Intelligent scheduling system for hot-dip production process of steel plate

Abstract

The present application relates to the technical field of steel plate hot galvanizing production, and discloses an intelligent scheduling system for steel plate hot galvanizing production process.The system comprises a working condition sensing module, which continuously collects the steel strip inlet temperature of the galvanizing pot area, zinc liquid composition concentration fluctuation data, air knife pressure time sequence parameters and roller system operation state signals; a feature extraction module receives the working condition data stream, analyzes the steel strip surface temperature field distribution characteristics and the viscosity variation law of the zinc liquid, and generates a process feature map; a parameter optimization module dynamically adjusts the membership function parameters and rule base weight coefficients of the fuzzy reinforcement learning controller according to the process feature map, and outputs an optimized coating control parameter set; an adaptive coating control module generates a coating thickness control instruction based on the real-time zinc liquid temperature parameters and the optimized coating control rule set; and a dynamic roller system scheduling module realizes global collaborative control of the roller system speed based on the instruction.The system can realize intelligent scheduling of the steel plate hot galvanizing production process.

Description

Technical Field

[0001] This invention relates to the field of hot-dip galvanizing production technology for steel plates, specifically to an intelligent scheduling system for the hot-dip galvanizing production process of steel plates. Background Technology

[0002] In the hot-dip galvanizing process of steel sheets, balancing coating quality and production efficiency remains a significant challenge for the industry. In traditional production models, the operating parameters of the galvanizing pot area rely heavily on manual experience for adjustment, making it difficult to cope with complex and ever-changing production environments. Fluctuations in the steel strip inlet temperature directly affect the wetting effect of the zinc bath; excessively high or low temperatures can easily lead to defects such as incomplete coating and blistering. Furthermore, instability in the zinc bath's composition concentration alters its physical properties, thereby affecting the uniformity of the coating.

[0003] The accuracy of air knife pressure control plays a crucial role in coating thickness. Existing technologies often employ fixed parameter control methods, which cannot be dynamically adjusted according to real-time operating conditions, resulting in significant coating thickness deviations. Insufficient coordination of the roller system's operating status can cause fluctuations in steel strip tension, affecting not only production continuity but also potentially causing quality problems such as scratches on the steel strip surface.

[0004] There are complex coupling relationships among various process parameters, and optimizing a single parameter is insufficient to improve overall production efficiency. Traditional scheduling systems lack the ability to perform global collaborative optimization of multiple parameters, resulting in high energy consumption during production and difficulty in maintaining a stable product qualification rate at an ideal level, thus failing to meet the requirements of modern industrial production for high efficiency and high quality. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent scheduling system for the hot-dip galvanizing production process of steel plates to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides an intelligent scheduling system for the hot-dip galvanizing production process of steel plates, the system comprising:

[0007] The working condition sensing module continuously collects working condition data streams from the galvanizing pot area, including steel strip inlet temperature, zinc liquid composition concentration fluctuation data, air knife pressure timing parameters, and roller system operating status signals.

[0008] The feature extraction module receives the operating condition data stream, analyzes the temperature field distribution characteristics on the steel strip surface and the viscosity variation law of the zinc liquid, and generates a process feature map.

[0009] The parameter optimization module dynamically adjusts the membership function parameters and rule base weight coefficients of the preset fuzzy reinforcement learning controller based on the process feature map, and outputs the optimized coating control parameter set.

[0010] The adaptive coating control module, based on real-time zinc bath temperature parameters and an optimized coating control rule set, performs dual-objective collaborative decision-making on air knife opening and zinc ingot addition amount, and generates coating thickness control instructions.

[0011] The dynamic roller system scheduling module, based on the coating thickness control command, solves the speed constraint boundary of the tension roller group, and performs mixed integer programming on the roller gap parameters by combining the original dual interior point method, so as to realize the global coordinated control of the roller system speed.

[0012] Preferably, the continuous acquisition of the operating condition data stream of the galvanizing pot area includes:

[0013] Based on historical coating defect data and zinc bath temperature fluctuation curves, a correlation mapping table between steel strip running speed and sampling frequency was established.

[0014] Based on the aforementioned correlation mapping table, the sampling interval between the temperature sensor and the component analyzer is dynamically adjusted. Sliding window filtering is performed on the collected raw zinc liquid temperature data, and outlier correction and compensation are performed on the component concentration data.

[0015] Preferably, the step of analyzing the surface temperature field distribution characteristics of the steel strip and the viscosity variation law of the zinc liquid to generate a process characteristic map includes:

[0016] The preprocessed time-series data of steel strip surface temperature is divided into continuous subsequence blocks;

[0017] Initialize the weight matrix and learning rate decay coefficient, and calculate the local temperature prediction bias of each subsequence block using the loss function;

[0018] The weight matrix parameters are updated according to the gradient direction of the temperature prediction deviation, and the iteration terminates when the rate of change of the prediction deviation of five consecutive sub-sequence blocks is less than the convergence threshold.

[0019] The updated weight matrix is ​​used to predict the viscosity change trend of zinc liquid, and the process feature map containing temperature field gradient vector and viscosity change rate is generated by combining the air knife pressure fluctuation characteristics.

