PLC automatic control system for drilling bentonite production

Abstract

The application relates to the technical field of industrial automation control, and particularly discloses a PLC automatic control system for drilling bentonite production, which comprises a real-time acquisition module for acquiring operation current spectrum data of a material conveying stage and temperature gradient change rate and stirring torque dynamic data in the initial reaction stage; a feature extraction module for extracting dominant feature components of raw material reaction activity through nonlinear mapping and feature space transformation; a parameter generation module for establishing a conversion relationship from raw material characteristics to process parameters based on dynamic correlation analysis and generating an optimized process parameter set in real time; a process control module for dynamically adjusting heating power and additive flow through a predictive control mechanism according to the optimized parameter set; and a self-adaptive optimization module for correcting feature space transformation parameters according to quality index analysis results; the application realizes closed-loop control from raw material characteristic identification to self-adaptive generation of process parameters, and solves the problem of unstable product quality caused by raw material characteristic fluctuation.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, specifically to a PLC automatic control system for drilling bentonite production. Background Technology

[0002] Drilling bentonite is a key material in oil drilling, and its quality directly affects the performance of drilling fluids and wellbore stability. In traditional production processes, sodium modification is a core step, and the reaction process control largely relies on manual experience and semi-automated equipment. Existing technologies typically employ simple programmable logic controllers (PLCs) or instrument monitoring to achieve point-to-point adjustment of single parameters such as temperature and stirring speed. However, this control method has significant limitations: First, it cannot effectively address the fluctuations in the physicochemical properties of bentonite raw materials from different sources. Due to the lack of online identification and adaptive adjustment mechanisms for raw material characteristics, fixed process parameters cannot guarantee that different batches of raw materials will achieve optimal reaction results, leading to instability in key indicators of the final product such as pulping rate and colloid content.

[0003] This invention solves the core technical problem in existing drilling bentonite production: due to the inherent differences in mineral composition, particle size and reactivity of raw materials from different mineral sources and batches, the traditional PLC control system can only execute fixed process parameters and cannot perceive and adapt to such raw material variations, resulting in unstable sodium modification reaction process and large fluctuations in key performance indicators of the final product. Summary of the Invention

[0004] The purpose of this invention is to provide a PLC automatic control system for drilling bentonite production to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: A PLC automatic control system for drilling bentonite production includes: The real-time acquisition module is used to acquire the operating current spectrum data of the conveying power unit during the material conveying stage in the production of drilling bentonite, as well as the temperature gradient change rate and stirring torque dynamic data of the material in the reaction vessel during the reaction initialization stage. The feature extraction module is used to process the collected multidimensional state parameters by combining nonlinear mapping and feature space transformation in order to extract the dominant feature components that characterize the reactivity of the raw materials. The parameter generation module is used to perform dynamic correlation analysis on the dominant feature components, establish the transformation relationship from raw material characteristics to process parameters, and generate an optimized set of process parameters for the current production batch in real time based on the transformation relationship. The process control module is used to dynamically adjust the reaction process based on the optimized process parameter set through a predictive control mechanism, including adjusting the power output of the heating system and the flow distribution of the additive dosing system based on the feedforward adjustment of the physical evolution trajectory. The adaptive optimization module is used to collect product quality indicators after production is completed, perform correlation analysis between the quality indicators and the dominant feature components and control effects, and adaptively adjust the parameter settings of the feature space transformation based on the analysis results. The PLC controller coordinates the timing of each module to achieve closed-loop control from raw material characteristic identification to adaptive generation of process parameters.

[0006] As a further aspect of the present invention: the extraction of the dominant characteristic components characterizing the reactivity of the raw materials specifically includes: The collected multidimensional state parameters are projected into a high-dimensional space to map the operating current spectrum data, temperature gradient change rate and stirring torque dynamic data into a high-dimensional feature vector of a unified scale. Principal component screening is performed on high-dimensional feature vectors. By calculating the variance contribution rate of each feature component, the top few principal components whose contribution rates exceed a set threshold are selected as the feature principal set. The feature set is concatenated with the real-time acquired acid and alkalinity parameters of the material, and the concatenated feature sequence is normalized to generate the final dominant feature component characterizing the reactivity of the raw material.

[0007] As a further aspect of the present invention: the screening process of the feature set is as follows: Construct the covariance matrix of high-dimensional eigenvectors, calculate the variance contribution rate corresponding to each eigenvalue of the covariance matrix, and arrange them in descending order to form a contribution rate sequence; Calculate the cumulative contribution rate of the contribution rate sequence. When five consecutive increments of the cumulative contribution rate are less than a set threshold, use the feature value position corresponding to the current cumulative contribution rate as the filtering threshold. Select all principal components whose variance contribution rate is greater than the screening threshold, and form the feature principal set by their corresponding feature vectors.

