PLC-based sand-fixing and dust-suppressing agent spray drying process control system

Abstract

The application relates to the technical field of spray drying control, and discloses a sand-fixing dust suppressant spray drying process control system based on a PLC. The system comprises the following steps: collecting an inlet material liquid flow signal, a hot air temperature signal and an outlet powder moisture content signal of a spray drying tower, and separating a pulsation component and a steady-state component from the inlet material liquid flow signal; then acquiring phase coupling characteristics of the pulsation component and the hot air temperature signal, and inhibition intensity characteristics of the outlet powder moisture content signal on the pulsation component; then generating process characteristics representing drying stability according to the characteristics, and calculating a drying stage identification probability distribution based on the process characteristics; and finally compensating and correcting the hot air temperature signal according to the probability distribution, and decoupling a target temperature control curve of a target drying stage. The system realizes self-adaptive adjustment of the hot air temperature, helps to improve the stability and controllability of a sand-fixing dust suppressant spray drying process, and improves product quality consistency.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of spray drying control, and particularly relates to a sand-fixing dust suppressant spray drying process control system based on a PLC. BACKGROUND

[0002] In the production process of sand-fixing dust suppressants, spray drying is a key process link that determines the product quality. In this process, the liquid feed is atomized and then contacted with a hot air flow, so that the water in the liquid feed is rapidly evaporated, and finally a powder product meeting the water content requirement is formed. The effect of spray drying directly affects the particle size distribution, compressive strength and water resistance of the sand-fixing dust suppressant, so accurate control of the drying process is crucial.

[0003] The control of the sand-fixing dust suppressant spray drying process mainly depends on traditional PLC control systems. Such systems usually adjust based on preset steady-state parameters, for example, fixed inlet liquid flow and hot air temperature are set, and the equipment operating state is adjusted through a simple feedback loop. However, in actual production, there are many interference factors, resulting in significant dynamic characteristics of the drying process. The inlet liquid flow is easily affected by fluctuations in the feed pump, changes in the liquid concentration and other factors, resulting in non-steady-state pulsating components; the hot air temperature may fluctuate due to unstable heat source output, changes in environmental temperature and humidity, etc. If these dynamic changes are not effectively captured and processed, the heat and mass exchange efficiency of the liquid and hot air will be directly affected.

[0004] The limitations of existing control systems mainly manifest in three aspects. Firstly, the pulsating components and steady-state components in the liquid flow are not effectively distinguished, and the two are controlled as a whole, resulting in a lag in the response of the control signal to the actual process fluctuations. Secondly, the dynamic correlation between the liquid flow pulsation and the hot air temperature is ignored, and the phase coupling relationship between the two cannot be identified, making it difficult for temperature adjustment to adapt to the instantaneous changes in the flow. Thirdly, the control of the outlet powder water content mainly uses simple threshold feedback, lacks quantitative analysis of the inhibitory relationship between the water content and the liquid flow pulsation, and it is difficult to predict the changes in the drying state in advance, often resulting in the water content of the product exceeding the standard.

[0005] The traditional control system has a fuzzy division of stages in the drying process, and cannot identify the current drying stage (such as the preheating stage, the constant-speed drying stage, and the falling-speed drying stage) according to real-time process characteristics, resulting in a lack of pertinence in the hot air temperature control curve and a large parameter fluctuation during stage transition. These problems together lead to insufficient stability of the sand-fixing dust suppressant product quality, high energy consumption in the production process, and increased frequency and difficulty of manual intervention, which restricts the improvement of production efficiency. Therefore, it is an urgent need in the industry to develop a control system that can accurately capture the dynamic characteristics of the drying process and achieve adaptive adjustment. SUMMARY

[0006] The present application aims to provide a PLC-based sand-fixing and dust-suppressing agent spray drying process control system to solve the problems in the background art.

[0007] To achieve the above-mentioned purpose, the present application provides a PLC-based sand-fixing and dust-suppressing agent spray drying process control system, which comprises:

[0008] The inlet material liquid flow signal, hot air temperature signal and outlet powder moisture content signal of the spray drying tower are collected, and the pulsating component and steady-state component are separated from the inlet material liquid flow signal;

[0009] The phase coupling characteristics of the pulsating component and the hot air temperature signal, and the suppression intensity characteristics of the outlet powder moisture content signal on the pulsating component are obtained;

[0010] According to the phase coupling characteristics and the suppression intensity characteristics, process characteristics representing drying stability are generated, and based on the process characteristics, drying stage identification probability distribution is calculated;

[0011] According to the drying stage identification probability distribution, the hot air temperature signal is compensated and corrected to decouple the target temperature control curve of the target drying stage from the hot air temperature signal.

[0012] Preferably, the phase coupling characteristics of the pulsating component and the hot air temperature signal, and the suppression intensity characteristics of the outlet powder moisture content signal on the pulsating component are obtained, comprising:

[0013] The hot air temperature fluctuation trajectory and the material liquid-temperature response lag coefficient in the pulsating period are calculated to obtain the phase coupling characteristics;

[0014] The moisture content-pulsating suppression correlation criterion is established, and the attenuation gradient and steady-state recovery rate of the pulsating component before and after the mutation of the outlet powder moisture content signal are analyzed to obtain the suppression intensity characteristics.

[0015] Preferably, a bimodal decomposition algorithm is used to separate the pulsating component and the steady-state component in the inlet material liquid flow signal, comprising:

[0016] A steady-state component main channel based on sparse coding is constructed, and the 0.01-0.1 Hz frequency band component is reserved through base function screening;

[0017] A pulsating component auxiliary channel based on transient feature extraction is established, and a nonlinear filter is used to amplitude demodulate the 0.5-2 Hz frequency band;

[0018] The signal independence of the main channel and the auxiliary channel is evaluated through cross-correlation analysis, and the decomposition threshold is dynamically adjusted to realize signal separation.

