Functional gypsum board processing control system

Abstract

The invention relates to the technical field of plasterboard processing control, in particular to a functional plasterboard processing control system which comprises a real-time data acquisition module, a dynamic flow adaptation module, a multi-nozzle cooperative control unit, a curing quality closed-loop regulation and control unit, a self-adaptive compensation algorithm module and a multi-process cooperative execution unit. The dynamic flow adaptation module generates an initial flow reference value based on a preset spraying coverage rate model, calculates an atomization efficiency attenuation coefficient in combination with the humidity compensation sub-model, and generates a dynamic flow instruction set by superposing a base material porosity weight. The multi-nozzle cooperative control unit generates a continuous flow distribution diagram through a spatial interpolation algorithm and a nozzle trajectory equation, and dynamically adjusts the atomization pressure and the opening and closing time sequence. Under abnormal working conditions, a parameter rollback mechanism calls historical snapshots to recover a control state, and an elastic weight solidification technology balances model stability and real-time adaptability. According to the method, dynamic adaptation of spraying parameters, collaborative optimization of working procedures and abnormal self-healing are achieved, and the uniformity of the coating is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gypsum board processing control, and particularly relates to a functional gypsum board processing control system. BACKGROUND

[0002] In the functional gypsum board surface treatment process, the prior art adopts a manual spraying mode to rely on the experience of an operator to adjust the spraying amount of functional reagents such as negative oxygen ion liquid and bacteriostatic liquid, and there is a technical problem that the spraying uniformity and dose stability are difficult to accurately control, which easily leads to a decrease in the adhesion rate of raw materials and is accompanied by aerosol diffusion, thereby increasing the production cost and affecting the safety of the working environment and the product quality.

[0003] Although the existing automatic spraying equipment adopts a multi-nozzle structure to realize mechanical transmission coverage, a dynamic flow adaptation mechanism for each spraying unit is still lacking, real-time closed-loop regulation and control cannot be performed according to the board running speed, environmental temperature and humidity and substrate adsorption characteristics, and the coating curing quality consistency is insufficient. Therefore, it is urgent to develop an intelligent control system integrating multi-process collaborative sensing and adaptive compensation algorithms to solve the technical problems of accurate control spraying and production efficiency collaborative optimization in the composite board processing process. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a functional gypsum board processing control system to solve the problems of unstable coating quality and insufficient production efficiency caused by the lack of dynamic flow adaptation and real-time closed-loop regulation and control in the existing automatic spraying equipment.

[0005] To solve the above technical problems, the specific technical solutions of the present application are as follows:

[0006] The functional gypsum board processing control system provided by the present application comprises a real-time data acquisition module, a dynamic flow adaptation module, a multi-nozzle collaborative control unit, a curing quality closed-loop regulation and control unit, an adaptive compensation algorithm module and a multi-process collaborative execution unit.

[0007] The real-time data acquisition module is arranged at the inlet of the spraying machine in the feeding conveying section, and is used to acquire the board running speed, environmental temperature and humidity data, and substrate surface porosity and adsorption characteristic parameters, and transmit them to the dynamic flow adaptation module.

[0008] The dynamic flow adaptation module generates initial flow reference values of each nozzle based on the received board speed data matched with the preset spraying coverage rate model, and modifies the initial flow reference values by humidity compensation in combination with the environmental temperature and humidity data, and outputs the compensated dynamic flow instructions to the multi-nozzle collaborative control unit.

[0009] The multi-jet cooperative control unit generates a dynamic flow distribution map according to the substrate porosity distribution data, controls the differential adjustment of the atomization pressure and the opening and closing time length of the jet during movement, and feeds back the execution parameters to the curing quality closed-loop regulation unit;

[0010] The curing quality closed-loop regulation unit generates quality deviation data by detecting the coating curing degree and thickness uniformity, reversely corrects the spraying flow parameters by using a PID algorithm, and triggers an adaptive compensation algorithm module;

[0011] The adaptive compensation algorithm module calls the environmental variable and quality deviation associated data in the historical process database, optimizes the compensation coefficient matrix by using a genetic algorithm to dynamically correct the control parameters, and synchronizes the optimization results to the multi-process cooperative execution unit;

[0012] The multi-process cooperative execution unit coordinates the timing of the spraying machine, drying furnace and conveying equipment based on the optimized control parameters, and realizes the cooperative optimization of production rhythm and process quality by dynamically adjusting the drying temperature curve and discharging speed.

[0013] Further, the functional gypsum board processing control system of the present application, the dynamic flow adaptation module comprises:

[0014] The temperature and humidity data transmitted by the real-time data acquisition module is input into the humidity compensation sub-model in the dynamic flow adaptation module, and the atomization efficiency attenuation coefficient is calculated;

[0015] The initial flow reference value is weighted and corrected based on the atomization efficiency attenuation coefficient, and a dynamic flow instruction set after humidity compensation is generated;

[0016] The dynamic flow instruction set and the substrate porosity distribution data obtained by the real-time data acquisition module are matrix superimposed in the spatial coordinate system to generate jet control parameters with porosity weight factors, and are output to the multi-jet cooperative control unit.

[0017] Further, the functional gypsum board processing control system of the present application, the multi-jet cooperative control unit comprises:

[0018] Receives the jet control parameters with porosity weight factors output by the dynamic flow adaptation module; converts the discrete jet control parameters into a continuous flow distribution map by a spatial interpolation algorithm;

[0019] Based on the jet mechanical motion trajectory equation, the real-time atomization pressure gradient value on each jet moving path is calculated in combination with the continuous flow distribution map;

[0020] According to the real-time atomization pressure gradient value and the current position coordinates of the jet, the opening and closing time sequence pulse width of the corresponding jet is dynamically adjusted to generate regional spraying control instructions.

[0021] Further, the functional gypsum board plate processing control system of the present application, the self-adaptive compensation algorithm module comprises:

[0022] The deviation vector is extracted from the quality deviation data sent by the curing quality closed-loop regulation unit, and the environmental variables collected by the real-time data acquisition module are combined to construct a multi-dimensional parameter space;

[0023] The compensation coefficient candidate set is generated by cross mutation, and the fitness score of each candidate set in the dynamic flow adaptation module is calculated based on the sliding time window mechanism;

[0024] The compensation coefficient with the highest fitness score is injected into the humidity compensation sub-model of the dynamic flow adaptation module, and the mapping relationship between the environmental variables and the compensation coefficient in the historical process database is updated.

[0025] Further, the functional gypsum board plate processing control system of the present application, the multi-process collaborative execution unit comprises:

[0026] The spraying flow correction signal and the equipment state parameter generated by the curing quality closed-loop regulation unit are received;

[0027] The spraying flow correction signal and the drying furnace temperature control instruction are time-aligned by dynamic time warping algorithm to construct a collaborative control digital twin model;

[0028] The parameters of the drying temperature and conveying speed correlation model in the digital twin model are updated by using an incremental fine-tuning algorithm;

[0029] According to the updated model parameters, a device collaborative control sequence is generated to synchronously adjust the reciprocating frequency of the spraying machine, the speed of the drying fan, and the acceleration of the discharging conveyor.

