Intelligent control method for evaporation crystallization system based on internet of things

Abstract

This invention discloses an intelligent control method for an evaporation crystallization system based on the Internet of Things (IoT), relating to the field of evaporation crystallization control technology. By constructing a coupled variable model of the dynamic evolution process of crystal particle size distribution based on scenario characteristics, it analyzes the influence of transient changes in process parameters on crystal particle size distribution, quantifies the fluctuation trend of crystal particle size distribution under coupling effects, dynamically adjusts control variables based on the fluctuation trend of crystal particle size distribution, and uses a genetic algorithm to optimize the combination of adjustment strategies. This control system relies on IoT sensing technology to achieve real-time acquisition and dynamic monitoring of multi-dimensional process parameters of the evaporation crystallization system. Combined with scenario characteristics, it constructs a multivariate coupled model of the dynamic evolution process of crystal particle size distribution, fully exploring the dynamic coupling relationship between the adaptive fluctuation of evaporation rate and precipitation rate on the evolution law of crystal particle size distribution, thereby improving the prediction accuracy and descriptive ability of complex nonlinear dynamic processes.

Description

Technical Field

[0001] This invention relates to the field of evaporation and crystallization control technology, and more specifically to an intelligent control method for an evaporation and crystallization system based on the Internet of Things. Background Technology

[0002] Evaporation crystallization technology is widely used in various industries such as chemical, salt production, pharmaceutical, metallurgy, and environmental protection. It is mainly used to achieve supersaturation of solute in solution by evaporating water, thereby promoting crystal precipitation. With the rapid development of Internet of Things (IoT) technology, its application in evaporation crystallization systems provides strong support for realizing equipment interconnection, real-time data acquisition, and intelligent decision control. The control system can collect key process parameters such as evaporator temperature, pressure, feed concentration, liquid level, steam flow, cooling water flow, and crystal slurry concentration in real time. Through real-time data upload and cloud storage, combined with edge computing and intelligent algorithms, predictive control, anomaly warning, and adaptive optimization of the evaporation crystallization process can be achieved.

[0003] The existing technology has the following shortcomings:

[0004] 1. Traditional evaporation crystallization process control generally suffers from problems such as coarse adjustment, slow response, and insufficient understanding of the coupling relationship of process parameters. Specifically, traditional control methods mainly rely on manually setting empirical parameters and single-variable feedback adjustment, lacking a systematic understanding of the dynamic coupling relationship of multiple variables such as evaporation rate, precipitation rate, and crystal particle size distribution. In particular, when the evaporation rate and precipitation rate undergo adaptive fluctuations, existing methods cannot identify the strength and trend of the coupling effect of their coordinated changes on the crystal particle size distribution in real time, resulting in large fluctuations in the crystal particle size distribution range and uneven particle size distribution, which ultimately affects product purity, crystal morphology, and subsequent processing performance.

[0005] 2. Existing control methods mostly operate with static setpoint regulation or open-loop mode, lacking a dynamic adaptive regulation mechanism based on a combination of predictive models and disturbance identification. Especially in complex disturbance scenarios, such as fluctuations in evaporation load, changes in feed concentration, or abnormal crystallization precipitation rate, the system often fails to dynamically sense the coupled changes of multiple variables, resulting in a lag in control response, causing over-regulation, amplified fluctuations, or even system oscillation. At the same time, it is unable to comprehensively optimize particle size distribution uniformity, energy consumption efficiency, and production capacity balance under multi-objective and multi-constraint conditions, resulting in defects such as low overall system performance, poor crystal quality consistency, and poor operating economy.

[0006] Based on this, this application proposes an intelligent control method for an evaporation crystallization system based on the Internet of Things (IoT). By relying on IoT sensing technology, the method realizes real-time acquisition and dynamic monitoring of multi-dimensional process parameters of the evaporation crystallization system. Combined with the characteristics of the scenario, a multivariate coupled model of the dynamic evolution process of crystal particle size distribution is constructed. The method fully explores the dynamic coupling relationship between the adaptive floating of evaporation rate and precipitation rate and the evolution law of crystal particle size distribution, thereby improving the prediction accuracy and description ability of complex nonlinear dynamic processes. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent control method for an evaporation and crystallization system based on the Internet of Things, so as to overcome the shortcomings of the prior art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control method for an evaporation crystallization system based on the Internet of Things, the control method comprising the following steps:

[0009] Based on the characteristics of the scene, the control system constructs a coupled variable model of the dynamic evolution process of crystal grain size distribution;

[0010] The influence of transient changes in process parameters on crystal grain size distribution was analyzed, and the fluctuation trend of crystal grain size distribution under the coupling effect was quantified.

[0011] The control variables are dynamically adjusted based on the fluctuation trend of the crystal grain size distribution, and a genetic algorithm is used to find the optimal combination of adjustment strategies.

[0012] In a preferred embodiment, the process parameters of the evaporation crystallization system are monitored in real time using IoT sensors, including evaporation rate, precipitation rate, crystal slurry concentration, temperature, vacuum degree, and feed flow rate.

[0013] In a preferred embodiment, the genetic algorithm is used to optimize the combination of regulation strategies, including the following steps:

[0014] Evaporation rate, precipitation rate, and cooling rate were converted into a string of chromosome gene fragments, and several individuals were generated under constraints using a random generation tool.