[0008] As a further aspect of the present invention: the real-time generation of an optimized process parameter set for the current production batch based on the transformation relationship specifically includes: A dynamic correlation matrix is ​​constructed by linking the dominant feature components with the corresponding time-series process data in the historical production cycle. A sliding time window was used to perform local correlation analysis on the dynamic correlation matrix to calculate the correlation strength between the characteristic components and process parameters at different production stages. Based on the correlation strength, a nonlinear mapping relationship between characteristic components and process parameters is established, and the optimal combination of process parameters is determined by solving the extreme points of the nonlinear mapping relationship. The optimized process parameter set is calculated and output in real time based on the dominant characteristic components of the current production batch.

[0009] As a further aspect of the present invention: the calculation of the correlation strength between the characteristic components and process parameters at different production stages specifically includes: Each production stage is divided into several equally long analysis periods, and a corresponding dynamic correlation matrix subset is extracted within each analysis period. For each subset of the dynamic correlation matrix within an analysis period, the correlation coefficients between each feature component and various process parameters are calculated to form a local correlation strength sequence for that period. Perform a continuity test on the local correlation strength sequences of adjacent analysis periods, and retain the correlation strength data that pass the test; All the correlation strength data that passed the test in each production stage were combined in chronological order to form a complete distribution of the correlation strength between the characteristic components and the process parameters in different production stages.

[0010] As a further aspect of the present invention: the calculation process of the correlation coefficient is as follows: Calculate the corrected covariance of each feature component and each process parameter during the analysis period. The corrected covariance is the mean of their product minus the product of their means. The mean of the product of the absolute differences between each characteristic component and each process parameter during the analysis period is calculated and used as the stability factor. Dividing the corrected covariance by the stability factor yields the correlation coefficients between the characteristic components and the process parameters during the analysis period.

[0011] As a further aspect of the present invention: the method for dynamically adjusting the reaction process based on an optimized set of process parameters through a predictive control mechanism specifically includes: Based on the physical evolution trajectory of the optimized process parameters, calculate the gradient of material state change within a future set time period and generate corresponding predictive control commands. Predictive control commands are decomposed into power regulation of the heating system and flow regulation of the additive dosing system, which serve as the basis for feedforward regulation. The material state parameters inside the reaction vessel are collected in real time, and their real-time deviation from the material evolution trajectory is calculated. The feedforward adjustment is then compensated and corrected based on the real-time deviation. The compensated and corrected feedforward adjustment is converted into a control signal and synchronously output to the heating system and the additive dosing system.

[0012] As a further aspect of the present invention: the compensation and correction of the feedforward adjustment amount based on the real-time deviation specifically includes: Simultaneously collect temperature distribution data and viscosity change data of materials in the reaction vessel, compare them with the expected values ​​of the physical evolution trajectory, and generate a real-time deviation vector containing amplitude differences and trend differences. The initial compensation amount is calculated using a nonlinear compensation function based on the norm of the real-time deviation vector. The intensity of the initial compensation amount is dynamically adjusted according to the rate of change of the real-time deviation vector, and the weight of the compensation amount is automatically increased when the rate of change exceeds the set threshold. The dynamically adjusted compensation amount is superimposed on the original feedforward adjustment amount to generate the final compensated and corrected feedforward adjustment amount.

[0013] As a further aspect of the present invention: the parameter settings for adaptively correcting the feature space transformation based on the analysis results specifically include: Construct an influence factor matrix based on product quality indicators. The influence factor matrix includes the interaction terms between the dominant characteristic components and the control effect. Calculate the correlation weights between each element in the influence factor matrix and the product quality indicators, and obtain the overall adjustment requirements for the feature space transformation parameters by weighted summation; The adjustment direction and adjustment step size of the nonlinear mapping parameters in the feature space transformation are determined based on the overall adjustment requirements. The parameters of the feature space transformation are iteratively updated based on the adjustment direction and step size, while retaining the historical optimal parameter configuration.

[0014] As a further aspect of the present invention: the PLC controller coordinates the timing execution of each module to achieve closed-loop control from raw material characteristic identification to adaptive generation of process parameters, specifically including: Once the real-time acquisition module completes data acquisition, the feature extraction module is automatically triggered to start running. After the feature extraction module generates the dominant feature components, the parameter generation module is immediately activated to perform dynamic correlation analysis, and the feature component data is stored in the cache queue. Based on the optimized process parameter set output by the parameter generation module, the execution flow of the process control module is started synchronously, and the parameter set data is transmitted to the adaptive optimization module for monitoring. After the production process is completed, the adaptive optimization module updates the feature space transformation parameters based on the quality analysis results and feeds the updated parameters back to the initial settings of the feature extraction module.