[0019] Preferably, the process feature is generated according to the phase coupling feature and the suppression intensity feature, comprising:

[0020] The temperature fluctuation trajectory in the phase coupling feature is fused with the attenuation gradient in the suppression intensity feature to generate a drying parameter coupling feature;

[0021] Based on the drying parameter coupling feature, the conditional probability density of different drying stages is calculated;

[0022] According to the matching degree of the conditional probability density distribution and the preset process database, a target drying stage identifier is output.

[0023] Preferably, the hot air temperature signal is compensated and corrected according to the drying stage identifier probability distribution, comprising:

[0024] The drying stage identifier probability distribution is converted into a hot air temperature compensation weight coefficient, and the original hot air temperature signal is weighted and reconstructed using the hot air temperature compensation weight coefficient;

[0025] The reconstructed signal is processed for time-frequency domain feature alignment to generate a target temperature control curve.

[0026] Preferably, after generating the target temperature control curve, the system further comprises:

[0027] The exhaust pressure fluctuation signal of the spray drying tower is monitored in real time;

[0028] The energy envelope correlation between the exhaust pressure fluctuation signal and the target temperature control curve is calculated;

[0029] When the energy envelope correlation is lower than a dynamic tolerance threshold, a process feature re-generation instruction is triggered.

[0030] Preferably, after the process feature re-generation instruction is triggered, the system further comprises:

[0031] Based on the updated outlet powder moisture content signal, the suppression intensity feature is recalculated;

[0032] The latest time sequence segment in the phase coupling feature is intercepted using a sliding time window;

[0033] The updated suppression intensity feature is fused with the intercepted phase coupling feature to generate an optimized process feature.

[0034] Preferably, after the optimized process feature is generated, the system further comprises:

[0035] The confidence of the optimized process feature in the drying stage identifier probability distribution is verified;

[0036] When the confidence exceeds the reliability threshold, the optimized process feature is adopted to cover the original process feature;

[0037] The target temperature control curve is updated according to the covered process feature.

[0038] Preferably, when the outlet powder moisture content signal is collected, the system further comprises:

[0039] The powder particle size distribution signal is synchronously acquired;

[0040] The joint probability density function of the powder particle size distribution signal and the outlet powder moisture content signal is established;

[0041] When the joint probability density function appears a multi-peak distribution, a pulsation component secondary separation mechanism is activated.

[0042] Preferably, the activated pulsation component secondary separation mechanism comprises:

[0043] The decomposition threshold is adjusted according to the peak distance of the joint probability density function;

[0044] The inlet feed liquid flow signal is re-decomposed by using an adaptive bandwidth algorithm, and the re-decomposed pulsation component is input into the phase coupling feature calculation process.

[0045] Compared with the prior art, the present application has the beneficial effects that:

[0046] The sand-fixing and dust-suppressing agent spray drying process control system based on PLC significantly improves the control efficiency of the drying process through an innovative signal processing and feature analysis method. In the signal acquisition and processing link, the system separates the inlet feed liquid flow signal into pulsation component and steady-state component, which breaks through the limitation of the overall processing of the flow signal in traditional control. The pulsation component can reflect the instantaneous fluctuation of the feed liquid flow, and the steady-state component reflects its long-term change trend. This subdivision processing enables the system to adopt different control strategies for signals with different characteristics, avoiding the insufficient adaptation of a single control logic to complex dynamic processes.

[0047] By acquiring the phase coupling feature of the pulsation component and the hot air temperature signal, the system can clearly identify the dynamic correlation of the two in the time dimension. The pulsation of the feed liquid flow and the fluctuation of the hot air temperature do not exist in isolation, and their phase relationship directly affects the heat and mass exchange efficiency. The extraction of the phase coupling feature enables the control system to grasp this internal relationship, so that the change rule of the flow pulsation can be referred to when adjusting the hot air temperature, making the temperature adjustment more in line with the actual process requirements and reducing the decrease in drying efficiency caused by the imbalance between the two.

[0048] The acquisition of the suppression intensity characteristic of the outlet powder moisture content signal to the pulsating component provides a response quantization basis of the drying effect to the dynamic interference for the system. Under different working conditions, the sensitivity of the moisture content to the pulsation of the feed liquid flow is different, and the suppression intensity characteristic can reflect this difference, so that the system can adjust the control strength according to the actual sensitivity during regulation, to avoid excessive regulation or insufficient regulation, thereby more accurately maintaining the stability of the moisture content.

[0049] The process characteristic generated based on the phase coupling characteristic and the suppression intensity characteristic can comprehensively represent the stability state of the drying process. The process characteristic converts the dispersed signal characteristics into an overall state description, providing a reliable basis for the identification of the drying stage. The drying stage identification probability distribution calculated based on the process characteristic realizes dynamic judgment of the current drying stage, overcoming the limitations of traditional fixed stage division, so that the system can perceive the stage change in real time, providing accurate stage basis for subsequent temperature control curve correction.

[0050] According to the drying stage identification probability distribution, the hot air temperature signal is compensated and corrected, which can effectively decouple the target temperature control curve of the target drying stage. This process makes the hot air temperature regulation no longer rely on fixed parameters, but dynamically adapt to real-time stage characteristics, ensuring that the hot air temperature can match the state of the feed liquid and the evaporation demand of water in different drying stages (such as preheating, constant speed drying, and speed reduction drying), reducing parameter fluctuations during stage conversion, improving the continuity and stability of the entire drying process, and thus improving the consistency of product quality. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 The timing diagram of the PLC-based sand fixation dust suppressant spray drying process control system described in the present application;

[0052] Figure 2 The flowchart for phase coupling and suppression intensity characteristic acquisition;

[0053] Figure 3 The flowchart for process characteristic generation;

[0054] Figure 4 The flowchart for exhaust pressure fluctuation monitoring and feedback. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0056] Please refer toFigure 1 The application provides a PLC-based sand-fixing and dust-suppressing agent spray drying process control system, which comprises:

[0057] By collecting the inlet material liquid flow signal, hot air temperature signal and outlet powder moisture content signal of the spray drying tower, accurate control of the drying process is realized. The inlet material liquid flow signal is processed by signal separation to extract the pulsating component and the steady-state component. By analyzing the phase coupling characteristics of the pulsating component and the hot air temperature signal, and the suppression intensity characteristics of the outlet powder moisture content signal on the pulsating component, a process characteristic representing drying stability is generated. Based on the process characteristic, the drying stage identification probability distribution is calculated, and the hot air temperature signal is compensated and corrected accordingly, finally decoupling the target temperature control curve of the target drying stage. The system can dynamically adapt to the parameter changes in the drying process, improving the drying efficiency and product quality.