[0030] Further, the functional gypsum board plate processing control system of the present application, the incremental fine-tuning algorithm comprises:

[0031] The neuron weights associated with the historical process database in the collaborative control digital twin model are locked by using an elastic weight solidification technology;

[0032] The momentum gradient descent fine-tuning is performed on the newly added real-time spraying quality data feature layer to generate fine-tuned model parameters;

[0033] The fine-tuned model parameters are mapped to the drying furnace temperature setting value correction amount and the discharge conveyor acceleration adjustment threshold.

[0034] Further, the functional gypsum board plate processing control system of the present application further comprises:

[0035] When the input feature data is out of the confidence interval of the historical process database, a parameter rollback mechanism of the collaborative control digital twin model is triggered to call a model parameter snapshot of the last stable production cycle to replace the current parameters;

[0036] The compensation coefficient optimized by the genetic algorithm in the adaptive compensation algorithm module is activated for temporary correction; after the system recovers to a steady state, the learning process of the incremental fine-tuning algorithm is reset, and the neuron weight locked by the elastic weight solidification technology in the collaborative control digital twin model is reloaded.

[0037] The present application has the following advantages:

[0038] The present application has the following advantages: BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, for those skilled in the art, other drawings can also be obtained from the drawings without creative labor.

[0040] Figure 1 The system architecture diagram of the functional gypsum board processing control system provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0041] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application. The technical solutions provided by the embodiments of the present application will be described in detail below in combination with the drawings. In order to better understand the objects of the present application, the present application will be further described in detail below.

[0042] Please refer to Figure 1 The present application provides a functional gypsum board processing control system, comprising a real-time data acquisition module, a dynamic flow adaptation module, a multi-nozzle collaborative control unit, a curing quality closed-loop regulation unit, a self-adaptive compensation algorithm module and a multi-process collaborative execution unit.

[0043] The real-time data acquisition module is deployed at the feeding conveying section and the entrance of the spraying machine, used to acquire the board running speed, environmental temperature and humidity data, and substrate surface porosity and adsorption characteristic parameters, and transmit them to the dynamic flow adaptation module.

[0044] The real-time data acquisition module is deployed with a rotary encoder at the feeding conveying section, which captures the rotational speed pulse signal of the conveying roller in real time, and converts it into board running speed data through differential operation. A humidity sensor array is installed at the entrance of the spraying machine, which generates calibrated temperature and humidity values after collecting the original temperature and humidity signals and eliminating high-frequency noise interference by using Kalman filtering algorithm. The substrate surface porosity and adsorption characteristic parameters are obtained by a multispectral imaging unit, which integrates a high-resolution optical lens and a near-infrared spectrometer. After scanning the surface of the gypsum board, the original image is processed by a convolutional neural network to output a porosity distribution heat map and adsorption characteristic quantitative values.

[0045] The collected board running speed data and calibrated temperature and humidity values are synchronized by a time stamp alignment module to form a data stream with consistent time dimension. The substrate porosity distribution heat map is associated with the board position information of the feeding conveying section by a spatial coordinate mapping algorithm to construct a three-dimensional feature matrix containing speed, environmental state and substrate characteristics. The matrix is transmitted to the dynamic flow adaptation module through industrial Ethernet, and the data integrity is guaranteed during transmission by using a cyclic redundancy check mechanism, and the time sequence reconstruction at the receiving end is realized by marking the data packet sequence number.

[0046] The speed sensor and the humidity and temperature array capture physical signals respectively, and generate structured data after noise reduction and calibration; the multispectral imaging technology analyzes the microstructure of the substrate and outputs quantitative parameters; the timestamp alignment and spatial coordinate mapping integrate multi-source data to form a unified input matrix. The data flow from physical signal acquisition, noise reduction processing to multi-dimensional fusion provides accurate working condition feature input for the downstream module.

[0047] The dynamic flow adaptation module matches the preset spraying coverage rate model based on the received plate speed data to generate initial flow reference values of each nozzle, and combines with the environmental temperature and humidity data to correct the initial flow reference values for humidity compensation, and outputs the compensated dynamic flow instructions to the multi-nozzle collaborative control unit;

[0048] The dynamic flow adaptation module receives the plate running speed data transmitted by the real-time data acquisition module, and analyzes the mapping relationship between speed and spraying coverage rate through the preset spraying coverage rate model. The model is generated based on historical process data training, which maps different speed intervals to theoretical flow requirements to form initial flow reference values of each nozzle. The initial reference values are stored in matrix form, and the matrix dimensions correspond to the physical arrangement positions of the nozzles to ensure that the nozzle flow instructions are synchronized with the plate movement.

[0049] When the initial flow reference values are corrected for humidity compensation combined with the environmental temperature and humidity data, the humidity compensation sub-model receives the calibrated temperature and humidity values, and calculates the atomization efficiency decay coefficient through a nonlinear regression algorithm. The coefficient quantifies the influence of environmental humidity on the particle size of the paint, reflecting the deviation of the actual working condition from the standard condition. The decay coefficient and the initial flow reference value are weighted and multiplied by each nozzle to generate a set of dynamic flow instructions after humidity compensation. The weighted operation introduces a piecewise linear correction strategy, which increases the compensation weight in high humidity intervals to suppress the uneven coating problem caused by the decline in atomization efficiency.

[0050] The set of dynamic flow instructions after compensation is further subjected to spatial superposition operation with the substrate porosity distribution data. The substrate porosity data is analyzed into two-dimensional grid data by multispectral imaging technology, and each grid cell is associated with a corresponding porosity weight factor. The flow values in the dynamic flow instruction set are multiplied by the porosity weight factor point by point according to the spatial coordinates to generate nozzle control parameters with porosity correction. The parameters are transmitted to the multi-nozzle collaborative control unit through the industrial bus to drive the nozzle actuator to adjust the atomization pressure and on-off timing according to the regional characteristics, completing the fine flow adaptation.

[0051] The initial flow reference value is generated based on a speed and coverage model, a humidity compensation modification introduces environmental variable quantitative adjustment, and porosity superposition operation integrates substrate characteristic data. Three-layer modification gradually refines from basic theoretical value to environmental adaptation value and then to substrate adaptation value, forming a data-driven dynamic control link. Instruction transmission and execution feedback constitute a closed loop, supporting real-time and accurate regulation and control of the spraying process.

[0052] The multi-jet coordinated control unit generates a dynamic flow distribution map according to substrate porosity distribution data, controls the differential adjustment of atomization pressure and on-off duration of the jet during movement, and feeds back the execution parameters to the curing quality closed-loop control unit;

[0053] The multi-jet coordinated control unit receives the jet control parameters with porosity weight factors transmitted by the dynamic flow adaptation module, which includes the target flow value of each jet in the preset spatial coordinate system and the corresponding porosity correction weight. The control parameters are transmitted in the form of data packets through the industrial bus, and each data packet is associated with a specific jet number and position code, realizing accurate matching of parameters and physical jets. The porosity correction weight is generated based on the substrate surface porosity distribution thermogram, reflecting the differentiated demand for coating penetration depth in different areas.