[0020] Preferably, the step of dynamically adjusting the membership function parameters and rule base weight coefficients of the preset fuzzy reinforcement learning controller based on the process feature map, and outputting the optimized coating control parameter set, includes:

[0021] Using the coating thickness uniformity index as the optimization objective function, the particle swarm position vector and velocity vector are initialized;

[0022] Each particle position vector is decoded into the rule trigger threshold and membership function width parameter of the fuzzy reinforcement learning controller;

[0023] The thickness control performance of each particle's corresponding parameter set is evaluated in a coating simulation environment, and the particle fitness score is calculated.

[0024] The particle velocity vector is iteratively updated based on the global optimal particle position and the individual's historical optimal position. When the optimal fitness score has not improved for ten consecutive generations, the parameter set corresponding to the current optimal particle is output.

[0025] Preferably, the dual-objective collaborative decision-making on air knife opening and zinc ingot addition amount based on real-time zinc bath temperature parameters and an optimized coating control rule set includes:

[0026] Construct a decision variable space that includes air knife opening, zinc ingot addition rate, and steel strip running speed;

[0027] Randomly generate a neighborhood solution set within the decision variable space based on the current zinc liquid temperature parameters;

[0028] Calculate the coating thickness prediction bias and zinc consumption cost weighted evaluation value for each neighborhood solution;

[0029] Based on the annealing temperature curve, the neighboring solutions are received or discarded. When the annealing temperature drops to the termination threshold, the air knife opening command and zinc ingot addition command corresponding to the Pareto optimal solution are output.

[0030] Preferably, the step of calculating the speed constraint boundary of the tension roller group based on the coating thickness control command includes:

[0031] Establish the speed constraint equations for the tension roller group and the set of inequalities for the response delay constraint of the coating thickness;

[0032] Calculate the subgradient projection vector of the objective function under the current roll speed configuration;

[0033] The initial solution of the roll gap parameters is updated along the subgradient projection direction, and the projection operator is used to map the new solution to the velocity constraint feasible region.

[0034] When the roller speed adjustment range is less than the displacement threshold for three consecutive iterations, the roller system coordinated control parameters are locked.

[0035] Preferably, the system further includes:

[0036] The coating quality feedback module uses a Kalman filter to fuse coating thickness distribution data and feeds the fused thickness distribution features back to the feature extraction module, parameter optimization module, and adaptive coating control module, triggering dynamic reconfiguration of the coating control strategy.

[0037] Preferably, the operation of the coating quality feedback module includes:

[0038] Simultaneously collect data on coating thickness distribution at the air knife outlet and images of the cooling tower outlet surface quality;

[0039] A Kalman filter is used to fuse thickness gauge measurements with image recognition thickness distribution features to generate a thickness distribution compensation matrix;

[0040] Input the thickness distribution compensation matrix into the parameter correction interface of the feature extraction module to update the temperature field gradient vector calculation rules;

[0041] Simultaneously, the air knife pressure compensation coefficient is sent to the adaptive coating control module, triggering real-time recalculation of the air knife opening command.

[0042] Preferably, the generation of the thickness distribution compensation matrix includes:

[0043] Extract the standard deviation of the lateral fluctuation and the slope of the vertical trend from the thickness distribution data;

[0044] Identify the location distribution characteristics of zinc nodules in surface quality images;

[0045] By fusing the lateral fluctuation standard deviation and zinc nodule distribution density data using the Kalman gain coefficient, a weighted distribution vector for the thickness distribution compensation matrix is ​​generated.

[0046] Preferably, the global coordinated control of the roller system speed includes:

[0047] Analysis of the tension roller acceleration constraint condition in the coating thickness control command;

[0048] The mixed integer programming problem of roll gap parameters is decomposed into a velocity allocation subproblem and a delay compensation subproblem.

[0049] The primitive dual interior point method is used to alternately solve the integer relaxation solutions of the dual gap and delay compensation subproblems of the velocity allocation subproblem;

[0050] When the difference between the objective functions of the two subproblems is less than the cooperative error threshold, the output roller speed cooperative control instruction set is used.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] By continuously collecting various key operating condition data streams from the galvanizing pot area through the operating condition sensing module, comprehensive real-time monitoring of the production process is achieved. It can promptly capture subtle changes in steel strip inlet temperature, zinc liquid composition concentration, air knife pressure, and roller system operating status, providing comprehensive and accurate basic information for subsequent parameter optimization and control decisions.

[0053] The feature extraction module performs in-depth analysis of the collected operating condition data stream, generates process feature maps, and clearly presents the temperature field distribution characteristics of the steel strip surface and the viscosity variation law of the zinc liquid. This helps to deeply understand the intrinsic relationship between various parameters and the mechanism of their influence on coating quality, and provides a scientific basis for parameter optimization.

[0054] Based on the process feature map, the parameter optimization module dynamically adjusts the membership function parameters and rule base weight coefficients of the fuzzy reinforcement learning controller, enabling the controller to better adapt to complex and ever-changing production conditions. The optimized coating control parameter set output is more in line with actual production needs, enhancing the system's adaptability to different production conditions.

