Multivariable control system based on organic matter aqueous solution spray drying

Abstract

The invention relates to the technical field of spray drying control, and discloses a spray drying multivariable control system based on an organic matter aqueous solution. The system comprises a data acquisition module, a drying parameter determination module, a drying state sensing module, a drying abnormity intervention module and a multivariable control module. The data acquisition module acquires a to-be-dried solution, acquires equipment operation parameters and identifies an environment state and solution characteristics; a drying parameter determination module sets target parameters, positions the relative position of the equipment and determines an initial drying formation; the dry state sensing module senses a real-time state in combination with the initial formation and a preset rule, analyzes an environment change trend and sets an anomaly sensing mode; the drying abnormity intervention module identifies abnormity based on the mode, and sets an intervention mechanism in combination with a preset track prediction network; and the multivariable control module synthesizes the information to carry out multivariable control to obtain a result. The system can enhance the accuracy and stability of spray drying.

Description

Technical Field

[0001] This invention relates to the field of spray drying control technology, specifically to a multivariable control system for spray drying of organic aqueous solutions. Background Technology

[0002] In the production processes of chemical, food, and pharmaceutical industries, spray drying of organic aqueous solutions is a commonly used process. It involves atomizing a solution and then exposing it to a hot air stream, causing the water to evaporate rapidly and resulting in a solid powder product. However, the spray drying process of organic aqueous solutions is subject to a complex interplay of factors, including the solution's concentration, viscosity, surface tension, and other characteristics, as well as the operating parameters of the spray drying equipment such as inlet and outlet air temperatures, air velocity, atomization pressure, and material flow rate. It is also affected by external conditions such as ambient temperature and humidity.

[0003] Currently, traditional spray drying control methods mostly employ single-variable control or simple multi-variable correlation control, making it difficult to comprehensively consider the interactions and dynamic changes among various factors. For example, when solution characteristics fluctuate, such as a sudden increase in concentration, adjusting only the inlet air temperature may lead to over- or under-drying, affecting product quality. Conversely, adjusting multiple parameters simultaneously without a precise understanding of their synergistic relationships may introduce new instabilities. Furthermore, traditional control systems are slow to detect anomalies during the drying process, often intervening only after significant product quality issues arise, hindering real-time and effective adjustments, resulting in reduced production efficiency and significant raw material waste.

[0004] Different organic aqueous solutions exhibit significantly different characteristics during spray drying, requiring varying drying parameters. Existing control systems lack the ability to adaptively adjust to different solution properties and environmental conditions, making it difficult to maintain stable drying results under complex and variable operating conditions. For example, excessively high temperatures can cause heat-sensitive organic compounds to decompose and deteriorate, while poor atomization of high-viscosity solutions can affect drying efficiency. Traditional control systems struggle to automatically optimize drying parameters based on these characteristics, requiring manual adjustments based on operator experience. This not only increases labor costs but also makes it difficult to guarantee the accuracy and consistency of control. Summary of the Invention

[0005] The purpose of this invention is to provide a multivariable control system for spray drying of organic aqueous solutions to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a multivariable control system for spray drying of organic aqueous solutions, the system comprising: The data acquisition module is used to acquire the aqueous solution of the organic matter to be spray-dried and to collect the operating parameter data of the spray drying equipment. Based on the operating parameter data, the current environmental state and solution characteristics of the spray drying process are identified. The drying parameter determination module is used to set the target drying parameters for spray drying based on the current environmental conditions and the solution characteristics, locate the relative position parameters of the spray drying equipment during the drying process, and determine the initial drying array for spray drying based on the operating parameter data and the relative position parameters. The drying status sensing module is used to combine the initial drying array and preset drying rules to sense the real-time drying status of spray drying, analyze the environmental change trend of the current environmental state based on the operating parameter data, and set the abnormal sensing mode of spray drying by combining the real-time drying status and the environmental change trend. The drying anomaly intervention module is used to identify drying anomalies in spray drying based on the anomaly perception mode; and to set up anomaly intervention mechanisms for spray drying based on the drying anomalies and a preset trajectory prediction network. The multivariable control module is used to perform multivariable control processing on the spray drying process according to the abnormality perception mode, the abnormality intervention mechanism and the preset drying rules, so as to obtain the spray drying control result.

[0007] Preferably, determining the initial drying array for spray drying based on the operating parameter data and the relative position parameters includes: Based on the aforementioned operating parameter data, the drying trajectory of the spray dryer is identified; Based on the drying trajectory, schedule the drying road condition information and target drying parameters for spray drying; Based on the dry road condition information and the relative position parameters, the target dryness parameters are adaptively adjusted to obtain adaptive dryness parameters; By combining the dry road condition information and the adaptive drying parameters, the initial drying pattern for spray drying is determined.

[0008] Preferably, identifying drying anomalies in spray drying based on the anomaly perception mode includes: Based on the aforementioned anomaly detection mode, historical drying data of spray drying is collected. Based on the historical drying data, identify the normal drying parameters for spray drying; Identify the current drying state of spray drying, and combine the current drying state with the historical drying data to identify the current behavioral deviation of spray drying; Based on the normal drying parameters, set the threshold for abnormal parameters in spray drying; Based on the abnormal parameter threshold and the current behavior deviation, identify the drying abnormality of spray drying.

[0009] Preferably, the step of analyzing the environmental change trend of the current environmental state based on the operating parameter data includes: The operating parameter data is classified by attributes to obtain classified parameter data; Extract the key features of the classification parameter data, and collect historical environmental data of the current environmental state based on the key features; Identify temperature and humidity data from the historical environmental data; Based on the temperature data, analyze the temperature change rate and airflow behavior pattern of the current environmental state; Based on the humidity data, analyze the humidity change level of the current environmental state; By combining the temperature change rate, the airflow behavior pattern, and the humidity change level, the environmental change trend of the current environmental state is analyzed.