[0054] When processing discrete jet control parameters through spatial interpolation algorithm, the discrete target flow value is expanded into continuous two-dimensional flow distribution map by using bicubic spline interpolation method. The distribution map covers all the areas to be sprayed on the gypsum board surface, and each grid node is associated with a flow target value and a porosity weight. In the interpolation process, an edge smoothing algorithm is introduced to eliminate the flow step change when the moving direction of the jet is switched, and the spatial continuity of the spraying coverage is maintained. The continuous flow distribution map and the jet mechanical motion trajectory equation are aligned in coordinates, and the spraying area is mapped to the real-time motion path of the jet.

[0055] Based on the jet mechanical motion trajectory equation, combined with the continuous flow distribution map, the real-time atomization pressure gradient value is calculated, according to the instantaneous moving speed and acceleration parameters of the jet, a position-time function relationship is established. According to the target flow value of each position in the flow distribution map, the required atomization pressure value is back calculated, and a pressure gradient curve along the motion path is generated. The pressure gradient curve is converted into real-time atomization pressure adjustment instruction through differential operation, driving the jet to dynamically adjust the atomization pressure output according to the preset gradient during movement.

[0056] According to the real-time atomization pressure gradient value and the current position coordinates of the nozzle, the opening and closing timing pulse width of the corresponding nozzle is dynamically adjusted, the pressure gradient value is converted into a pulse width modulation signal. The signal is synchronized with the nozzle movement position, and the opening and closing duration of the electromagnetic valve is controlled through a high-precision timer to realize the real-time matching of the spraying flow and the nozzle moving speed. The regional spraying control instruction set integrates the pressure adjustment parameters and the pulse timing information, which is sent to each nozzle actuator through the driving circuit to complete the differentiated spraying operation based on the substrate characteristics. The executed nozzle movement parameters and the actual spraying flow data are fed back to the curing quality closed-loop control unit through the industrial bus to form a feed-forward-feedback composite control loop.

[0057] The logic connection statement: the control parameters realize accurate positioning of the nozzle through coding matching; the spatial interpolation algorithm eliminates the discontinuity of discrete data to generate smooth flow distribution; the motion trajectory equation is combined with the flow distribution to derive the pressure gradient, which converts the flow demand into pressure control instructions; dynamic pulse width modulation realizes accurate timing control, and feedback data closed loop optimizes the spraying quality. Each step realizes the complete control link from parameter analysis to execution driving through data conversion and coordinate mapping, supporting the improvement of coating uniformity and process stability.

[0058] The curing quality closed-loop control unit generates quality deviation data by detecting the coating curing degree and thickness uniformity, and uses the PID algorithm to correct the spraying flow parameters in reverse and trigger the adaptive compensation algorithm module;

[0059] The adaptive compensation algorithm module calls the environmental variable and quality deviation associated data in the historical process database, optimizes the compensation coefficient matrix through genetic algorithm to dynamically correct the control parameters, and synchronizes the optimization results to the multi-process collaborative execution unit;

[0060] The curing quality closed-loop control unit deploys an infrared spectrometer at the outlet of the drying furnace to detect the coating curing degree and thickness uniformity data. Through Fourier transform, the energy distribution of the coating characteristic frequency band is extracted and compared with the standard frequency range of the preset quality threshold to generate a multi-dimensional quality deviation vector. The vector is converted into a spraying flow correction coefficient by a PID controller to adjust the initial flow reference value of the dynamic flow adaptation module in reverse, forming a feed-forward-feedback composite control loop. The corrected flow parameters trigger the optimization process of the adaptive compensation algorithm module through the industrial bus.

[0061] The adaptive compensation algorithm module calls the environmental variable data and quality deviation vector stored in the historical process database to construct a multi-dimensional parameter space of environmental temperature, humidity and coating quality deviation. The parameter space is organized in time series form, associated with historical compensation coefficients and control effect records. The genetic algorithm initializes the compensation coefficient population, performs multi-point crossover and Gaussian mutation operations to generate candidate parameter sets, uses a sliding time window to intercept current period process data, calculates the deviation convergence rate and stability index of each candidate set in the dynamic flow adaptation module, and uses the fitness score to select the optimal compensation coefficient.

[0062] The optimized compensation coefficient is injected into the humidity compensation sub-model of the dynamic flow adaptation module through matrix multiplication operation, adjusting the weight distribution in the atomization efficiency decay coefficient calculation logic. The historical process database is updated synchronously with the current environmental variables, compensation coefficients and quality detection results after control, forming a closed-loop training data set. The updated parameter mapping relationship is used as the population initialization boundary for subsequent genetic algorithm iteration, improving the parameter search efficiency. The optimization result is synchronized to the multi-process collaborative execution unit through the industrial bus, driving the parameter fine-tuning process of the collaborative control digital twin model.

[0063] Infrared spectrum detection generates quality deviation data to drive PID correction, forming a preliminary closed-loop control; deviation data and historical environmental variables construct a parameter space, and genetic algorithm mines implicit process rules to generate compensation coefficients; coefficients are injected into the humidity compensation model and the database is updated to realize the interaction of parameter optimization and experience accumulation; optimized parameters are synchronized to the collaborative execution unit to support cross-process control. Data flow from quality detection to algorithm optimization, and then to equipment control layer, forming a self-learning full-process closed-loop system.

[0064] The multi-process collaborative execution unit coordinates the timing of the spraying machine, drying oven and conveying equipment based on the optimized control parameters, and realizes the collaborative optimization of production rhythm and process quality by dynamically adjusting the drying temperature curve and discharge speed.

[0065] The multi-process collaborative execution unit receives the optimized control parameters from the adaptive compensation algorithm module, which include the spraying flow correction coefficient, the compensation coefficient optimized by the genetic algorithm, and the mapping relationship updated by the historical process database. The parameters are transmitted to the input end of the collaborative control digital twin model through the industrial bus, and are time-aligned with the real-time collected drying oven temperature gradient data and conveying equipment acceleration parameters. The dynamic time warping algorithm eliminates the communication delay between devices, maps multi-source data to a unified time axis, and constructs a time sequence correlation model of spraying, drying and conveying processes.

[0066] Based on the output parameters of the collaborative control digital twin model, a device collaborative control instruction sequence is generated. The instruction sequence is parsed by a priority scheduling algorithm, and the adjustment priorities of the spraying machine reciprocating frequency, drying oven temperature zone set value and conveying roller acceleration are dynamically allocated. The spraying machine reciprocating frequency is adjusted according to the real-time flow correction coefficient to match the change of the substrate advancing speed; the drying oven temperature zone set value is dynamically calculated based on the coating curing rate prediction model and is changed in linkage with the discharge conveying acceleration to prevent the un-cured board from entering the next process. The conveying acceleration parameter is calibrated in real time according to the drying oven outlet temperature gradient to balance the production rhythm and the curing quality demand.