[0015] The beneficial effects of this invention are: (1) By collecting multi-dimensional dynamic data (such as operating current spectrum, temperature gradient change rate, and stirring torque) of raw material transportation and the initial stage of reaction in real time, and by using nonlinear mapping and feature space transformation to accurately extract the dominant characteristic components that characterize the reactivity of raw materials, the system can accurately identify the characteristic differences of different batches of raw materials. Then, based on the transformation relationship established by dynamic correlation analysis, a customized set of optimal process parameters is generated for each batch of raw materials in real time, overcoming the problem of product performance fluctuation caused by raw material variability.

[0016] (2) This invention constructs a complete closed-loop control system from perception, decision-making, execution to learning. The process control module performs feedforward-feedback composite control based on the material evolution trajectory, dynamically adjusting reaction conditions to ensure that the process accurately follows the preset optimal path. More importantly, the adaptive optimization module continuously optimizes the core parameters of feature extraction by correlating the final product quality with process data, enabling the system to learn and evolve autonomously from production results. This intelligent adaptive mechanism throughout the entire process not only reduces the dependence on operator experience but also proactively adapts to changes in raw material characteristics, environmental conditions, etc., ensuring quality while continuously improving the intelligence level and overall efficiency of the production process. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0019] 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.

[0020] Please see Figure 1 As shown, this invention is a PLC automatic control system for drilling bentonite production, comprising: The real-time acquisition module is used to acquire the operating current spectrum data of the conveying power unit during the material conveying stage in the production of drilling bentonite, as well as the temperature gradient change rate and stirring torque dynamic data of the material in the reaction vessel during the reaction initialization stage. The feature extraction module is used to process the collected multidimensional state parameters by combining nonlinear mapping and feature space transformation in order to extract the dominant feature components that characterize the reactivity of the raw materials. The parameter generation module is used to perform dynamic correlation analysis on the dominant feature components, establish the transformation relationship from raw material characteristics to process parameters, and generate an optimized set of process parameters for the current production batch in real time based on the transformation relationship. The process control module is used to dynamically adjust the reaction process based on the optimized process parameter set through a predictive control mechanism, including adjusting the power output of the heating system and the flow distribution of the additive dosing system based on the feedforward adjustment of the physical evolution trajectory. The adaptive optimization module is used to collect product quality indicators after production is completed, perform correlation analysis between the quality indicators and the dominant feature components and control effects, and adaptively adjust the parameter settings of the feature space transformation based on the analysis results. The PLC controller coordinates the timing of each module to achieve closed-loop control from raw material characteristic identification to adaptive generation of process parameters.

[0021] In the real-time acquisition module, during the material conveying stage, the operating current spectrum data of the conveying power unit is acquired as follows: A Hall effect current sensor is installed in the power supply circuit of the motor driving the conveyor belt or screw feeder. This sensor detects the current signal during motor operation in real time. The current signal is transmitted to a dynamic signal analyzer connected to the sensor. The analyzer performs a fast Fourier transform on the current signal to extract the spectrum data composed of the fundamental component and its harmonic components of the current. This spectrum data reflects the changes in load characteristics of the material during conveying due to differences in composition, humidity, and particle size.

[0022] During the initial reaction phase, the rate of change of the temperature gradient of the material within the reaction vessel is obtained as follows: multiple platinum resistance temperature sensors are arranged on the inner wall of the reaction vessel or on the stirring blades, forming a temperature sensor array. These sensors are distributed at fixed spatial intervals to synchronously measure the material temperature at different locations. The temperature readings are transmitted to the analog input module of the PLC controller, where a built-in program calculates the temperature difference between adjacent sensor measurement points per unit time, and the rate of change of the material's temperature gradient is calculated based on the trend of this difference sequence. This parameter characterizes the heat transfer and absorption efficiency during the initial stage of the reaction.

[0023] Simultaneously, during the reaction initialization phase, dynamic data on stirring torque is acquired through the following methods: a rotary torque sensor is installed on the stirring shaft of the reaction vessel, or the instantaneous power of the stirring motor is detected and converted into a torque value based on the motor's characteristic curve. The torque sensor or power detection device monitors in real time the resistance torque experienced by the stirring mechanism as it rotates within the material. This resistance torque signal is transmitted to the PLC controller, forming a dynamic data sequence of stirring torque reflecting the material's viscosity and mixing uniformity. This data, along with the temperature gradient change rate, constitutes the core observation index of the material's state in the initial stage of the reaction.