[0010] Preferably, the step of setting an anomaly detection mode for spray drying by combining the real-time drying status and the environmental change trend includes: Based on the real-time drying status, the surrounding environmental factors of spray drying are identified; The interaction between the surrounding environmental factors and spray drying was analyzed; Based on the environmental change trends and the interaction relationships, identify potential risk factors for spray drying and determine the risk levels of these potential risk factors. Based on the real-time drying status and the risk level, a risk response mode for spray drying is set. Based on the risk level and the risk response mode, an anomaly detection mode for spray drying is set.

[0011] Preferably, the step of setting an abnormal intervention mechanism for spray drying based on the drying anomaly and the preset trajectory prediction network includes: Based on the described drying anomalies, locate the abnormal drying points in the spray dryer; Analyze the predicament type of the abnormal dry point, and based on the predicament type, set the predicament strain layer of the abnormal dry point; Based on the preset trajectory prediction network, identify the side drying trajectory of the abnormal dry point; Based on the side drying trajectory, an intervention path is set for the abnormal drying point; By combining the aforementioned stress response layer and the aforementioned intervention path, an abnormal intervention mechanism for spray drying is established.

[0012] Preferably, setting the target drying parameters for spray drying based on the current environmental conditions and the solution characteristics includes: Based on the current environmental conditions, extract the temperature distribution data for spray drying; The impact of the temperature distribution data on drying stability was analyzed, and the correlation between temperature and drying efficiency was identified. Based on the aforementioned solution characteristics, calculate the rate of change of the solution's concentration; By combining the temperature distribution data and the concentration change rate, the distribution balance of drying parameters is adjusted; Based on the requirements for drying stability and distribution uniformity, the target drying parameters for spray drying are set.

[0013] Preferably, the adaptive adjustment of the target drying parameters to obtain adaptive drying parameters includes: Based on the dry road condition information, the parameter offset during the drying process is identified; Analyze the degree of interference of the parameter offset on the drying array; Based on the relative position parameters, a parameter compensation mechanism is set; Based on the degree of interference and the parameter compensation mechanism, the distribution ratio of the target drying parameters is adjusted; Based on the adjusted distribution ratio, adaptive drying parameters are generated.

[0014] Preferably, identifying the drying trajectory of spray drying includes: Based on the aforementioned operating parameter data, real-time data points for spray drying are collected; The real-time data points are analyzed using a pattern recognition algorithm, and the data points are assigned to corresponding drying modes. Based on the distribution of drying patterns, identify the clustering characteristics of data points; The drying trajectory of spray drying is determined based on the clustering characteristics and the changes in drying patterns.

[0015] Preferably, the multivariate control processing of the spray drying process to obtain the spray drying control result includes: Based on the aforementioned anomaly perception mode, the drying difficulties of spray drying are identified; Based on the aforementioned drying predicament, the triggering conditions for the abnormal intervention mechanism are set; Based on the triggering conditions and the preset drying rules, adjust the multivariate control parameters; Based on the adjusted multivariate control parameters, spray drying control is executed; Generate spray drying control results.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By comprehensively acquiring the characteristics of the solution to be dried and the operating parameters of the equipment through the data acquisition module, the system can accurately identify the current environmental conditions and solution characteristics, providing comprehensive and reliable basic information for subsequent parameter setting and control adjustment. Compared with traditional control methods, it is no longer limited to monitoring a single or a few parameters, but achieves comprehensive perception of multiple factors affecting the drying process, enabling the system to have a more comprehensive understanding of the initial conditions and dynamic changes of the drying process.

[0017] The drying parameter determination module sets target drying parameters and determines the initial drying configuration based on the acquired environmental conditions and solution characteristics, fully considering the individualized requirements of different solution characteristics and environmental conditions for the drying process. This approach avoids the limitations of using fixed or simple empirical parameters in traditional control, and can formulate a more suitable initial plan for the drying process based on actual conditions. This ensures that the spray drying equipment is in a more reasonable operating state during the start-up phase, helping to reduce the problem of poor initial drying effect caused by parameter mismatch.

[0018] The drying status sensing module, combining the initial drying array and preset drying rules, senses the real-time drying status and analyzes environmental change trends. It then sets an anomaly sensing mode, enabling dynamic monitoring and early warning of anomalies in the drying process. By tracking the drying status in real time, the system can promptly detect subtle changes during the drying process. Furthermore, setting an anomaly sensing mode based on environmental change trends enhances the ability to predict potential anomalies, overcoming the lag in response to anomalies in traditional control systems. This allows the system to prepare for intervention at the initial stage or even before anomalies occur.

[0019] The drying anomaly intervention module identifies abnormal situations based on anomaly perception patterns and combines this with a preset trajectory prediction network to set up anomaly intervention mechanisms. This allows for rapid and targeted intervention when anomalies are detected. This intervention mechanism is not simply a matter of parameter adjustment, but rather a systematic adjustment plan based on accurate identification of anomalies and prediction of their development trends. This approach more effectively curbs the development of anomalies, preventing them from having a greater impact on product quality and the production process.