[0058] Example 1: see Figure 2 In the PLC-based sand-fixing and dust-suppressing agent spray drying process control system, signal separation is one of the core links. The inlet material liquid flow signal contains high-frequency pulsating components and low-frequency steady-state components, which need to be effectively separated by a dual-mode decomposition algorithm. The construction of the steady-state component main channel adopts the sparse coding method, and the components in the frequency band of 0.01-0.1 Hz are reserved through the selection of basis functions, which can reflect the long-term trend changes of the liquid flow. The selection of basis functions is based on the energy distribution characteristics of the signal, which preferentially retains the components with higher contribution degree while suppressing noise interference. The auxiliary channel focuses on extracting the pulsating components in the frequency band of 0.5-2 Hz, and uses a nonlinear filter for amplitude demodulation to capture the instantaneous fluctuation characteristics in the flow signal. The design of the nonlinear filter combines envelope detection and dynamic threshold adjustment to ensure accurate extraction of high-frequency components.

[0059] During the signal separation process, the independence of the main channel and the auxiliary channel is evaluated by cross-correlation analysis. The cross-correlation coefficient of the two channel signals is calculated, and if it exceeds the preset threshold, the decomposition parameters, including the number of basis functions and the filter cutoff frequency, are dynamically adjusted. The decomposition threshold is adjusted using an iterative optimization strategy to gradually approach the optimal separation effect until the cross interference of the two channel signals is reduced to an acceptable range.

[0060] The acquisition of phase coupling characteristics depends on the time series analysis of the pulsating component and the hot air temperature signal. Within the pulsating period, the hot air temperature fluctuation trajectory is extracted by a sliding time window, and the window length is synchronized with the pulsating period. The extreme points and mean value offset of the temperature signal are used to describe its fluctuation characteristics, and the material liquid-temperature response lag coefficient is used to reflect the dynamic coupling relationship between them. The lag coefficient is calculated using the cross-correlation function, which determines the response time of the pulsating component to temperature changes by finding the time delay corresponding to the maximum correlation.

[0061] The analysis of the suppression intensity feature is based on the influence of the outlet powder moisture content signal on the pulsation component. The establishment of the moisture content-pulsation suppression correlation criterion is achieved by monitoring the change in the pulsation component before and after the moisture content mutation. The attenuation gradient is used to quantify the degree of suppression of the pulsation component by the moisture content change, and the slope is calculated using curve fitting to reflect the strength of the suppression effect. The steady-state recovery rate is used to evaluate the self-regulating ability of the system by measuring the time for the pulsation component amplitude to recover to the steady-state level.

[0062] The overall process of signal separation and feature extraction adopts a modular design, facilitating parameter adjustment and algorithm optimization. The separation accuracy of the steady-state component and the pulsation component directly affects the reliability of the subsequent phase coupling and suppression intensity features, so real-time monitoring and feedback mechanisms are needed to ensure the separation effect. The dynamic update frequency of the phase coupling feature and the suppression intensity feature is matched with the real-time requirements of the drying process to ensure that the control system can quickly respond to process changes.

[0063] In the signal processing process, noise suppression cannot be ignored. The inlet liquid flow signal may be affected by sensor noise or environmental interference, so pre-processing is needed before bimodal decomposition. The pre-processing methods include moving average filtering and outlier rejection to improve signal quality. The collection of hot air temperature signals also needs to consider noise factors, using temperature sensors with strong anti-interference ability and combining digital filtering techniques to eliminate high-frequency noise.

[0064] The calculation of the phase coupling feature not only focuses on the temperature fluctuations within a single pulsation cycle, but also analyzes the statistical properties over multiple cycles. By calculating the standard deviation and mean drift of the temperature fluctuation trajectory, the stability of the hot air temperature is evaluated. The dynamic change of the liquid-temperature response lag coefficient can reflect the nonlinear characteristics of the drying process. When the lag time significantly increases, it may indicate that the system has potential instability risks.

[0065] The quantification of the suppression intensity feature needs to be combined with the actual needs of the drying process. The calculation of the attenuation gradient uses piecewise linear fitting to adapt to the dynamic characteristics of different drying stages. The evaluation of the steady-state recovery rate needs to consider the sampling frequency of the moisture content signal to ensure the accuracy of the time measurement. The establishment of the moisture content-pulsation suppression correlation criterion is based on historical data training, and the criterion parameters are optimized through machine learning methods to improve their adaptability to different working conditions.

[0066] The real-time performance of signal separation and feature extraction is achieved through parallel computing technology. The extraction process of the steady-state component and the pulsation component runs independently, reducing algorithm delay. The calculation of the phase coupling feature and the suppression intensity feature uses a pipeline processing mode to ensure the continuity of data processing. The hardware configuration of the control system needs to meet the real-time computing requirements, including high-performance PLC and dedicated signal processing modules.

[0067] The dynamic characteristics of the drying process require that the signal separation and feature extraction have strong robustness. When the inlet liquid flow fluctuates greatly, the dual-mode decomposition algorithm needs to adaptively adjust the parameters to avoid signal distortion. The calculation of the phase coupling feature needs to introduce a fault-tolerant mechanism to output reasonable results when the temperature signal is abnormal. The evaluation of the suppression intensity feature needs to combine multi-dimensional data to avoid misjudgment caused by a single indicator.