[0024] In the feature extraction module, the collected multidimensional state parameters are first subjected to high-dimensional spatial projection processing. The original feature vector is composed of three types of parameters: the operating current spectrum data, the temperature gradient change rate, and the stirring torque dynamic data. This vector is then mapped to a high-dimensional feature space using a nonlinear kernel function. This mapping process is implemented using a Gaussian kernel function, and its mathematical expression is: ; in, and These represent two sample points in the original feature space. This represents the Euclidean distance between two sample points. The width parameter of the kernel function controls the scale of the mapping. Represents the Gaussian kernel function. This represents an exponential function with base e. This mapping transforms the original parameters of different dimensions and distributions into high-dimensional feature vectors of a uniform scale.

[0025] Secondly, principal component screening is performed on the obtained high-dimensional eigenvectors. This process first constructs the covariance matrix of the high-dimensional eigenvectors, calculated using the following formula: ; in, Indicates the first A high-dimensional feature vector sample, This represents the vector of means for all samples. For the sample size, This represents the matrix transpose operation. Let the covariance matrix be represented. Next, the eigenvalues ​​and eigenvectors of the covariance matrix are calculated, and the eigenvalues ​​are arranged in descending order. For each eigenvalue, its corresponding variance contribution rate is calculated by dividing that eigenvalue by the sum of all eigenvalues. The variance contribution rates of all eigenvalues ​​are arranged in descending order to form a contribution rate sequence, and then the cumulative contribution rate of this sequence is calculated by summing the contribution rates sequentially. A stable growth threshold for the cumulative contribution rate is set when the increment of the cumulative contribution rate of five consecutive eigenvalues ​​is less than 1%. The position of the eigenvalue corresponding to this cumulative contribution rate is used as a screening threshold. All principal components with variance contribution rates greater than this screening threshold are selected, and their corresponding eigenvectors are used to form a principal feature set.

[0026] Then, the obtained principal feature set is concatenated with the real-time acquired material pH parameters. This process concatenates the principal component feature vectors in the principal feature set with the real-time material pH parameters acquired by the pH sensor according to the feature dimensions, forming an extended feature vector. The extended feature vector is then processed using the min-max normalization method to scale the feature values ​​of each dimension to the range of zero to one. During normalization, for each feature value, the minimum value of that feature dimension among all samples is subtracted, and then divided by the difference between the maximum and minimum values ​​of that feature dimension among all samples. This process eliminates the dimensional differences between the feature dimensions, generating the final dominant feature components characterizing the reactivity of the raw materials.

[0027] Finally, the obtained dominant feature components are output to subsequent processing stages for the optimization and generation of process parameters. The entire process is completed in the arithmetic unit of the programmable logic controller. By executing the above steps sequentially, the real-time performance and accuracy of feature extraction are ensured, providing a reliable feature data foundation for subsequent process parameter optimization. The output of each step serves as the input for the next step, forming a complete feature extraction flow and ensuring the coherence and systematic nature of the data processing.

[0028] In the parameter generation module, firstly, a dynamic association matrix is ​​constructed by linking the dominant feature components with the corresponding time-series process data from historical production cycles. This process extracts data from multiple past production batches from the historical production database. Each production batch contains dominant feature component data arranged chronologically within the complete production cycle and corresponding process parameter data. The dominant feature component data for each production batch is arranged into a feature vector sequence by time point, and the process parameter data is arranged into a parameter vector sequence by the same time point. For each time point, the feature vector and parameter vector are combined to form a feature-parameter association pair for that time point. All feature-parameter association pairs for all time points are arranged chronologically to construct a dynamic association matrix. The rows of this matrix correspond to different time points, the columns correspond to different feature components and process parameters, and the matrix elements are the specific values ​​of each feature component and process parameter at the corresponding time point.

[0029] Secondly, a sliding time window is used to perform local correlation analysis on the dynamic association matrix. Each production stage is divided into several equally long analysis periods, each containing data from multiple consecutive time points. The length of the time window is set to 1 / 10 of the total production stage duration, but not less than five time points. The sliding time window is moved sequentially along the time axis, moving the distance of one time point at a time. Within each analysis period, a subset of data within the corresponding time window in the dynamic association matrix is ​​extracted. This subset contains the characteristic component data and process parameter data for all time points within that period. For each subset of the dynamic association matrix within each analysis period, the correlation coefficient between each characteristic component and each type of process parameter is calculated. When calculating the correlation coefficient, firstly, the corrected covariance between each characteristic component and each process parameter within that analysis period is calculated. The corrected covariance is calculated by subtracting the product of the two means from the mean of their products. Then, the mean of the product of the absolute differences between each characteristic component and each process parameter within that analysis period is calculated as a stability factor. Finally, the corrected covariance is divided by the stability factor to obtain the correlation coefficient between the characteristic component and the process parameter within that analysis period. The correlation coefficients of all characteristic components and process parameters are arranged in time sequence to form a local correlation strength sequence for that period.