[0020] The multivariable control module integrates anomaly detection modes, anomaly intervention mechanisms, and preset drying rules to perform multivariable control processing on the drying process, achieving coordinated adjustment of multiple influencing parameters. This multivariable control method fully considers the interactions and influences between parameters, avoiding the chain reactions that may be caused by adjusting a single parameter. It optimizes the efficiency of the drying process while ensuring product quality, reducing energy consumption and raw material waste. Simultaneously, the system can adapt to the differences in the characteristics of different organic aqueous solutions and changes in environmental conditions, maintaining stable drying effects under various complex operating conditions. It requires less reliance on manual operation, reducing the need for operator experience and improving the automation and intelligence level of the spray drying process. Attached Figure Description

[0021] Figure 1 This is a timing diagram of the multivariable control system for spray drying of organic aqueous solution described in this invention; Figure 2 Flowchart for determining the initial drying array; Figure 3 A flowchart for analyzing environmental change trends; Figure 4 Flowchart for setting target drying parameters; Figure 5 A flowchart generated for adaptive drying parameters. Detailed Implementation

[0022] 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, and 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.

[0023] Please see Figure 1 The present invention provides a multivariable control system for spray drying of organic aqueous solution, the system comprising: a data acquisition module, a drying parameter determination module, a drying state sensing module, a drying anomaly intervention module, and a multivariable control module.

[0024] The data acquisition module obtains the physical properties of the organic aqueous solution and the operating parameters of the spray drying equipment, identifying the current environmental state and solution characteristics. The drying parameter determination module sets target drying parameters based on the environmental state and solution characteristics, and determines the initial drying configuration by combining operating parameters and relative position parameters. The drying status sensing module monitors the drying status and environmental change trends in real time, setting an anomaly sensing mode. The drying anomaly intervention module identifies drying anomalies based on the anomaly sensing mode and formulates intervention mechanisms through a preset trajectory prediction network. The multivariate control module integrates the anomaly sensing mode, intervention mechanisms, and preset drying rules to perform multivariate control of the spray drying process, outputting the final control results.

[0025] Example 1: See Figure 2 The network comprises a temperature sensor array, a humidity detection unit, an ultrasonic flow meter, and a laser anemometer, covering all key areas of the drying tower. The temperature sensors employ platinum resistance probes, acquiring hot air distribution data every 0.5 seconds; the humidity detection unit, based on capacitive sensing principles, monitors the moisture content of the airflow in real time; the ultrasonic flow meter tracks the supply rate of the organic aqueous solution; and the laser anemometer records airflow vector information. All data is transmitted to the central processing unit via an RS-485 bus. After outliers are removed by a data cleaning module, data alignment is performed using a sliding time window mechanism.

[0026] The system standardizes the three-dimensional data of temperature, humidity, and flow rate, and eliminates time skew in the time series data using a dynamic time warping algorithm. Then, a Gaussian mixture model is applied to assign data points, and the optimal number of clusters is automatically determined based on the Bayesian information criterion. Each cluster center corresponds to a drying mode, including a rapid evaporation zone, a constant-rate drying zone, and a falling-rate drying zone. The system identifies the main and branch paths of the drying trajectory by calculating the spatial distribution density of the cluster centers. The evolution trend of the drying trajectory is predicted using a hidden Markov model, with the model's state transition probabilities generated based on historical drying data.

[0027] When analyzing the uniformity of airflow distribution, the dry road condition analysis engine uses computational fluid dynamics simulation results as a benchmark. It quantifies the road condition fluctuation coefficient by comparing the difference between the real-time airflow velocity distribution and the ideal model. Hot air temperature gradient analysis is based on infrared thermal imaging data to establish a three-dimensional thermographic map of the temperature field. The atomization pressure setpoint, inlet air temperature threshold, and solution flow rate range in the target drying parameters are dynamically corrected using an adaptive filtering algorithm. The core of the algorithm employs a Kalman filter, whose process noise covariance matrix is ​​adjusted in real-time according to the relative position parameters.

[0028] The parameter offset calculation module analyzes the deviation between actual parameters and target values ​​in each region, and assesses the interference intensity of the offset on the drying array through variance analysis. The parameter compensation mechanism is designed as a two-level control: the primary compensation uses a proportional-integral controller to adjust the compensation gain in real time according to the interference intensity; the advanced compensation introduces a feedforward control loop to correct the parameters in advance based on the dry road condition prediction model. The compensated drying parameters are normalized to form an adaptive drying parameter set.

[0029] The internal space of the drying tower is divided into 1 cm³ voxel units, each associated with an adaptive drying parameter set. A three-dimensional parameter distribution matrix is ​​generated through tensor operations, where each element contains the temperature setpoint, humidity threshold, and airflow velocity range. This matrix is ​​stored in a real-time database as the basic control framework for the drying process.

[0030] In the data preprocessing stage, median filtering was used for smoothing, and the probability distribution of normal drying parameters was calculated using kernel density estimation. The normal parameter range was defined as the distribution band with a 95% confidence interval, including the atomization pressure fluctuation threshold, particle moisture content variation curve, and standard evaporation rate value. The real-time monitoring system acquired particle moisture content through an online laser particle size analyzer and calculated the drying rate through a differential pressure sensor, with the sampling frequency set to 10Hz.

[0031] The deviation quantification index includes shape difference and numerical offset, which are weighted and fused to generate a comprehensive deviation coefficient. The abnormal parameter threshold setting system is based on a reinforcement learning model, which continuously optimizes the threshold boundary through the Q-learning algorithm. When the comprehensive deviation coefficient exceeds the dynamic threshold, the system marks an abnormal drying situation and triggers a three-level early warning mechanism: Level 1 early warning adjusts the parameter compensation intensity, Level 2 early warning activates the backup drying plan, and Level 3 early warning executes the emergency shutdown procedure.

[0032] After each drying process, the system automatically adds the parameter data to the training set and retrains the Gaussian mixture model and the Hidden Markov Model. Model updates employ incremental learning, preserving historical model features while incorporating new data patterns. The false alarm rate of the anomaly detection module is controlled through confusion matrix analysis, continuously optimizing the splitting threshold of the Isolation Forest algorithm. The entire perception process forms a closed-loop control system, ensuring that anomaly identification accuracy continuously evolves over runtime.