[0015] The fitness value of each individual is calculated using a fitness function;

[0016] A subset of individuals is selected using the roulette wheel selection method, and then crossover / mutation operations are performed to create a set.

[0017] Repeat the fitness calculation, roulette wheel selection, and crossover / mutation operations for all individuals in the set;

[0018] When the convergence condition is met, output all sets and select the individual with the largest fitness value among all sets as the optimal adjustment strategy combination.

[0019] In a preferred embodiment, dynamically adjusting the control variables based on the fluctuation trend of the crystal grain size distribution includes the following steps:

[0020] A change in the mean particle size distribution greater than 0 indicates an increase in the mean particle size distribution, while a change in the mean particle size distribution less than 0 indicates a decrease in the mean particle size distribution.

[0021] The obtained mean change in particle size distribution The changes are compared with preset first change thresholds Z1 and Z2, wherein the first change threshold Z1 < 0, the second change threshold Z2 > 0, and ;

[0022] like If the particle size is too large, reduce the precipitation rate and increase the evaporation rate.

[0023] like If the particle size is too small, increase the precipitation rate and decrease the evaporation rate.

[0024] In a preferred embodiment, the fluctuation trend of crystal grain size distribution under quantitative coupling includes the following steps:

[0025] By using IoT sensors and industrial buses to collect process parameters in real time and record them as time series data, the transient change rates of evaporation rate and precipitation rate are calculated.

[0026] Based on the perturbation identification algorithm, the dynamic response relationship between rate change and crystal size distribution change is established, and the fluctuation trend of crystal size distribution under coupling effect is obtained.

[0027] In a preferred embodiment, analyzing the impact of transient changes in process parameters on crystal grain size distribution includes the following steps:

[0028] The perturbation identification algorithm is given in matrix multiplication form: In the formula, This is a vector of particle size distribution variation values. It is a matrix composed of evaporation rate, precipitation rate, and the product of evaporation rate and precipitation rate. It is a coefficient vector;

[0029] The parameters of the disturbance identification algorithm are estimated using the least squares method, and the solution is obtained. , , The expression is: In the formula, Describes the transpose of matrix X. Let covariance be the coefficient matrix. is the inverse of the covariance matrix.

[0030] In a preferred embodiment, a dynamic response relationship between rate change and crystal size distribution change is established based on a perturbation identification algorithm. The algorithm expression is as follows: In the formula, This represents the change in the mean of particle size distribution. The coefficient representing the intensity of the effect of evaporation rate perturbation on particle size distribution. The intensity coefficient representing the effect of precipitation rate perturbation on particle size distribution. This is the coefficient representing the effect of the coupling effect between evaporation rate and precipitation rate on particle size distribution changes. The transient rate of change of evaporation rate. The transient rate of change of precipitation rate;

[0031] The transient rate of change of evaporation rate and precipitation rate is expressed as follows:

[0032] In the formula, The transient rate of change of evaporation rate. The transient rate of change of precipitation rate, The sampling time interval, The evaporation rate is obtained at time t. The precipitation rate is obtained at time t.

[0033] In a preferred embodiment, the vector expression for the particle size distribution variation value is: The matrix consisting of the evaporation rate, the precipitation rate, and the product of the evaporation rate and the precipitation rate is expressed as: In the formula, This represents the number of sampling times.

[0034] In a preferred embodiment, constructing a coupled variable model of the dynamic evolution process of crystal grain size distribution includes the following steps:

[0035] The process parameters are divided into training, validation and test sets. The coupled variable model is trained using historical data to fit the influence of fluctuations in evaporation rate and precipitation rate on crystal size distribution. The loss function is used to measure the difference between the predicted and actual particle size distribution. The Adam optimization algorithm is used to optimize the coupled variable model parameters.

[0036] The coupled variable model is trained using a training set, and the weights and biases are gradually optimized to capture the dynamic relationship between evaporation rate and precipitation rate and the change in crystal grain size distribution.

[0037] In a preferred embodiment, the coupled variable model includes an input layer, an LSTM layer, and a fully connected layer:

[0038] Input layer: Used to receive time-series data of historical evaporation rates, precipitation rates, and other relevant process parameters;

[0039] LSTM layer: Captures the temporal dependence of evaporation rate and precipitation rate through multiple LSTM units;

[0040] Fully connected layer: Used to map the features output by the LSTM layer to the predicted results of the crystal grain size distribution.

[0041] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0042] 1. This invention constructs a coupled variable model of the dynamic evolution process of crystal particle size distribution based on scenario characteristics, analyzes the influence of transient changes in process parameters on crystal particle size distribution, quantifies the fluctuation trend of crystal particle size distribution under coupling effect, dynamically adjusts control variables based on the fluctuation trend of crystal particle size distribution, and uses a genetic algorithm to optimize the combination of adjustment strategies. This control system relies on IoT sensing technology to realize real-time acquisition and dynamic monitoring of multi-dimensional process parameters of the evaporation crystallization system. Combining scenario characteristics, a multivariate coupled model of the dynamic evolution process of crystal particle size distribution is constructed. This model adopts the LSTM recurrent neural network method to fully explore the dynamic coupling relationship between the adaptive fluctuation of evaporation rate and precipitation rate on the evolution law of crystal particle size distribution, thereby improving the prediction accuracy and description ability of complex nonlinear dynamic processes.