[0068] In the drying process control system, the generation and temperature compensation correction of process features are core links. The construction of process features is based on the fusion processing of phase coupling features and suppression intensity features. The temperature fluctuation trajectory in the phase coupling feature reflects the periodic variation law of the hot air temperature, and the decay gradient in the suppression intensity feature reflects the suppression ability of the moisture content to the liquid pulse. The two features are aligned in dimension through vectorization processing, and the common mode is extracted by principal component analysis to generate a drying parameter coupling feature vector. The vector contains the correlation information of hot air temperature stability and pulse suppression ability, representing the comprehensive dynamic characteristics of the drying system. Figure 3 The conditional probability density calculation of the drying stage is based on the historical process database. The database stores feature vector samples of typical drying stages, including the parameter distribution of the constant speed drying period, the falling speed drying period and the equilibrium period. The current drying parameter coupling feature is input into the Bayesian classifier to calculate the matching probability with each drying stage sample. The probability density estimation uses kernel density estimation method to calculate the similarity distribution in the feature space through Gaussian kernel function. The output of the drying stage identification is based on the maximum a posteriori probability principle. When there is a significant peak in the probability distribution, the corresponding stage identification is directly output; if it is a multi-peak distribution, the most likely stage is selected according to the time sequence continuity constraint.

[0069] The execution of temperature compensation correction converts the drying stage identification probability distribution into operable weight coefficients. The mapping of probability value to weight uses S-shaped function transformation, and the stage with probability higher than the preset threshold obtains linearly growing weight coefficient, and the stage with probability lower than the threshold tends to zero. The weight coefficients act on the dimensions of the original hot air temperature signal, including time domain mean shift adjustment, frequency domain energy distribution correction and phase offset compensation. The original signal is superimposed in three-dimensional space (time, frequency, amplitude) with weight, and the reconstruction process introduces orthogonal decomposition to eliminate interference between dimensions.

[0070]

[0071] ​The time-frequency domain feature alignment of the reconstructed signal is a key step in generating the target temperature control curve. The short-time Fourier transform converts the continuous temperature signal into a time-frequency matrix, and the dynamic time warping algorithm corrects the matrix. The correction process is based on the reference time-frequency template of the target drying stage, and by stretching or compressing the local time-frequency spectrum, the instantaneous frequency distribution of the reconstructed signal is consistent with the template. The elimination of frequency aliasing uses a multiple filter bank design, which adaptively selects the filter bandwidth according to the local signal-to-noise ratio of the signal, retains the effective frequency band and suppresses the high-frequency noise.

[0072] The generation of the target temperature control curve uses a phased output strategy. When the system is in the stable drying stage, the curve is directly output as a continuous smooth function; if a transition stage is detected, a segmented continuous curve is generated and a gradual transition zone is set at the stage switching point. The time domain accuracy of the curve is guaranteed by the interpolation algorithm, which inserts a cubic spline function between the sampling points to maintain high-order derivative continuity. The frequency domain characteristics are achieved through least squares spectral fitting, ensuring that the power spectral density distribution meets the energy distribution requirements of the target drying stage.

[0073] The update mechanism of the drying parameter coupling characteristics has adaptive ability. When the newly collected temperature fluctuation trajectory and the decay gradient data are input, the system dynamically adjusts the covariance matrix of the principal component analysis and recalculates the projection direction of the characteristic vector. The prior distribution of the Bayesian classifier is also updated as the historical database expands, using an incremental learning method to avoid full data retraining. This mechanism ensures that the process characteristics always maintain the best representation of the current working conditions.

[0074] The weighting process of signal reconstruction introduces an anti-interference design. When transient interference pulses are detected in the original temperature signal, the pulse suppression algorithm is automatically enabled to eliminate the influence of abnormal values through median filtering of neighboring sampling points. The frequency domain compensation section sets a band-stop filter bank to selectively attenuate the known equipment vibration frequency range, avoiding the pollution of mechanical vibration to the temperature signal.

[0075] The template matching in the time-frequency alignment process uses a multi-scale strategy. In the early stage of drying, the temperature changes rapidly, and a short-time window high-resolution template is used to capture rapid transients; in the later stage of drying, a long-time window template is switched to enhance the overall stability. The adaptive step size limit is set for the path constraint of dynamic time warping to avoid global distortion caused by local distortion.

[0076] The system implementation layer uses a hierarchical processing architecture. The process characteristic generation layer is deployed on a dedicated signal processor, and the temperature compensation correction layer runs on the PLC control core. The two layers exchange data through shared memory, with a time synchronization accuracy of microseconds. The output of the target temperature control curve uses a double buffer mechanism to ensure continuous and stable transmission of control commands to the actuator, avoiding control interruption caused by calculation delay.

[0077] The adjustment of process control parameters is included in the closed-loop management. The actual temperature value and the difference data between the target curve fed back by the actuator are used to optimize the weight distribution strategy of the reconstruction algorithm. When the difference exceeds the limit, the feature fusion parameter is recalibrated, and the contribution rate threshold of principal component analysis is adjusted to optimize the representation ability of the coupling features of the drying parameters. This mechanism enables the control system to have the ability of continuous self-optimization.

[0078] The application scenario adaptability design of the temperature control curve considers different equipment characteristics. According to the differences in heat transfer characteristics of different types of spray drying towers, the system pre-stores a basic time-frequency template library of various equipment types. In the initialization stage, the corresponding template is automatically loaded according to the equipment identification, and the template parameters are dynamically corrected based on the actual heat exchange efficiency during the running process. This design improves the portability of the system in different drying equipment.

[0079] Through the fine feature fusion and intelligent signal reconstruction technology, accurate identification in the drying stage and dynamic optimization of temperature control are realized. The scientific construction of process features and the combination of Bayesian decision mechanism provide accurate stage identification for different drying states. Multi-dimensional signal compensation and time-frequency domain joint correction technology ensure that the target temperature control curve meets the requirements of the process nature and has engineering realizability. The whole processing chain adopts a fully digital process, and the hierarchical architecture and adaptive algorithm ensure the practicality and reliability in industrial environment.

