Quick response time sequence control method for high-speed array type injection air valve

Abstract

The invention discloses a quick response time sequence control method for a high-speed array type injection air valve, and the method comprises the following steps: S1, collecting and preprocessing multi-source operation data, and constructing a discrete data set; s2, constructing a decision table and performing attribute reduction to generate a blowing control rule set; s3, bee colony algorithm optimization is executed, and an optimal injection control strategy is obtained; s4, a strategy is issued, an air valve is controlled to execute, and injection feedback data is collected; and S5, feeding back the data to update the data set and the rule, and performing closed-loop optimization on the control flow. The method is used for realizing intelligent time sequence scheduling and closed-loop optimization control of the high-speed array type injection air valve so as to improve the response efficiency, reduce the air consumption and enhance the self-adaptive capability of the system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial automation control, and particularly relates to a high-speed array type jet air valve fast response timing control method. BACKGROUND

[0002] In the fields of industrial dust treatment, dust removal equipment, pneumatic transmission system and jet control device, jet air valve as a core executive element is widely used in automatic control devices such as bag type dust collector, jet loom, powder conveying system, etc. The traditional jet system mostly relies on time relay, PLC logic sequence or fixed parameter pressure threshold control mode, and the response sequence, time window and air pressure control of the jet valve are preset or statically adjusted. This control mode has the advantages of simple realization and low cost in early industrial application, but as the system scale expands, the number of air valves increases and the requirement for jet effect precision improves, the original control method gradually exposes obvious shortcomings.

[0003] The existing jet control system mainly has the following problems. Firstly, in the face of a large number of high-speed air valves working cooperatively, the traditional control system usually adopts serial or fixed polling mode for jet operation, which is difficult to meet the real-time scheduling requirements of fast synchronization and high response precision. When the number of jet valves reaches dozens or even hundreds, the execution time lag and sequence mismatch of the control system will significantly affect the dust removal effect or the uniformity of material jetting. Secondly, most of the current control systems cannot dynamically generate jet strategies based on real-time data, and lack of sensing and adaptive ability to working conditions. In the case of pressure difference fluctuation, local dust concentration or system load variation, the jet command issued under fixed rules often does not match the actual demand, resulting in uneven jet effect, waste of compressed air source and even accelerated equipment wear.

[0004] In terms of scheduling algorithm, the generation of existing jet strategies mostly adopts static logic table or simple heuristic sorting algorithm, such as air valve number round-robin based on priority, fixed time window setting based on time threshold, etc. These methods are difficult to balance the multi-objective trade-off between jet response efficiency, gas energy consumption and cleaning effect under complex conditions. Especially in the face of non-linear dynamic system working conditions, abnormal state of part of air valves, gas source fluctuation and other complex actual scenes, these static control strategies are easy to fall into local optimum or invalid jet, further reducing the system reliability and energy saving level.

[0005] In addition, in terms of feedback loop, most current systems have data collection capability, but often only for post-statistical analysis or manual correction, and fail to build a truly intelligent feedback optimization mechanism. The system lacks the ability to feedback the actual effect (such as response delay, pressure difference recovery, gas consumption change) after blowing to the control strategy to form the basis for decision-making, and the control process is still mainly preset by humans, and automatic learning and strategy iteration cannot be realized, resulting in the system unable to continuously optimize the control behavior according to the historical execution.

[0006] On the other hand, although existing scheduling optimization algorithms such as genetic algorithm and particle swarm algorithm have been tried to be introduced into the blowing optimization problem in some scenarios, the algorithm structure is complex, the parameter dependence is strong, and the convergence speed is unstable, and in the multi-gas valve high-frequency control scene, there are problems such as large operation overhead, high deployment threshold. At the same time, these algorithms rarely combine domain knowledge for control guidance, resulting in blind search direction and low efficiency, and it is difficult to directly interface with industrial controller structure, which restricts its actual landing ability.

[0007] Therefore, how to provide a high-speed array type blowing gas valve fast response timing control method is a problem that those skilled in the art need to solve. SUMMARY

[0008] One object of the present application is to provide a high-speed array type blowing gas valve fast response timing control method. The present application combines rough set theory and bee swarm intelligence algorithm, builds a blowing control rule set based on multi-source blowing feedback data, and generates an optimal gas valve response sequence, time window and blowing pressure control scheme through an optimization strategy to realize intelligent scheduling and closed-loop optimization of the blowing system. The present application describes in detail the whole process from discrete data set construction, rule mining, strategy optimization to feedback iteration update, has the advantages of fast response speed, high control precision, high gas source utilization efficiency and strong self-adaptive scheduling ability, and is suitable for intelligent control scenarios of large-scale high-frequency blowing systems.

[0009] The high-speed array type blowing gas valve fast response timing control method according to the embodiment of the present application comprises the following steps:

[0010] S1, collecting multi-source running data in a high-speed array type blowing gas valve system, and pre-processing the multi-source running data to construct a discrete data set;

[0011] S2, constructing a decision table based on the discrete data set, eliminating redundant information by using a rough set attribute reduction method, extracting core attribute features affecting the blowing response efficiency of the gas valve, and generating a blowing control rule set;

[0012] S3, under the guidance of the injection control rule set, initializing the population size, the maximum number of iterations, the search space boundary and the fitness function of the swarm intelligence algorithm, constructing the coding structure of the injection scheme, executing the population iteration, fitness evaluation and global optimization search of the swarm intelligence algorithm, and obtaining the optimal injection control strategy;

[0013] S4, constructing the injection control system, issuing the optimal injection control strategy to the injection control system, controlling each gas valve to perform injection operation according to the optimal order and response time, and collecting actual injection feedback data;

[0014] S5, incorporating the actual injection feedback data into the discrete data set, updating the decision table and the injection control rule set, re-executing the swarm intelligence algorithm optimization process, constructing a closed-loop scheduling control process, and realizing dynamic optimization and iterative update of the injection timing control scheme.

[0015] Optionally, the multi-source operation data specifically includes gas valve number, injection control signal, pressure difference signal, response delay, injection duration and dust concentration data, which are used to construct a discrete data set reflecting the correlation characteristics of injection state and effect.