[0067] When the drying temperature curve is dynamically adjusted, the digital twin model predicts the required hot air energy distribution of each temperature zone based on real-time coating thickness uniformity data. The weight parameters of the temperature-speed correlation model are updated by an incremental fine-tuning algorithm to generate a staged temperature rising curve. The discharge conveying speed is dynamically adjusted according to the temperature zone set value and the coating curing degree detection result, and a feedforward-feedback composite control strategy is adopted to simultaneously optimize the production efficiency and process stability. The adjusted device parameters are issued to the actuators through the industrial bus to drive the spraying machine electromagnetic valve, drying fan frequency converter and conveying motor to act in collaboration.

[0068] The executed process parameters and device state data are collected in real time through the sensor network and fed back to the curing quality closed-loop regulation unit and the adaptive compensation algorithm module. The quality deviation data triggers a new round of parameter optimization, and the compensation coefficients updated by the genetic algorithm and the fine-tuning results of the collaborative model form a two-way iteration to realize the continuous collaborative optimization of production rhythm and process quality.

[0069] The optimized parameters are input into the digital twin model after time alignment to generate device control instructions; the priority scheduling parses the instructions and allocates the execution priority to ensure that the key process parameters are adjusted in priority; the temperature curve and the discharge speed are dynamically linked based on model prediction to suppress process conflicts; the feedback data drive parameter iteration to form a closed-loop control link of optimization, execution and feedback. Each link realizes cross-process collaboration through data conversion and instruction transmission to improve the overall control accuracy and stability of the system.

[0070] The real-time data acquisition module is deployed at the feeding conveying section and the spraying machine inlet, and the speed sensor is used to capture the board advancing speed pulse signal in real time. The temperature and humidity sensors are used to collect environmental data synchronously, and the calibrated values are generated after Kalman filter denoising. The substrate characteristic detection unit integrates a multi-spectral imaging device to perform optical scanning on the surface of the gypsum board. The original optical signal is processed by a convolutional neural network to output a porosity distribution thermogram and a quantitative value of adsorption characteristics. The calibrated speed, temperature and humidity data and porosity thermogram are aligned by time stamp and mapped by spatial coordinates to form a three-dimensional feature matrix containing motion parameters, environmental state and substrate characteristics, providing structured input for the dynamic flow adaptation module.

[0071] After receiving the three-dimensional feature matrix, the dynamic flow adaptation module calls the preset spraying coverage rate model to perform matrix decomposition, extracts the speed and coverage rate mapping relationship, and generates the initial flow reference value of each spray head. The temperature and humidity calibration value is input into the humidity compensation sub-model to calculate the atomization efficiency attenuation coefficient. Based on the weighted correction algorithm, the initial flow reference value is dynamically adjusted, and the dynamic flow instruction set after humidity compensation is output. The instruction set and the porosity distribution thermograph of the substrate are superimposed in the spatial coordinate system to generate the spray head control parameter with the porosity weight factor, which is transmitted to the multi-spray head cooperative control unit.

[0072] After receiving the spray head control parameter, the multi-spray head cooperative control unit converts the discrete parameter into a continuous flow distribution graph through a spatial interpolation algorithm. Combined with the spray head mechanical motion trajectory equation, the real-time atomization pressure gradient value of each spray head on the transverse movement path is calculated. According to the current position coordinates of the spray head, the on-off timing pulse width is dynamically adjusted to generate a regional differentiated spraying instruction. The actual motion parameters of the spray head and the spraying quality detection data are fed back to the curing quality closed-loop control unit through the industrial bus to form a feed-forward-feedback composite control loop.

[0073] The curing quality closed-loop control unit deploys an infrared spectrometer at the outlet of the drying furnace to detect the curing degree and thickness uniformity of the coating. Through Fourier transform, the energy value of the characteristic frequency band is extracted, and deviation analysis is performed with the preset quality threshold to generate a deviation vector. The PID controller converts the deviation vector into a spraying flow correction coefficient to trigger the adaptive compensation algorithm module to start the parameter optimization process. The corrected flow parameter is injected back into the dynamic flow adaptation module through the industrial bus to form a closed-loop control link.

[0074] The adaptive compensation algorithm module calls the environmental variable and quality deviation associated data stored in the historical process database to construct a multi-dimensional parameter space. Through genetic algorithm, the population space of the compensation coefficient matrix is initialized, and the cross variation operation is performed to generate candidate parameter combinations. The sliding time window mechanism is used to evaluate the fitness score of each candidate set in real-time control, and the optimal compensation coefficient is injected into the humidity compensation sub-model to update the parameter mapping relationship in the historical database. The optimized compensation coefficient is input into the multi-process cooperative execution unit as a boundary condition to trigger the cooperative control model fine-tuning.

[0075] The multi-process cooperative execution unit constructs a cooperative control digital twin model, updates the drying temperature and conveying speed correlation model parameters based on real-time spraying quality data and equipment state parameters using an incremental fine-tuning algorithm. The spraying flow correction signal and the drying temperature control instruction time sequence are aligned through a dynamic time warping algorithm to eliminate communication delays between devices. The fine-tuned model parameters are mapped to the drying oven partition temperature set values and the discharge conveying acceleration adjustment threshold values to generate a device cooperative control sequence. The spraying machine reciprocating frequency, drying fan speed, and conveying roller start-stop time sequence are coordinated through a priority scheduling algorithm to suppress coating defects caused by parameter conflicts between processes, achieving dynamic balance between production rhythm and process quality.

[0076] Each module forms a closed-loop control link through data flow - real-time collected environmental and substrate data drive dynamic flow calculation, spray head control parameters generate execution instructions after spatial interpolation, coating quality detection results reverse correct flow parameters and trigger genetic algorithm optimization. The optimized compensation coefficients and cooperative model fine-tuning parameters jointly act on the device control layer to realize cross-process parameter coupling through time sequence alignment and priority scheduling. Under abnormal conditions, historical parameter snapshots and elastic weight solidification technology ensure the system quickly recovers to steady state, forming a full-process technical closed loop covering data acquisition, dynamic control, quality optimization, and cooperative execution.

[0077] Specifically, the functional gypsum board processing control system described in the application comprises a dynamic flow adaptation module, a multi-spray head cooperative control unit, a real-time data acquisition module, a dynamic flow calculation module, a coating quality detection module, a fine-tuning parameter optimization module, and a priority scheduling module.

[0078] The humidity compensation sub-model in the dynamic flow adaptation module inputs the temperature and humidity data transmitted by the real-time data acquisition module to calculate the atomization efficiency decay coefficient;

[0079] The initial flow reference value is weighted and corrected based on the atomization efficiency decay coefficient to generate a set of dynamic flow instructions after humidity compensation;

[0080] The dynamic flow instruction set and the substrate porosity distribution data obtained by the real-time data acquisition module are matrix superimposed in the spatial coordinate system to generate spray head control parameters with porosity weight factors, and output to the multi-spray head cooperative control unit.