[0030] In the calculation of the correlation coefficient, the formula for calculating the corrected covariance is: ; in, This indicates the correction of covariance. Indicates the first Each feature component Indicates the first One process parameter, This represents the number of time points within the analysis period. Indicates the first The eigencomponent in the th th eigenvalue 1 Values ​​at each point in time. Indicates the first The process parameter at the ... The values ​​at each point in time.

[0031] The formula for calculating the stability factor is: ; in, Indicates the stability factor. Indicates the first The average value of each characteristic component during the analysis period. Indicates the first The average value of each process parameter during the analysis period.

[0032] Then, a continuity test is performed on the local correlation strength sequences of adjacent analysis periods. This test calculates the change in the correlation coefficient between the corresponding feature component and the process parameter combination in the local correlation strength sequences of two adjacent analysis periods. A threshold of 0.1 is set for the change; when the change in the correlation coefficient between the same feature component and the process parameter combination between adjacent periods is less than this threshold, the correlation coefficient is considered continuous. Correlation coefficient data that pass the continuity test are retained, while data that fail the test are discarded. All correlation strength data that pass the continuity test within each production stage are combined in chronological order to form a complete distribution of the correlation strength between the feature component and the process parameter in different production stages.

[0033] Next, a nonlinear mapping relationship between feature components and process parameters is established based on the correlation strength distribution. This mapping relationship is realized through a weighted polynomial function, which is a polynomial expression of the process parameters with respect to the feature components, where the weight coefficients of each term are determined by the correlation strength between the corresponding feature component and the process parameter. For each process parameter, a mapping relationship expression with all feature components is established. The optimal combination of process parameters is determined by solving for the extreme points of the nonlinear mapping relationship. The extreme points are solved using the gradient descent method, searching for the points in the feature component value space that enable the process parameters to reach their optimal values.

[0034] Finally, the optimized process parameter set is calculated and output in real time based on the dominant feature components of the current production batch. The dominant feature components of the current production batch are input into the established nonlinear mapping relationship to calculate the corresponding process parameter values. The calculated process parameter values ​​are then checked for reasonableness to ensure that each value is within the allowable range. The checked process parameter values ​​are combined into an optimized process parameter set and output to the production process control system. Throughout the process, changes in the correlation between feature components and process parameters are monitored in real time. When the correlation strength changes significantly, the nonlinear mapping relationship update mechanism is automatically triggered to ensure the accuracy of the generated process parameters.

[0035] In the process control module, firstly, based on the material evolution trajectory contained in the optimized process parameter set, the gradient of material state change within a future set time period is calculated. This process reads the expected values ​​of the material evolution trajectory at five future sampling time points and uses the central difference method to calculate the rate of change at each time point. Specifically, the difference in material state values ​​between adjacent time points is taken and divided by the corresponding time interval to obtain the instantaneous rate of change at each time point. These instantaneous rates of change are connected to form a gradient curve, and corresponding predictive control instructions are generated based on the characteristics of this curve. The predictive control instructions contain information on the direction and magnitude of control quantity adjustments for each future sampling time point.

[0036] Secondly, the generated predictive control command is decomposed into power regulation of the heating system and flow regulation of the additive dosing system. The decomposition process is based on equipment characteristic parameters stored in the process knowledge base. The power regulation of the heating system is obtained by multiplying the temperature control component in the predictive control command by a power conversion coefficient, which is determined based on the thermal efficiency characteristics of the heater. The flow regulation of the additive dosing system is obtained by multiplying the component control component in the predictive control command by a flow conversion coefficient, which is determined based on the flow characteristic curve of the dosing pump. The calculated power and flow regulation values ​​are used as the basis for feedforward regulation; these basis values ​​reflect the expected control requirements based on the material evolution trajectory.

[0037] Next, the material state parameters within the reaction vessel are acquired in real time, and their real-time deviation from the material evolution trajectory is calculated. The acquired parameters include material temperature distribution data measured by a temperature sensor array and material viscosity change data measured by an online viscometer. The acquired actual data are compared with the expected values ​​at the corresponding time points in the material evolution trajectory to calculate a real-time deviation vector that includes amplitude differences and trend differences. The amplitude difference is calculated using the algebraic difference between the actual and expected values, and the trend difference is calculated using the algebraic difference between the actual and expected rates of change. The real-time deviation vector simultaneously contains deviation information at the current moment and deviation change information over the past three sampling times.