[0043] 2. This invention analyzes the intensity of the disturbance effect of transient changes in process parameters on crystal particle size distribution in real time, and quantifies the fluctuation trend of particle size distribution under coupling effect. The system can quickly identify and dynamically predict abnormal trends in particle size distribution under complex floating scenarios. Combining disturbance identification and model prediction results, it dynamically adjusts key control variables such as evaporation rate, vacuum degree, and feed flow rate, and introduces a multi-objective genetic algorithm (MOGA) to optimize the combination of adjustment strategies to ensure that the global optimal solution is obtained between particle size distribution uniformity, system energy consumption and production efficiency. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0045] Figure 1 This is a flowchart of the method of the present invention.

[0046] Figure 2 This is a system architecture diagram of the present invention.

[0047] Figure 3This is a mind map of the method of the present invention.

[0048] Figure 4 This is a timing diagram of the method of the present invention.

[0049] Figure 5 This is a mind map of the method of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1: Please refer to Figure 1 and Figures 3-5 As shown in this embodiment, the intelligent control method for an evaporation crystallization system based on the Internet of Things includes the following steps:

[0052] The control system utilizes IoT sensors to monitor key process parameters in the evaporation crystallization system in real time, including feed temperature, feed concentration, steam pressure, crystallizer vacuum, circulation flow rate, and crystal slurry density. High-precision online particle size analyzers (such as laser particle size analyzers) are used to acquire crystal particle size distribution data, enabling full-cycle, multi-dimensional data acquisition of the crystallization system's dynamic process.

[0053] Based on the characteristics of the scenario, evaporation rate and precipitation rate are defined as adaptive floating variables, and a coupled variable model is constructed to investigate their influence on the dynamic evolution of crystal grain size distribution. This coupled variable model, based on the LSTM recurrent neural network method, fits the influence of evaporation rate-precipitation rate fluctuations on the dynamic process of crystal grain size distribution.

[0054] Real-time monitoring of process parameters, analysis of transient changes in evaporation rate and precipitation rate, application of perturbation identification algorithm to identify the intensity of influence of individual or synchronous fluctuations of evaporation rate and precipitation rate on crystal size distribution under different operating conditions, and quantification of the fluctuation trend of crystal size distribution under coupling effect.

[0055] The control variables are dynamically adjusted based on the fluctuation trend of crystal particle size distribution to achieve coordinated regulation of evaporation and precipitation rates. This ensures that the crystallization process remains within the optimal particle size distribution range under fluctuating conditions. A genetic algorithm is used to optimize the combination of adjustment strategies, bringing the crystal particle size distribution closer to the target range. Simultaneously, energy consumption, production capacity, and equipment load are balanced, achieving real-time solution for the optimal dynamic performance of the system. All process data is uploaded to an IoT cloud platform for real-time visualization of system operating status, particle size distribution curves, and the effectiveness of the adjustment strategies.

[0056] This application constructs a coupled variable model of the dynamic evolution process of crystal particle size distribution based on scenario characteristics, analyzes the influence of transient changes in process parameters on crystal particle size distribution, quantifies the fluctuation trend of crystal particle size distribution under coupling effect, dynamically adjusts control variables based on the fluctuation trend of crystal particle size distribution, and uses a genetic algorithm to optimize the combination of adjustment strategies. This control system relies on IoT sensing technology to realize real-time acquisition and dynamic monitoring of multi-dimensional process parameters of the evaporation crystallization system. Combining scenario characteristics, a multivariate coupled model of the dynamic evolution process of crystal particle size distribution is constructed. This model adopts the LSTM recurrent neural network method to fully explore the dynamic coupling relationship between the adaptive fluctuation of evaporation rate and precipitation rate on the evolution law of crystal particle size distribution, thereby improving the prediction accuracy and description ability of complex nonlinear dynamic processes.

[0057] This application analyzes the intensity of the disturbance effect of transient changes in process parameters on crystal particle size distribution in real time, and quantifies the fluctuation trend of particle size distribution under the coupling effect. The system can quickly identify and dynamically predict abnormal trends in particle size distribution under complex floating scenarios. Combining the disturbance identification and model prediction results, it dynamically adjusts key control variables such as evaporation rate, vacuum degree, and feed flow rate, and introduces a multi-objective genetic algorithm (MOGA) to optimize the combination of adjustment strategies to ensure that the global optimal solution is obtained between particle size distribution uniformity, system energy consumption and production efficiency.

[0058] like Figure 2 As shown: The control system includes a processing layer, a data layer, a regulation layer, and an IoT cloud platform;

[0059] Processing layer: Based on the characteristics of the scene, a coupled variable model of the dynamic evolution process of crystal grain size distribution is constructed, and the coupled variable model is sent to the data layer and the adjustment layer;

[0060] Data layer: Analyzes the impact of transient changes in process parameters on crystal grain size distribution, quantifies the fluctuation trend of crystal grain size distribution under coupling effect, and sends the fluctuation trend of crystal grain size distribution to the adjustment layer;

[0061] The regulation layer dynamically adjusts control variables based on the fluctuation trend of crystal particle size distribution and uses a genetic algorithm to optimize the combination of regulation strategies to make the crystal particle size distribution tend to the target range. At the same time, it balances energy consumption, production capacity and equipment load, and completes the real-time solution of the optimal solution for the dynamic performance of the system. All process data are uploaded to the Internet of Things cloud platform to visualize the system operation status, particle size distribution curve and the execution effect of regulation strategies in real time.