[0033] Example 2: See Figure 3The attribute classification engine employs a random forest-based classification model, dynamically dividing the original data stream into temperature, humidity, airflow velocity, and solution property datasets. The temperature dataset includes multi-point temperature records from the drying tower's inlet, constant temperature, and outlet zones; the humidity dataset covers inlet and outlet humidity, as well as the relative humidity gradient within the tower; the airflow velocity dataset records axial wind speed, radial wind speed, and eddy current intensity; and the solution property dataset includes concentration, viscosity, and surface tension parameters. The classified datasets then proceed to feature extraction, using principal component analysis to reduce data dimensionality and extract key feature vectors characterizing the environmental state. These key feature vectors include the temperature distribution dispersion coefficient, humidity change slope, airflow uniformity index, and solution concentration fluctuation amplitude.

[0034] After receiving key feature vectors, the environmental change trend analysis module initiates a historical environmental database search. The database stores snapshots of the environmental conditions over the past 72 hours, organized into a circular buffer structure based on time series. Temperature change rate analysis employs the sliding window difference method to calculate the magnitude of temperature gradient changes within adjacent time windows. Simultaneously, a computational fluid dynamics simulation model compares real-time airflow data with a standard turbulence model to identify characteristic parameters of airflow behavior patterns, including vortex core location, recirculation zone range, and boundary layer separation point. Humidity change level analysis uses an adaptive threshold method to establish a correlation curve between humidity change rate and time variable, and the second derivative of the curve is used to determine the accelerating trend of humidity change.

[0035] Temperature change rate, airflow behavior pattern characteristic parameters, and humidity change level are input into a time series prediction model, which is built based on a gated recurrent unit network. The network input layer receives standardized feature vectors, the hidden layer contains 128 neurons, and the output layer generates a comprehensive score of environmental change trends. The score is mapped to a range of 0-100, with a score above 60 indicating a significant risk of environmental change. The score results are transmitted in real time to the anomaly perception mode configuration unit as a benchmark for adjusting the mode sensitivity.

[0036] A laser dust sensor is installed at the air inlet of the drying tower to detect changes in PM2.5 concentration in real time; a silicon photodiode array monitors ambient light intensity with a resolution of 0.1 lux; and a network of ultrasonic anemometers is deployed around the equipment to capture external airflow interference vectors. The interaction between environmental factors and spray drying is analyzed using a multiple regression model to establish the correlation matrix between dust concentration and airflow cleanliness, the response surface between light intensity and solution photosensitivity, and the coupling coefficient between external airflow and the flow field inside the drying tower. The quantitative results of the interaction are stored as correlation coefficient tensors, with the tensor dimensions corresponding to the combined influence weights of different environmental factors.

[0037] The potential risk factor identification engine loads a correlation coefficient tensor and calculates the risk value of each factor through a risk probability model. The model input consists of real-time environmental factor readings and an interaction tensor, and the output is a risk level matrix. Risk levels are divided into four levels: Level 1 risk corresponds to slight fluctuations in environmental factors, Level 2 risk indicates significant changes in a single factor, Level 3 risk represents the synergistic deterioration of multiple factors, and Level 4 risk predicts systemic environmental out-of-control. The risk response mechanism is configured using a fuzzy logic controller. Input variables are real-time drying status parameters and risk levels, and output variables are drying parameter adjustment strategies. The controller defines 49 fuzzy rules; for example, when Level 3 risk is detected and particle moisture content increases, an inlet air temperature increase strategy is triggered; when encountering Level 2 risk accompanied by airflow turbulence, a deflector angle adjustment program is initiated.

[0038] The first level of decision-making is based on a risk level matrix, generating a basic monitoring frequency: Level 1 risk corresponds to routine monitoring (1Hz sampling), Level 2 risk triggers enhanced monitoring (5Hz sampling), and Level 3 and above risks activate high-frequency monitoring (20Hz sampling). The second level of decision-making combines risk response methods to configure monitoring parameter combinations: temperature-related risks focus on monitoring hot air distribution, humidity risks emphasize dew point detection, and airflow risks enhance eddy current monitoring. The third level of decision-making uses a neural network prediction model to dynamically adjust the weight coefficients of each sensor. The final anomaly perception model includes a three-layer early warning mechanism: primary early warning prompts parameter fine-tuning, intermediate early warning requires intervention verification, and advanced early warning triggers system review.

[0039] After each drying process, the system compares the predicted environmental change trend with the actual recorded data to calculate the trend prediction error. This error data is input into the training algorithm of the gated recurrent unit network, and the network weights are updated using backpropagation. Simultaneously, the interaction relationship analysis model undergoes offline optimization monthly, recalculating the correlation coefficient tensor using accumulated environmental data. The risk probability model is continuously updated through online learning, automatically adjusting the risk value calculation parameters whenever a new drying anomaly case is added.

[0040] The system constructs a virtual model of the spray drying process and periodically compares real-time operational data with the simulation results of the virtual model. When a significant deviation is found between the sensing mode and the actual operating conditions, a mode reconstruction program is initiated. The reconstruction process retains valid parameters from historical configurations and performs local optimizations on the deviation points to ensure that the sensing mode always remains synchronized with the actual state of the equipment. The entire system establishes a configuration version management mechanism, recording the timestamp and changes of each mode update, forming a complete mode evolution archive.

[0041] Example 3: See Figure 4The positioning system loads a real-time drying status data stream and employs an improved isolated forest algorithm to construct an anomaly detection model. This model uses 100 isolated trees, each randomly selecting 32 feature subsets. An anomaly score matrix is ​​generated by calculating the path length differences between data points. When the anomaly score of a data point exceeds a dynamic threshold, the system marks it as a candidate anomaly. After spatial clustering analysis, the three-dimensional coordinates of the abnormal drying points are confirmed. The coordinate data includes the axial position, radial distance, and height information of the drying tower, with millimeter-level accuracy.