[0038] Then, the feedforward adjustment is compensated and corrected based on the real-time deviation. This process first calculates the norm of the real-time deviation vector using the L2 norm method, taking the square root of the sum of the squares of each element. Based on the norm value, a preliminary compensation amount is calculated using a nonlinear compensation function. This nonlinear compensation function is a piecewise function; when the norm value is less than a set threshold, the compensation amount is linearly related to the norm; when the norm value is greater than or equal to the set threshold, the compensation amount is proportional to the square of the norm. Simultaneously, the rate of change of the real-time deviation vector is calculated by dividing the difference between the norm at the current moment and the norm at the previous moment by the sampling time interval. When the rate of change exceeds a set threshold of 0.5 units per second, the preliminary compensation amount is multiplied by an enhancement coefficient of 1.5. Finally, the dynamically adjusted compensation amount is algebraically superimposed with the original feedforward adjustment amount to generate the final compensated and corrected feedforward adjustment amount.

[0039] Finally, the compensated and corrected feedforward regulation is converted into control signals and synchronously output to the heating system and the additive dosing system. The conversion process is based on the actuator input signal specifications. The heating system's control signal converts the power regulation into a 4-20 mA analog current signal, while the additive dosing system's control signal converts the flow regulation into a pulse-width modulation (PWM) signal. The two control signals are synchronized to ensure that the heating power regulation and additive flow regulation are coordinated in time. After the control signals are output, the next sampling cycle begins, and the above steps are repeated to achieve continuous dynamic adjustment of the reaction process.

[0040] In the adaptive optimization module, firstly, an influence factor matrix is ​​constructed based on the product quality index. The rows of this matrix correspond to different production batches, and the columns correspond to different dominant characteristic components, control effect parameters, and their interaction terms. The interaction term, obtained by multiplying the dominant characteristic component and the control effect parameter pairwise, characterizes the synergistic influence between them. The matrix elements are the specific values ​​of each dominant characteristic component, control effect parameter, and interaction term in the corresponding production batch. The product quality index serves as a benchmark reference value, corresponding one-to-one with the data in each row of the influence factor matrix.

[0041] Secondly, the correlation weights between each element in the influence factor matrix and the product quality indicators are calculated. Grey relational analysis is used to calculate the correlation degree between the numerical sequence of each element in the influence factor matrix and the numerical sequence of the product quality indicators. Specifically, the two sequences are first dimensionless, then the correlation coefficient for each corresponding point is calculated, and finally, the average of all correlation coefficients is taken to obtain the correlation degree. The calculated correlation degrees are normalized so that the sum of all correlation degrees is 1, yielding the correlation weight of each element. Based on the correlation weights, the corresponding elements in the influence factor matrix are weighted and summed to obtain the overall adjustment requirement for the feature space transformation parameters. The overall adjustment requirement is a vector with the same dimension as the number of feature space transformation parameters.

[0042] Then, based on the overall adjustment requirements, the adjustment direction and step size of the nonlinear mapping parameters in the feature space transformation are determined. The adjustment direction is determined by the sign of the overall adjustment requirement vector, with a positive sign indicating an increase in parameter value and a negative sign indicating a decrease. The adjustment step size is determined by the product of the magnitude of the overall adjustment requirement vector and the step size coefficient, which is dynamically set based on historical adjustment effects, with an initial value of 0.1. For each nonlinear mapping parameter, its adjustment direction and step size are determined according to the sign and magnitude of its corresponding component in the overall adjustment requirement vector. The adjustment step size is also limited by the maximum step size, and the single adjustment step size cannot exceed 20% of the parameter's current value.

[0043] Finally, the parameters of the feature space transformation are iteratively updated based on the adjustment direction and step size. Gradient descent is used for parameter updates; the new parameter value equals the original parameter value plus the product of the adjustment step size and the adjustment direction. During the update process, historical optimal parameter configurations are retained, i.e., the parameter combinations that achieve the best product quality indicators. After each update, the new parameter configuration is compared with the historical optimal configuration. If the new configuration improves the product quality indicators, the historical optimal configuration is updated. Simultaneously, a parameter update record is established to record the parameter changes and corresponding quality indicator changes for each adjustment, used to analyze the effectiveness of parameter adjustments.