[0062] Example 2: Based on the characteristics of the scenario, the evaporation rate and precipitation rate are defined as adaptive floating variables, and a coupled variable model is constructed to investigate their influence on the dynamic evolution of crystal grain size distribution. This coupled variable model is based on the LSTM recurrent neural network method to fit the influence of the evaporation rate-precipitation rate fluctuations on the dynamic process of crystal grain size distribution.

[0063] The core process parameters of the evaporation crystallization system are monitored in real time using IoT sensors, including but not limited to evaporation rate, precipitation rate, crystal slurry concentration, temperature, vacuum level, and feed flow rate. The data acquisition frequency needs to be high enough to ensure that real-time dynamic changes are reflected. The collected raw data is often affected by noise and outliers, therefore data cleaning and preprocessing are necessary. This includes: outlier removal: eliminating erroneous data caused by sensor malfunctions or external factors; normalization: normalizing or standardizing variables of different dimensions to eliminate the influence of dimensions; and time window sliding: setting an appropriate time window based on the characteristics of the time series for data smoothing and feature extraction.

[0064] The process parameters were divided into training, validation, and test sets. A coupled-variable model was trained using historical data to fit the effects of fluctuations in evaporation and precipitation rates on crystal size distribution. A loss function (e.g., mean squared error, MSE) was used to measure the difference between the predicted and actual particle size distributions. The Adam optimization algorithm was used to optimize the coupled-variable model parameters, ensuring that the model could fit the effects of fluctuations in evaporation and precipitation rates as accurately as possible.

[0065] An LSTM (Long Short-Term Memory) recurrent neural network is employed, which, due to its advantages in processing time series data, can effectively learn the long-term dependency between evaporation rate and precipitation rate. The coupled variable model includes the following main components:

[0066] Input layer: Receives time-series data of historical evaporation rate, precipitation rate and other relevant process parameters.

[0067] LSTM layer: The temporal dependence of evaporation rate and precipitation rate is captured through multiple LSTM units.

[0068] Fully connected layer: Used to map the features output by the LSTM layer to the predicted results of the crystal grain size distribution.

[0069] First, it is necessary to collect relevant process parameters from the evaporation crystallization system in real time, including evaporation rate, precipitation rate, temperature, pressure, and feed concentration, and record the crystal particle size distribution at the corresponding time.

[0070] The collected data should be organized into a time series format, as follows:

[0071] Input variables (x): [evaporation rate, precipitation rate, temperature, pressure];

[0072] Output variable (y): [Crystal grain size distribution];

[0073] In the code, we use randomly generated data to simulate this part:

[0074] #Simulated input data (number of samples, time steps, number of features)

[0075] num_samples=1000

[0076] time_steps=10

[0077] num_features=4# Evaporation rate, precipitation rate, temperature, pressure

[0078] x=np.random.rand(num_samples,time_steps,num_features)

[0079] #Simulation output data (crystal grain size distribution)

[0080] y=np.random.rand(num_samples,1)

[0081] Here, x is a three-dimensional array representing 1000 samples, each sample has 10 time steps, and each time step has 4 features. y is the corresponding crystal grain size distribution value.

[0082] Because different process parameters have significantly different dimensions and numerical ranges, directly inputting them into the neural network will affect the training effect. Therefore, it is necessary to standardize all input and output data. The code example is as follows:

[0083] from_sklearn.preprocessing_import_MinMaxScaler

[0084] #Data Normalization

[0085] scaler_X=MinMaxScaler()

[0086] X_reshaped=X.reshape(-1,num_features)

[0087] X_scaled=scaler_X.fit_transform(X_reshaped).reshape(num_samples,time_steps,num_features)

[0088] scaler_y=MinMaxScaler()

[0089] y_scaled=scaler_y.fit_transform(y)

[0090] MinMaxScaler linearly maps the data to the [0,1] interval, ensuring that features are on the same scale and avoiding certain variables from dominating the model. Special attention should be paid to X, which needs to be reshaped into two dimensions first, then fitted_transformed, and finally restored to a three-dimensional input structure.

[0091] To ensure the model's generalization ability, the data needs to be divided into a training set and a test set:

[0092] from_sklearn.model_selection_import_train_test_split

[0093] # Divide the dataset into training and test sets

[0094] X_train,X_test,y_train,y_test=train_test_split(X_scaled,y_scaled,test_size=0.2,random_state=42)

[0095] Here, the dataset is split into 80% training and 20% testing, and random_state is fixed with a random seed to ensure that the experiment is reproducible.

[0096] Coupled variable model structure code:

[0097] from_tensorflow.keras.models_import_Sequential

[0098] from_tensorflow.keras.layers_import_LSTM,Dense

[0099] # Constructing a Coupled LSTM Model

[0100] model=Sequential()

[0101] model.add(LSTMunits=64,input_shape=time_steps,num_features,return_sequences=False))

[0102] model.add(Dense32,activation='relu')

[0103] model.add(Dense1) # Outputs the predicted particle size distribution value

[0104] LSTM (units=64): 64 units, extracting the temporal dependencies of multidimensional inputs.