[0055] The adaptive coating control module combines real-time zinc bath temperature parameters with an optimized coating control rule set to make dual-objective collaborative decisions on air knife opening and zinc ingot addition amount. This enables the module to balance multiple objectives while ensuring coating quality, production efficiency, and resource consumption.

[0056] The dynamic roll system scheduling module, based on the coating thickness control command, solves the speed constraint boundary of the tension roll group and uses the original dual interior point method to perform mixed integer programming on the roll gap parameters, realizing global coordinated control of the roll system speed. This effectively reduces the tension fluctuation of the steel strip, ensures the continuity and stability of production, and reduces the probability of quality problems such as scratches on the steel strip surface. Attached Figure Description

[0057] Figure 1 This is a timing diagram of the intelligent scheduling system for the hot-dip galvanizing production process of steel plates described in this invention.

[0058] Figure 2 Flowchart generated for process feature map;

[0059] Figure 3 A flowchart for optimizing the parameters of a fuzzy reinforcement learning controller;

[0060] Figure 4 A flowchart for solving the speed constraints of the tension roller assembly;

[0061] Figure 5 This is a flowchart for the coating quality feedback operation. Detailed Implementation

[0062] 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.

[0063] Please see Figure 1 This invention provides an intelligent scheduling system for the hot-dip galvanizing production process of steel plates. The system includes the coordinated operation of a working condition sensing module, a feature extraction module, a parameter optimization module, an adaptive coating control module, and a dynamic roller system scheduling module.

[0064] The working condition sensing module uses a multi-source sensor network distributed throughout the galvanizing pot area to collect real-time data on steel strip inlet temperature, zinc liquid composition concentration fluctuations, air knife pressure timing parameters, and roller system operating status signals, forming a continuous working condition data stream. The feature extraction module receives this data stream and uses time series analysis to analyze the temperature field distribution characteristics of the steel strip surface, while simultaneously generating a process feature map based on a zinc liquid viscosity change model. The parameter optimization module adjusts the structural parameters of the fuzzy reinforcement learning controller through an online learning mechanism based on the dynamic changes in the process feature map, outputting an optimized coating control parameter set. The adaptive coating control module integrates real-time zinc liquid temperature monitoring data with optimized control rules to execute multi-objective decisions regarding air knife opening and zinc ingot addition. The dynamic roller system scheduling module coordinates the operating parameters of the tension roller group using a mixed integer programming algorithm based on coating control commands, achieving precise control of the coating thickness.

[0065] Example 1: See Figure 2 This study focuses on the dynamic acquisition mechanism of operating condition data streams and the generation process of process characteristic maps. In the galvanizing pot area, nine sets of infrared temperature sensors are distributed in a staggered layout on the transverse section of the steel strip running channel. The sensor spacing is adaptively adjusted according to the strip width, with a minimum spacing of 50 mm. A nonlinear correlation mapping is established between the sampling frequency of the temperature array and the running speed of the steel strip. The basic sampling frequency is set to 10 Hz. When the steel strip speed exceeds 120 m / min, the sampling frequency is increased by 1.5 Hz for every 10 m / min increase, with the maximum sampling frequency limited to 25 Hz. The zinc liquid composition analyzer uses an immersion probe to complete a full elemental spectrum scan every 20 seconds. Outlier correction processing is performed on the collected aluminum, lead, and tin concentration data: a historical 30-minute concentration distribution model is established, and data points exceeding three standard deviations are replaced with the median value of three adjacent sampling points within a sliding window.

[0066] After receiving the continuous temperature data stream, the temperature field analysis unit of the feature extraction module first performs data alignment. The steel strip surface temperature sequence is divided into continuous blocks at 500-millisecond intervals, with each block containing 75 original sampling points. During the initialization phase, a 9×9 weight matrix is ​​constructed. The initial values ​​of the matrix elements are assigned different weights based on the positional relationship of the temperature measurement points, with the weight coefficient for the central region set to 0.15 and the coefficient for the edge region set to 0.08. The initial learning rate is 0.7, and a step-wise decay strategy is adopted. After processing every 20 data blocks, the decay coefficient is adjusted to 95% of the original value. When calculating the mean square error, a 1.2-fold weight is applied to the rate of change of the temperature gradient. When the prediction deviation change of five consecutive blocks is less than 0.15℃ / s, the system freezes the current weight matrix parameters and starts the update cycle timer.

[0067] The viscosity characteristic calculation unit synchronously receives the processed zinc liquid composition data and establishes a concentration-viscosity conversion model. This model includes a negative correlation curve between aluminum content and viscosity and a positive correlation compensation function for lead. After the weight matrix is ​​updated, the viscosity prediction program is automatically activated: the component values ​​of the temperature gradient vector are input into the viscosity baseline model, and the composition compensation factor acts on the second derivative term. Every 5 seconds, the system outputs a temperature field gradient vector and viscosity change characteristic values ​​containing nine dimensions, with viscosity values ​​accurate to 0.1 mPa·s and temperature gradient values ​​retained to three decimal places. The encoding structure of the process characteristic map is a two-dimensional matrix format. The first column records the timestamp information, and the following ten columns store the nine components of the temperature gradient vector and the viscosity change rate values, respectively. The matrix row indexes strictly correspond to the time series.