[0042] The network input layer receives temperature gradient, humidity distribution, and airflow vector data within a 5cm³ area surrounding the anomaly point. The hidden layer contains three dedicated analysis channels: a temperature channel analyzes thermal inertia and heat transfer efficiency; a humidity channel calculates moisture diffusion rate and latent heat of phase change; and an airflow channel assesses Reynolds stress and vortex intensity. The output layer generates a probability distribution of distress types; a valid diagnosis is considered valid when the probability of a certain type exceeds 85%. The system pre-defines three main distress types: temperature imbalance corresponds to a heat transfer efficiency decrease exceeding a baseline value of 30%; humidity exceeding the standard is characterized by a sustained local relative humidity exceeding a set threshold of 15%; and airflow turbulence is characterized by a sudden change in vortex intensity exceeding twice the standard deviation of the historical mean.

[0043] The temperature imbalance strain unit comprises three levels of response: the primary response adjusts the heater power of the zone where the anomaly point is located, controlling the change within ±10%; the intermediate response initiates thermal compensation in adjacent areas, redistributing hot air to balance the temperature field; and the advanced response triggers intermittent spraying from cooling nozzles to prevent material overheating and degradation. The humidity exceedance strain unit deploys a humidity gradient control system, automatically selecting a dehumidification scheme based on the degree of exceedance: mild exceedance activates the molecular sieve adsorption module, moderate exceedance activates the condensation dehumidifier unit, and severe exceedance executes an airflow replacement procedure. The airflow turbulence strain unit integrates a guide vane matrix, with 128 intelligent guide vanes independently controlled by servo motors. The guide vane angle adjustment algorithm is based on fluid dynamics simulation, generating the optimal angle combination to restore the airflow to a laminar state.

[0044] The network input includes equipment geometric topology data, real-time operating parameters, and historical drying trajectories. A graph convolutional layer constructs a 3D mesh model of the drying tower, where nodes represent spatial locations and edge weights represent mass transfer relationships between locations. Convolutional kernels extract local features in the spatial dimension, while a temporal recursive layer captures the evolution patterns of the trajectories. The network output is a set of drying trajectories adjacent to abnormal drying points, with each trajectory containing a coordinate sequence and a confidence score. The trajectory selection mechanism uses non-dominated sorting, retaining the top 5 Pareto optimal candidate trajectories.

[0045] Intervention path optimization introduces multi-objective functions :

[0046] in: The energy consumption function representing time t. Represents the drying efficiency function. and These are dynamic weighting coefficients. Path optimization employs an improved ant colony algorithm, using 200 virtual ants for parallel search. The pheromone update rule incorporates a simulated annealing mechanism to avoid local optima. The algorithm outputs a sequence of key nodes along the intervention path, with the node spacing adaptively adjusted based on the drying rate.

[0047] The model defines seven states: standby monitoring, anomaly identification, predicament classification, trajectory generation, path optimization, intervention execution, and effect evaluation. State transition conditions are dynamically set based on real-time sensor data; for example, the condition for transitioning from the predicament classification state to the trajectory generation state is a predicament diagnosis confidence level exceeding 90%. The intervention execution phase employs a distributed actuator network, with 32 control terminals simultaneously receiving path commands. Temperature intervention terminals control 128 heating elements, humidity intervention terminals manage 24 dehumidification units, and airflow intervention terminals operate the deflector matrix. The execution process uses a gradual adjustment strategy, adjusting the intervention intensity by 10% every 5 seconds to avoid sudden parameter changes.

[0048] An infrared thermal imaging system scans the drying tower at a frequency of 10 Hz to generate point cloud data of temperature distribution. After voxelization, the point cloud is reconstructed using a heat conduction inversion algorithm to reconstruct the three-dimensional temperature field. The solution characteristic monitoring unit analyzes concentration changes using an online ultraviolet-visible spectrometer, covering a spectral range of 200-800 nm. The concentration change rate is calculated using the sliding window derivative method, with the window width adaptively adjusted according to the solution viscosity.

[0049] A coupling matrix between the temperature and concentration fields is established, and the optimal parameter distribution is solved using the Jacobi iteration method. The objective function considers both the drying stability coefficient and the distribution uniformity index, with constraints including equipment safety thresholds and material thermosensitivity. The final target drying parameter set contains 256 control variables, which are distributed to each execution unit via a control bus. The parameter update mechanism employs a version rolling strategy, retaining the three most recent parameter configurations for rapid rollback.

[0050] The initial validation uses a network of miniature sensors installed in the intervention area to collect real-time data on temperature recovery rate, humidity decrease slope, and airflow stability. The secondary validation initiates parallel simulations using a digital twin system, comparing the actual intervention effects with the virtual model's predictions. When the actual effect deviates from the predicted value by more than 15%, an online optimization loop for the intervention parameters is triggered. Each intervention process generates a complete operation log, including timestamps, intervention type, execution parameters, and effect indicators, used to continuously improve the accuracy of the prediction network.

[0051] The state space is defined as the dryness anomaly feature vector, the action space corresponds to the combination of intervention strategies, and the reward function is based on a comprehensive evaluation of intervention effect and energy consumption. The agent adopts a dual-deep Q-network architecture, updating network parameters after every 10 interventions. The experience replay buffer stores the most recent 1000 intervention records, prioritizing the sampling of cases with abnormal effects for training. The network update adopts a soft synchronization strategy, synchronizing the target network parameters every 50 updates.