[0081] The dynamic flow adaptation module receives the temperature and humidity data transmitted by the real-time data acquisition module, and analyzes the influence of environmental parameters on atomization efficiency through the humidity compensation sub-model. The humidity compensation sub-model is based on a nonlinear regression algorithm to map calibrated temperature and humidity data to atomization efficiency decay coefficients, quantifying the effect of increased droplet size caused by increased environmental humidity. The atomization efficiency decay coefficient represents the degree of deviation of atomization efficiency under current environmental conditions relative to standard conditions, providing a quantitative basis for subsequent flow correction.

[0082] When the initial flow reference value is dynamically corrected based on the atomization efficiency attenuation coefficient, a segmented linear weighting algorithm is used to adjust the flow of each nozzle. The initial flow reference value is calculated by a pre-set spraying coverage rate model according to the plate advancing speed, and reflects the theoretical spraying demand without environmental interference. The attenuation coefficient and the reference value are multiplied by each nozzle to generate a dynamic flow instruction set after humidity compensation, which includes the flow target value and the allowable fluctuation range of each nozzle after compensation.

[0083] When the dynamic flow instruction set after humidity compensation is subjected to spatial superposition operation with the substrate porosity distribution data, a mapping relationship between the nozzle position and the substrate surface area is established based on the spatial coordinate system. The substrate porosity distribution data is analyzed into a two-dimensional matrix by multispectral imaging technology, and each matrix element corresponds to the porosity value of the gypsum board surface grid unit. The flow target value in the dynamic flow instruction set and the corresponding grid porosity value are weighted and summed element by element to generate a nozzle control parameter with a porosity weight factor. The weight factor is determined according to the porosity-adsorption characteristic correlation curve, and the high porosity area corresponds to a higher flow weight to ensure that the coating penetration depth meets the process requirements. The finally generated nozzle control parameter is transmitted to the multi-nozzle cooperative control unit through the industrial bus to drive the nozzle actuator to adjust the spraying operation according to the regional characteristics.

[0084] The humidity compensation sub-model converts environmental parameters into quantifiable attenuation coefficients to provide input for the dynamic correction of the flow reference value; the weighted correction process combines the theoretical flow demand with environmental impact factors to generate a flow instruction that adapts to actual working conditions; the spatial superposition operation further integrates substrate characteristic data to accurately match the flow parameter to the surface microstructure. The three layers of data processing are closely linked to form a full-link adaptation mechanism from environmental perception to execution control.

[0085] Specifically, the functional gypsum board processing control system disclosed by the present application comprises:

[0086] The nozzle control parameter with the porosity weight factor output by the dynamic flow adaptation module is received; and the discrete nozzle control parameter is converted into a continuous flow distribution map through a spatial interpolation algorithm;

[0087] Based on the nozzle mechanical motion trajectory equation, the real-time atomization pressure gradient value on the moving path of each nozzle is calculated in combination with the continuous flow distribution map;

[0088] According to the real-time atomization pressure gradient value and the current position coordinates of the nozzle, the on-off timing pulse width of the corresponding nozzle is dynamically adjusted to generate a regional spraying control instruction.

[0089] The multi-nozzle coordinated control unit receives the nozzle control parameters with the porosity weight factor transmitted by the dynamic flow adaptation module, and the parameters include the target flow value of each nozzle in the preset coordinate system and the corresponding porosity weight information. The nozzle control parameters are transmitted in the form of data packets through the industrial bus, and each data packet is associated with a specific nozzle number and position code to ensure that the parameters correspond to the physical nozzles one by one. The porosity weight information is generated based on the substrate surface porosity distribution data and reflects the different requirements of different regions for paint adsorption capacity.

[0090] When processing the discrete nozzle control parameters through the spatial interpolation algorithm, the bilinear interpolation method is used to expand the discrete nozzle target flow value into a continuous two-dimensional flow distribution map. The flow distribution map covers the entire spraying area of the gypsum board, and each pixel point corresponds to a target flow value and a porosity weight. In the interpolation process, an edge smoothing algorithm is introduced to eliminate the flow discontinuity when the nozzle moving direction is switched, and the continuity of the spraying coverage is maintained. The continuous flow distribution map is aligned with the nozzle mechanical motion trajectory equation in the coordinate system, and the spraying area is mapped to the motion path coordinate system of the nozzle.

[0091] Based on the nozzle mechanical motion trajectory equation and the continuous flow distribution map, the real-time atomization pressure gradient value is calculated, the function relationship between the nozzle position and the time variable is established according to the moving speed and acceleration parameters of the nozzle, the target flow value at each position in the continuous flow distribution map is used to back-calculate the required atomization pressure value, and the pressure gradient curve along the motion path is generated. The pressure gradient curve is converted into real-time atomization pressure adjustment instructions through differential operation to control the nozzle to adjust the atomization pressure output according to the preset gradient during movement.

[0092] According to the real-time atomization pressure gradient value and the current position coordinates of the nozzle, the opening and closing timing pulse width of the corresponding nozzle is dynamically adjusted, the pressure gradient value is converted into a pulse width modulation signal, the modulation signal is synchronized with the nozzle motion position, the opening and closing duration of the electromagnetic valve is accurately controlled through the hardware timer, and the real-time matching of the spraying flow and the nozzle moving speed is realized. The regional spraying control instruction set integrates the pressure adjustment parameters and the pulse timing information, which is sent to each nozzle actuator through the driving circuit to complete the regional differential spraying operation.

[0093] The received nozzle control parameters are associated with physical devices through coding to provide structured input for subsequent processing; the spatial interpolation algorithm expands the discrete parameters into a continuous distribution map to eliminate the flow discontinuity on the motion path; the motion trajectory equation and the flow distribution map are combined to generate a pressure gradient curve, which converts the flow requirement into a pressure control instruction; the dynamic pulse adjustment converts the pressure instruction into a timing signal to drive the nozzle to perform accurate spraying. Each step realizes the layer-by-layer refinement of the control parameters through data conversion and coordinate mapping to form a complete control link from parameter reception to execution driving.

[0094] Specifically, the functional gypsum board processing control system comprises a self-adaptive compensation algorithm module, a dynamic flow adjustment module, a curing quality closed-loop regulation unit, a real-time data acquisition module, a historical process database and a humidity compensation sub-model.

[0095] The deviation vector is extracted from the quality deviation data sent by the curing quality closed-loop regulation unit, and combined with the environmental variables collected by the real-time data acquisition module to construct a multi-dimensional parameter space.

[0096] The compensation coefficient candidate set is generated by crossover and mutation, and the fitness score of each candidate set in the dynamic flow adjustment module is calculated based on the sliding time window mechanism.

