] 11) Construct an initial mechanistic model. Based on the physical structural parameters of the concentration equipment and the physicochemical properties of the materials, establish a mechanistic model describing the mass transfer, heat transfer, and phase change laws during the concentration process. The mechanistic model is used to describe the mass transfer, heat transfer, and phase change laws of the materials during the concentration process.

[0027] (1) Mechanism model.

[0028] In this embodiment, the steps for constructing the mechanism model are as follows.

[0029] 111) Calculate the pipeline pressure based on the valve opening, and then calculate the steam temperature based on the pipeline pressure:

[0030]

[0031] in: For pipeline pressure; The steam temperature; and Upper and lower limits for pipeline pressure design; Valve opening degree (dimensionless, 0~100); , and is a coefficient.

[0032] 112) Using the Antoine equation, based on the vacuum degree Calculate the boiling point of pure water The actual liquid temperature was further calculated using the boiling point elevation constant. :

[0033]

[0034] in: To convert the vacuum level into pressure values ​​in millimeters of mercury; In this embodiment, the pressure is atmospheric pressure. ; As a pressure unit conversion constant, in this embodiment, ; This is a parameter for boiling point elevation; , and is a coefficient.

[0035] 113) Based on the temperature difference between steam and liquid feed Calculate the actual net heat transfer power absorbed. :

[0036]

[0037] in: The heat transfer coefficient; The heat transfer area of ​​the evaporator; This represents the percentage of heat loss from the equipment.

[0038] 114) Based on the net heat transfer rate Calculating the instantaneous evaporation rate using latent heat of vaporization :

[0039]

[0040] in: It is the latent heat of vaporization. and These are all engineering thermodynamic constants, obtained by looking up tables. For example... The latent heat of vaporization of water at 0°C (kJ / kg) is an engineering thermodynamic constant. It is the coefficient that decreases latent heat for every 1°C increase in temperature, and is also an engineering thermodynamic constant.

[0041] 115) Based on the initial concentration of the incoming material Calculate the total evaporation water volume and required concentration time based on the feed rate and final concentration. :

[0042]

[0043] in: This represents the total amount of water evaporated. For the quality of incoming materials; The target endpoint concentration.

[0044] 116) Calculate the total steam energy consumption of the concentration process. :

[0045]

[0046]

[0047] in: This refers to the equivalent energy consumption of steam. For vacuum pump power consumption; This refers to standby power. Vacuum pump power as a function of vacuum degree A changing function; In order to be in The degree of vacuum at any given moment; As a reference vacuum level; is the power-pressure coupling coefficient; n is the power-pressure curve fitting exponent.

[0048] (2) Data acquisition and preprocessing.

[0049] In this embodiment, the physical structural parameters of the equipment include: steam valve opening, pipeline pressure, heating temperature, vacuum degree, heat exchange area of ​​the evaporator, heat transfer coefficient of the evaporator, and the proportion of heat loss of the equipment (related to steam temperature and ambient temperature; the higher the steam temperature, the greater the proportion of loss). The physicochemical properties of the material include: boiling point of pure water, actual liquid temperature, latent heat of vaporization, initial density of the incoming material, and feed rate.

[0050] In this embodiment, process data is collected in real time by sensors installed at key locations in the concentration equipment. For the collected process data, this embodiment employs the Hampel outlier removal algorithm and the linear interpolation time alignment algorithm to preprocess the sensor-collected data on steam flow rate, steam temperature, vacuum degree, liquid temperature, initial density, real-time density, concentration time, and energy consumption, obtaining a batch-segmented effective sample set. The process involves resampling each batch of data according to its timestamp and aligning all sensor sequences to a unified time axis; removing or replacing outliers that exceed the preset deviation range of the middle value in the sliding window; and then dividing the data into batches based on the start time of material feeding, the time of vacuum establishment, and the time of paste collection to obtain the cleaning dataset required for subsequent parameter calibration.

[0051] 12) Temperature calculation module parameter calibration method: establish the mapping relationship between steam flow rate and steam pressure, and the mapping relationship between steam pressure and steam temperature; calculate the boiling point of pure water based on vacuum degree, and correct the boiling point rise value based on real-time concentration to obtain the predicted value of liquid temperature; construct the weighted sum of squares of steam temperature prediction deviation and the weighted sum of squares of liquid temperature prediction deviation respectively, and minimize the sum of the two as the temperature calibration objective function to obtain the calibration values ​​of temperature mapping parameters and boiling point rise parameters.

[0052] Specifically, in this embodiment, using steam flow rate, vacuum degree, measured steam temperature, measured liquid temperature, and initial concentration as inputs, a bounded nonlinear least squares algorithm is employed to construct a temperature calibration objective function in the form of a weighted sum of steam temperature prediction deviation and liquid temperature prediction deviation. This objective function is then solved to calibrate and obtain a set of temperature parameters, which includes at least: flow-pressure mapping coefficients a and b; and a steam temperature mapping coefficient. , , Antoine equation coefficients , , ; and the boiling point elevation benchmark (BPR).