[0016] Optionally, the preprocessing of the multi-source operation data specifically includes data cleaning, standardization and discretization processing, which is used to improve the consistency and calculability of the multi-source operation data, support rough set rule mining and injection strategy optimization.

[0017] Optionally, S2 specifically includes:

[0018] S21, calling the constructed discrete data set to form an attribute-decision pair set D={(x i ,d i )}, as the data basis for rule extraction, the discrete data set specifically includes dust concentration fluctuation and pressure difference response abnormal frequency, wherein x i is the attribute vector of the i-th sample, d i is the decision attribute value corresponding to the i-th sample;

[0019] S22, constructing a decision table T=(U,A∪{d}) based on the discrete data set, wherein U is the object domain, A is the condition attribute set, and d is the decision attribute representing the injection effect level, the effect level is encoded with three values: excellent, good and poor;

[0020] S23, introducing an adaptive weighted rough set attribute reduction mechanism, comprehensively considering the dependency, attribute redundancy rate and execution cost, constructing a joint evaluation function to determine the optimal attribute subset

[0021] S24, after the attribute reduction is completed, a boundary sample set is constructed for a sample area with a decision value intersection or uncertainty in the decision table, and a rule confidence evaluation mechanism is set, all candidate rules are screened, and a blowing control rule with a confidence higher than a set threshold is retained, which is used to enhance the adaptability and effectiveness of the blowing control rule set;

[0022] S25, a blowing interference sensitivity factor is introduced, a high sensitivity working condition sample set is identified according to dust concentration fluctuation and differential pressure response abnormal frequency, and a disturbance sensitivity label is added in the decision table as an auxiliary attribute to participate in rule extraction;

[0023] S26, during operation, the decision table is extended for the updated sample in combination with the collected new blowing feedback data, the rule content and attribute information are supplemented for the new sample, the original rule set is kept unchanged, and the continuous expansion and dynamic improvement of the blowing control rule set are realized;

[0024] S27, the extracted blowing control rule set is reorganized in a blowing window structure, and a blowing control rule output form is constructed: a condition attribute combination corresponds to a recommended air valve number, a recommended response time window and a control configuration of a blowing pressure level;

[0025] S28, the blowing control rule set is converted into a blowing strategy structure template, which is used as a generation basis of a blowing scheme coding structure, and a rule driven parameter space for optimization of a bee colony algorithm is formed.

[0026] Optionally, the S3 specifically comprises:

[0027] S31, based on the generated blowing control rule set, a recommended air valve number, a recommended response time window and a blowing pressure level in the blowing control rule set are extracted, an initial coding structure of a blowing scheme is constructed, and each group of scheme coding is expressed as a parameter vector Wherein, q i is the recommended air valve number, is the start time of the blowing action, is the end time of the blowing action, and p i is the blowing pressure level;

[0028] S32, the global parameters of the bee colony intelligent algorithm are initialized, including the population size N b , the maximum iteration number G max , the food source individual coding set and the boundary range of each coding structure, each individual is composed of a group of parameter vectors E i , wherein X j is the jth blowing strategy individual, j∈{1,2,...,N b};

[0029] S33, set a multi-objective fitness function F(X j ):

[0030] F(X j )=ω1T resp (X j )+ω2E gas (X j )-ω3Q eff (X j );

[0031] Wherein, T resp is the total response delay time, E gas is the total gas consumption, Q eff is the blowing effect score, ω1, ω2 and ω3 are weight coefficients;

[0032] S34, in each iteration, the weight coefficient in the fitness function is dynamically adjusted according to the total response delay time, the total gas consumption and the blowing effect score, and when the actual blowing response exceeds the expectation, the corresponding weight is improved, forming a target weight self-adaptive mechanism driven by operation feedback;

[0033] S35, the initial population is divided into multiple sub-populations, each sub-population independently performs the blowing strategy search task, and the local optimal solution sharing is performed between the sub-populations based on the preset coordination cycle, forming a multi-population parallel coordination search mechanism;

[0034] S36, in the worker phase, the individual position is updated by using the asymmetric Gaussian disturbance strategy, the search direction is generated by using the Gaussian distribution with the current solution as the mean value, and the disturbance variance is automatically adjusted according to the historical fitness, realizing the variable neighborhood search strategy;

[0035] S37, in the observer phase, the selection probability of each individual is calculated according to the multi-objective fitness value, and the individual with high selection probability will have a greater opportunity to be searched by the observer, which is used to enhance the directionality of local search;

[0036] S38, in each iteration, a historical elite solution memory mechanism is introduced, and an elite solution set A backtracking probability threshold P mem is set for each iteration, when the candidate individual meets the random triggering condition, the individual with the optimal fitness is selected from the elite solution set as the guide direction of the current individual, guiding the current individual to approach the historical excellent blowing scheme, wherein, is the kth optimal individual in the elite solution set;

[0037] S39, if a certain food source solution has no fitness improvement in continuous iterations, it is considered to be trapped in local optimum, and the scout bee performs global reset operation, re-generates the blowing coding structure and replaces the original position;

[0038] S310, completing the maximum iteration number G max or meeting the error convergence threshold, output the optimal injection control strategy X * , the optimal injection control strategy consists of recommended gas valve number, recommended response time window and injection pressure level, as the final scheduling instruction of the injection control system.

[0039] Optionally, the S4 specifically comprises:

[0040] S41, the obtained optimal injection control strategy is issued to the injection control system, which specifically comprises a gas valve driving module, a timing scheduling unit, a pressure regulating module and a sensor acquisition module, for receiving the optimal injection control strategy and executing gas valve response, pressure control and feedback data acquisition, to realize precise scheduling and closed-loop control of injection action;

[0041] S42, according to the recommended gas valve number q i of each optimal parameter vector in the optimal injection control strategy, the corresponding gas valve driving module in the injection controller is activated in turn, and the injection control instruction is issued according to the established optimal order;

[0042] S43, during the injection control process, the opening and closing of each gas valve are accurately controlled according to the start time of the injection action and the end time of the injection action to maintain the synchronization of injection timing and scheduling;

[0043] S44, according to the injection pressure level p i set in the optimal injection control strategy, the injection output pressure parameter of each gas valve is set in the controller, and the pressure control unit is called to execute pressure regulation;

[0044] S45, during the injection process, the actual response start time response end time actual injection pressure and pressure difference value ΔP i of each gas valve are collected by the sensor acquisition module;

[0045] S46, all the collected actual injection feedback data are combined into actual injection feedback data and temporarily stored in the data cache area of the controller as the basis for updating and optimizing the injection control strategy.