[0068] The processing of air knife pressure fluctuation characteristics is performed independently. The pressure sensor acquires the raw signal at a frequency of 100Hz, and the signal is then filtered by a 0.1-15Hz bandpass filter to extract the dominant pressure oscillation frequency component. The energy spectral density distribution of the pressure fluctuation is calculated every 2 seconds, and the frequency values ​​corresponding to the first three significant peaks are selected as feature parameters. During the generation of the process feature map, the pressure feature is used as an independent channel for spatiotemporal alignment with the temperature-viscosity data: when a temperature field vector and viscosity value matching the time interval are received, the system automatically associates the pressure feature values ​​within that time period. The final map data package contains a fourteen-dimensional process feature vector comprising a timestamp, nine temperature gradient components, viscosity change, and three components: the dominant pressure frequency parameter. The data update cycle is synchronized with the temperature field feature at 5 seconds.

[0069] The abnormal data handling mechanism has an independent workflow within the system. When the component analyzer detects a sudden jump in aluminum concentration in the zinc liquid exceeding 1.5%, a verification sampling program is automatically initiated, performing three repeated measurements within 200 milliseconds. If the two verification results confirm the initial abnormal value, the viscosity calculation module immediately activates the emergency calculation model, replacing the current value with the moving average of the aluminum concentration from the previous 5 minutes, and adding a data anomaly marker to the process feature map. Simultaneously, if the weight matrix fails to converge during the temperature field feature calculation stage, the system automatically switches to the historical optimal weight configuration library, prioritizing the use of successful parameter combinations from the production of similar products, and sends a model reset alarm signal to the main control console. This nested anomaly handling architecture maintains the continuity of the feature extraction process and eliminates the impact of single-point failures on map generation.

[0070] Example 2: See Figure 3This document covers the optimization process of coating control parameters and the specific operation of a dual-objective decision-making mechanism. The parameter optimization module incorporates a particle swarm optimization (PSO) algorithm framework. In the initialization phase, 50 individual particles are set, each with a position vector having 19 dimensions. The first 7 dimensions correspond to the control parameters of the Gaussian membership function, while the last 12 dimensions store the rule base weight coefficients. Real-number encoding is used for vector encoding, and the decoding range of the membership function width parameter is limited to 0.2 to 1.5. The rule triggering threshold is mapped to a standardized range of input variables. Particle velocity vectors are initialized using a uniform random distribution, with the maximum velocity constraint set at 15% of the search space range.

[0071] The particle fitness evaluation stage initiates a simulation environment for the coating process, which is synchronized with real-time data from the production line. Each evaluation run lasts for 5 production cycles (approximately 300 seconds), during which measured values ​​of the current zinc bath temperature and steel strip speed are loaded. The control system imports the set of parameters to be measured into a fuzzy reinforcement learning controller architecture, recording the instantaneous fluctuations in the output coating thickness. The data processing unit calculates the root mean square error between the actual coating thickness and the target thickness, adding the standard deviation of the rule base trigger frequency as a complexity penalty term, and finally weighting the results to form the fitness evaluation value. The evaluation value dynamically maps to a score between 0 and 100, with lower scores indicating better control performance.

[0072] The particle update mechanism employs a dynamic balancing strategy. Each iteration calculates the distance function between the individual's historical best solution and the group's best solution, dynamically adjusting the ratio of the social learning factor and the individual learning factor based on this distance. The inertia weight in the velocity update formula uses a linear decay mechanism, gradually decreasing from an initial value of 0.9 to 0.4. After position updates, boundary checks are performed, and mirror reflection is applied to dimension values ​​exceeding the parameter allowable range. The optimization process monitoring unit tracks the historical change curve of the optimal fitness; when the fluctuation range of the optimal evaluation value over ten consecutive generations is less than 0.01 points, it outputs the controller parameter set corresponding to the current optimal particle to the decision module. The parameter conversion process retains three decimal places of precision and uses binary verification to ensure data transmission integrity.

[0073] After receiving the optimized parameter set, the adaptive coating control module activates the simulated annealing decision-making process. The decision variable space is constructed as a three-dimensional cube structure: the air knife opening variable ranges from 0-100%, with a resolution of 1%; the zinc ingot addition rate is measured in kg / min with a step size of 0.5 kg; the steel strip speed adjustment range is controlled within ±15% of the baseline value. The initial solution generation uses the current production state as the baseline point, and the initial neighborhood radius is set to 15% of the solution space diameter. The initial annealing temperature is 1200℃, and the temperature decay coefficient is fixed at 0.92, decreasing once after each round of neighborhood search.

[0074] The neighborhood solution generation mechanism employs a Gaussian perturbation strategy to randomly generate 20 candidate solutions. The evaluation of each solution involves two parallel processes: the coating thickness prediction model loads the current zinc bath temperature parameters and calculates the theoretical coating thickness based on the air knife opening and steel strip speed; the cost accounting unit calculates the total zinc ingot consumption and converts it into economic cost. In the dual-objective evaluation value calculation formula, the thickness deviation term has a weight coefficient of 0.7, and the zinc consumption cost term has a weight of 0.3. The acceptance probability calculation for new solutions considers the quality difference between the current annealing temperature and the solution, and uses an exponential probability distribution function to control the random acceptance probability.