[0080] Example 3: see Figure 4 In the spray drying process control system, after the target temperature control curve is generated, it needs to be continuously monitored for its matching degree with the actual working condition. The exhaust pressure fluctuation signal, as an important parameter reflecting the internal state of the drying tower, is collected in real time through a high-frequency pressure sensor, and the sampling frequency is not less than 100 Hz to ensure the integrity of the transient characteristics. The signal is immediately pre-processed after being collected, including power frequency interference elimination and baseline drift correction, and a zero-phase digital filter bank is used to realize distortionless filtering. The pre-processed pressure signal is extracted by Hilbert transform to calculate the instantaneous amplitude envelope line. The envelope line reflects the macro fluctuation trend of the exhaust pressure, and its mathematical expression is:

[0081]

[0082] Wherein: represents the envelope amplitude of the exhaust pressure at time t, is the original pressure signal, is the corresponding Hilbert transform result. In the envelope calculation process, sliding average smoothing processing is introduced, and the window width is an integer multiple of the basic period of the target temperature control curve to avoid spectrum leakage.

[0083] The calculation of energy envelope correlation uses an improved cross-correlation algorithm. The target temperature control curve The envelope is obtained by the same Hilbert transform The dynamic correlation coefficient between them is calculated by:

[0084]

[0085] where: is the time delay parameter, is the number of sampling points within the analysis window, and represent the mean values of the corresponding envelopes, respectively. The correlation calculation adopts an overlapping sliding window strategy, and the window length is dynamically adjusted according to the current drying stage: a longer window is used in the constant-speed drying period to enhance stability, and the window is shortened in the transition period to improve sensitivity. The dynamic tolerance threshold is set based on the drying stage identification probability distribution, and the probability value is converted into a threshold boundary through a nonlinear mapping function. When , the process feature regeneration instruction is triggered.

[0086] The process feature updating process adopts an incremental adjustment strategy. The sudden change detection of the outlet powder moisture content signal is realized through the CUSUM control chart, and when the cumulative deviation exceeds the statistical control limit, the suppression intensity feature recalculation is immediately started. The generation of the new suppression intensity feature uses a sliding time window to intercept the latest 30 pulsation period data, and reanalyzes the decay gradient and steady-state recovery rate. The starting point of the sliding window is automatically aligned according to the moisture content mutation time, ensuring that the analysis data contains a complete transition process.

[0087] The local update of the phase coupling feature focuses on the latest time sequence segment. By dynamically time warping algorithm matching the historical temperature fluctuation trajectory and the current segment, the most representative 5 complete fluctuation periods are identified for feature extraction. The updated phase coupling feature and the suppression intensity feature are fused to generate the optimized process feature through feature-level fusion: the feature component with the newer timestamp obtains a higher weight. The fusion process adopts a tensor decomposition method, which organizes the two types of features into a three-dimensional tensor and extracts the core feature matrix through Tucker decomposition.

[0088] The verification of the optimized process feature uses the Monte Carlo sampling method. 1000 groups of perturbation samples are randomly generated in the feature space, and the Mahalanobis distance distribution between them and the original process feature is calculated. When the confidence of the optimized feature meets:

[0089]

[0090] where: is the total number of samples, is the perturbation sample, is the optimized feature vector, is a similarity radius, is an indicator function. When the confidence exceeds a preset reliability threshold, the new feature vector replaces the corresponding entry in the original feature library, and a control curve update process is triggered.

[0091] The auxiliary analysis of the exhaust pressure signal also includes spectrum feature monitoring. The power spectral density of the pressure signal is calculated by the Welch method, and the change in the energy proportion of the 5-15 Hz frequency band is monitored. An abnormal increase in the energy of this frequency band often indicates a turbulent flow field in the tower, so even if the envelope correlation meets the standard, preventive feature updating needs to be started. The spectrum monitoring result and the envelope correlation analysis form a double-checking mechanism to improve the system's ability to identify hidden abnormal conditions.

[0092] The implementation level of the control system adopts an event-driven architecture. Process feature update events are divided into three priorities: emergency update triggered by envelope correlation, preventive update triggered by spectrum anomaly, and regular maintenance update triggered by timing. Different priority events correspond to different computing resource allocation strategies. Emergency updates can be interrupted and executed immediately, while regular updates are processed in batches using idle computing resources. The event queue management uses a pre-emptive scheduling algorithm based on time stamps to ensure that update requests with high time efficiency requirements are responded to quickly.

[0093] The storage and retrieval of historical data are optimized using a time series database. All process feature versions and corresponding operating parameters are stored in chronological order, and a multi-dimensional index structure is established. When the performance of a feature under a specific operating condition needs to be traced back, a composite query can be used to quickly locate the historical record. The database uses a ring buffer mechanism to automatically discard low-value data that exceeds the retention period, maintaining storage efficiency.

[0094] The system also has an abnormal recovery mechanism. When the confidence does not meet the standard after continuous feature updates, it automatically reverts to the last stable version and increases the sensor verification frequency. The rollback operation is accompanied by detailed diagnostic log records, including signal quality indicators, feature matching degree curves, and environmental parameter snapshots, providing complete data support for subsequent offline analysis. This mechanism effectively prevents control instability caused by sensor failure or extreme operating conditions.

[0095] The timing control of the entire dynamic adjustment process is accurate to the millisecond level. The end-to-end delay from exhaust pressure signal collection to completion of control curve update is controlled within 200 ms, and time-consuming operations such as Hilbert transform and correlation calculation are accelerated by FPGA. Real-time measures include memory pre-allocation, interrupt priority optimization, and computing pipeline design, ensuring that the hard real-time requirements of the control period are still met even under high load.