[0046] Optionally, the S5 specifically comprises:

[0047] S51, the collected actual injection feedback data are formatted and input as new sample data;

[0048] S52, incorporate the formatted actual blowing feedback data into the existing discrete data set to form an extended data set, for enhancing the diversity and timeliness of the data base;

[0049] S53, based on the updated discrete data set, reconstruct the decision table while retaining the original attribute structure, perform local rule reconstruction according to the new samples, and maintain the continuous updating of the blowing control rule set;

[0050] S54, call the updated blowing control rule set as the strategy guiding condition and fitness evaluation basis of the swarm intelligence algorithm, reinitialize the population and perform the optimization search process;

[0051] S55, generate a new blowing control strategy in the new round of optimization iteration as the candidate solution for the current round of control execution;

[0052] S56, construct the S51-S55 process into a closed-loop scheduling control process executed in a loop, dynamically adjust the strategy according to the feedback data after each blowing, and realize the continuous optimization and adaptive evolution of the blowing control system.

[0053] Optionally, the formatting processing of the collected actual blowing feedback data specifically includes field standardization, time alignment and attribute discretization of response time, blowing pressure and pressure difference change before and after blowing.

[0054] The beneficial effects of the present application are:

[0055] The present application realizes significant functional improvement and control optimization on the basis of the existing technology by constructing a rapid response time sequence control method for high-speed array type blowing air valve. First, attribute reduction and decision rule generation are performed on multi-source discrete operation data based on rough set theory, solving the problem of rigid control logic and inadaptability to working condition changes in traditional blowing systems, so that the blowing control strategy has data-driven self-learning ability. By introducing a structured blowing control rule set, the system can dynamically generate blowing decision conditions with strong adaptability according to actual dust concentration, pressure difference fluctuation and response lag, etc., effectively improving the pertinence and real-time performance of the blowing command.

[0056] Secondly, the present application uses an improved swarm intelligence algorithm to globally optimize the blowing strategy, which integrates multi-objective fitness function, asymmetric neighborhood search mechanism, elite solution memory guiding strategy and multi-population parallel collaborative architecture, significantly improving the optimization accuracy and convergence speed. The system can automatically adjust the weights between control targets (such as response time, air source consumption and cleaning effect), realizing the collaborative optimization of multiple performance indicators. By accurately controlling the opening sequence, response time window and blowing pressure level of each blowing air valve, the present application effectively avoids the problems of sequence disorder, air valve response overlap and air pressure waste in traditional control.

[0057] In addition, the application constructs a feedback-driven closed-loop optimization process. The system can collect real-time spraying execution data, including response time, actual pressure, and cleaning effect, and other key feedback indicators, and re-integrate them into a discrete data set for updating control rules and optimization strategies. This mechanism ensures that the control model evolves continuously with the system state, providing high adaptability and iterative optimization capability. The entire spraying control process changes from static presetting to dynamic evolution, significantly improving the robustness and efficiency of the system in complex, variable, and high-speed spraying scenarios.

[0058] In summary, the application realizes a fundamental change in the spraying control system from traditional fixed logic to intelligent scheduling, closed-loop control, and autonomous optimization, achieving significant benefits in response accuracy, control flexibility, resource utilization, and system intelligence. It is particularly suitable for multi-valve array, high-speed frequency control, and energy-saving and cost-reducing industrial spraying scenarios, and has a wide range of engineering application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0059] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, together with the embodiments of the application, to explain the application, and do not constitute a limitation on the application. In the drawings:

[0060] Figure 1 Flowchart of the high-speed array type spraying air valve fast response time sequence control method proposed by the application;

[0061] Figure 2 Improved flowchart of the improved spraying control strategy optimized by the bee colony intelligent algorithm of the high-speed array type spraying air valve fast response time sequence control method proposed by the application. DETAILED DESCRIPTION

[0062] The application will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams that only schematically illustrate the basic structure of the application, and therefore only show the components related to the application.

[0063] Reference Figure 1 and Figure 2 The high-speed array type spraying air valve fast response time sequence control method includes the following steps:

[0064] S1, collect multi-source operation data in the high-speed array type spraying air valve system, and pre-process the multi-source operation data to construct a discrete data set;

[0065] S2, construct a decision table based on the discrete data set, use the rough set attribute reduction method to eliminate redundant information, extract core attribute features that affect the spraying response efficiency of the air valve, and generate a spraying control rule set;

[0066] S3, under the guidance of the injection control rule set, initializing the population size, the maximum number of iterations, the search space boundary and the fitness function of the swarm intelligence algorithm, constructing the coding structure of the injection scheme, executing the population iteration, fitness evaluation and global optimization search of the swarm intelligence algorithm, and obtaining the optimal injection control strategy;

[0067] S4, constructing the injection control system, issuing the optimal injection control strategy to the injection control system, controlling each gas valve to perform injection operation according to the optimal order and response time, and collecting actual injection feedback data;

[0068] S5, integrating the actual injection feedback data into the discrete data set, updating the decision table and the injection control rule set, re-executing the optimization process of the swarm intelligence algorithm, constructing a closed-loop scheduling control process, and realizing dynamic optimization and iterative update of the injection timing control scheme.

[0069] The present application realizes precise control and adaptive optimization of the high-speed array type injection gas valve system in a multi-target scheduling scene by constructing an injection control process, which has significant technical effects. First, by collecting multi-source operation data and constructing a discrete data set, the real-time working condition of the injection system can be fully described, providing a data basis for subsequent rule extraction. Second, the rough set attribute reduction method effectively removes redundant information and extracts key attributes that affect injection response efficiency, ensuring that the control rules are targeted and concise. Further, the injection scheme coding structure is constructed by combining the swarm intelligence algorithm, and the optimization search based on the control rule set is executed, realizing global optimization of injection order, response time and pressure parameters, and improving injection efficiency and gas source utilization. During the control execution process, the system collects injection feedback data in real time and integrates it into the discrete data set, continuously updating the control rules and optimization strategies, and constructing a feedback-driven closed-loop control mechanism. The overall process has the advantages of fast response, precise scheduling, low energy consumption, and strong self-optimization capability, and is suitable for complex industrial control scenes such as multi-valve parallel and high-frequency injection.