[0075] An elite retention strategy is implemented during the search process. The optimal solution is recorded at each temperature stage and compared horizontally with the optimal solution from the previous temperature stage. A composite judgment mechanism is used for annealing termination: the search process terminates when the temperature drops to the 50℃ threshold or when the optimal solution has not been updated for three consecutive temperature stages. Pareto front analysis is performed in the output stage, selecting the solution with the smallest thickness deviation and zinc consumption cost below a set threshold as the final decision result. The output data packet includes the percentage value of the air knife opening command (accuracy 0.1%), the kg / min value of the zinc ingot addition rate (accuracy 0.1kg), and a status identifier indicating whether the steel strip speed needs adjustment. The decision cycle is controlled within 40 seconds, and operation commands are synchronously sent to the actuator before the system times out.

[0076] Example 3: See Figure 4 This involves the calculation of the speed constraint boundary of the tension roll group and the generation of the roll system collaborative control strategy. The constraint modeling unit of the dynamic roll system scheduling module establishes a set of dynamic equations for 12 tension rolls, each equation including the inter-roll tension balance condition and the coating response delay limit. The construction of the roll system speed constraint equations is based on the theory of strip elastic deformation, considering the relationship between the speed difference between the front and rear tension rolls and the strip elongation. The delay constraint inequality quantifies the response time of the coating thickness to the roll speed change, converting the thickness control accuracy required by the process into an upper limit of time delay. The mathematical expression of the constraint conditions is as follows:

[0077]

[0078] in: Indicates the first One constraint condition. For the first The linear velocity of the tension roller, It is the first The response delay time of the strip steel section The influence coefficient of speed difference. It is the roller spacing. As a delay-sensitive factor, This represents the maximum constraint value allowed by the system.

[0079] The subgradient projection calculation uses the Barzilai-Borwein step size selection method, with an initial step size of... Set to 0.05. Calculate the subgradient vector of the objective function in each iteration. The projection direction is determined as the angle bisector of the angle between the negative subgradient direction and the tangent plane of the feasible region. Projection operator When mapping the new solution to the feasible region of the velocity constraint, the tolerance threshold for handling equality constraints. Set to 0.001 mm / s. Monitor the change in the solution vector during the iteration process. When three times in a row Value less than displacement threshold When the speed is in mm, lock the current roller speed configuration parameters.

[0080] The hybrid integer programming solver discretizes the roll gap adjustment into integer multiples of 0.1 mm and uses a branch-and-bound method to handle integer variables. During the relaxation phase, the speed variables of 9 non-critical rolls out of the 12 tension rolls are treated as continuous, while the discrete characteristics of the 3 critical rolls (entry section, middle control section, and exit section main roll) are retained. The initial value of the obstacle parameter in the original dual interior point method is set to 1.0, and it decays by a factor of 0.5 after every 5 iterations. Dual gap... The calculation includes two parts: primary feasibility and dual feasibility. The convergence criterion is... And integer solution deviation .

[0081] The delay compensation subproblem is solved using a prediction-correction strategy. The prediction step calculates the theoretical delay time under ideal roll speed configuration. The calibration step adjusts the delay compensation amount based on feedback from the actual coating thickness. The discretization of the compensation amount employs an up-rounding principle to ensure sufficient adjustment margin. During the collaborative solution of the velocity allocation subproblem and the delay compensation subproblem, the original and dual variables are updated alternately, and the difference in the objective functions of the two subproblems is checked after each iteration. .when Three consecutive times less than the cooperative error threshold When the system reaches a state of coordinated equilibrium, it is determined that the system has achieved this state.

[0082] The generation of roller control commands adopts a hierarchical coding structure. The base layer contains the reference speed values ​​for each tension roller. The accuracy is 0.01 m / s; the adjustment layer stores the acceleration limit curve. Sampling is performed at 0.1s intervals; the synchronization layer records the start timestamp of each roller. Time synchronization accuracy is controlled within ±5ms. The control cycle is fixed at 100ms, and the readiness status of the actuator is checked at the beginning of each cycle. An anomaly handling mechanism monitors the deviation between the actual speed of each roller and the command value in real time. ,when Exceeding the tolerance limit After three consecutive cycles, the speed recalibration process is triggered.

[0083] The system maintains a dynamically updated roller system status database, recording historical optimal parameter combinations. After each successful coating control task, the current roller speed configuration parameters are stored in the database using the product specification code as an index. When processing products of the same specification, historical parameters are preferentially used as the initial solution to shorten the solution time. The database uses a binary tree structure to organize data, and the query response time is controlled within 50ms. The parameter transmission process uses a CRC-16 checksum mechanism to ensure data integrity.

[0084] The online adjustment of roll gap parameters is equipped with multiple protection mechanisms. The amount of roll gap change each time... Both mechanical limit constraints and strip plastic deformation limits must be met simultaneously. Mechanical limit constraints check the relationship between the current roll gap and the equipment's maximum allowable adjustment, while deformation limits verify that the strip elongation does not exceed 70% of the material's yield strength. The protection mechanism operates on a 10ms cycle, issuing an emergency stop signal within 2ms when a potential risk is detected. During normal adjustment, the roll gap change rate is limited to below 0.5mm / s to avoid impact loads.