[0052] Example 4: See Figure 5 The collection of dry road condition information is achieved through a multi-sensor fusion system. A monitoring network of 32 wireless sensor nodes is deployed within a spray drying tower with a diameter of 2 meters and a height of 5 meters. The nodes are spaced 0.5 meters apart along the axial direction, with 8 nodes evenly distributed around the circumference of each layer. Each node integrates a sensor for temperature, humidity, and airflow, with a sampling frequency of 20Hz. The dry road condition analysis engine processes the sensor data in real time, generating a 3D road condition heat map. The heat map divides the drying tower into 8 independent control zones, each labeled with an airflow uniformity level (1-5) and a temperature stability index (0-1.0).

[0053] Parameter offset measurement employs ultra-wideband positioning technology. Positioning tags are installed on the atomizing nozzles, and four base stations are placed at the four corners of the drying tower. The system records the nozzle spatial coordinates with a time resolution of 100ms and calculates the Euclidean distance difference between the actual position and the theoretical trajectory. The offset analysis module converts the distance difference into a parameter influence factor: the influence factor is 0.1 when the offset is less than 5mm, 0.3 for 5-10mm, 0.6 for 10-20mm, and 1.0 for greater than 20mm. Interference level assessment uses a regional correlation model to establish a weighted relationship matrix between the offset position and each control area.

[0054] The parameter compensation mechanism is based on an adaptive neural network. The network input layer contains three data items: real-time position coordinates (x, y, z), interference influence factor (δ), and region weight matrix (W). The hidden layer is designed with a dual-channel structure: the spatial channel processes the relationship between position and weights, and the interference channel analyzes the transmission characteristics of the influence factor. The output layer generates a compensation coefficient matrix C for eight control regions, with coefficients ranging from 0.8 to 1.2. The network is trained using real-time incremental learning; after each drying process, the network weights are adjusted in reverse based on the actual drying effect.

[0055] The distribution ratio of the target drying parameters is adjusted through tensor operations. The initial target parameters are stored as an 8×3 matrix: each row represents a region, and the three columns represent the temperature setpoint, humidity threshold, and airflow velocity range, respectively. The compensation coefficient matrix C is multiplied by the target parameter matrix using a Hadamard product to generate the compensation parameter matrix. The adjusted parameters are normalized: temperature parameters are scaled proportionally, humidity parameters are taken as the regional mean, and airflow parameters maintain their range. The final adaptive drying parameter set is written to the control actuator and updated every 30ms.

[0056] The drying trajectory recognition system employs an improved K-means clustering algorithm. The algorithm dynamically sets the number of cluster centers, initially with three core patterns (atomization zone, heat transfer zone, and settling zone). Real-time data points are collected at a frequency of 50Hz, and each data point includes four-dimensional features: spatial coordinates, temperature, humidity, and airflow velocity. Data preprocessing uses sliding window normalization, with the window width dynamically adjusted according to the drying stage: 10 seconds for the atomization stage, 20 seconds for the heat transfer stage, and 30 seconds for the settling stage.

[0057] The clustering process incorporates dynamic time warping. The algorithm establishes two similarity metric pools: a spatial distance pool calculates Euclidean distance, and a pattern similarity pool evaluates the matching degree of drying features. When a new data point arrives, both distances to each cluster center are calculated simultaneously, and the weighted fusion determines its category. Weight allocation follows a spatial priority principle: spatial weight 0.7 / pattern weight 0.3 in the atomization stage, 0.5 / 0.5 in the heat transfer stage, and 0.3 / 0.7 in the settling stage.

[0058] Cluster feature extraction employs the density peak detection method. The system scans the clustering results every 5 seconds and calculates the following for each category: core density (number of data points per unit volume); boundary gradient (steepness of the boundary between clusters); and stability index (reciprocal of the distance the cluster centers have moved).

[0059] The system constructs a pattern transition matrix to record the migration path of each cluster center in consecutive time slices. When a new cluster center is detected or an old center disappears, a trajectory reconstruction program is initiated. The reconstruction algorithm is based on Bayesian filtering, fusing the current cluster distribution with historical transition probabilities to generate the optimal dry trajectory prediction. Trajectory data is stored as an ordered coordinate sequence, outputting 10 trajectory point coordinates per second. See Table 1.

[0060] Table 1: Examples of regional parameter compensation effects.

[0061]

[0062] The system maintenance module performs periodic self-checks. A sensor calibration procedure is initiated daily at midnight: a standard temperature source is heated to 200°C, a standard humidity gas is introduced, and the measurement deviations of 32 nodes are recorded. Calibration data is used to compensate for subsequent measurements. Clustering algorithm parameters are optimized weekly, and the initial cluster centers are recalculated using the week's dry data. The neural network is trained offline monthly, loading historical data to enhance the model's generalization ability. All maintenance operations are recorded in a blockchain log, forming an immutable equipment health record.

[0063] The visualization of trajectory recognition results is achieved through an augmented reality interface. Operators wearing AR glasses can observe a 3D trajectory cloud map inside the drying tower, with different colors marking atomization, heat transfer, and settling trajectory segments. The thickness of the trajectory indicates the density of data points, and dynamic flashing indicates areas of abnormal fluctuation. The interface supports gesture operations to zoom and rotate the model, and key parameters are displayed in real time overlaid at the corresponding spatial locations.

[0064] Example 5: Anomaly Perception Mode Identifies Drying Difficulties Through a Multi-Dimensional Scanning Mechanism. The system loads a real-time drying status data stream and uses a convolutional neural network to analyze the three-dimensional parameter distribution map. The network input layer receives fused data from the temperature field thermogram, humidity gradient cloud map, and airflow vector field. The feature extraction layer contains eight parallel convolutional channels, each capturing spatial patterns at different scales. The classification output layer generates a difficulty probability distribution: local overheating corresponds to a continuous expansion pattern of high-temperature areas; uneven drying is manifested as an excessive standard deviation in humidity distribution; and particle agglomeration is characterized by a bimodal particle size distribution. The difficulty identification result is accompanied by a confidence score; when the score exceeds 90%, subsequent control procedures are triggered.