[0105] return_sequences=False: This means that only the output of the last time step is returned, which is suitable for regression prediction.

[0106] Dense(32,activation='relu'): Non-linear mapping, enhancing feature representation capabilities.

[0107] Dense(1): Outputs the predicted crystal grain size distribution.

[0108] The coupled variable model is trained using the training set, and the weights and biases are gradually optimized to enable the model to effectively capture the dynamic relationship between evaporation rate and precipitation rate and the change in crystal size distribution. The model performance is evaluated using the validation set and the test set to observe the error between the predicted crystal size distribution and the actual situation. If the error is too large, the model hyperparameters (such as the number of LSTM layers) are adjusted for further optimization to improve the model's prediction accuracy.

[0109] Once the coupled variable model has been trained and validated, it can be embedded into the real-time control framework of the evaporation crystallization system. Real-time process parameters (such as evaporation rate and precipitation rate) are acquired through IoT sensors, and the coupled variable model is used to predict the changing trend of the current crystal particle size distribution.

[0110] Choose a suitable loss function and optimizer, and begin training:

[0111] #Model compilation

[0112] model.compile(optimizer='adam',loss='mse',metrics=['mae'])

[0113] #Training the model

[0114] history=model.fit(X_train,y_train,epochs=50,batch_size=32,validation_split=0.2)

[0115] In the code above, the Adam optimizer is suitable for modeling complex nonlinear systems, exhibiting good robustness and fast convergence. MSE (mean squared error) measures the sum of squared errors between predicted and true values, suitable for regression tasks. `validation_split=0.2` further divides the training set by 20% for validation, monitoring for overfitting.

[0116] Using the trained model, predictions are made on the test data, and the standardized results are inversely normalized to calculate the predicted actual particle size distribution. A code example is shown below:

[0117] #Model Prediction

[0118] y_pred = model.predict(X_test)

[0119] #Inverse normalization prediction results

[0120] y_pred_real=scaler_y.inverse_transform(y_pred)

[0121] y_test_real=scaler_y.inverse_transform(y_test)

[0122] #Print the first 5 predicted results

[0123] print("Predicted value:", y_pred_real[:5].flatten)

[0124] print("True value:", y_test_real[:5].flatten)

[0125] In the code above, the `predict` method obtains the standardized prediction result. The `inverse_transform` method restores the standardized result to the true particle size distribution value. By comparing the predicted values ​​with the true values, the model performance and prediction accuracy can be evaluated.

[0126] The evaporation crystallization process is influenced by a variety of factors, including material properties and operating conditions. Among these, the evaporation rate and precipitation rate are two key factors affecting crystal size distribution. The evaporation rate refers to the speed at which the solvent evaporates from the solution, while the precipitation rate refers to the speed at which the solute precipitates from the solution. These rates are not only closely related to operating parameters (such as temperature and pressure) but are also affected by multiple factors, including feed concentration and seed crystal morphology.

[0127] Different crystallization targets require different adaptive adjustments to the evaporation rate and precipitation rate:

[0128] When controlling the particle size distribution of crystals is required, the dynamic adjustment of evaporation and precipitation rates must be based on the crystal growth. For example, higher evaporation and precipitation rates may lead to rapid crystal growth, resulting in a wider or larger particle size distribution; while lower rates may result in smaller crystals and a narrower particle size distribution. In such cases, it is necessary to adjust the evaporation and precipitation rates based on real-time monitoring of the particle size distribution to optimize the product particle size.

[0129] By combining process requirements and objectives (such as particle size distribution width and purity requirements), the fluctuation range of evaporation rate and precipitation rate at different stages can be set. For example, if the objective is to improve crystal uniformity, the crystal growth rate can be controlled by reducing the evaporation rate in the early stage of crystallization; while at the end of crystallization, the precipitation rate can be appropriately increased to promote faster crystal precipitation and reach the final size. In the actual evaporation crystallization process, there are many other application scenarios, which will not be listed here without further explanation.

[0130] Real-time monitoring of process parameters, analysis of transient changes in evaporation rate and precipitation rate, application of perturbation identification algorithm to identify the intensity of influence of individual or synchronous fluctuations of evaporation rate and precipitation rate on crystal size distribution under different operating conditions, and quantification of the fluctuation trend of crystal size distribution under coupling effect.

[0131] In the evaporation and crystallization process, the evaporation rate and precipitation rate are dynamically changing process variables. Their individual or coupled changes can significantly affect the crystal particle size distribution. To achieve precise control, it is necessary to monitor key process parameters in real time and use a perturbation identification algorithm to identify the intensity of the dynamic impact of variable changes on particle size distribution under different operating conditions. This allows for the quantification of the particle size distribution fluctuation trend under coupled effects. The specific steps are as follows:

[0132] By utilizing IoT sensors and industrial buses, key process parameters are collected in real time and recorded as time-series data. The transient rate of change of evaporation rate and precipitation rate is calculated, defined as the ratio of the rate change before and after a certain moment to the time change, expressed as: In the formula, The transient rate of change of evaporation rate. The transient rate of change of precipitation rate, The sampling time interval, The evaporation rate is obtained at time t. The precipitation rate is obtained at time t. By capturing the rapid fluctuation characteristics of the evaporation rate and precipitation rate, it is used as the input feature for subsequent perturbation identification.