[0085] Example 4: See Figure 5 This system encompasses the acquisition and fusion of coating quality data, as well as the dynamic adjustment of closed-loop control strategies. The coating quality feedback module is equipped with an XR-5000 X-ray thickness gauge, with 21 measuring probes evenly distributed along the strip width. The probe spacing adaptively adjusts with the strip width: an 80mm spacing is used for strips wider than 1500mm, while narrower strips are divided into 5% intervals. The thickness gauge sampling frequency automatically matches the production line speed, with a baseline sampling rate of 5 times per second. When the production line accelerates to over 120m / min, a high-speed mode is triggered, increasing the sampling rate to 10 times per second. A synchronously operating high-speed industrial camera system, model CVS-8000H, is installed 3.5 meters above the cooling tower outlet, with a lens tilt angle adjusted to 15 degrees. The camera is equipped with a ring LED fill light, a stable color temperature of 5600K, and captures 2048×1536 pixel RGB images at 200fps with a fixed exposure time of 2ms.

[0086] Table 1: Coating thickness data fusion processing (unit: μm).

[0087]

[0088] The surface defect identification algorithm first performs image preprocessing: converting the original RGB image to the HSV color space, and performing histogram equalization in the saturation channel to enhance the contrast of zinc nodules. An adaptive threshold segmentation method is used to extract suspected defect areas, combined with morphological opening operations to eliminate noise. The criteria for determining zinc nodules consider both the area factor (minimum 20 pixels) and the shape factor (aspect ratio greater than 0.6). When finally generating the defect distribution map, the strip width is divided into 100 equal segments, and the defect density value (unit: nodules / cm²) for each segment is calculated. The density value is quantified as a normalized exponent between 0 and 1.

[0089] The data fusion core employs a discrete Kalman filter architecture. The state vector is defined as the transverse thickness distribution matrix of the strip. The observation matrices are derived from direct measurement data from the thickness gauge. Image feature transformation data In the prediction phase, the thickness distribution change is extrapolated based on the strip speed, and the material flow coefficient in the state transition matrix is ​​fixed at 0.98. In the observation update phase, a dual-channel fusion mechanism is adopted: the covariance matrix of the thickness gauge measurement noise is set as a diagonal matrix with diagonal element values ​​of 0.25, and the image recognition observation noise is dynamically adjusted according to the image sharpness. When the sharpness is below 80, the noise figure is increased by 30%.

[0090] The thickness distribution compensation matrix generation process comprises three parallel computation threads: a lateral fluctuation calculation thread, based on thickness gauge data, employs a sliding window standard deviation algorithm with a window width of 15% of the current strip length, mapping the standard deviation calculation results to a 0-10 level fluctuation index; a longitudinal trend analysis thread performs linear regression calculations, retaining the slope value to three decimal places; and a defect feature quantification thread integrates the zinc nodule density distribution from the current and previous two frames to generate a spatial location-weighted index. Finally, the three types of features are fused using the Kalman gain coefficient to output the weight distribution vector of the compensation matrix. Each element corresponds to a correction coefficient at a specific position in the bandwidth direction.

[0091] The control strategy reconfiguration mechanism has dual trigger conditions: when the transverse thickness standard deviation exceeds the target value by 20%, or when the zinc nodule defect density exceeds 0.8 in the critical area (15% of the bandwidth on both sides), the system immediately activates the parameter correction process. The feature extraction module's interface supports parameter hot update: after receiving the compensation matrix, the temperature field gradient vector calculator performs a step-by-step multiplication operation with the original weight matrix, and the new matrix overwrites the original configuration within 2 seconds. The adaptive coating control module receives the air knife pressure compensation coefficient, which is calculated and generated by the central controller according to the thickness deviation direction: when the thickness deviation on the left side of the strip is greater than that on the right side, the compensation coefficient is distributed with a gradient of high on the left and low on the right, and the deviation value is scaled by 0.5% of the difference.

[0092] The system employs a three-tiered fault handling mechanism to ensure data continuity. When a single thickness gauge probe fails, Lagrange interpolation of data from adjacent probes is used for supplementation. In case of image recognition failure, the system automatically switches to historical fluctuation trend prediction of the thickness data from the previous 5 seconds. In extreme cases (such as simultaneous loss of thickness gauge and image data), the system generates a virtual compensation matrix based on a zinc liquid temperature change model and triggers an alarm signal. All data channels are equipped with timestamp verification, with a maximum allowable time difference of 50ms; expired data is automatically discarded and filled by the prediction module. The compensation command transmission adopts a dual-channel redundancy design: the main channel transmits the compensation matrix via CAN bus, while the backup channel sends control coefficient change commands separately via industrial Ethernet.

[0093] Example 5: This example covers the implementation process of a global collaborative control strategy for the roll system, specifically analyzing the tension roll motion constraints in the coating thickness control command and performing optimization calculations. After receiving the coating thickness control command, the analysis module identifies the implicit tension roll acceleration constraint parameters within the command. These constraints are manifested as differential acceleration restriction curves at the strip entry section, middle control section, and exit section. The central controller automatically decomposes the mixed-integer programming model of the roll gap parameters into two interrelated sub-problems: the speed allocation sub-problem coordinates the relative speed relationship between the tension rolls, and the delay compensation sub-problem handles the response time difference due to coating thickness changes.