[0096] The linkage through the feature version management is achieved in Example 2. After each temperature control curve update, the system automatically records the process feature version number followed and embeds the feature fingerprint in the control parameters. When the Bayesian classifier of Example 2 needs to update the prior distribution, the specific feature generation condition can be traced back through the fingerprint to ensure data consistency between modules. The version management adopts a block chain type hash chain structure, and any feature modification will generate a new tamper-proof record.

[0097] This implementation enables the control system to continuously adapt to changing drying conditions through multi-level dynamic monitoring and intelligent adjustment mechanisms. Correlation analysis based on energy envelope captures macro matching, combined with frequency spectrum monitoring to identify micro abnormalities, forming a complete condition evaluation system. The incremental feature update strategy ensures timeliness while maintaining system stability, and the abnormal recovery mechanism provides safety protection for long-term reliable operation. The entire scheme deeply integrates signal processing, feature engineering, and control theory to form an adaptive industrial solution.

[0098] Example 4: In the spray drying process control system, optimizing the confidence verification of process features is a key link to ensure control reliability. When the system generates new optimized process features, it needs to pass through a strict statistical verification process to confirm its effectiveness. Taking the feature update in a real drying process as an example, the system collected the outlet powder moisture content signal, which suddenly increased from the initial 5.2% to 7.8%, and the powder particle size distribution signal showed that the D50 value fluctuated from 45 μm to 52 μm. These changes triggered the recalculation of the joint probability density function, which showed a clear bimodal distribution characteristic, indicating that there were two possible operating states in the drying process.

[0099] Table 1: The distribution characteristic analysis results of the joint probability density function in this event are as follows.

[0100]

[0101] According to the data in the table, the system automatically activates the secondary separation mechanism of the pulsation component. The adjustment of the decomposition threshold considers the peak distance of each parameter and uses a weighted average algorithm to determine the final adjustment amount. For the re-decomposition of the inlet liquid flow signal, the adaptive bandwidth algorithm selects the optimal analysis window according to the current signal-to-noise ratio, effectively suppressing noise interference while preserving the true pulsation component. The re-decomposed pulsation component is sent to the phase coupling feature calculation process, at which time the system detects that the liquid-temperature response lag time has extended from the original 9 seconds to 14 seconds, reflecting changes in the heat transfer efficiency inside the drying tower.

[0102] The confidence verification stage uses the Markov Chain Monte Carlo method for probability evaluation. The system generates a large number of simulation samples in the feature space, and statistically optimizes the similarity distribution between the process features and the historical stable features. When the evaluation result shows that the confidence of the new features reaches 92% (exceeding the pre-set reliability threshold of 85%), the system performs feature coverage operation. The coverage process uses a gradual replacement strategy: the first round of updates replaces 30% of the feature components, and after two control period verifications show no abnormalities, the remaining part is replaced. This phased updating method can avoid control oscillation caused by feature mutations.

[0103] The synchronous acquisition of powder particle size distribution signals uses the laser diffraction principle, and a complete particle size distribution spectrum is obtained every 30 seconds. When the joint probability density function of particle size distribution and moisture content is detected to have a multi-peak distribution, the system automatically enhances the signal sampling frequency to once per second and starts the abnormal mode recording function. The data recorded during a certain run shows that when the D90 particle size jumps from 68 μm to 82 μm, the joint probability density function evolves from a single peak to a double peak and then to a triple peak within 15 minutes. This nonlinear change prompts the system to increase the decomposition threshold adjustment amount to 1.8 times the regular value.

[0104] The feature version management uses a tree structure storage scheme. Each feature update generates a new branch node, recording the complete modification trajectory and decision basis. When historical state backtracking is needed, the system can quickly locate the feature version at a specific time point and its associated operating parameters. During the restart process after a device maintenance, the system automatically loads the feature sets of the last three stable versions, compares their matching degrees with the current signal through parallel calculation, and finally selects the version with the highest matching degree as the initialization reference.

[0105] The dynamic visualization interface of the probability distribution of the drying stage helps the operator understand the system decision. The interface uses a heat map to show the time sequence changes of the probabilities of different stages, and when the optimized process features are verified, the probability blocks of the corresponding stages will be highlighted in a pulse manner. The operator can click on any time point to view the detailed feature composition, including the fitting curve of the temperature fluctuation trajectory, the quantitative indicators of the pulsation suppression intensity, etc. This transparent display method helps to establish a human-machine trust relationship.

[0106] The abnormal handling mechanism sets a multi-level response strategy. For borderline cases with a confidence between 80% and 85%, the system does not immediately perform feature coverage, but starts an enhanced monitoring mode: shortens the control period to 1 / 3 of the original value, increases the sensor checking frequency, and records more detailed operation logs. Only when the confidence is consistently higher than the threshold for three consecutive enhanced periods, the update validity is confirmed. In a certain borderline case, the system successfully identified the false feature change caused by temporary drift of the pressure sensor through this method, avoiding unnecessary control parameter adjustment.

[0107] The linkage with Example 3 is achieved through an event bus. When the envelope correlation analysis of Example 3 detects an anomaly, a feature check request is issued to the bus. Example 4 immediately suspends the current non-critical task upon receiving the request and prioritizes the abnormal check requirement. The data exchange between the two modules uses memory-mapped files to ensure the rapid transfer of large amounts of feature data. In one linkage processing, this mechanism takes only 150 milliseconds from the correlation anomaly alarm to the completion of feature updating, far lower than the 500-millisecond interval of the regular control cycle.

[0108] The version compatibility design of control parameters ensures continuous system operation. When a new optimized process feature is confirmed to be effective, the system automatically generates a corresponding control parameter conversion matrix, ensuring smooth transition of control instructions under the old and new feature systems. The conversion matrix is based on feature space mapping theory and preserves the continuity of key control characteristics. During a major process adjustment, this mechanism successfully achieved seamless switching between different drying formulations, with the standard deviation of product moisture content maintained within 0.3% during the process.