[0133] Based on the perturbation identification algorithm, a dynamic response relationship between rate change and crystal size distribution change is established. The algorithm expression is as follows: In the formula, This represents the change in the mean of the grain size distribution, i.e., the fluctuation trend of the crystal grain size distribution. The coefficient representing the intensity of the effect of evaporation rate perturbation on particle size distribution. The intensity coefficient representing the effect of precipitation rate perturbation on particle size distribution. This is the coefficient representing the effect of the coupling effect between evaporation rate and precipitation rate on particle size distribution changes. The transient rate of change of evaporation rate. The transient rate of change of precipitation rate.

[0134] The perturbation identification algorithm is given in matrix multiplication form: In the formula, This is a vector of particle size distribution variation values. It is a matrix composed of evaporation rate, precipitation rate, and their product. This is the coefficient vector we are looking for.

[0135] The vector expression for the change in particle size distribution is: The matrix consisting of the evaporation rate, the precipitation rate, and their product is expressed as: In the formula, This represents the number of sampling times.

[0136] The parameters of the disturbance identification algorithm are estimated using the least squares method, and the solution is obtained. , , The expression is: In the formula, Describes the transpose of matrix X. Let covariance be the coefficient matrix. is the inverse of the covariance matrix.

[0137] Assume there are N=3 sampling times, and the data collected is shown in Table 1:

[0138] Table 1

[0139]

[0140] Construct the X matrix according to: Substituting the values, we get: The Y vector can be constructed as follows: .

[0141] calculate , The calculation process is as follows:

[0142] First line:

[0143] (1×1)+(0.5×0.5)+(2×2)=1+0.25+4=5.25;

[0144] (1×2)+(0.5×1.5)+(2×1)=2+0.75+2=4.75;

[0145] (1×2)+(0.5×0.75)+(2×2)=2+0.375+4=6.375;

[0146] Second line:

[0147] (2×1)+(1.5×0.5)+(1×2)=2+0.75+2=4.75;

[0148] (2×2)+(1.5×1.5)+(1×1)=4+2.25+1=7.25;

[0149] (2×2)+(1.5×0.75)+(1×2)=4+1.125+2=7.125;

[0150] Third line:

[0151] (2×1)+(0.75×0.5)+(2×2)=2+0.375+4=6.375;

[0152] (2×2)+(0.75×1.5)+(2×1)=4+1.125+2=7.125;

[0153] (2×2)+(0.75×0.75)+(2×2)=4+0.5625+4=8.5625.

[0154] calculate , The calculation process is as follows:

[0155] First row: (1×5)+(0.5×3)+(2×6)=5+1.5+12=18;

[0156] Second line: (2×5)+(1.5×3)+(1×6)=10+4.5+6=20.5;

[0157] Third line: (2×5)+(0.75×3)+(2×6)=10+2.25+12=24.25.

[0158] Multiply the vector by the inverse matrix to find The result can be obtained using a calculator or software, which will not be elaborated upon in this application. Then multiply by... have:

[0159] First line: 4.646×18.5+(-2.493)×20.5+(-1.825)×24.25≈85.961-51.107-44.288=-9.434

[0160] Second line: -2.493×18.5+1.983×20.5+0.640×24.25≈-46.120+40.651+15.520=10.051

[0161] The third line: -1.825×18.5+0.640×20.5+1.171×24.25-33.763+13.120+28.379=7.736, which means... .

[0162] Therefore, after calculation, This indicates the individual effect of evaporation rate on particle size distribution. A negative value indicates that increasing the evaporation rate will decrease the particle size distribution. This indicates the individual effect of precipitation rate on particle size distribution. A positive value suggests that an increased precipitation rate is beneficial for increasing particle size. This indicates the coupling effect of evaporation rate and precipitation rate. A positive value indicates that when both increase simultaneously, there is a tendency to enhance particle size distribution.

[0163] The control variables are dynamically adjusted based on the fluctuation trend of crystal particle size distribution to achieve coordinated regulation of evaporation and precipitation rates. This ensures that the crystallization process remains within the optimal particle size distribution range under fluctuating conditions. A genetic algorithm is used to optimize the combination of adjustment strategies, bringing the crystal particle size distribution closer to the target range. Simultaneously, energy consumption, production capacity, and equipment load are balanced, achieving real-time solution for the optimal dynamic performance of the system. All process data is uploaded to an IoT cloud platform for real-time visualization of system operating status, particle size distribution curves, and the effectiveness of the adjustment strategies.

[0164] The change in the mean of particle size distribution represents the fluctuation trend of the crystal particle size distribution. A change in the mean of particle size distribution greater than 0 indicates that the mean of particle size distribution is increasing, meaning that the crystal particle size is getting larger and the particles are becoming coarser. A change in the mean of particle size distribution less than 0 indicates that the mean of particle size distribution is decreasing, meaning that the crystal particle size is getting smaller and the particles are becoming finer. A change in the mean of particle size distribution equal to 0 indicates no fluctuation or a small fluctuation effect.

[0165] The obtained mean change in particle size distribution The changes are compared with preset first change thresholds Z1 and Z2, wherein the first change threshold Z1 < 0, the second change threshold Z2 > 0, and .