[0094] In the implementation of the primal dual interior-point method, an independent solution environment is set up for each subproblem during the initialization phase. When constructing the initial feasible solution for the velocity allocation subproblem, the current roll system speed configuration is used as the reference point, and the objective function is defined as minimizing strip tension fluctuation. The dual gap calculation phase adopts a dual iterative mechanism: the inner loop processes the Lagrange multiplier update of the equality constraints, and the outer loop adjusts the obstacle parameters to achieve path tracking within the feasible region. The calculation process records the original objective value and the dual objective value for each iteration, and the subproblem is considered to have converged when the difference between the two converges to within a set threshold range.

[0095] The delay compensation subproblem solving module introduces an integer relaxation mechanism, transforming discrete time delay variables into continuous constraints. Integer characteristics are retained for key rolls (inlet looper roll, main roll of the process section, and outlet tension roll), while the parameters of other auxiliary rolls are processed as continuous. Upper and lower bounds are set for relaxation variables to prevent excessive deviation from physical reality. The compensation amount calculation adopts a predictive correction mode: based on the current roll speed, the coating thickness change trend is predicted, and a correction vector is generated by comparing it with the actual detected values. The delay compensation subproblem solver outputs a hybrid solution containing both continuous adjustment amounts and discrete step values.

[0096] The collaborative optimization controller establishes a feedback connection mechanism between the two subproblems. After each iteration, the intermediate solution output by the velocity allocation subproblem is automatically input into the delay compensation subproblem as a constraint boundary; the time parameters generated by the delay compensation subproblem are fed back to modify the equality constraints of the velocity allocation subproblem. The dual-loop structure exchanges intermediate variables through a shared memory area, with a fixed exchange period of once every three minutes. A tolerance monitoring mechanism is set during the alternating solution process; when the difference between the objective functions of the two subproblems stabilizes within a set fluctuation range for six consecutive iterations, the system automatically generates a convergence status indicator.

[0097] After receiving the optimization results, the roller control command generation unit constructs a hierarchical command structure. The basic control layer contains reference speed command values ​​for twelve tension rollers, with command accuracy reaching the thousandth of a meter level per second. The dynamic adjustment layer generates speed adjustment curves distributed according to timestamps, defining acceleration change sequences with an interval of 0.1 seconds. The safety protection layer sets speed deviation tolerance parameters, and the upper limit of the dynamic error between the actual speed and the command value of each roller is set to a specific proportion of the theoretical speed. The control cycle is fixed at one hundred milliseconds, and the command issuing unit refreshes the control queue at the beginning of each cycle.

[0098] The system is equipped with an anomaly handling module to prevent control failure. It monitors the load fluctuation data of the tension roll drive motor in real time. When a sudden change in the load of a roll reaches a warning threshold, it automatically activates the independent PID control mode for that roll and reduces the speed gradient of adjacent rolls. In case of abnormal coating thickness response, a safety mechanism initiates parameter rollback, recalculating the historical optimized parameters from the previous five minutes and simultaneously sending an anomaly code to the main control console. After each parameter adjustment, the system automatically executes a simulation verification program to virtually verify the execution effect of control commands, with the verification cycle covering the entire strip steel processing stage.

[0099] The roller system collaborative control database adopts a hierarchical storage architecture. The basic layer retains detailed optimization records for the most recent 72 hours, including boundary parameters, intermediate variables, and output instructions for each solution; the index layer constructs a mapping relationship between strip steel specification features and control parameters, using thirteen dimensions such as material thickness and zinc coating target value as classification labels; the strategy layer stores the feature vectors of the optimal solution under typical working conditions. When the matching degree between the currently processed strip steel specification and the database record reaches a certain threshold, the system preferentially selects historical optimization solutions as initial values, reducing the optimization startup time to 40% of the conventional process.

[0100] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0101] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent scheduling system for the hot-dip galvanizing production process of steel plates, characterized in that, include: The working condition sensing module continuously collects working condition data streams from the galvanizing pot area, including steel strip inlet temperature, zinc liquid composition concentration fluctuation data, air knife pressure timing parameters, and roller system operating status signals. The feature extraction module receives the operating condition data stream, analyzes the temperature field distribution characteristics on the steel strip surface and the viscosity variation law of the zinc liquid, and generates a process feature map. The parameter optimization module dynamically adjusts the membership function parameters and rule base weight coefficients of the preset fuzzy reinforcement learning controller based on the process feature map, and outputs the optimized coating control parameter set. The adaptive coating control module performs dual-objective collaborative decision-making on air knife opening and zinc ingot addition amount based on real-time zinc liquid temperature parameters and optimized coating control parameter set, and generates coating thickness control instructions. The dynamic roller system scheduling module, based on the coating thickness control command, solves the speed constraint boundary of the tension roller group, and performs mixed integer programming on the roller gap parameters by combining the original dual interior point method, so as to realize the global coordinated control of the roller system speed.

2. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, The continuously acquired data stream of the galvanizing pan area includes: Based on historical coating defect data and zinc bath temperature fluctuation curves, a correlation mapping table between steel strip running speed and sampling frequency was established. Based on the aforementioned correlation mapping table, the sampling interval between the temperature sensor and the component analyzer is dynamically adjusted. Sliding window filtering is performed on the collected raw zinc liquid temperature data, and outlier correction and compensation are performed on the component concentration data.

3. The intelligent scheduling system for the hot-dip galvanizing production process of steel plates according to claim 1, characterized in that, The analysis of the surface temperature field distribution characteristics of the steel strip and the viscosity variation law of the zinc liquid generates a process characteristic map, including: The preprocessed time-series data of steel strip surface temperature is divided into continuous subsequence blocks; Initialize the weight matrix and learning rate decay coefficient, and calculate the local temperature prediction bias of each subsequence block using the loss function; The weight matrix parameters are updated according to the gradient direction of the temperature prediction deviation, and the iteration terminates when the rate of change of the prediction deviation of five consecutive sub-sequence blocks is less than the convergence threshold. The updated weight matrix is ​​used to predict the viscosity change trend of zinc liquid, and the process feature map containing temperature field gradient vector and viscosity change rate is generated by combining the air knife pressure fluctuation characteristics.

4. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, The process involves dynamically adjusting the membership function parameters and rule base weight coefficients of the preset fuzzy reinforcement learning controller based on the process feature map, and outputting an optimized set of coating control parameters, including: Using the coating thickness uniformity index as the optimization objective function, the particle swarm position vector and velocity vector are initialized; Each particle position vector is decoded into the rule trigger threshold and membership function width parameter of the fuzzy reinforcement learning controller; The thickness control performance of each particle's corresponding parameter set is evaluated in a coating simulation environment, and the particle fitness score is calculated. The particle velocity vector is iteratively updated based on the global optimal particle position and the individual's historical optimal position. When the optimal fitness score has not improved for ten consecutive generations, the parameter set corresponding to the current optimal particle is output.

5. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, The dual-objective collaborative decision-making process for air knife opening and zinc ingot addition amount based on real-time zinc bath temperature parameters and an optimized coating control parameter set includes: Construct a decision variable space that includes air knife opening, zinc ingot addition rate, and steel strip running speed; Randomly generate a neighborhood solution set within the decision variable space based on the current zinc liquid temperature parameters; Calculate the coating thickness prediction bias and zinc consumption cost weighted evaluation value for each neighborhood solution; Based on the annealing temperature curve, the neighboring solutions are received or discarded. When the annealing temperature drops to the termination threshold, the air knife opening command and zinc ingot addition command corresponding to the Pareto optimal solution are output.

6. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, The calculation of the tension roller group speed constraint boundary based on the coating thickness control command includes: Establish the speed constraint equations for the tension roller group and the set of inequalities for the response delay constraint of the coating thickness; Calculate the subgradient projection vector of the objective function under the current roll speed configuration; The initial solution of the roll gap parameters is updated along the subgradient projection direction, and the projection operator is used to map the new solution to the velocity constraint feasible region. When the roller speed adjustment range is less than the displacement threshold for three consecutive iterations, the roller system coordinated control parameters are locked.

7. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, Also includes: The coating quality feedback module uses a Kalman filter to fuse coating thickness distribution data and feeds the fused thickness distribution features back to the feature extraction module, parameter optimization module, and adaptive coating control module, triggering dynamic reconfiguration of the coating control strategy.

8. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 7, characterized in that, The operation of the coating quality feedback module includes: Simultaneously collect data on coating thickness distribution at the air knife outlet and images of the cooling tower outlet surface quality; A Kalman filter is used to fuse thickness gauge measurements with image recognition thickness distribution features to generate a thickness distribution compensation matrix; Input the thickness distribution compensation matrix into the parameter correction interface of the feature extraction module to update the temperature field gradient vector calculation rules; Simultaneously, the air knife pressure compensation coefficient is sent to the adaptive coating control module, triggering real-time recalculation of the air knife opening command.

9. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 8, characterized in that, The generated thickness distribution compensation matrix includes: Extract the standard deviation of the lateral fluctuation and the slope of the vertical trend from the thickness distribution data; Identify the location distribution characteristics of zinc nodules in surface quality images; By fusing the lateral fluctuation standard deviation and zinc nodule distribution density data using the Kalman gain coefficient, a weighted distribution vector for the thickness distribution compensation matrix is ​​generated.

10. The intelligent scheduling system for hot-dip galvanizing production of steel plates according to claim 1, characterized in that, The global coordinated control of the roller system speed includes: Analysis of the tension roller acceleration constraint condition in the coating thickness control command; The mixed integer programming problem of roll gap parameters is decomposed into a velocity allocation subproblem and a delay compensation subproblem. The primitive dual interior point method is used to alternately solve the integer relaxation solutions of the dual gap and delay compensation subproblems of the velocity allocation subproblem; When the difference between the objective functions of the two subproblems is less than the cooperative error threshold, the output roller speed cooperative control instruction set is used.

Images