[0070] In the present embodiment, the multi-source operation data specifically includes gas valve number, injection control signal, pressure difference signal, response delay, injection duration and dust concentration data, which are used to construct a discrete data set reflecting the correlation characteristics of injection state and effect.

[0071] In the embodiment, the preprocessing of the multi-source operation data specifically includes data cleaning, standardization and discretization processing. The data cleaning mainly eliminates abnormal values, repeated data and incomplete records by setting a reasonable threshold range and a missing value judgment mechanism. The standardization processing adopts the minimum-maximum normalization method to map the index data collected by various sensors to the interval [0, 1] to improve the comparability between attributes. The discretization processing divides the continuous variables into several intervals according to the distribution characteristics of different indicators and assigns symbolic codes to provide a formal input basis for the construction of the rough set decision table, which improves the consistency and computability of the multi-source operation data and supports the rough set rule mining and injection strategy optimization.

[0072] In the embodiment, S2 specifically includes:

[0073] S21, calling the constructed discrete data set to form an attribute-decision pair set D = {(x i ,d i )}, which is used as the data basis for rule extraction. The discrete data set specifically includes dust concentration fluctuation and differential pressure response abnormal frequency, wherein x i is the attribute vector of the ith sample, and d i is the decision attribute value corresponding to the ith sample.

[0074] S22, constructing a decision table T = (U, A∪{d}) based on the discrete data set, wherein U is the object domain, A is the condition attribute set, and d is the decision attribute representing the injection effect level. The effect level adopts three-value coding: excellent, good and poor.

[0075] S23, introducing an adaptive weighted rough set attribute reduction mechanism to comprehensively consider the dependency, attribute redundancy rate and execution cost, construct a joint evaluation function, and determine the optimal attribute subset

[0076] S24, after the attribute reduction is completed, a boundary sample set is constructed for the sample region with intersecting or uncertain decision values in the decision table, and a rule confidence evaluation mechanism is set. All candidate rules are screened, and the injection control rules with a confidence higher than a set threshold are retained to enhance the adaptability and effectiveness of the injection control rule set.

[0077] S25, introducing an injection disturbance sensitivity factor, which specifically includes dust concentration fluctuation amplitude, differential pressure response abnormal frequency and injection control deviation rate, to comprehensively measure the sensitivity of the injection working condition to disturbance factors and assist in identifying high sensitivity samples. According to the dust concentration fluctuation and the differential pressure response abnormal frequency, a high sensitivity sample set is identified, and a disturbance sensitivity label is added in the decision table as an auxiliary attribute to participate in rule extraction.

[0078] S26, in the running process, the newly added injection feedback data is combined to extend the decision table of the updated sample, the rule content and attribute information of the newly added sample are supplemented, the original rule set is unchanged, and the continuous expansion and dynamic improvement of the injection control rule set are realized;

[0079] S27, the extracted injection control rule set is reorganized into an injection window structure, and the injection window structure reorganization specifically refers to adjusting and optimizing the start time and end time of the recommended response time window in each rule, calibrating the time boundary combined with historical response data, setting reasonable minimum injection interval and maximum response delay, and simultaneously binding the corresponding air valve number and injection pressure level in the rule structure to form a unified and standardized three-element control configuration, i.e., "condition attribute→recommended air valve number+response time window+injection pressure level", so as to improve the execution efficiency and scheduling accuracy of the rule, and construct the injection control rule output form as: the control configuration of the condition attribute combination corresponding to the recommended air valve number, the recommended response time window and the injection pressure level;

[0080] S28, the injection control rule set is converted into an injection strategy structure template, and the injection control rule set is converted into an injection strategy structure template specifically refers to extracting the air valve number, response time window and injection pressure level corresponding to each rule in the rule set, and uniformly encoding into a parameter vector format as an input structure unit recognizable and optimized by the bee colony intelligent algorithm, as the generation basis of the injection scheme coding structure, forming a rule-driven parameter space for the optimization of the bee colony algorithm.

[0081] The injection control rule generation process constructed in the application significantly improves the intelligent decision-making ability and control adaptability of the high-speed array type injection air valve system, and has significant beneficial effects. By calling the constructed discrete data set and establishing a decision table, the system can extract high-value attributes from actual working conditions and construct an interpretable decision basis combined with injection effect levels. The self-adaptive weighted rough set attribute reduction mechanism is introduced, which fully considers the dependency, redundancy and execution cost while ensuring that the extracted attributes have optimal information content. Further, through the introduction of credibility evaluation and interference sensitive factors, the stability and key samples of the rules are realized, and the robustness and accuracy of the rule set under complex working conditions are improved. Through the incremental rule expansion mechanism, the system can continuously iterate the rule set after collecting new samples, and keep the knowledge base updated in real time. The finally constructed injection control rule set not only has a clear structure, covers air valve number, response time window and pressure level, but also is converted into an optimization input parameter space of the bee colony algorithm in a template way, laying a standardized and structured foundation for subsequent optimization. The application significantly improves the efficiency, flexibility and practicality of rule generation, and is suitable for large-scale, dynamic and multi-constrained industrial injection control scenes.