[0044] During the PLC controller's coordination of the timing execution of each processing step, the specific operations proceed as follows: Once the real-time data acquisition step completes the data acquisition for the current batch, the feature extraction step is automatically triggered. After the feature extraction step generates the dominant feature components, the parameter generation step is immediately activated for dynamic correlation analysis, and the feature component data is stored in the data buffer. Based on the optimized process parameter set output by the parameter generation step, the execution flow of the process control step is synchronously started, and the parameter set data is transmitted to the adaptive optimization step for real-time monitoring. After the entire production process is completed, the adaptive optimization step updates the feature space transformation parameters based on the quality analysis results and feeds the updated parameters back to the initial parameter settings of the feature extraction step.

[0045] The working principle of this invention is as follows: A real-time acquisition module acquires the operating current spectrum data during the material conveying stage and the temperature gradient change rate and stirring torque dynamic data during the reaction initialization stage. A feature extraction module processes this data using nonlinear mapping and feature space transformation to extract the dominant characteristic components characterizing the reactivity of the raw materials. A parameter generation module establishes a conversion relationship from raw material characteristics to process parameters through dynamic correlation analysis of the dominant characteristic components, generating an optimized set of process parameters in real time. A process control module dynamically adjusts the heating power and additive flow rate based on the material evolution trajectory using a predictive control mechanism according to the optimized parameter set. An adaptive optimization module adaptively corrects the feature space transformation parameters based on the correlation analysis results between product quality indicators, dominant characteristic components, and control effects. All modules form a closed-loop control under the timing coordination of the PLC controller, realizing a complete optimization process from raw material characteristic identification to adaptive generation of process parameters.

[0046] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A PLC automatic control system for drilling bentonite production, characterized in that, include: The real-time acquisition module is used to acquire the operating current spectrum data of the conveying power unit during the material conveying stage in the production of drilling bentonite, as well as the temperature gradient change rate and stirring torque dynamic data of the material in the reaction vessel during the reaction initialization stage. The feature extraction module is used to process the collected multidimensional state parameters by combining nonlinear mapping and feature space transformation in order to extract the dominant feature components that characterize the reactivity of the raw materials. The parameter generation module is used to perform dynamic correlation analysis on the dominant feature components, establish the transformation relationship from raw material characteristics to process parameters, and generate an optimized set of process parameters for the current production batch in real time based on the transformation relationship. The process control module is used to dynamically adjust the reaction process based on the optimized process parameter set through a predictive control mechanism, including adjusting the power output of the heating system and the flow distribution of the additive dosing system based on the feedforward adjustment of the physical evolution trajectory. The adaptive optimization module is used to collect product quality indicators after production is completed, perform correlation analysis between the quality indicators and the dominant feature components and control effects, and adaptively adjust the parameter settings of the feature space transformation based on the analysis results. The PLC controller coordinates the timing of each module to achieve closed-loop control from raw material characteristic identification to adaptive generation of process parameters.

2. The PLC automatic control system for drilling bentonite production according to claim 1, characterized in that, The extraction of the dominant characteristic components representing the reactivity of the raw materials specifically includes: The collected multidimensional state parameters are projected into a high-dimensional space to map the operating current spectrum data, temperature gradient change rate and stirring torque dynamic data into a high-dimensional feature vector of a unified scale. Principal component screening is performed on high-dimensional feature vectors. By calculating the variance contribution rate of each feature component, the top few principal components whose contribution rates exceed a set threshold are selected as the feature principal set. The feature set is concatenated with the real-time acquired acid and alkalinity parameters of the material, and the concatenated feature sequence is normalized to generate the final dominant feature component characterizing the reactivity of the raw material.

3. The PLC automatic control system for drilling bentonite production according to claim 2, characterized in that, The selection process for the feature set is as follows: Construct the covariance matrix of high-dimensional eigenvectors, calculate the variance contribution rate corresponding to each eigenvalue of the covariance matrix, and arrange them in descending order to form a contribution rate sequence; Calculate the cumulative contribution rate of the contribution rate sequence. When five consecutive increments of the cumulative contribution rate are less than a set threshold, use the feature value position corresponding to the current cumulative contribution rate as the filtering threshold. Select all principal components whose variance contribution rate is greater than the screening threshold, and form the feature principal set by their corresponding feature vectors.

4. The PLC automatic control system for drilling bentonite production according to claim 1, characterized in that, The process parameter set for the current production batch is generated in real time based on the transformation relationship, specifically including: A dynamic correlation matrix is ​​constructed by linking the dominant feature components with the corresponding time-series process data in the historical production cycle. A sliding time window was used to perform local correlation analysis on the dynamic correlation matrix to calculate the correlation strength between the characteristic components and process parameters at different production stages. Based on the correlation strength, a nonlinear mapping relationship between characteristic components and process parameters is established, and the optimal combination of process parameters is determined by solving the extreme points of the nonlinear mapping relationship. The optimized process parameter set is calculated and output in real time based on the dominant characteristic components of the current production batch.