[0166] like If the particle size is determined to be too large, it is necessary to reduce the precipitation rate (e.g., reduce it by 5% from the current precipitation rate) and increase the evaporation rate (e.g., increase it by 5% from the current evaporation rate).

[0167] like If the particle size is too small, it is necessary to increase the precipitation rate (e.g., increase it by 5% from the current precipitation rate) and decrease the evaporation rate (e.g., decrease it by 5% from the current evaporation rate).

[0168] Furthermore, a first warning threshold J1 and a second warning threshold J2 can be preset, wherein the first warning threshold J1 is less than the first change threshold Z1, and the second warning threshold J2 is greater than the second change threshold Z2;

[0169] like or The control system stops the evaporation and crystallization system and sends a warning signal to the relevant management personnel.

[0170] Evaporation rate, precipitation rate, and cooling rate are converted into a string of chromosomal gene fragments. For example, for an individual: [75,60,12], it represents an evaporation rate of 75%, a precipitation rate of 60%, and a cooling rate of 12℃ / h. Several individuals are generated under constraints using a random generation tool. The fitness value of each individual is calculated using a fitness function. A subset of individuals is selected using a roulette wheel selection method, and crossover / mutation operations are performed to establish a set. The fitness value calculation, roulette wheel selection, and crossover / mutation operations are repeated for all individuals in the set. When the convergence condition is met, all sets are output, and the individual with the largest fitness value in all sets is selected as the optimal combination of adjustment strategies.

[0171] In this application, the constraints include: evaporation rate, ranging from 50% to 100%; precipitation rate, ranging from 40% to 90%; and cooling rate, ranging from 5 to 20℃ / h. Examples of individual codes generated under these constraints using a random generation tool are shown below:

[0172] import_random

[0173] # Define the range of constraints

[0174] Re_min,Re_max=50,100#Evaporation rate%

[0175] Rc_min,Rc_max=40,90#precipitation rate%

[0176] Cr_min,Cr_max=5,20# Cooling rate ℃ / h

[0177] #Number of individuals generated

[0178] population_size=10

[0179] #Randomly generate individual groups

[0180] population=[]

[0181] for_inrange(population_size):

[0182] Re = random.uniform(Re_min, Re_max) # Evaporation rate, floating-point number

[0183] Rc = random.uniform(Rc_min, Rc_max) # Extraction rate, floating-point number

[0184] Cr = random.uniform(Cr_min, Cr_max) # Cooling rate, floating point

[0185] individual=[round(Re,2),round(Rc,2),round(Cr,2)]

[0186] population.append(individual)

[0187] # Output results

[0188] for_idx,individual_in_enumerate(population):

[0189] print(f"Individual {idx+1}:{individual}")

[0190] Example output (random value):

[0191] Individual 1: [72.43, 64.19, 15.67];

[0192] Individual 2: [85.56, 43.28, 9.12];

[0193] Individual 3: [60.72, 70.13, 14.89];

[0194] Individual 4: [98.27, 84.55, 7.34]; ...

[0196] In the code example: each individual is a set of ternary arrays, representing a set of adjustment parameter combinations. random.uniform(a,b) ensures that the generated values ​​are continuous floating-point values ​​within the range of a to b, which are closer to the actual control values. round(x,2) retains two decimal places for easy display.

[0197] The calculation logic of the fitness function is as follows: obtain the historical production capacity, production energy consumption, and particle size distribution deviation of each individual, normalize the production capacity, production energy consumption, and particle size distribution deviation so that the value range of production capacity, production energy consumption, and particle size distribution deviation is mapped to [0,1]. Subtract the production energy consumption and particle size distribution deviation from the normalized production capacity to obtain the fitness value of the individual. The larger the fitness value, the better the overall performance of the individual.

[0198] The selection method for roulette is as follows:

[0199] Summing up the fitness values ​​of all individuals yields the denominator. Dividing the fitness value by the denominator gives the selection probability of each individual. Mapping the selection probability onto a sector of a virtual roulette wheel results in a larger sector. The virtual roulette wheel is then started. When the virtual roulette wheel stops, the pointer is moved to select the individual corresponding to the sector. The roulette wheel selection ends when the number of selected individuals equals a preset threshold.

[0200] The genes of individuals selected by roulette are exchanged to form new offspring, and the first two crossover points are set to 1, which is the exchange cooling rate (Cr).

[0201] Parent generation: P1: [72.43, 64.19, 15.67]; P2: [76.18, 58.22, 11.05];

[0202] Offspring: C1:[72.43,64.19,11.05]; C2:[76.18,58.22,15.67].

[0203] A gene is randomly perturbed to increase diversity, and the phenotyping rate Rc=64.19 of the offspring C1 is mutated within a range of ±5%. Assuming a random number of 0.03 is generated, the mutated offspring will be [72.43, 66.12, 11.05], calculated as: 64.19 × (1 + 0.03) = 66.12.

[0204] Final set (new generation): [72.43,66.12,11.05] (C1-after mutation); [76.18,58.22,15.67] (C2); [85.56,43.28,9.12] (preserving superior individuals).

[0205] Specifically, the convergence condition is determined when the number of iterations equals the threshold, or when any iteration obtains an individual in the set whose fitness value is greater than or equal to the fitness threshold.