[0053] The specific method is as follows: First, determine the set of undetermined coefficients to be optimized (including 3 coefficients in the heating temperature calculation formula, 4 coefficients related to the Antoine equation and boiling point rise constant in the actual liquid temperature calculation, and the heat transfer coefficient and equipment heat loss ratio in the net heat transfer rate calculation). Compare the key process data collected in real time by sensors, such as steam valve opening, pipeline pressure, heating temperature, vacuum degree, actual liquid temperature, initial density of incoming material, feed rate, evaporation rate, and total steam energy consumption at different times, with the predicted values ​​from the digital twin mechanism model to construct a loss function with the goal of minimizing the deviation between the two. Then, use an optimization algorithm to optimize the above undetermined coefficients. During the optimization process, the sensor-collected data needs to be preprocessed, including outlier detection and removal (e.g., based on the 3σ principle or isolated forest algorithm) and data normalization (unifying parameters of different magnitudes to the [0,1] or [-1,1] interval) to improve the convergence speed and optimization accuracy of the algorithm.

[0054] In this embodiment, the steam flow rate is used as the reference. Vacuum degree Measured steam temperature Measured temperature of the medicine solution and initial concentration As input, the steam-side temperature mapping parameters and boiling point elevation parameters are used. For the output, a bounded nonlinear least squares algorithm is used to calibrate the temperature-related parameters, resulting in a set of temperature parameters. The steps are as follows.

[0055] 121) Normalize the steam flow rate:

[0056] in: This is the normalized value for steam flow rate; Valve opening degree; This represents the maximum valve opening.

[0057] 122) The steam-side pressure is calculated using a quadratic mapping function:

[0058] in: It is the steam pressure; and These are the minimum and maximum values ​​of the steam pressure; and This is the flow-pressure mapping coefficient.

[0059] 123) Calculate the predicted steam temperature based on the steam pressure:

[0060] in: This is a predicted value for steam temperature. , and is a coefficient.

[0061] 124) Based on vacuum degree Calculate the boiling point of pure water :

[0062] in: Atmospheric pressure; , and is a coefficient.

[0063] 125) Calculate the effective boiling point of the drug solution based on the concentration-corrected boiling point elevation parameter:

[0064]

[0065] in: This is the effective boiling point of the drug solution; This is a parameter for boiling point elevation; This is the baseline value for the boiling point elevation at the initial concentration; This represents the real-time concentration at the current moment.

[0066] 126) Construct the objective function for temperature calibration. And solve for it, and obtain , , , , , , , and The calibration values. That is, the set of temperature parameters in this embodiment. .

[0067] In this embodiment, the constructed temperature calibration objective function is:

[0068] in: and The first Predicted and measured steam temperature values ​​at each time step; and The first Predicted and measured values ​​of the boiling point of the liquid at each time step; and These are the weighting coefficients.

[0069] In this embodiment, the calibration yielded: ; ; ; ; ; ; ; ; .

[0070] 13) Parameter identification for the heat transfer calculation module: Calculate the real-time concentration based on the current mass along the time axis of the concentration process, and then calculate the current heat transfer temperature difference and net heat transfer power; calculate the current instantaneous evaporation rate based on the corrected latent heat of vaporization, and update the current batch mass; repeat the above steps until the current mass reaches the target endpoint mass, and obtain the predicted values ​​of concentration time and total energy consumption; construct the weighted sum of squares of the predicted deviations of instantaneous evaporation rate, concentration time, and total energy consumption, respectively, and minimize the sum of the three as the objective function for heat transfer identification to obtain the identified values ​​of heat transfer coefficient, heat loss ratio, and latent heat of vaporization correction coefficient.

[0071] Specifically, in this embodiment, the calibrated set of temperature parameters, heat exchange area, steam flow rate, vacuum degree, liquid temperature, and initial batch mass are used as inputs. The Euler explicit integral algorithm combined with the weighted nonlinear least squares algorithm is used to construct a heat transfer identification objective function in the form of a weighted summation of evaporation rate prediction deviation, concentration time prediction deviation, and total energy consumption prediction deviation. The heat transfer parameter set is then identified, which includes at least the heat transfer coefficient, heat loss ratio, and latent heat of vaporization correction coefficient. This allows for the construction of a digital twin model that accurately reflects the dynamic characteristics of the concentration process.

[0072] Specifically, the heat transfer calculation module parameter identification uses the temperature parameter set obtained in step 12). Heat exchange area Steam flow rate Vacuum degree Temperature of the medicine solution and initial batch quality As input, with heat transfer coefficient heat loss ratio and latent heat of vaporization correction factor For the output, the Euler explicit integration algorithm combined with the weighted nonlinear least squares algorithm is used to identify the heat transfer and evaporation parameters, thus obtaining the set of heat transfer parameters. .

[0073] Specifically, within each time step dt, the following steps are performed.