[0082] In the embodiment, the S3 specifically includes:

[0083] S31, based on the generated injection control rule set, extracting the recommended gas valve number, recommended response time window and injection pressure level in the injection control rule set, and constructing the initial coding structure of the injection scheme. Each group of scheme coding is represented as a parameter vector wherein q i is the recommended gas valve number, is the start time of the injection action, is the end time of the injection action, p i is the injection pressure level;

[0084] S32, initializing the global parameters of the swarm intelligence algorithm, including the population size N b , the maximum number of iterations G max , the food source individual coding set and the boundary range of each coding structure, each individual consisting of a group of parameter vectors E i , wherein X j is the jth injection strategy individual, j∈{1,2,...,N b};

[0085] S33, setting a multi-objective fitness function F(X j ):

[0086] F(X j ) = ω1T resp (X j ) + ω2E gas (X j ) - ω3Q eff (X j );

[0087] wherein T resp is the total response delay time, E gas is the total gas consumption, Q eff is the injection effect score, and ω1, ω2 and ω3 are weight coefficients;

[0088] The multi-objective fitness function is used to comprehensively evaluate the performance of each individual purging strategy across different performance indicators, reflecting the multi-dimensional optimality requirements of the control strategy. This function uses three key performance indicators—total purging response time, total gas consumption, and purging cleaning effect—as core evaluation dimensions, and quantifies them into a unified fitness value by setting weighting coefficients, thus providing a global search direction for the bee colony intelligent algorithm. Total response time reflects the speed of the purging system, gas consumption represents energy efficiency, and the cleaning score measures the achievement of dust removal or purging objectives. Through the construction of the fitness function, the algorithm can accurately determine in each iteration which purging strategies are more advantageous in balancing time efficiency, energy utilization, and cleaning effect, and accordingly retain and strengthen high-quality solutions while eliminating inefficient ones. This mechanism enables the purging control scheme to maximize the conservation of compressed air resources while meeting production cycle time and ensuring purging quality, thereby achieving efficient, low-consumption, and highly adaptable purging control strategy generation under dynamic operating conditions. This fitness function is the core evaluation basis for the bee colony algorithm to achieve intelligent scheduling optimization, directly affecting the strategy evolution direction and optimization effect.

[0089] S34. In each iteration, the weight coefficients in the fitness function are dynamically adjusted based on the total response delay time, total gas consumption, and blowing effect score. When the actual blowing response exceeds the expectation, the corresponding weight is increased, thus forming a target weight adaptive mechanism driven by runtime feedback.

[0090] S35. Divide the initial population into multiple subpopulations. Each subpopulation independently executes the spraying strategy search task. Subpopulations share local optimal solutions based on a preset coordination period, forming a multi-population parallel cooperative search mechanism.

[0091] S36. In the worker bee phase, an asymmetric Gaussian perturbation strategy is adopted to update the individual position. Specifically, the asymmetric Gaussian perturbation strategy is adopted to update the individual position. Specifically, the new position is generated by introducing a Gaussian random perturbation with mean shift based on the current individual and according to the set offset direction and mutation intensity. The search direction is generated using a Gaussian distribution with the current solution as the mean. The perturbation variance is automatically adjusted according to the historical fitness. Specifically, when the individual fitness has not improved for a long time, the perturbation variance is appropriately increased to enhance the global search capability. When the fitness continues to improve, the perturbation variance is decreased to improve the local search accuracy. This realizes the dynamic adaptive adjustment of the perturbation intensity, thereby realizing the variable neighborhood search strategy.

[0092] S37, in the observation bee stage, the selection probability of each individual is calculated according to the multi-objective fitness value, and the multi-objective fitness value of each spraying strategy individual is normalized, and the selection probability of the observation bee is calculated by using the proportional distribution method, specifically: the fitness values of all individuals are inversely mapped, then the inverse fitness value of each individual is divided by the sum of the inverse fitness values of all individuals, to obtain the probability of the current individual being selected for local search in this round, so as to guide the observation bee to preferentially select high-quality spraying strategy individuals for further optimization, and the individual with high selection probability will have a greater opportunity to be searched by the observation bee, so as to enhance the directionality of local search;

[0093] S38, a historical elite solution memory mechanism is introduced in each iteration to construct an elite solution set A backtracking probability threshold P is set for each iteration mem When the candidate individual meets the random triggering condition, the individual with the optimal fitness value is selected from the elite solution set as the guide direction of the current individual, so as to guide the current individual to approach the historical excellent spraying scheme, wherein, The kth optimal individual in the elite solution set is;

[0094] S39, if the food source solution has no fitness improvement in continuous iterations, it is considered to be trapped in a local optimum, and a global reset operation is performed by the scout bee to generate a spraying coding structure and replace the original position;

[0095] S310, the maximum iteration number G is completed max Or when the error convergence threshold is met, the optimal spraying control strategy X is output * The optimal spraying control strategy consists of recommended air valve number, recommended response time window and spraying pressure level, and serves as the final scheduling instruction of the spraying control system.

[0096] The constructed swarm intelligence algorithm optimization process based on the injection control rule set in the application realizes intelligent evolution of the injection control strategy from rule driving to global optimization, and has remarkable beneficial effects. First, the system extracts key control parameters in the injection control rule through structuring, constructs a standardized injection scheme coding vector, and forms a parameter input space suitable for the optimization algorithm. On this basis, a multi-objective fitness function is introduced, which uniformly considers three core indexes of response time, gas energy consumption and cleaning effect, and can comprehensively reflect the injection performance. By dynamically adjusting the weight coefficient, the system can adaptively optimize the target direction according to the actual feedback, and enhance the matching of the scheduling strategy and the running state. In addition, through the population division and collaborative updating mechanism, the search efficiency and parallelism of the solution space are significantly improved. The asymmetric disturbance strategy and the elite solution memory mechanism jointly enhance the local search accuracy and the global jump-out ability, ensuring that the optimization result is accurate and stable. The finally output injection control strategy maintains consistency in structure and considers performance and resource utilization rate in effect, and is suitable for high-speed array gas valve systems under complex working conditions and multiple constraint conditions. The overall process has the advantages of clear structure, strong robustness, fast convergence and good scalability, and is significantly superior to the traditional static scheduling mode.