[0109] Data persistence uses a layered storage architecture. High-frequency collected raw signals are saved in the cache area, retaining the last 2 hours of data; feature vectors and process parameters are stored in the medium-speed database, retaining 30 days of records; and system configurations and model parameters are written to non-volatile memory. This architecture optimizes storage efficiency while ensuring data traceability. When long-term analysis is needed, the system supports queries of historical data in various dimensions such as time range, process stage, or abnormal type.

[0110] The system also has an expert intervention interface. When the automatic verification process fails to reach a definitive conclusion multiple times, a feature report to be reviewed is generated for manual confirmation. The report includes a time sequence animation of feature changes, a statistical summary of key parameters, and the system's recommended treatment plan. Experts can adjust feature weights by dragging or manually mark abnormal data segments. These intervention information will be learned by the system and used to improve subsequent automatic decision-making algorithms.

[0111] The multi-peak distribution phenomenon of the joint probability density function in the spray drying process indicates that the material drying state has non-stationary transition characteristics. At this time, the system activates the secondary separation mechanism of the pulsation component, and enhances the adaptability to abnormal working conditions by dynamically adjusting the separation parameters. The identification of the multi-peak distribution feature relies on the change of the joint probability density surface of the moisture content and the particle size distribution. When more than two significant probability concentration areas are detected, the multi-dimensional Euclidean distance between the peak tops is automatically calculated as the basis for adjusting the decomposition threshold. The measurement of the peak distance is not limited to a single dimension of moisture content or particle size, but a high-dimensional feature space composed of multiple parameters such as temperature fluctuation amplitude and pulsation suppression response time. The principal coordinate analysis method is used to reduce the dimension to obtain the distance scalar with the most discriminant degree. The scalar is normalized to map to the adjustment amount of the decomposition threshold, and the mapping function is designed as a piecewise linear relationship. In the small peak distance interval, the adjustment is slow, and when the distance exceeds the critical value, the threshold is greatly increased.

[0112] The re-decomposition process of the inlet liquid flow signal introduces an adaptive bandwidth technology. Instead of using a pre-set fixed frequency band, the signal frequency band is dynamically configured based on the local characteristics of the current signal. The local signal-to-noise ratio evaluation is achieved by comparing the energy ratio of the target pulsation component and the environmental noise of the signal in the 0.5-2Hz frequency band. When the ratio is below a certain level, the bandwidth is automatically expanded to capture wider frequency domain features, otherwise the bandwidth is reduced to improve the resolution. The frequency band boundary setting has hysteresis characteristics to avoid frequent switching near the critical value. The wavelet packet transform is responsible for the adaptive segmentation of the frequency band as the core algorithm, which optimizes the selection of the optimal sub-band combination through a tree structure to ensure that the pulsation main energy is concentrated in a specific sub-band. The basis function library contains 12 mother wavelets with different time-frequency localization characteristics, and the most suitable basis function type is automatically matched according to the signal instantaneous frequency content.

[0113] The re-decomposed pulsation component undergoes feature enhancement processing. Morphological filters are used to smooth burr interference, and the size of the structural element is adjusted synchronously with the pulsation period. The pulsation envelope extraction combined with the Teager energy operator enhances the transient component identification accuracy and significantly improves the sensitivity to sudden fluctuations in the liquid. The processing flow also includes a false pulsation discrimination section, which verifies the authenticity of the pulsation by correlating multiple sensor data: if the liquid flow pulsation does not cause simultaneous temperature or pressure changes, it is marked as an interference signal and removed.

[0114] The updated pulsation component input phase coupling feature calculation process. The time delay analysis of temperature signal and pulsation component uses an improved instantaneous phase difference method to determine the delay time by cross-correlation detection of signal phase angle. This method has better robustness for non-steady state signals compared to traditional cross-correlation techniques. The coupling strength calculation introduces the concept of multi-scale analysis, which evaluates the coupling characteristics on the time scales of seconds, tens of seconds, and minutes, and finally generates a comprehensive coupling index by fusion. The calculation process monitors the phase drift phenomenon in real time, and when the phase difference changes by more than the limit in three consecutive pulsation periods, the sampling sequence is automatically restarted.

[0115] The system sets multiple protection strategies for the reseparation mechanism. The decomposition threshold adjustment amplitude is set with an upper limit value to avoid excessive response to temporary disturbances and cause control instability. A two-minute stable observation period is started after each threshold modification, during which further adjustments are suspended and system responses are recorded. For sustained multi-peak distribution caused by equipment failure, when stable pulsation characteristics cannot be obtained after five consecutive re-decompositions, a system-level maintenance alarm is triggered and the system automatically switches to fixed parameter mode operation.

[0116] The feature reconstruction quality control module verifies the re-decomposition effect. The improvement degree of signal order is evaluated by calculating the change of signal information entropy before and after re-decomposition, and the orthogonality index of pulsation component and steady component is checked. The quality control report is embedded in the feature vector transmission to the downstream module for decision algorithm weighted reference. When the quality evaluation does not meet the standard, the backup strategy is automatically enabled: switch to the historical optimal decomposition parameter combination under the same working condition, and make limited adaptive adjustment combined with the current signal characteristics.

[0117] The signal processing flow implements resource optimization configuration. The re-decomposition task is marked as a high-priority calculation job, which can preempt the processing resources of regular analysis tasks. The memory management system implements a double-copy mechanism for signal buffer area to ensure the integrity of the original data during re-decomposition. The core part of the algorithm uses fixed-point acceleration calculation to compress the calculation delay to hundreds of milliseconds while ensuring accuracy.

[0118] The interface design with the drying equipment control system considers the real-time requirements. The re-decomposition results are transmitted through a dedicated data channel, and the update package is pushed immediately after each round of analysis is completed, without waiting for the complete control cycle. The control parameter modification uses an incremental update protocol, only transmitting the difference to reduce bandwidth pressure. The interface includes a data verification mechanism, which uses a cyclic redundancy check code to ensure transmission integrity, and automatically requests retransmission of the latest valid data package when an error is found.