[0074] 131) Calculate the current concentration:

[0075] in: This is the current concentration; For the quality of incoming materials; For the first The current quality of the time step; This is the initial concentration.

[0076] 132) Based on the boiling point elevation parameter Calculate the temperature of the feed liquid and heat transfer temperature difference :

[0077]

[0078] in: For the first Steam temperature at the time step; This is the boiling point of pure water.

[0079] 133) Calculate the net heat transfer power:

[0080] in: For the first Net heat transfer power at the time step; The heat transfer coefficient; The heat transfer area of ​​the evaporator; This represents the percentage of heat loss from the equipment.

[0081] 134) Calculate the instantaneous evaporation rate based on the evaporation heat consumption corrected for latent heat of vaporization:

[0082]

[0083] in: For the first Corrected latent heat of vaporization for the time step; For the first Instantaneous evaporation rate at the time step; This is the latent heat of vaporization correction factor; and All of these are engineering thermodynamic constants.

[0084] 135) Update batch quality:

[0085] in: For the first The quality of the time step.

[0086] when Less than or equal to the target mass Stop integration at that point to obtain the predicted concentration time for that batch. Simultaneously, the total energy consumption is obtained by accumulating both steam energy consumption and vacuum energy consumption. .

[0087] 136) Construct the heat transfer identification objective function. The heat transfer coefficient is then obtained by solving the problem. heat loss ratio and latent heat of vaporization correction factor The identification results. In this embodiment, the constructed heat transfer identification objective function is:

[0088] in: and The first Instantaneous evaporation rate at time step Predicted and measured values; and The first Time step condensation time Predicted and measured values; and The first Total energy consumption per time step Predicted and measured values; , and All are weighting coefficients.

[0089] Among them, the Instantaneous evaporation rate at time step The measured value can be calculated by collecting the liquid level of the medicine liquid using a level gauge, or the density of the medicine liquid in the tank can be predicted in real time using an independently deployed visual density soft measurement model. The result is obtained through reverse calculation. Specifically, the first value is obtained using an independently deployed visual density soft measurement model. Instantaneous evaporation rate at time step The method for measuring the values ​​is as follows.

[0090] First, obtain the real-time concentration at time step k using the density-concentration curve:

[0091] Then, the real-time mass of the liquid medicine in the tank is obtained by solute conservation:

[0092] Finally, the instantaneous evaporation rate is obtained by taking the mass difference between adjacent time steps:

[0093] in: This is the density-concentration calibration coefficient matrix (including slope a and intercept b). This refers to the initial solution density upon startup. The density of the solution at the final discharge point; This represents the total mass of the solution at the initial startup. The total mass of the solution when it is discharged from the boiler at the final point; The density of the solvent (water, take) ); This is the initial concentration (mass fraction). The endpoint concentration (mass fraction); This represents the time difference between two adjacent time steps.

[0094] The calibration was performed based on the collected data. Since the initial drug concentration was low, the concentration could be approximated by density. According to the law of conservation of solute, we can have Calibration coefficient matrix , utilize points and Establish the equation of the straight line Analysis: In this embodiment, , .

[0095] No. Total energy consumption per time step Measured values The measurement method is as follows: measured value The result is obtained by integrating the steam-side energy consumption and the vacuum pump power consumption:

[0096] in: This is the sum of the instantaneous steam mass flow rates of the two vortex flow meters; This represents the enthalpy difference between steam and feedwater. Vacuum pump power as a function of vacuum degree A changing function.

[0097] In this embodiment, the calibration is obtained ; ; .

[0098] Step Two: Multi-Objective Optimization and Decision-Making. Based on the aforementioned digital twin model, with the optimization objectives of achieving the target concentration, minimizing concentration time, and minimizing total energy consumption, and combined with process parameter constraints, a multi-objective optimization algorithm is used to generate the optimal Pareto process parameter combination, and the optimal process parameter setpoints are determined based on the current production status.

[0099] In this embodiment, the calibrated and online-corrected digital twin model and the initial quality of the current batch are used. The system takes real-time vacuum level, real-time steam flow rate, and preset process boundaries as inputs to recommend a steam flow rate. Recommended vacuum level The process parameters are optimized using a constrained hierarchical grid search algorithm, with the output being the allowable fluctuation range. This yields a recommended combination of process parameters that meets the concentration endpoint requirements while also considering time and energy consumption.

[0100] (1) Optimization of process parameters.

[0101] Specifically, this embodiment uses a constrained hierarchical grid search algorithm to optimize process parameters, as follows.

[0102] The first layer employs a coarse-grained search to generate candidate parameter points within the process parameter range. The digital twin model is then used for evaluation, eliminating candidate points that do not meet the constraints. Specifically, candidate parameter points are generated within the steam flow rate and vacuum degree ranges. For each candidate point, the digital twin model is used to calculate the feed temperature, evaporation rate, concentration completion time, and total energy consumption, eliminating candidate points that do not meet the feed temperature constraint, minimum effective heat transfer temperature difference constraint, and endpoint requirements.