5. The PLC automatic control system for drilling bentonite production according to claim 4, characterized in that, The calculation of the correlation strength between the characteristic components and process parameters at different production stages specifically includes: Each production stage is divided into several equally long analysis periods, and a corresponding dynamic correlation matrix subset is extracted within each analysis period. For each subset of the dynamic correlation matrix within an analysis period, the correlation coefficients between each feature component and various process parameters are calculated to form a local correlation strength sequence for that period. Perform a continuity test on the local correlation strength sequences of adjacent analysis periods, and retain the correlation strength data that pass the test; All the correlation strength data that passed the test in each production stage were combined in chronological order to form a complete distribution of the correlation strength between the characteristic components and the process parameters in different production stages.

6. The PLC automatic control system for drilling bentonite production according to claim 5, characterized in that, The correlation coefficient is calculated as follows: Calculate the corrected covariance of each feature component and each process parameter during the analysis period. The corrected covariance is the mean of their product minus the product of their means. The mean of the product of the absolute differences between each characteristic component and each process parameter during the analysis period is calculated and used as the stability factor. Dividing the corrected covariance by the stability factor yields the correlation coefficients between the characteristic components and the process parameters during the analysis period.

7. The PLC automatic control system for drilling bentonite production according to claim 1, characterized in that, The method for dynamically adjusting the reaction process based on an optimized set of process parameters through a predictive control mechanism specifically includes: Based on the physical evolution trajectory of the optimized process parameters, calculate the gradient of material state change within a future set time period and generate corresponding predictive control commands. Predictive control commands are decomposed into power regulation of the heating system and flow regulation of the additive dosing system, which serve as the basis for feedforward regulation. The material state parameters inside the reaction vessel are collected in real time, and their real-time deviation from the material evolution trajectory is calculated. The feedforward adjustment is then compensated and corrected based on the real-time deviation. The compensated and corrected feedforward adjustment is converted into a control signal and synchronously output to the heating system and the additive dosing system.

8. The PLC automatic control system for drilling bentonite production according to claim 7, characterized in that, The compensation and correction of the feedforward adjustment based on the real-time deviation specifically includes: Simultaneously collect temperature distribution data and viscosity change data of materials in the reaction vessel, compare them with the expected values ​​of the physical evolution trajectory, and generate a real-time deviation vector containing amplitude differences and trend differences. The initial compensation amount is calculated using a nonlinear compensation function based on the norm of the real-time deviation vector. The intensity of the initial compensation amount is dynamically adjusted according to the rate of change of the real-time deviation vector, and the weight of the compensation amount is automatically increased when the rate of change exceeds the set threshold. The dynamically adjusted compensation amount is superimposed on the original feedforward adjustment amount to generate the final compensated and corrected feedforward adjustment amount.

9. The PLC automatic control system for drilling bentonite production according to claim 1, characterized in that, The parameter settings for adaptively correcting the feature space transformation based on the analysis results specifically include: Construct an influence factor matrix based on product quality indicators. The influence factor matrix includes the interaction terms between the dominant characteristic components and the control effect. Calculate the correlation weights between each element in the influence factor matrix and the product quality indicators, and obtain the overall adjustment requirements for the feature space transformation parameters by weighted summation; The adjustment direction and adjustment step size of the nonlinear mapping parameters in the feature space transformation are determined based on the overall adjustment requirements. The parameters of the feature space transformation are iteratively updated based on the adjustment direction and step size, while retaining the historical optimal parameter configuration.

10. A PLC automatic control system for drilling bentonite production according to claim 1, characterized in that, The PLC controller coordinates the timing execution of each module to achieve closed-loop control from raw material characteristic identification to adaptive generation of process parameters, specifically including: Once the real-time acquisition module completes data acquisition, the feature extraction module is automatically triggered to start running. After the feature extraction module generates the dominant feature components, the parameter generation module is immediately activated to perform dynamic correlation analysis, and the feature component data is stored in the cache queue. Based on the optimized process parameter set output by the parameter generation module, the execution flow of the process control module is started synchronously, and the parameter set data is transmitted to the adaptive optimization module for monitoring. After the production process is completed, the adaptive optimization module updates the feature space transformation parameters based on the quality analysis results and feeds the updated parameters back to the initial settings of the feature extraction module.

Images