[0097] In the embodiment, the S4 specifically includes:

[0098] S41, the obtained optimal injection control strategy is issued to the injection control system, and the injection control system specifically includes a gas valve driving module, a timing scheduling unit, a pressure regulating module and a sensor acquisition module, for receiving the optimal injection control strategy and executing gas valve response, pressure control and feedback data acquisition, realizing accurate scheduling and closed-loop control of the injection action;

[0099] S42, according to the recommended gas valve number q of each optimal parameter vector in the optimal injection control strategy i , the corresponding gas valve driving module in the injection controller is activated in turn, and the injection control instruction is issued according to the established optimal order;

[0100] S43, during the injection control process, the opening and closing of each gas valve are accurately controlled according to the starting time of the injection action and the ending time of the injection action , so as to maintain the synchronization of the injection timing and scheduling;

[0101] S44, according to the set injection pressure level p in the optimal injection control strategy i , the injection pressure level specifically includes three levels of low pressure, medium pressure and high pressure, corresponding to light cleaning, regular cleaning and strong cleaning demand respectively, for matching the injection intensity control under different working conditions, setting the injection output pressure parameter of each gas valve in the controller, and calling the gas pressure control unit to execute pressure regulation;

[0102] S45, collecting the actual response start time of each gas valve by the sensor acquisition module during the blowing process response end time actual blowing pressure and the pressure difference ΔP before and after blowing i ;

[0103] S46, forming all the collected actual blowing feedback data into actual blowing feedback data and temporarily storing in the data buffer area of the controller as the basis for blowing control strategy updating and optimization.

[0104] The blowing control execution and feedback acquisition process constructed in the present application realizes a complete closed-loop control mechanism from optimal strategy execution to running feedback acquisition of the high-speed array type blowing gas valve system, which has significant beneficial effects. By issuing the optimal blowing control strategy generated by optimization to the blowing control system, the system can accurately control the opening sequence and action duration of each gas valve according to the gas valve number, response time window and blowing pressure level set in the control parameters. The gas valve driving module, timing scheduling unit and pressure regulation module work cooperatively to ensure the accuracy, synchronization and pressure stability of the blowing response, effectively avoiding the timing misalignment and pressure drift problems in traditional blowing. At the same time, the sensor acquisition module acquires the response start and end time, actual blowing pressure and pressure difference change of each gas valve in real time during the blowing process, forming comprehensive running feedback data. The feedback data is structured and saved, providing real and complete performance basis for the subsequent optimization strategy. This mechanism not only realizes the high integration of control and sensing, but also provides a data-driven adaptive evolution basis for the control system, with the advantages of high-precision execution, high-real-time feedback and high-controllability, significantly improving the reliability, energy saving and intelligent response capability of the blowing system.

[0105] In the present embodiment, the S5 specifically comprises:

[0106] S51, formatting the collected actual blowing feedback data and inputting as new sample data;

[0107] S52, incorporating the formatted actual blowing feedback data into the existing discrete data set to form an expanded data set for enhancing the diversity and timeliness of the data base;

[0108] S53, based on the updated discrete data set, reconstructing the decision table while retaining the original attribute structure, performing local rule reconstruction according to the new sample, and maintaining the continuous updating of the blowing control rule set;

[0109] S54, calling the updated injection control rule set as the strategy guidance condition and fitness evaluation basis of the swarm intelligence algorithm, reinitializing the population and performing the optimization search process;

[0110] S55, generating a new injection control strategy in a new round of optimization iteration as a candidate solution for the current round control execution;

[0111] S56, constructing the S51-S55 process into a closed-loop scheduling control process executed in a loop, dynamically adjusting the strategy according to the feedback data after each injection, and realizing continuous optimization and adaptive evolution of the injection control system.

[0112] The injection feedback-driven closed-loop optimization process constructed in the application realizes the intelligent control mechanism transformation of the injection control system from single optimization to continuous iteration and autonomous evolution, and has significant beneficial effects. By formatting the actual injection feedback data and incorporating the original discrete data set, the system can expand the breadth and timeliness of the data sample in real time, effectively improving the accuracy and representativeness of data-driven modeling. On this basis, the system locally updates the decision table and control rule set using the new samples, so that the injection control rule set has the ability to dynamically grow and continuously adapt to different working conditions. The updated rule set not only maintains structural stability, but also improves its generalization ability and discrimination accuracy in complex environments. Combined with the optimization strategy iteration mechanism triggered by feedback data, the swarm intelligence algorithm can reinitialize the search population based on the latest rule set and generate the optimal injection control strategy that adapts to the current system state. This mechanism enables the injection system to be adjusted according to the actual performance after each round of execution, forming a self-closed-loop collaborative path of data-rules-optimization-control. The overall process has high adaptability, feedback responsiveness and control continuity, significantly improving the intelligent decision-making level and operation efficiency of the injection system, and is suitable for dynamic, multi-constrained and multi-objective optimization industrial application scenarios.

[0113] In the embodiment, the formatted processing of the collected actual injection feedback data specifically includes field standardization, time alignment and attribute discretization of the response time, injection pressure and pressure difference change before and after injection. Field standardization is to uniformly name and structure the originally collected response start time, end time, injection pressure, pressure difference and other data to ensure data field consistency. Time alignment is to compare and correct the actual response data according to the target response time window in the control strategy, and record the response deviation and time sequence synchronization. Attribute discretization divides continuous variables such as response duration, pressure deviation and pressure difference recovery value into discrete levels such as "normal", "mild abnormality" and "severe abnormality" by interval division method, which provides structured input for subsequent rule reconstruction and strategy optimization, and improves model processing efficiency and decision explanation ability.

[0114] Example 1:

[0115] In order to verify the feasibility of the application in implementation, the application is applied to the bag type dust removal system of a sintering workshop of a certain steel group, which is equipped with 64 groups of high-speed electromagnetic blowing air valves. The original system adopts a traditional PLC sequential timing blowing mode, the blowing logic is fixed, and the blowing strategy cannot be adjusted in real time according to the pressure difference change and dust accumulation, which often causes problems such as blowing not in time, serious gas source waste, filter bag easy to wear and tear, etc., affecting the dust removal efficiency and increasing the operation and maintenance cost.

[0116] In view of the above problems, the high-speed array type blowing air valve fast response timing control method proposed by the application is introduced into the workshop. After the system is deployed, firstly, the running state, blowing response time, actual blowing pressure, pressure difference change before and after blowing, filter bag ventilation resistance and other multi-source running data of each air valve are collected, and a discrete data set is constructed after preprocessing. The system generates an initial blowing control rule set by attribute reduction and rule extraction of the sampling data through the rough set method. On this basis, the system uses the bee colony intelligent algorithm to construct the blowing strategy coding structure and optimize the air valve blowing sequence, response time window and pressure level parameters.