[0097] The compensation coefficient with the highest fitness score is injected into the humidity compensation sub-model of the dynamic flow adjustment module, and the mapping relationship between the environmental variables and the compensation coefficient in the historical process database is updated.

[0098] The self-adaptive compensation algorithm module receives the quality deviation data sent by the curing quality closed-loop regulation unit, which includes the coating thickness uniformity and curing degree detection results. The deviation vector is separated from the quality deviation data by a feature extraction algorithm, which represents the deviation direction and amplitude of the current spraying parameters from the process standard. At the same time, the environmental variable data transmitted by the real-time data acquisition module is normalized and aligned with the deviation vector by timestamp to construct a multi-dimensional parameter space containing environmental parameters, quality deviation and historical compensation coefficients. The parameter space is stored in the historical process database in the form of a matrix, providing a training sample set for genetic algorithm optimization.

[0099] When generating the compensation coefficient candidate set by crossover and mutation, the genetic algorithm is used to initialize the population, and each set of candidate coefficients corresponds to a compensation strategy for environmental- quality deviation. The crossover operation selects parent parameter combinations based on fitness scores, and generates offspring candidate sets through single-point crossover; the mutation operation randomly adjusts some coefficients in the offspring parameters to increase population diversity. The sliding time window mechanism intercepts process data from a preset time period before the current time, calculates the quality deviation improvement rate of each candidate set after execution in the dynamic flow adjustment module, and uses it as the fitness score. The fitness score is generated by weighting and accumulating the convergence speed and stability indicators of each component of the deviation vector, and the compensation coefficient combination with the best overall performance is selected.

[0100] When the compensation coefficient with the highest fitness score is injected into the humidity compensation sub-model of the dynamic flow adjustment module, the weight parameters of the sub-model are dynamically adjusted in proportion to the compensation coefficient. The compensation coefficient is embedded in the humidity compensation algorithm through matrix multiplication operation to correct the calculation logic of the atomization efficiency decay coefficient. The historical process database is updated synchronously to map the relationship between the environmental variables and the compensation coefficient, and the new data record includes the current environmental parameters, the compensation coefficient and the quality detection results after injection, forming a closed-loop feedback training data set. The updated mapping relationship serves as the initial population boundary condition for subsequent genetic algorithm iteration, improving the parameter search efficiency.

[0101] The mass deviation data is fused with environmental variables to construct a parameter space, which provides an optimization basis for a genetic algorithm; a candidate set is generated through crossover and mutation, and the actual control effect is evaluated based on a time window to ensure the dynamic adaptability of the compensation strategy; the optimal compensation coefficient is injected into a humidity compensation sub-model and the database is updated, forming a two-way interaction of parameter optimization and historical experience accumulation. The data flow is from quality detection feedback to algorithm optimization, and then to dynamic flow control, forming a self-learning closed-loop regulation system to realize continuous iterative optimization of the spraying process parameters.

[0102] Specifically, the functional gypsum board processing control system comprises a multi-process cooperative execution unit.

[0103] The spraying flow correction signal generated by the solidification quality closed-loop regulation unit and the equipment state parameters are received.

[0104] A cooperative control digital twin model is constructed, and the spraying flow correction signal and the drying furnace temperature control instruction are time-aligned through a dynamic time warping algorithm.

[0105] An incremental fine-tuning algorithm is used to update the parameters of the drying temperature and conveying speed correlation model in the digital twin model.

[0106] According to the updated model parameters, a device cooperative control sequence is generated to synchronously adjust the reciprocating frequency of the spraying machine, the speed of the drying fan, and the acceleration of the discharging conveyor.

[0107] The multi-process cooperative execution unit receives the spraying flow correction signal and the equipment state parameters transmitted by the solidification quality closed-loop regulation unit. The spraying flow correction signal contains the flow adjustment coefficient generated by the PID controller, and the equipment state parameters integrate the real-time temperature of each temperature zone of the drying furnace, the conveying roller speed, and the motion position data of the spraying machine. The signals and parameters are transmitted through an industrial bus, and a data verification mechanism is used to ensure the integrity of the transmission, and a time stamp is used to realize the time sequence alignment of multi-source data.

[0108] When constructing the cooperative control digital twin model, the spraying flow correction signal and the drying furnace temperature control instruction are associated at the input end of the model. The dynamic time warping algorithm analyzes the time sequence difference between the two types of signals, and eliminates the time sequence misalignment caused by the communication delay between devices by stretching or compressing the time axis. The aligned signal sequence is mapped to the input layer of the digital twin model, and the dynamic correlation between the drying temperature and the conveying speed is constructed. The model is pre-trained based on historical steady-state process data to generate initial parameters, and the control instruction interface of the spraying machine, the drying furnace, and the conveying equipment is associated at the output layer of the model.

[0109] When updating the model parameters by using the incremental fine-tuning algorithm, the model loss function is calculated based on the real-time spraying quality detection data and the equipment state parameters. The momentum gradient descent algorithm is used to iteratively update the weights of the fully connected layer in small steps, so as to adjust the associated weights of the drying temperature gradient and the conveying acceleration. The elastic weight solidification technology is used to lock the neuron parameters in the model that are strongly associated with the historical process, so as to prevent overfitting caused by sudden working conditions. The fine-tuned model parameters are mapped to the correction amount of the temperature setting value of each temperature zone of the drying furnace and the adjustment threshold of the discharge conveying acceleration, and a parameter adjustment instruction set is generated.

[0110] When generating the equipment cooperative control sequence according to the updated model parameters, the parameter adjustment instruction set is analyzed by using the priority scheduling algorithm. The spraying machine reciprocating frequency is dynamically adjusted according to the corrected flow parameter, the drying fan speed and the temperature zone setting value are changed in linkage, and the discharge conveying acceleration is matched with the coating curing progress. The control sequence is synchronously sent to each actuator through the industrial bus, the action time sequence of the equipment is coordinated, and the uncured plate is prevented from being prematurely fed into the conveying section. The execution effect is detected again by using an infrared spectrometer, and the initial parameters of the model are optimized in a closed loop.

[0111] After the spraying flow correction signal and the equipment state parameters are checked and aligned, they are input into the digital twin model. The dynamic time warping eliminates the time sequence deviation. The incremental fine-tuning updates the model by combining the historical steady-state constraints and real-time data, and generates accurate control instructions. The priority scheduling algorithm coordinates the actions of multiple equipment, and guarantees the parameter coupling between processes. The closed-loop verification data continuously optimize the model, and form a whole-process cooperative control link from parameter adjustment to execution feedback.

[0112] Specifically, the functional gypsum board processing control system provided by the application comprises an incremental fine-tuning algorithm.

[0113] An elastic weight solidification technology is used to lock the neuron weights in the cooperative control digital twin model that are associated with the historical process database.

[0114] The momentum gradient descent fine-tuning is performed on the newly added real-time spraying quality data feature layer, and fine-tuned model parameters are generated.