[0103] The second layer performs a fine-grained search centered on the optimal region obtained from the coarse search, scores candidate points based on the comprehensive evaluation function, and selects the parameter combination with the best score as the recommended operation point.

[0104] Furthermore, perturbations are applied to the steam flow rate and vacuum near the optimal point, and the changes in concentration time and energy consumption are calculated to obtain the parameter sensitivity and allowable fluctuation range, which are then output as operational recommendations.

[0105] Of course, in other embodiments, a suitable multi-objective optimization algorithm can also be selected for optimization. Specifically, considering the characteristics of multi-objective optimization problems in the concentration process, such as high evaluation cost, numerous constraints, and mutual coupling between objectives, the multi-objective optimization algorithm in this embodiment can be a multi-objective Bayesian optimization method. Specifically, the method of optimization using the multi-objective Bayesian optimization method is as follows: a surrogate model of the digital twin model is constructed based on the initial samples; new candidate process parameter points are selected using the hypervolume lifting criterion; and the optimal solution region between time, energy consumption, and final mass is gradually approximated. After determining the optimal region, a local fine search method is further used to refine the optimization of this region, thereby improving the convergence speed and the accuracy of the parameter solution.

[0106] (2) Process parameter constraints. In this embodiment, the process parameter constraints include: steam flow rate constraint, vacuum degree constraint, liquid temperature constraint, heat transfer temperature difference constraint, single batch running time constraint, and final concentration constraint.

[0107] Specifically, the physical meaning and constraints of each parameter are explained as follows: When the temperature increases: the boiling point of the liquid increases, the heat transfer temperature difference decreases, and the evaporation rate decreases; When the concentration is increased: net heat transfer power increases, evaporation rate rises, and concentration time is shortened; When the volume increases: the effective heat transfer decreases; When the value increases: the heat consumption per unit of evaporation increases, and the total energy consumption increases; Steam flow rate and vacuum level: together affect steam temperature, boiling point of liquid feed, heat transfer temperature difference, and concentration completion time; , and Used for paste collection determination and linkage control, but not for thermodynamic parameter identification. Among them: The target endpoint density (or target endpoint concentration) refers to the set value of the density (or concentration) that the material is expected to reach at the end of the concentration process. It is the core process objective for determining whether the concentration process is complete. The lower limit of density is the minimum threshold for determining the concentration of the concentrate. When the real-time density is lower than this value, the concentration is considered to have failed. It can also be used as a safety constraint to prevent premature termination of concentration due to excessively low density. The upper limit of the allowable density refers to the upper limit threshold when the paste is collected. When the real-time density exceeds this value, it indicates that the material may be over-concentrated (such as the material becoming thick or clumping), which may trigger an alarm or automatically stop the operation to ensure product quality and equipment safety.

[0108] (3) Evaluation function for production optimization objectives. In this embodiment, the evaluation function is composed of a weighted sum of the endpoint quality deviation function, the enrichment time function, and the energy consumption function. The weight coefficients of the evaluation function are determined by dynamically fusing and normalizing the preset initial weights of the production mode with the CRITIC weights and entropy weights calculated based on real-time operating data, so that the weight coefficients can adapt to the priority changes of different production scenarios.

[0109] Specifically, based on production needs, optimization objectives for the concentration process are set, typically including achieving product density (or concentration) targets, maximizing production efficiency, minimizing energy consumption, and ensuring product quality stability (such as concentration fluctuation range). Based on the digital twin model constructed in step one, and combined with intelligent optimization algorithms such as genetic algorithms and particle swarm optimization, multi-objective optimization is performed on key process parameters affecting the concentration process (such as heating steam flow rate, material temperature, vacuum level, etc.) to obtain one or more sets of Pareto optimal process parameter combinations. The specific method is as follows.

[0110] First, clarify the quantitative indicators and constraints for each optimization objective. The optimization objectives mainly include: achieving the target product density (or concentration) requires setting an allowable fluctuation range; maximizing production efficiency can be quantified as minimizing the concentration completion time; minimizing energy consumption is measured by the steam consumption per unit product (kg steam / kg water evaporation); and product quality stability is characterized by the real-time density (or concentration) deviation during the concentration process, requiring this deviation to not exceed a set threshold. Constraints include: actual physical constraints on process parameters, such as the heating steam flow rate not exceeding the equipment's maximum rated value, the vacuum level being maintained within the equipment's safe operating range, and the material temperature being below the destructive temperature of its heat-sensitive components.

[0111] Secondly, a fitness function for multi-objective optimization is constructed based on the digital twin model. The fitness function is expressed as:

[0112]

[0113]

[0114]

[0115] in, Let be the vector of process parameters to be optimized. For the first One process parameter, ,like For hot steam flow rate, For vacuum degree, For valve opening, etc.; The endpoint quality deviation function; For condensed time function, It is an energy consumption function; The endpoint density or concentration predicted by the digital twin model; The target endpoint density or concentration; The maximum allowable range of endpoint fluctuations; The predicted condensed completion time for the digital twin model; As a time normalization benchmark; Total energy consumption predicted by the digital twin model; As a benchmark for energy consumption normalization; , and These are the weighting coefficients.