[0065] The trigger conditions for the abnormal intervention mechanism are configured using a fuzzy rule engine. The engine's input variables include the type of distress, severity index, and diffusion trend coefficient. The severity index is calculated using weighted averages of indicators such as the proportion of high-temperature areas and the standard deviation multiple of humidity; the diffusion trend coefficient is based on the output of a time-series prediction model. The rule base has 128 preset trigger logics. For example, when a local overheating type is accompanied by a severity index > 0.7 and a diffusion trend coefficient > 1.2, the highest-level intervention response is activated. An adaptive trigger threshold setting mechanism dynamically adjusts the threshold boundaries based on historical intervention effects to avoid over-response.

[0066] The adjustment of multivariable control parameters employs a model predictive control strategy. The controller establishes a state-space model containing 256 variables, covering key parameters such as temperature setpoint, airflow velocity, and atomization pressure. Rolling time-domain optimization is performed every 50 milliseconds: within the current control cycle, based on the characteristics of the drying dilemma and preset drying rules, the optimal parameter trajectory for the next 10 seconds is solved. The optimization objective function balances drying efficiency and energy consumption, with constraints including equipment physical limits and material safety thresholds. The solution process uses a parallel gradient descent algorithm, iterating synchronously across 8 computational cores.

[0067] The spray drying control employs a zoned collaborative strategy. The drying tower is divided into eight control zones along its axial direction, each equipped with an independent execution unit. The central controller decomposes the adjusted parameter set into zone instruction sets and distributes them via real-time industrial Ethernet. The temperature control unit uses pulse width modulation technology to precisely regulate the power output of 128 heating elements; the airflow control unit operates 64 variable frequency fans, achieving wind speed adjustment with a precision of 0.1 m / s; the atomization system controls the nozzle orifice diameter via piezoelectric ceramic actuators, with an adjustment range of 50-200 μm. A cross-validation mechanism is implemented during execution, with sensors in adjacent zones monitoring each other's performance.

[0068] The generation of spray drying control results integrates multi-source evaluation data. An online laser particle size analyzer scans 1000 particles per second to generate a particle size distribution spectrum; a near-infrared moisture analyzer measures particle moisture content in real time, with a sampling depth of up to 3 mm; a high-speed camera captures particle motion trajectories and calculates the drying uniformity index. The data fusion engine uses evidence theory to handle uncertain information, and when data from different sensors conflict, a confidence-weighted arbitration mechanism is activated. The final control results are output as a structured report, including a three-dimensional spatiotemporal distribution map and statistical characteristic values ​​of key indicators.

[0069] A visual monitoring interface creates an immersive operating environment. A 55-inch curved display presents a transparent model of the drying tower, with colored streamlines indicating airflow movement, a thermal map overlaid on the temperature field, and a semi-transparent cloud effect rendering the humidity distribution. Operators can control the rotation angle via gestures, and hovering over any point displays real-time parameters for that point. The alarm system uses tiered visual cues: Level 1 alerts display a flashing yellow border, Level 2 alerts activate a 3D arrow pointing to the abnormal area, and Level 3 alerts trigger a full-screen red pulsating warning. Historical data playback supports 50x speed simulation of the drying process.

[0070] The self-optimization mechanism of the control system is achieved through a dual-loop system. The inner loop monitors the deviation between the control results and the expected target in real time. When the key indicators deviate from the set range by more than 5%, the weight parameters of the model predictive control are automatically fine-tuned. The outer loop starts after each drying process, analyzes the overall control effect, and updates the structural parameters of the state-space model. The model update uses transfer learning technology, retaining the weights of the general feature layer and only fine-tuning the adaptation layer for specific operating conditions.

[0071] The operation log system implements holographic recording. The timestamp, decision basis, execution parameters, and feedback data for each control decision are written to a time-series database. The logs employ a hierarchical storage strategy: real-time data is retained for 7 days, feature data for 3 months, and statistical summaries are permanently stored. The log analysis engine automatically detects abnormal operating patterns; when it detects an abnormally high frequency of parameter adjustments or actuator overload, it triggers a system self-check procedure.

[0072] The version control module manages the iterative updates of the control algorithm. Each algorithm modification generates an independent version number, recording the changes and test cases. Three parallel versions are maintained during system runtime: a stable version for the production environment, a test version for verification in the virtual drying tower, and a development version for algorithm engineers to debug. Version switching employs a canary release mechanism, first running it for 24 hours in a single control area, and then rolling it out across the entire tower after confirmation. All version changes are mapped onto a timeline, visually demonstrating the evolution path of the control strategy.

[0073] The fault-tolerant mechanism features multiple layers of protection. In the event of a main control loop malfunction, it automatically switches to the backup PLC system; in the event of a communication interruption, each zone's execution unit continues operation according to the last valid parameters; when the emergency stop button is triggered, the system executes a three-step safety procedure: cutting off the heating power supply, starting the cooling fan, and injecting inert gas. The safety status monitoring board displays the real-time operating status of 16 key components, with each status verified by a dual-channel independent sensor.