[0115] The fine-tuned model parameters are mapped to the correction amount of the temperature setting value of each temperature zone of the drying furnace and the adjustment threshold of the discharge conveying acceleration.

[0116] The incremental fine-tuning algorithm applies an elastic weight solidification technology in the collaborative control digital twin model, quantifies the influence degree of neuron weights on the historical process database, and screens out key neurons with strong correlation with the steady-state process. An L2 regularization constraint is applied to the screened neurons based on the Fisher information matrix weight importance evaluation method to limit the weight update amplitude and prevent parameter drift caused by newly added real-time data. The technology ensures the stability of historical process knowledge during model updating and avoids degradation of model prediction ability under sudden working condition interference.

[0117] When fine-tuning the momentum gradient descent of the newly added real-time spraying quality data feature layer, the real-time collected coating thickness uniformity deviation data is used as the loss function input. The momentum term accumulates historical gradient direction information, smooths the parameter update trajectory, and suppresses the interference of high-frequency noise on the fine-tuning process. The fully connected layer weight parameters are updated iteratively with small steps to gradually adjust the response characteristics of the drying temperature and conveying speed associated model to new working conditions. The neuron weights locked by the elastic weight solidification technology are updated at a low learning rate during the fine-tuning process, balancing the contribution proportion of historical experience and real-time data.

[0118] When mapping the fine-tuned model parameters to device control instructions, the multi-dimensional parameter vector of the model output layer is decomposed into drying furnace temperature setting value correction amounts and discharge conveying acceleration adjustment thresholds through linear transformation. The temperature setting value correction amount is dynamically matched with the real-time detected coating solidification rate, and the discharge conveying acceleration adjustment threshold is calculated based on the temperature gradient of the drying furnace outlet plate. The mapped control instructions are sent to the device execution layer through the industrial bus to drive the drying fan speed, conveying roller motor, and spraying machine reciprocating mechanism to realize dynamic coupling of cross-process parameters.

[0119] The elastic weight solidification locks the key neurons, maintains the modeling ability of the model for historical processes, and fine-tunes the newly added feature layer under constraints through momentum gradient descent to adapt to real-time working condition changes. The parameter mapping converts the model output into device control quantities to form a closed-loop link from data updating to execution and control. The three-layer technology progresses in sequence, retains historical process knowledge, dynamically responds to real-time data, and finally realizes multi-device collaborative optimization through precise control instructions.

[0120] Specifically, the functional gypsum board processing control system disclosed by the present application further comprises:

[0121] When the input feature data exceeds the confidence interval of the historical process database, the parameter rollback mechanism of the collaborative control digital twin model is triggered, and the model parameter snapshot of the last stable production period is called to replace the current parameters.

[0122] The compensation coefficients optimized by the genetic algorithm in the adaptive compensation algorithm module are activated for temporary correction; after the system recovers to steady state, the learning process of the incremental fine-tuning algorithm is reset, and the neuron weights locked by the elastic weight solidification technology in the collaborative control digital twin model are reloaded.

[0123] When the input feature data exceeds the confidence interval of the historical process database, the system identifies feature space outliers based on the anomaly detection module constructed by the isolation forest algorithm, triggering the parameter rollback mechanism of the collaborative control digital twin model. The historical process database stores multiple model parameter snapshots of stable production cycles, including verified drying temperature and conveying speed associated model weights and device collaborative control parameters. The rollback mechanism calls the stable parameter snapshot closest to the current working condition timestamp, replaces the model parameters in the abnormal state, and restores to the historical reliable control state.

[0124] When the compensation coefficients optimized by the genetic algorithm in the adaptive compensation algorithm module are activated for temporary correction, a subset of compensation coefficients matching the current environmental variables is loaded from the historical process database. The compensation coefficients are selected by a sliding time window mechanism to obtain the parameter combination with the highest recent fitness score, and are injected into the humidity compensation sub-model of the dynamic flow adaptation module. The compensation coefficients adjust the calculation logic of the atomization efficiency decay coefficient through matrix operation, generate temporary flow correction instructions, and suppress the expansion of coating quality deviation in abnormal conditions.

[0125] When the infrared spectrometer detects that the coating curing degree and thickness uniformity data have regressed to the preset quality threshold range, it is determined that the system has recovered to steady state. The learning process of the incremental fine-tuning algorithm is reset, the gradient update data accumulated during the abnormal period is cleared, and the initial learning rate of the momentum gradient descent algorithm is restored. The neuron weights locked by the elastic weight solidification technology in the collaborative control digital twin model are reloaded, the temporary parameter rollback state is removed, and the modeling ability of the model on historical process data is restored. The elastic weight quantifies the importance score through the Fisher information matrix, re-imposes L2 regularization constraints, and ensures the stability of subsequent incremental learning.

[0126] After the anomaly detection module identifies input data anomalies, the parameter rollback mechanism quickly restores the historical reliable control state; the compensation coefficients optimized by the genetic algorithm temporarily correct the flow parameters to prevent quality deterioration; after the steady state is restored, the learning process is reset and the elastic weight is loaded to balance the needs of abnormal recovery and continuous optimization.

[0127] Real-time data acquisition module: This module is deployed at the inlet of the spraying machine and the feeding conveying section, integrating speed sensor, temperature and humidity sensor, and multi-spectral imaging unit. The speed sensor captures the conveying roller speed pulse signal through a rotary encoder, which is converted into real-time board running speed data through differential operation; the temperature and humidity sensor array collects environmental temperature and humidity raw signals, uses Kalman filtering algorithm to eliminate noise interference, and outputs calibrated temperature and humidity values; the multi-spectral imaging unit scans the surface of the gypsum board, processes the optical signals through convolutional neural network, and analyzes the porosity distribution thermal map and adsorption characteristic quantitative value. After time stamp alignment and spatial coordinate mapping, the data form a three-dimensional feature matrix, providing structured input for subsequent modules.

[0128] Dynamic flow adaptation module: based on the received board speed data, the preset spraying coverage model is matched to generate initial flow reference values for each spray head. The preset model reflects the theoretical spraying requirements at different speeds by establishing a speed-flow mapping relationship through historical process data training. The humidity compensation sub-model receives calibrated temperature and humidity data, calculates the atomization efficiency decay coefficient using a nonlinear regression algorithm, and performs weighted correction on the initial flow reference values to generate a humidity-compensated dynamic flow instruction set. Further, by integrating the substrate porosity distribution data and performing matrix superposition operation in the spatial coordinate system, the spray head control parameters with porosity weight factors are generated to accurately match the substrate surface characteristics.

[0129] The technical features of the technical solution of the present application are explained as follows:

[0130] Pre-set spraying coverage model:

[0131] Function positioning: deployed in the dynamic flow adaptation module, establishes the mapping relationship between the board running speed and the theoretical spraying flow.

[0132] Technical essence: based on historical process data training, a speed-flow correlation matrix is generated, which maps the real-time collected board speed data into initial flow reference values for each spray head. The model generates ideal spraying parameters without environmental interference by offline learning the coating coverage area and penetration depth data at different speed intervals, providing baseline input for subsequent dynamic correction.