[0117] After the blowing is executed, the control system automatically collects the actual response data of each air valve and generates a feedback data set, which is used to update the discrete data set and the control rule set, trigger the bee colony algorithm to optimize again, and construct a feedback driven closed loop scheduling process. After each round of optimization is completed, the blowing system can adjust the control strategy in real time according to the latest feedback data to adapt to the working condition change.

[0118] In order to verify the beneficial effects of the application, the researchers compared the 5 rounds of continuous operation data of the application and the original traditional control system under the dust removal working condition of the steel plant, and the results are shown in the following table:

[0119] From the data, it can be seen that in the continuous optimization process, the total blowing response time is gradually reduced from 1250ms to 910ms, with an average shortening of 27.2%; the total gas consumption is reduced from 42.3Nm 3 to 35.8Nm 3 , with an energy saving rate of 15.3%; the cleaning effect score is increased from 78 to 93, and the filter bag pressure difference recovery is more stable. The average response error is reduced to one third of the original, and the overall operation efficiency of the system is improved by more than 15%. Especially in multi-objective scheduling, feedback learning and response robustness, it shows significant advantages.

[0120] Table 1 comparison data table of traditional control method and the control method of the application

[0121]

[0122] According to the data in the "traditional control method and the control method of the application comparison data table", it can be clearly seen that the application has obvious advantages in the performance of injection control. First, in terms of total response time, the total response time under the traditional control method is 1250 milliseconds, and after the first round of optimization of the application, it is immediately reduced to 1120 milliseconds, and further optimized to 910 milliseconds in the fourth round, which shortens the overall response time by 340 milliseconds, and improves the response speed by 27.2%, significantly improving the real-time performance and rapid response ability of the injection system.

[0123] In terms of total gas consumption, each round of injection consumes about 42.3 cubic meters of compressed air under the traditional method, while the injection strategy optimized by the swarm intelligence makes the gas consumption decrease round by round, and finally stabilizes at 35.8 cubic meters, saving about 15.3% of the gas source, which shows that the application has obvious energy-saving advantages in the control of injection energy efficiency. At the same time, the cleaning effect score is also improved from 78 points under the traditional control to 93 points, which fully shows that the optimization strategy is closer to the demand of the on-site working condition, and can effectively improve the problems of filter bag residue and dust rebound after injection, and improve the dust removal efficiency.

[0124] In terms of control accuracy, the average response error is effectively reduced from 38 milliseconds under the traditional control to 12 milliseconds, the accuracy of injection control is obviously enhanced, the response ability of the system to the set time window is stronger, and the synchronization of injection is better. Through continuous learning and expansion of the rule set, different number of rules are updated after each round of optimization, which shows that the system has the ability to continuously adapt to new working conditions. Under the guidance of the swarm optimization strategy, the number of iterations is controlled within a reasonable range, which improves the performance while considering the algorithm running efficiency.

[0125] Overall, the application realizes the dynamic, closed-loop, self-learning injection control optimization by the deep integration of rough set decision driving and swarm intelligence optimization, and fully meets the comprehensive needs of intelligent control and energy efficiency improvement in industrial field. The data table verifies the reliability, practicality and engineering popularization value of the application in practical application scenarios.

[0126] The above is only the preferred specific embodiment of the application, but the protection scope of the application is not limited thereto, any skilled person in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the application within the technical range disclosed by the application, which should be covered within the protection scope of the application.

Claims

1. A method for timing control of a high speed arrayed air blast valve for fast response, characterized in that, The method comprises the following steps: S1, collecting multi-source operation data in a high-speed array type jet air valve system, and pre-processing the multi-source operation data to construct a discrete data set; S2, constructing a decision table based on the discrete data set, eliminating redundant information by using a rough set attribute reduction method, extracting core attribute features affecting the jet air valve jet response efficiency, and generating a jet control rule set; S3, under the guidance of the jet control rule set, initializing the population size, maximum iteration number, search space boundary and fitness function of the bee colony intelligent algorithm, constructing the coding structure of the jet scheme, executing population iteration, fitness evaluation and global optimization search of the bee colony intelligent algorithm, and obtaining the optimal jet control strategy; S4, constructing a jet control system, issuing the optimal jet control strategy to the jet control system, controlling each air valve to jet according to the optimal order and response time, and collecting actual jet feedback data; S5, incorporating the actual jet feedback data into the discrete data set, updating the decision table and the jet control rule set, re-executing the bee colony intelligent algorithm optimization process, constructing a closed-loop scheduling control process, and realizing dynamic optimization and iterative update of the jet timing control scheme.

2. The high speed arrayed air blast valve fast response timing control method of claim 1, wherein, The multi-source operation data specifically includes air valve number, jet control signal, pressure difference signal, response time delay, jet duration and dust concentration data, which are used to construct a discrete data set reflecting the correlation characteristics of jet state and effect.

3. The high speed arrayed air blast valve fast response timing control method of claim 1, wherein, The pre-processing of the multi-source operation data specifically includes data cleaning, standardization and discretization.