[0074] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multivariable control system for spray drying of organic aqueous solutions, characterized in that, include: The data acquisition module is used to acquire the aqueous solution of the organic matter to be spray-dried and to collect the operating parameter data of the spray drying equipment. Based on the operating parameter data, the current environmental state and solution characteristics of the spray drying process are identified. The drying parameter determination module is used to set the target drying parameters for spray drying based on the current environmental conditions and the solution characteristics, locate the relative position parameters of the spray drying equipment during the drying process, and determine the initial drying formation for spray drying based on the operating parameter data and the relative position parameters. The drying status sensing module is used to combine the initial drying array and preset drying rules to sense the real-time drying status of spray drying, analyze the environmental change trend of the current environmental state based on the operating parameter data, and set the abnormal sensing mode of spray drying by combining the real-time drying status and the environmental change trend. The drying anomaly intervention module is used to identify drying anomalies in spray drying based on the anomaly perception mode; and to set up anomaly intervention mechanisms for spray drying based on the drying anomalies and a preset trajectory prediction network. The multivariable control module is used to perform multivariable control processing on the spray drying process according to the anomaly perception mode, the anomaly intervention mechanism and the preset drying rules, so as to obtain the spray drying control result.

2. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, Determining the initial drying array for spray drying based on the operating parameter data and the relative position parameters includes: Based on the aforementioned operating parameter data, the drying trajectory of the spray dryer is identified; Based on the drying trajectory, schedule the drying road condition information and target drying parameters for spray drying; Based on the dry road condition information and the relative position parameters, the target dryness parameters are adaptively adjusted to obtain adaptive dryness parameters; By combining the dry road condition information and the adaptive drying parameters, the initial drying pattern for spray drying is determined.

3. The multivariable control system based on organic aqueous solution spray drying as described in claim 1, characterized in that, The step of identifying drying anomalies in spray drying based on the anomaly perception mode includes: Based on the aforementioned anomaly detection mode, historical drying data of spray drying is collected. Based on the historical drying data, identify the normal drying parameters for spray drying; Identify the current drying state of spray drying, and combine the current drying state with the historical drying data to identify the current behavioral deviation of spray drying; Based on the normal drying parameters, set the threshold for abnormal parameters in spray drying; Based on the abnormal parameter threshold and the current behavior deviation, identify the drying abnormality of spray drying.

4. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, The analysis of environmental change trends based on the operational parameter data includes: The operating parameter data is classified by attributes to obtain classified parameter data; Extract the key features of the classification parameter data, and collect historical environmental data of the current environmental state based on the key features; Identify temperature and humidity data from the historical environmental data; Based on the temperature data, analyze the temperature change rate and airflow behavior pattern of the current environmental state; Based on the humidity data, analyze the humidity change level of the current environmental state; By combining the temperature change rate, the airflow behavior pattern, and the humidity change level, the environmental change trend of the current environmental state is analyzed.

5. The multivariable control system for spray drying of organic aqueous solutions as described in claim 1, characterized in that, The method of setting an anomaly detection mode for spray drying by combining the real-time drying status and the environmental change trend includes: Based on the real-time drying status, identify the surrounding environmental factors of spray drying; The interaction between the surrounding environmental factors and spray drying was analyzed; Based on the environmental change trends and the interaction relationships, identify potential risk factors for spray drying and determine the risk levels of these potential risk factors. Based on the real-time drying status and the risk level, a risk response mode for spray drying is set. Based on the risk level and the risk response mode, an anomaly detection mode for spray drying is set.

6. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, The step of setting an abnormal intervention mechanism for spray drying based on the drying anomaly and the preset trajectory prediction network includes: Based on the described drying anomalies, locate the abnormal drying points in the spray dryer; Analyze the predicament type of the abnormal dry point, and based on the predicament type, set the predicament strain layer of the abnormal dry point; Based on the preset trajectory prediction network, identify the side drying trajectory of the abnormal dry point; Based on the side drying trajectory, an intervention path is set for the abnormal drying point; By combining the aforementioned stress response layer and the aforementioned intervention path, an abnormal intervention mechanism for spray drying is established.

7. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, The step of setting target drying parameters for spray drying based on the current environmental conditions and solution characteristics includes: Based on the current environmental conditions, extract the temperature distribution data for spray drying; The impact of the temperature distribution data on drying stability was analyzed, and the correlation between temperature and drying efficiency was identified. Based on the aforementioned solution characteristics, calculate the rate of change of the solution's concentration; By combining the temperature distribution data and the concentration change rate, the distribution balance of drying parameters is adjusted; Based on the requirements for drying stability and distribution uniformity, the target drying parameters for spray drying are set.

8. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, The adaptive adjustment of the target drying parameters to obtain adaptive drying parameters includes: Based on the dry road condition information, the parameter offset during the drying process is identified; Analyze the degree of interference of the parameter offset on the drying array; Based on the relative position parameters, a parameter compensation mechanism is set; Based on the degree of interference and the parameter compensation mechanism, the distribution ratio of the target drying parameters is adjusted; Based on the adjusted distribution ratio, adaptive drying parameters are generated.

9. The multivariable control system for spray drying of organic aqueous solutions as described in claim 1, characterized in that, The identification of the drying trajectory of spray drying includes: Based on the aforementioned operating parameter data, real-time data points for spray drying are collected; The real-time data points are analyzed using a pattern recognition algorithm, and the data points are assigned to corresponding drying modes. Based on the distribution of drying patterns, identify the clustering characteristics of data points; The drying trajectory of spray drying is determined based on the clustering characteristics and the changes in drying patterns.

10. The multivariable control system for spray drying organic aqueous solution as described in claim 1, characterized in that, The multivariate control process for the spray drying process, resulting in spray drying control results, includes: Based on the aforementioned anomaly perception mode, the drying difficulties of spray drying are identified; Based on the aforementioned drying predicament, the triggering conditions for the abnormal intervention mechanism are set; Based on the triggering conditions and the preset drying rules, adjust the multivariate control parameters; Based on the adjusted multivariate control parameters, spray drying control is executed; Generate spray drying control results.

Images