[0116] In this embodiment, the weighting coefficient , and The initial weights are determined using a combination of production mode initial weights and CRITIC-entropy weights: First, an initial weight vector is set according to the production mode that prioritizes energy saving, balance, or efficiency. Secondly, the endpoint deviation, concentration time, total energy consumption, and process variability within the sliding time window are dimensionless. The CRITIC method is used to calculate the comparative strength and conflict of each evaluation index, resulting in a weight vector. Simultaneously, the information entropy of each evaluation index is calculated using the entropy weight method to obtain the weight vector. Finally, , and Linear fusion and normalization are performed to obtain the real-time dynamic weight vector. ,satisfy This allows the evaluation function to reflect production mode preferences while also adaptively adjusting to real-time operating condition changes. When the endpoint concentration target and process safety boundary are introduced as strong constraints, It can be further transformed into penalty terms or constraint terms, with time targets and energy consumption targets being the main optimization objects in the comprehensive scoring.

[0117] In this embodiment, the process parameter constraints include: Steam flow constraint: ; Vacuum constraint: ; Boiling point constraint of the drug solution: ; Heat transfer temperature difference constraint: ; Single batch runtime constraints: ; endpoint concentration constraint: ; in: Steam flow rate; and These are the minimum and maximum values ​​of the steam flow rate, respectively. Vacuum degree; and These are the minimum and maximum values ​​of the vacuum degree, respectively; The boiling point of the liquid medicine; and These are the minimum and maximum values ​​of the boiling point of the medicinal solution, respectively. For heat transfer temperature difference; This is the minimum allowable heat transfer temperature difference; The steam temperature; This refers to the runtime of a single batch. This is the longest runtime in the history of batches. The endpoint concentration; The target concentration; This represents the allowable error for concentration.

[0118] Step 3: Closed-loop control: During batch operation, the optimal process parameter setpoints are applied to the actuators of the concentration equipment, and the prediction deviation is obtained by comparing the real-time collected measured data with the predicted values ​​of the digital twin model. Based on the prediction deviation, the steam flow, vacuum degree and feed flow of the concentration equipment actuators are adjusted in real time, so that the process parameter trajectory converges to the optimal process parameter curve determined by multi-objective optimization decision, thereby realizing closed-loop control.

[0119] Specifically, in the closed-loop control step of this embodiment, the heat transfer coefficient in the digital twin model is also incrementally corrected online. The method is as follows: the current heat transfer coefficient is back-calculated based on the real-time steam flow rate and the real-time heat transfer temperature difference. The back-calculated heat transfer coefficient is then written back to the digital twin model after being limited to update the model state.

[0120] In this embodiment, real-time steam temperature, real-time liquid medicine temperature, real-time steam flow rate, and heat exchange area are used. As input, the online heat transfer coefficient For the output, an inverse calculation update and limiting correction algorithm is used to dynamically correct the heat transfer coefficient online. Specifically, in this embodiment, the method for incrementally correcting the heat transfer coefficient in the digital twin model online includes the following steps.

[0121] 31) Based on real-time steam flow rate and steam temperature Calculate the steam-side heating power:

[0122] 32) Calculate the instantaneous heat transfer coefficient based on the real-time temperature difference:

[0123] 33) To Amplitude limiting is applied to constrain it within the allowable range of the project. Inside, and the revised Write back the digital twin model; where: and These are the lower and upper limits of the engineering allowable heat transfer coefficient, respectively.

[0124] Through precise digital twin modeling, the dynamic changes of various parameters during the actual concentration process can be accurately reproduced, thereby achieving a precise characterization of the dynamic characteristics of the concentration process. This provides a high-fidelity virtual mapping foundation for subsequent real-time monitoring and intelligent control, ensuring a high degree of consistency between model predictions and the actual production process. In terms of concentration process prediction, the digital twin model can dynamically predict the remaining concentration time and energy consumption required to reach the target endpoint density (or concentration), providing accurate time references for production scheduling and subsequent process connections. Regarding energy consumption optimization, the digital twin model can calculate the unit product energy consumption in real time based on the current combination of process parameters. Through the analysis of historical data, it can identify key sensitive parameters affecting energy consumption and their optimization ranges, providing data support for subsequent parameter optimization. For example, the model can simulate and compare the relationship between evaporation rate and steam consumption under different vacuum conditions to find the vacuum setting range that meets production efficiency requirements while achieving the lowest energy consumption.

[0125] Step 4: Online Update: Accumulate running data across batches. When the accumulated sample size reaches a set threshold or the prediction deviation of the digital twin model is detected to exceed the set range, trigger the re-identification of digital twin model parameters. Iterate and update the parameter sets of the temperature calculation module and the heat transfer calculation module to enable the digital twin model to adapt to long-term changes such as material batch differences and equipment aging.