[0206] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0207] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0208] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0209] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent control method for an evaporation and crystallization system based on the Internet of Things, characterized in that: The control method includes the following steps: Based on the characteristics of the scene, the control system constructs a coupled variable model of the dynamic evolution process of crystal grain size distribution; The influence of transient changes in process parameters on crystal grain size distribution was analyzed, and the fluctuation trend of crystal grain size distribution under the coupling effect was quantified. The control variables are dynamically adjusted based on the fluctuation trend of the crystal grain size distribution, and a genetic algorithm is used to find the optimal combination of adjustment strategies. The fluctuation trend of crystal grain size distribution under quantitative coupling includes the following steps: By using IoT sensors and industrial buses to collect process parameters in real time and record them as time series data, the transient change rates of evaporation rate and precipitation rate are calculated. Based on the perturbation identification algorithm, the dynamic response relationship between rate change and crystal size distribution change is established, and the fluctuation trend of crystal size distribution under coupling effect is obtained. Based on the perturbation identification algorithm, a dynamic response relationship between rate change and crystal size distribution change is established. The algorithm expression is as follows: In the formula, This represents the change in the mean of particle size distribution. The coefficient representing the intensity of the effect of evaporation rate perturbation on particle size distribution. The intensity coefficient representing the effect of precipitation rate perturbation on particle size distribution. This is the coefficient representing the influence of the coupling effect of evaporation rate and precipitation rate on particle size distribution changes. The transient rate of change of evaporation rate. The transient rate of change of precipitation rate; The transient rate of change of evaporation rate and precipitation rate is expressed as follows: In the formula, The transient rate of change of evaporation rate. The transient rate of change of precipitation rate, The sampling time interval, The evaporation rate is obtained at time t. The precipitation rate is obtained at time t.

2. The intelligent control method for an evaporation crystallization system based on the Internet of Things according to claim 1, characterized in that: The process parameters of the evaporation crystallization system are monitored in real time using IoT sensors, including evaporation rate, precipitation rate, crystal slurry concentration, temperature, vacuum degree, and feed flow rate.

3. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 2, characterized in that: Using genetic algorithms to find optimal combinations of regulatory strategies includes the following steps: Evaporation rate, precipitation rate, and cooling rate were converted into a string of chromosome gene fragments, and several individuals were generated under constraints using a random generation tool. The fitness value of each individual is calculated using a fitness function; A subset of individuals is selected using the roulette wheel selection method, and then crossover / mutation operations are performed to create a set. Repeat the fitness calculation, roulette wheel selection, and crossover / mutation operations for all individuals in the set; When the convergence condition is met, output all sets and select the individual with the largest fitness value among all sets as the optimal adjustment strategy combination.

4. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 3, characterized in that: The control variables are dynamically adjusted based on the fluctuation trend of crystal grain size distribution, including the following steps: A change in the mean particle size distribution greater than 0 indicates an increase in the mean particle size distribution, while a change in the mean particle size distribution less than 0 indicates a decrease in the mean particle size distribution. The obtained mean change in particle size distribution The changes are compared with preset first change thresholds Z1 and Z2, wherein the first change threshold Z1 < 0, the second change threshold Z2 > 0, and ; like If the particle size is too large, reduce the precipitation rate and increase the evaporation rate. like If the particle size is too small, increase the precipitation rate and decrease the evaporation rate.

5. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 4, characterized in that: The analysis of the impact of transient changes in process parameters on crystal grain size distribution includes the following steps: The perturbation identification algorithm is given in matrix multiplication form: In the formula, This is a vector of particle size distribution variation values. It is a matrix composed of evaporation rate, precipitation rate, and the product of evaporation rate and precipitation rate. It is a coefficient vector; The parameters of the disturbance identification algorithm are estimated using the least squares method, and the solution is obtained. , , The expression is: In the formula, Describes the transpose of matrix X. Let covariance be the coefficient matrix. is the inverse of the covariance matrix.

6. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 5, characterized in that: The vector expression for the change in particle size distribution is: The matrix consisting of the evaporation rate, the precipitation rate, and the product of the evaporation rate and the precipitation rate is expressed as: In the formula, This represents the number of sampling times.

7. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 6, characterized in that: Constructing a coupled-variable model of the dynamic evolution of crystal grain size distribution includes the following steps: The process parameters are divided into training, validation and test sets. The coupled variable model is trained using historical data to fit the influence of fluctuations in evaporation rate and precipitation rate on crystal size distribution. The loss function is used to measure the difference between the predicted and actual particle size distribution. The Adam optimization algorithm is used to optimize the coupled variable model parameters. The coupled variable model is trained using a training set, and the weights and biases are gradually optimized to capture the dynamic relationship between evaporation rate and precipitation rate and the change in crystal grain size distribution.

8. The intelligent control method for an evaporation and crystallization system based on the Internet of Things according to claim 7, characterized in that: The coupled variable model includes an input layer, an LSTM layer, and a fully connected layer: Input layer: Used to receive time-series data of historical evaporation rates, precipitation rates, and other relevant process parameters; LSTM layer: Captures the temporal dependence of evaporation rate and precipitation rate through multiple LSTM units; Fully connected layer: Used to map the features output by the LSTM layer to the predicted results of the crystal grain size distribution.

Images