4. The method of claim 1, wherein the method further comprises: The S2 specifically includes: S21, calling the constructed discrete data set to form an attribute-decision pair set D={(x i ,d i )}, as the data basis of rule extraction, the discrete data set specifically includes dust concentration fluctuation and differential pressure response abnormal frequency, wherein x i is an attribute vector of the i th sample, d i is the decision attribute value corresponding to the i th sample; S22, constructing a decision table T=(U,A∪{d}) based on the discrete data set, wherein U is the object domain, A is the condition attribute set, and d is the decision attribute representing the jet effect level, which is encoded in three values: excellent, good and poor; S23, introduce an adaptive weighted rough set attribute reduction mechanism, comprehensively consider the dependency, attribute redundancy rate and execution cost, construct a joint evaluation function, determine the optimal attribute subset S24, after attribute reduction, for sample regions with cross or uncertain decision values in the decision table, a boundary sample set is constructed, and a rule confidence evaluation mechanism is set, all candidate rules are screened, and jet control rules with a confidence higher than a set threshold are retained; S25, introducing a jet disturbance sensitivity factor, identifying a high sensitivity working condition sample set according to dust concentration volatility and pressure difference response abnormal frequency, and adding a disturbance sensitivity label as an auxiliary attribute in the decision table for rule extraction; S26, during operation, combined with the collected new jet feedback data, the updated samples are extended in the decision table, the rule content and attribute information of the new samples are supplemented, the original rule set is kept unchanged, and the continuous expansion and dynamic improvement of the jet control rule set are realized; S27, reorganizing the jet window structure of the extracted jet control rule set, and constructing the jet control rule output form as: the condition attribute combination corresponds to the recommended air valve number, the recommended response time window and the jet pressure level control configuration; S28, converting the jet control rule set into a jet strategy structure template as the basis for generating the jet scheme coding structure, forming a rule-driven parameter space for bee colony algorithm optimization.

5. The method of claim 4, wherein the method further comprises: The S3 specifically includes: S31, based on the generated injection control rule set, extracting the recommended gas valve number, the recommended response time window, and the injection pressure level in the injection control rule set, and constructing an initial coding structure of the injection scheme, with each group of scheme coding represented as a parameter vector wherein q i is the recommended gas valve number, is the start time of the injection action, is the end time of the injection action, and p i is the injection pressure level; S32, initialize global parameters of the swarm intelligence algorithm, including population size N b , maximum number of iterations G max , food source individual encoding set , and boundary range of each coding structure, each individual consisting of a set of parameter vectors E i , where X j is the jth injection strategy individual, j ∈ {1, 2,..., N b} S33, setting a multi-objective fitness function F(X j ): F(X j ) = ω1T resp (X j )+ ω2E gas (X j )- ω3Q eff (X j ); Wherein, T resp is the total response delay time, E gas is the total gas consumption, Q eff is the blowing effect score, and ω1, ω2 and ω3 are weight coefficients. S34, in each iteration, dynamically adjust the weight coefficients in the fitness function according to the response delay total time, total gas consumption and the score of the effect of the injection, when the actual injection response exceeds the expectation, the corresponding weight is increased, forming a target weight self-adaptive mechanism driven by operation feedback; S35, divide the initialized population into multiple sub-populations, each sub-population independently performs the injection strategy search task, and the local optimal solution is shared among the sub-populations based on the preset coordination period, forming a multi-population parallel coordination search mechanism; S36, in the worker phase, the individual position is updated by using the asymmetric Gaussian disturbance strategy, the search direction is generated by using the Gaussian distribution with the current solution as the mean, and the disturbance variance is automatically adjusted according to the historical fitness, realizing the variable neighborhood search strategy; S37, in the observer phase, the selection probability of each individual is calculated according to the multi-objective fitness value, and the individual with high selection probability will have a greater chance to be searched by the observer; S38, a historical elite solution memory mechanism is introduced in each iteration to construct an elite solution set Set a backtracking probability threshold P for each iteration mem When the candidate individual meets the random trigger condition, the individual with the best fitness in the elite solution set is selected as the guide direction of the current individual, guiding the current individual to approach the historical excellent blowing scheme, wherein, is the kth optimal individual in the elite solution set; S39, if the fitness of a certain food source solution does not improve in continuous iterations, it is considered to be trapped in local optimum, and the global reset operation is performed by the scout bee to replace the original position by generating a new injection coding structure; S310, completing the maximum iteration number G max or meeting the error convergence threshold, output the optimal injection control strategy X * , the optimal injection control strategy consists of recommended gas valve number, recommended response time window and injection pressure level, as the final scheduling instruction of the injection control system.

6. The method of claim 5, wherein the method further comprises: The S4 specifically comprises: S41, the optimal injection control strategy obtained is issued to the injection control system, the injection control system specifically comprises a gas valve driving module, a timing scheduling unit, a pressure regulating module and a sensor acquisition module, which is used to receive the optimal injection control strategy and perform gas valve response, pressure control and feedback data acquisition, realize precise scheduling and closed-loop control of injection action; S42, recommending the gas valve number q of each optimal parameter vector in the optimal injection control strategy i , sequentially activating the corresponding gas valve driving module in the injection controller, and issuing the injection control instruction according to the established optimal order; S43、in the process of injection control, according to the starting time of the injection action and the ending time of the injection action precise control of the opening and closing of each gas valve, keeping the injection timing and scheduling synchronization; S44, according to the blowing pressure level p set in the optimal blowing control strategy i In the controller, the blowing output pressure parameter of each air valve is set, and the air pressure control unit is called to perform pressure regulation; S45, in the blowing process, the actual response start time of each gas valve is collected by the sensor collection module response end time actual blowing pressure and the pressure difference value ΔP before and after blowing i ; S46, all the collected actual injection feedback data is grouped into actual injection feedback data and temporarily stored in the data buffer area of the controller as the basis for updating and optimizing the injection control strategy.

7. The method of claim 1, wherein the method further comprises: The S5 specifically comprises: S51, the collected actual injection feedback data is formatted and input as new sample data; S52, the formatted actual injection feedback data is integrated into the existing discrete data set to form an expanded data set; S53, based on the updated discrete data set, the decision table is reconstructed, the original attribute structure is retained, the local rule reconstruction is performed according to the new sample, and the continuous update of the injection control rule set is maintained; S54, the updated injection control rule set is called as the strategy guidance condition and fitness evaluation basis of the bee colony intelligent algorithm, the population is reinitialized and the optimization search process is performed; S55, a new injection control strategy is generated in a new round of optimization iteration, which is used as the candidate solution for the current round of control execution; S56, the processes of S51-S55 are constructed as a closed-loop scheduling control process for cyclic execution, the strategy is dynamically adjusted according to the feedback data after each injection, and the continuous optimization and adaptive evolution of the injection control system are realized.

8. The method of claim 7, wherein the method further comprises: The formatted processing of the collected actual injection feedback data specifically comprises field standardization, time alignment and attribute discretization of response time, injection pressure and pressure difference change before and after injection.

Images