[0126] Specifically, based on the Pareto optimal process parameter combination obtained in step two, and combined with the real-time status of the current concentration process (updated in real-time by the digital twin model) and production scheduling requirements, the decision-making system makes intelligent decisions to select the most suitable process parameter combination. Subsequently, these parameters are precisely applied to the concentration equipment through control actuators (such as steam regulating valves and vacuum control valves), and changes in process parameters are monitored in real time. If the actual process deviates from expectations due to fluctuations in raw material characteristics, changes in equipment status, or external disturbances, the digital twin model will quickly capture these deviations and trigger a new round of parameter optimization and control, forming a closed-loop control loop of "perception-decision-execution-feedback," realizing online updates and iterations of the model to adapt to the time-varying characteristics and disturbances of the process.

[0127] The specific implementation is as follows: First, the decision-making system automatically filters from the Pareto optimal solution set or allows the operator to interactively select an optimal combination of process parameters as the current execution plan based on the priority of the current production task (such as prioritizing energy consumption reduction while ensuring product quality, or prioritizing production efficiency when orders are urgent). For example, if the production objective is to minimize energy consumption and achieve the product concentration target, then the energy consumption function is selected. Minimum concentration scaling function The process involves several steps. First, the selected process parameters (such as heating steam flow rate and vacuum setpoint) are converted into control signals for the actuators. These signals are then sent to the corresponding execution units, such as steam regulating valves and vacuum system control valves, via a PLC (Programmable Logic Controller) or DCS (Distributed Control System), enabling precise regulation of the concentration process. During this regulation, sensors deployed on the equipment continuously collect real-time process data (such as actual steam flow rate, vacuum level, material temperature, and real-time concentration) and feed this data back to the digital twin model. The digital twin model updates its predictions of the concentration process status in real time based on the latest feedback data and evaluates the actual effect of the current combination of process parameters. If the actual operating state deviates from the expected target (such as concentration deviating from the set range or energy consumption exceeding expectations), the decision-making system automatically makes dynamic adjustments, re-invoking the multi-objective optimization algorithm to re-optimize the process parameters online based on the updated process data, generating a new regulation strategy until the process status returns to the ideal range. Through this closed-loop control mechanism of "perception-decision-execution-feedback," the concentration process is ensured to always operate in the optimal state, achieving a dynamic balance between product quality, production efficiency, and energy consumption.

[0128] This embodiment also proposes an intelligent measurement and control robot system for the concentration process, used to implement the intelligent measurement and control method for the concentration process described above in this embodiment. Figure 2 As shown, the intelligent measurement and control robot system for the concentration process in this embodiment includes a perception module, a data processing module, a decision control module, a feedback execution module, and a cross-batch learning module.

[0129] Specifically, the sensing module is configured on the concentration equipment to collect real-time data on material temperature, tank vacuum, material density, feed flow rate, steam flow rate, and valve opening during the concentration process. In this embodiment, the sensing module consists of multiple types of sensors deployed at key locations on the concentration equipment, including a temperature sensor (for collecting material temperature), a pressure sensor (for collecting vacuum in the concentration tank), an online concentration / density detector (for collecting feed density and real-time concentrate density), a flow sensor (for collecting feed flow rate and steam flow rate), and equipment status sensors (such as valve opening sensors). The core function of the sensing layer is to achieve comprehensive, real-time, and high-precision collection of physical quantities during the concentration process and to send the collected raw data to the data processing layer.

[0130] The data processing module includes a data preprocessing module, a digital twin model module, and a database module. The data preprocessing module is responsible for cleaning, denoising, outlier removal, data normalization, and time-stamp alignment of the raw data collected by the perception layer to ensure data quality. The digital twin model module is the core of the cognitive robot, integrating the digital twin mechanism model of the condensation process constructed in step one and its correction and optimization algorithms. It can update the model state in real time based on the preprocessed data, enabling dynamic simulation, state prediction, and energy consumption calculation of the condensation process. The data processing module also includes a database unit for storing raw perception data, preprocessed data, model parameters, optimization results, historical production data, and fault records. It can employ a combination of relational databases (such as MySQL) and time-series databases (such as InfluxDB) to meet the storage and query needs of different types of data.

[0131] The decision control module comprises a multi-objective optimization module, an intelligent decision-making module, and an execution module. The multi-objective optimization module implements the multi-objective optimization algorithm from step two, solving for the Pareto optimal process parameter combination based on a digital twin model and the set optimization objectives. The intelligent decision-making module, based on the current production status, the Pareto optimal solution set, and production scheduling requirements, selects the final optimal process parameter setpoints through comprehensive evaluation and decision-making algorithms, and generates control commands.