[0119] The commissioning support system records a complete re-separation event log. The stored content contains the snapshot of multi-peak distribution feature at the triggering moment, the decision chain of resolution threshold adjustment, the key intermediate results of re-resolution process and the final quality assessment report. The log adopts a structured storage scheme, supporting playback analysis process along the time axis. The system maintenance interface provides a channel for manual intervention, allowing engineers to manually fine-tune bandwidth selection parameters or directly inject specific resolution thresholds for test verification when necessary.

[0120] This embodiment establishes a systematic response mechanism for the unstable state of the complex drying process. The intelligent detection based on the joint probability multi-peak distribution accurately identifies the abnormal starting point of the process, and the dynamic resolution threshold adjustment strategy realizes fine signal separation control. The adaptive bandwidth technology and multi-base function configuration guarantee the re-resolution quality, and the closed-loop feature inspection mechanism maintains the reliability of the whole system. The whole process not only maintains the efficiency of automatic decision-making, but also reserves reasonable space for manual intervention, forming a practical solution in the industrial scene. By accurately reconstructing the pulsation characteristics and updating the coupling analysis, the system continuously maintains the adaptability and stability of the drying process control.

[0121] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0122] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application as defined by the appended claims and their equivalents.

Claims

1. A PLC based sand-fixing dust suppressant spray drying process control system characterized by, The method comprises the following steps: acquiring an inlet liquid flow signal, a hot air temperature signal and an outlet powder moisture content signal of a spray drying tower, and separating a pulsation component and a steady-state component in the inlet liquid flow signal; obtaining a phase coupling feature of the pulsation component and the hot air temperature signal, and a suppression intensity feature of the outlet powder moisture content signal on the pulsation component; generating a process feature representing drying stability according to the phase coupling feature and the suppression intensity feature, and calculating a drying stage identification probability distribution based on the process feature; compensating and correcting the hot air temperature signal according to the drying stage identification probability distribution, so as to decouple a target temperature control curve of a target drying stage from the hot air temperature signal; the obtaining of the phase coupling feature of the pulsation component and the hot air temperature signal, and the suppression intensity feature of the outlet powder moisture content signal on the pulsation component comprises: calculating a hot air temperature fluctuation trajectory in a pulsation period and a liquid-temperature response lag coefficient to obtain the phase coupling feature; establishing a moisture content-pulsation suppression correlation criterion, and analyzing an attenuation gradient and a steady-state recovery rate of the pulsation component before and after a sudden change in the outlet powder moisture content signal to obtain the suppression intensity feature; generating the process feature according to the phase coupling feature and the suppression intensity feature comprises: fusing the temperature fluctuation trajectory in the phase coupling feature and the attenuation gradient in the suppression intensity feature to generate a drying parameter coupling feature; calculating a conditional probability density of different drying stages based on the drying parameter coupling feature; outputting a target drying stage identification according to a matching degree of the conditional probability density distribution and a preset process database; the compensating and correcting of the hot air temperature signal according to the drying stage identification probability distribution comprises: converting the drying stage identification probability distribution into a hot air temperature compensation weight coefficient, and weighting and reconstructing an original hot air temperature signal by using the hot air temperature compensation weight coefficient; performing time-frequency domain feature alignment processing on the reconstructed signal to generate a target temperature control curve.

2. The PLC based sand dune dust suppressant spray drying process control system of claim 1, wherein, The pulsation component and the steady-state component are separated in the inlet liquid flow signal by using a dual-mode decomposition algorithm, which comprises: constructing a steady-state component main channel based on sparse coding, and retaining components in a 0.01-0.1 Hz frequency band by base function screening; establishing a pulsation component auxiliary channel based on transient feature extraction, and performing amplitude demodulation on a 0.5-2 Hz frequency band by using a nonlinear filter; evaluating signal independence of the main channel and the auxiliary channel by cross-correlation analysis, and dynamically adjusting a decomposition threshold to realize signal separation.

3. The PLC based sand dune dust suppressant spray drying process control system of claim 1, wherein, After the target temperature control curve is generated, the method further comprises: real-time monitoring of an exhaust pressure fluctuation signal of the spray drying tower; calculating an energy envelope correlation of the exhaust pressure fluctuation signal and the target temperature control curve; when the energy envelope correlation is lower than a dynamic tolerance threshold, triggering a process feature regeneration instruction.

4. The PLC based sand dune dust suppressant spray drying process control system of claim 3, wherein, After the process feature regeneration instruction is triggered, the method further comprises: re-calculating the suppression intensity feature based on an updated outlet powder moisture content signal; using a sliding time window to intercept a latest time sequence segment in the phase coupling feature; fusing the updated suppression intensity feature and the intercepted phase coupling feature to generate an optimized process feature.

5. The PLC based sand dune dust suppressant spray drying process control system of claim 4 wherein, The generating the optimized process feature further comprises: verifying the confidence of the optimized process feature in the probability distribution of the drying phase; when the confidence exceeds a reliability threshold, adopting the optimized process feature to cover the original process feature; updating the target temperature control curve according to the covered process feature.

6. The PLC based sand dune dust suppressant spray drying process control system of claim 1, wherein, When the outlet powder moisture content signal is collected, further comprising: synchronously acquiring a powder particle size distribution signal; establishing a joint probability density function of the powder particle size distribution signal and the outlet powder moisture content signal; when the joint probability density function appears as a multimodal distribution, activating a pulsation component secondary separation mechanism.

7. The PLC-based sand dune dust suppressant spray drying process control system of claim 6, wherein, The activated pulsation component secondary separation mechanism comprises: adjusting a decomposition threshold according to the peak distance of the joint probability density function; adopting an adaptive bandwidth algorithm to re-decompose the inlet liquid flow signal, and inputting the re-decomposed pulsation component into the phase coupling feature calculation process.

Images