[0132] The feedback execution module is used to apply control commands to the actuators of the concentration equipment and to compare the real-time measured data with the predicted values ​​of the digital twin model, making real-time feedback adjustments based on the comparison deviation. Specifically, the feedback execution module obtains the prediction deviation by comparing the real-time measured data with the predicted values ​​of the digital twin model, and adjusts the steam flow rate, vacuum degree, and feed flow rate of the concentration equipment actuators in real time based on the prediction deviation, so that the process parameter trajectory converges to the optimal process parameter curve determined by multi-objective optimization, thereby achieving closed-loop control.

[0133] The cross-batch recognition module is used to accumulate operational data across batches. When the accumulated sample size reaches a set threshold or the prediction deviation of the digital twin model is detected to exceed a set range, the digital twin model parameter re-identification is triggered. The parameter sets of the temperature calculation module and the heat transfer calculation module are iteratively updated to enable the digital twin model to adapt to long-term changes such as material batch differences and equipment aging.

[0134] The human-computer interaction module includes a visual monitoring interface and an alarm unit. The visual monitoring interface provides a graphical display of information such as real-time parameter trend curves of the condensed process, simulation results of the digital twin model, equipment operating status, and achievement of optimization goals. It supports operators in querying, setting, and manually intervening in key parameters. The alarm unit is used to issue early warning information to users through interface pop-ups, audible and visual alarms, SMS, or email when the real-time collected data deviates from the normal prediction range of the digital twin model, and displays alarm cause analysis and handling suggestions.

[0135] The human-machine interface module visualizes real-time data from the concentration process, simulation results from the digital twin model, optimized process parameters, and equipment operating status using charts and curves, facilitating real-time monitoring of the production process by operators. Notably, the model can predict the remaining time and energy consumption of the concentration process based on process parameters, enabling operators to adjust strategies accordingly. Simultaneously, the system includes an anomaly detection module. By comparing real-time collected data with the normal prediction range of the digital twin model, it can promptly identify equipment and process anomalies, issuing warnings to management via audible and visual alarms and SMS notifications. It also automatically generates preliminary fault analysis and handling suggestions, supporting rapid troubleshooting and production recovery.

[0136] Furthermore, the intelligent monitoring and control robot system for the concentration process in this embodiment also possesses data storage and analysis capabilities, automatically recording all key data during the concentration process to form a production database. Through statistical analysis and data mining of historical data, production process traceability, continuous improvement of process parameters, and equipment performance evaluation can be achieved, providing data support for enterprise production management optimization and decision support. For example, by analyzing energy consumption data from different batches, the combination of process parameters with the lowest energy consumption can be identified and used as a reference for new optimization targets.

[0137] In summary, the intelligent measurement and control robot system for the concentration process in this embodiment achieves intelligent perception, precise modeling, dynamic optimization, and autonomous control of the concentration process through the collaborative work of its various modules. It constructs an intelligent measurement and control system for the concentration process based on cognitive intelligence, which can significantly improve the automation level and intelligence of the concentration process. It effectively solves the problems of reliance on human experience, lagging process control, high energy consumption, and insufficient product quality stability in traditional concentration production, and provides strong technical support for achieving efficient, low-consumption, and high-quality production in the concentration process.

[0138] The embodiments described above are merely preferred embodiments for fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for intelligent measurement and control of a concentration process, characterized in that: Includes the following steps: Step 1: Constructing a digital twin model of the concentration process: Obtain the physical structure parameters and material property parameters of the concentration equipment, and construct an initial mechanism model based on the heat and mass transfer mechanism; collect historical operating data, and calibrate the parameters of the temperature calculation module in the initial mechanism model so that the predicted values ​​of steam temperature and liquid temperature are close to the measured values ​​respectively. Based on the calibrated temperature parameters, the heat transfer calculation module in the initial mechanism model is identified to make the predicted values ​​of evaporation rate, concentration time and total energy consumption close to the measured values, thus obtaining a digital twin model of the concentration process. Step Two: Multi-objective Optimization and Decision Making Based on the digital twin model, with the optimization objectives of achieving the target concentration, minimizing the concentration time, and minimizing the total energy consumption, and combined with the process parameter constraints, a multi-objective optimization algorithm is used to generate the optimal combination of Pareto process parameters, and the optimal process parameter settings are determined based on the current production status. Step 3: Closed-loop control: During batch operation, the actuators of the concentration equipment are controlled according to the optimal process parameter setpoints; Real-time process data is collected and compared with the predicted values ​​of the digital twin model. Based on the deviation obtained from the comparison, the steam flow rate, vacuum degree and feed flow rate are adjusted in real time to enable the actual process parameters to track the trajectory of the optimal process parameters. Step 4: Online Update When the accumulated sample size reaches a set threshold or the model prediction deviation exceeds the limit, the parameters of the digital twin model are re-identified, and the parameters of the temperature calculation module and the heat transfer calculation module are iteratively updated.

2. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step one, the parameter calibration method for the temperature calculation module is as follows: Establish the mapping relationship between steam flow rate and steam pressure, and the mapping relationship between steam pressure and steam temperature; The boiling point of pure water is calculated based on the vacuum degree, and the boiling point rise is corrected based on the real-time concentration to obtain the predicted value of the drug solution temperature. We construct the weighted sum of squares of the predicted deviations of steam temperature and liquid temperature, respectively, and minimize the sum of the two as the objective function for temperature calibration to obtain the calibration values ​​of the temperature mapping parameter and the boiling point elevation parameter.

3. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step one, the parameter identification method of the heat transfer calculation module is as follows: Along the time axis of the concentration process, the real-time concentration is calculated based on the current mass, and then the current heat transfer temperature difference and net heat transfer power are calculated. The current instantaneous evaporation rate is calculated based on the corrected latent heat of vaporization, and the current batch mass is updated. Repeat the above steps until the current quality reaches the target endpoint quality, and obtain the predicted enrichment time and total energy consumption. We construct the weighted sum of squares of the prediction errors of instantaneous evaporation rate, concentration time, and total energy consumption, respectively, and use the sum of these three as the objective function for heat transfer identification to minimize the solution, thereby obtaining the identification values ​​of heat transfer coefficient, heat loss ratio, and latent heat of vaporization correction coefficient.

4. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step two, a hierarchical search strategy is used for multi-objective optimization. The method is as follows: The first layer uses coarse-grained search to generate candidate parameter points within the process parameter range, calls the digital twin model for evaluation, and eliminates candidate points that do not meet the constraints. The second layer performs a fine-grained search centered on the optimal region obtained from the coarse search, scores candidate points based on the comprehensive evaluation function, and selects the parameter combination with the best score as the recommended operation point.

5. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step two, a multi-objective Bayesian optimization method is used for optimization. The method is as follows: a surrogate model of the digital twin model is constructed based on the initial sample, and new candidate process parameter points are selected using the hypervolume lifting criterion to gradually approach the optimal solution region; after determining the optimal region, a local fine search method is further used to refine and optimize the region.

6. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step two, the evaluation function is a weighted sum of the endpoint quality deviation function, the concentration time function, and the energy consumption function; The weight coefficients of the evaluation function are determined by dynamically fusing and normalizing the preset initial weights of the production mode with the CRITIC weights and entropy weights calculated based on real-time operating data, so that the weight coefficients can adapt to the priority changes of different production scenarios.

7. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: In step two, the process parameter constraints include: steam flow rate constraint, vacuum degree constraint, liquid temperature constraint, heat transfer temperature difference constraint, single batch running time constraint, and final concentration constraint.

8. The intelligent measurement and control method for the concentration process according to claim 1, characterized in that: Step three also includes online incremental correction of the heat transfer coefficient in the digital twin model: the current heat transfer coefficient is back-calculated based on the real-time steam flow rate and the real-time heat transfer temperature difference, and the back-calculated heat transfer coefficient is written back to the digital twin model after amplitude limiting to update the model state.

9. A smart monitoring and control robot system for a concentration process, used to implement the smart monitoring and control method for a concentration process as described in any one of claims 1 to 8, characterized in that: include: The sensing module, configured on the concentration equipment, is used to collect data on material temperature, tank vacuum, material density, feed flow rate, steam flow rate, and valve opening in real time during the concentration process. The data processing module includes a data preprocessing unit, a digital twin model unit, and a database unit; the data preprocessing unit is used to clean, denoise, and time-stamp align the raw data collected by the sensing module. The digital twin model unit has a built-in digital twin model of the concentration process, which is used to update the model state in real time based on the preprocessed data, so as to realize dynamic simulation, state prediction and energy consumption calculation of the concentration process; the database unit is used to store the original sensing data, preprocessed data, model parameters, optimization results and historical production data. The decision control module includes a multi-objective optimization unit and an intelligent decision-making unit; The multi-objective optimization unit is used to solve for the Pareto optimal process parameter combination based on the digital twin model and the set optimization objectives; the intelligent decision unit is used to determine the optimal process parameter set value from the Pareto optimal solution set according to the current production status and generate control commands. The feedback execution module is used to receive the control commands and drive the actuator to adjust the process parameters. And it is used to compare the real-time collected measured data with the predicted value of the digital twin model, and to make real-time feedback adjustments based on the comparison deviation; The cross-batch learning module is used to accumulate running data across batches. When the accumulated sample size reaches a set threshold or the model prediction deviation exceeds a set range, it triggers the parameter re-identification of the digital twin model and iteratively updates the parameters of the temperature calculation module and the heat transfer calculation module.

10. The intelligent monitoring and control robot system for the concentration process according to claim 9, characterized in that: It also includes a human-computer interaction module, which includes: The visual monitoring interface provides a graphical display of real-time parameter trend curves in the concentration process, simulation results of digital twin models, equipment operating status, and the achievement of optimization goals. It also supports operators in querying, setting, and manually intervening in key parameters. An alarm unit is used to issue a warning message when the real-time collected data deviates from the normal prediction range of the digital twin model.

Images