[0133] Humidity compensation sub-model:

[0134] Function positioning: the core algorithm unit of the dynamic flow adaptation module, quantifying the influence of environmental temperature and humidity on atomization efficiency.

[0135] Technical essence: a nonlinear regression algorithm is used to construct a temperature and humidity-atomization efficiency decay coefficient mapping table. The calibrated temperature and humidity data are input, and the decay coefficient representing the increase in atomized droplet size caused by environmental factors is output. This coefficient is weighted with the initial flow reference value to dynamically correct the spray head flow instruction, suppressing the uneven coating atomization problem in high humidity environments.

[0136] PID controller (proportional-integral-derivative controller):

[0137] Function orientation: The core control algorithm of the curing quality closed-loop regulation unit, realizing the closed-loop feedback of quality deviation to flow correction.

[0138] Technical essence: Receive the coating thickness uniformity deviation vector detected by the infrared spectrometer, generate the spraying flow correction coefficient through the proportional term to quickly respond to the deviation amplitude, the integral term to eliminate the steady-state error, and the differential term to predict the deviation trend. The correction coefficient is injected into the dynamic flow adaptation module in reverse, which adjusts the nozzle flow instruction in real time, forming a deviation dynamic suppression mechanism.

[0139] Genetic algorithm optimization model:

[0140] Function orientation: The core optimization engine of the adaptive compensation algorithm module, which excavates the implicit correlation between environmental variables and quality deviation.

[0141] Technical essence: With the environmental variables (temperature and humidity, substrate parameters) and quality deviation data in the historical process database as input, a multi-dimensional parameter space is constructed. Through cross variation, a candidate set of compensation coefficients is generated, and the fitness score (deviation convergence speed, stability) is evaluated by using a sliding time window to intercept real-time process data. The optimal compensation coefficient is injected into the humidity compensation sub-model to realize adaptive correction of control parameters.

[0142] Collaborative control digital twin model:

[0143] Function orientation: The core simulation model of the multi-process collaborative execution unit, realizing cross-process parameter coupling and timing coordination.

[0144] Technical essence: Based on the physical characteristics of the equipment, a virtual mapping model is constructed, inputting the spraying flow correction signal, drying temperature control instruction, and conveying speed data, and aligning the timing of multi-source signals through dynamic time warping algorithm. The model outputs the associated parameters of drying temperature and conveying speed, and uses an incremental fine-tuning algorithm (elastic weight consolidation + momentum gradient descent) to update the weights, generating a collaborative control sequence of the equipment, solving the coating defects caused by parameter conflicts between processes.

[0145] Elastic weight consolidation technology:

[0146] Function orientation: The constraint mechanism of the incremental fine-tuning algorithm, which guarantees the stability of model optimization.

[0147] Technical essence: Quantify the importance of neuron weights in the neural network model through the Fisher information matrix, impose L2 regularization constraints on the strongly correlated neurons in historical process data, and limit the weight update amplitude. This technology prevents the over-interference of new real-time data on the historical process knowledge model, balancing the contradiction between experience inheritance and dynamic adaptability.

[0148] Data-driven: A preset spray coverage model provides baseline parameters, a humidity compensation sub-model is superimposed with environmental corrections, a PID controller provides feedback on quality deviations, a genetic algorithm mines hidden patterns to optimize compensation coefficients, and a digital twin model coordinates parameters across multiple processes.

[0149] Anomaly tolerance: Elastic weight solidification technology locks in key parameters, genetic algorithm provides temporary compensation, and parameter rollback mechanism quickly restores steady state, forming a multi-layered anomaly response system.

[0150] Knowledge iteration: The historical process database continuously stores the optimized parameter mapping relationship, supports the initialization of the genetic algorithm population and the pre-training of the digital twin model, and realizes the accumulation of process knowledge and the self-evolution of the system.

[0151] Technical effect mapping: Through the hierarchical linkage of the above model, the problems of coating quality fluctuation and production cycle imbalance caused by the lag in dynamic adaptation of spraying flow, insufficient response to environmental interference, and parameter conflicts between processes are systematically solved.

[0152] On the functional gypsum board production line, a real-time data acquisition module deploys a rotary encoder in the feeding conveyor section to capture the rotational speed pulse signals of the conveyor rollers in real time. This data is then converted into board travel speed data through differential calculations. A temperature and humidity sensor array is installed at the spraying machine inlet. After collecting the raw ambient temperature and humidity signals, a Kalman filter algorithm is used to eliminate high-frequency noise interference, generating calibrated temperature and humidity values. The substrate characteristic detection unit integrates multispectral imaging equipment to optically scan the gypsum board surface. The raw images are processed using a convolutional neural network to output a porosity distribution heatmap and quantitative values ​​of adsorption characteristics. The calibrated speed, temperature, and humidity data are synchronized with the porosity heatmap via a timestamp alignment module, forming a three-dimensional feature matrix containing motion parameters, environmental conditions, and substrate characteristics. This matrix is ​​then transmitted to the dynamic flow adaptation module via industrial Ethernet.

[0153] The dynamic flow adaptation module calls a preset spray coverage model, matches the theoretical flow requirement based on the material's travel speed, and generates initial flow reference values ​​for each nozzle. The humidity compensation sub-model receives calibrated temperature and humidity data, calculates the atomization efficiency attenuation coefficient using a nonlinear regression algorithm, and weights the initial flow reference values ​​to generate a humidity-compensated dynamic flow instruction set. This instruction set is then matrix-superimposed with the substrate porosity distribution data in a spatial coordinate system to generate nozzle control parameters with porosity weighting factors, which are transmitted to the multi-nozzle collaborative control unit via an industrial bus. The multi-nozzle collaborative control unit uses a bicubic spline interpolation algorithm to convert discrete parameters into a continuous flow distribution map, calculates the real-time atomization pressure gradient value based on the nozzle's mechanical motion trajectory equation, and dynamically adjusts the solenoid valve opening and closing timing pulse width to achieve differentiated spraying in different regions. The spraying execution parameters are fed back to the curing quality closed-loop control unit via the industrial bus, forming a feedforward-feedback composite control loop.

[0154] The curing quality closed-loop control unit deploys an infrared spectrometer at the drying oven outlet to detect coating curing degree and thickness uniformity data. It extracts characteristic frequency band energy values ​​through Fourier transform and performs deviation analysis against a preset quality threshold to generate a multi-dimensional quality deviation vector. The PID controller converts the deviation vector into a spray flow correction coefficient, triggering the adaptive compensation algorithm module to initiate a genetic algorithm optimization process. Environmental variables and quality deviation correlation data stored in the historical process database construct a multi-dimensional parameter space. A candidate set of compensation coefficients is generated through cross-mutation, and a sliding time window mechanism is used to evaluate the fitness score, selecting the optimal compensation coefficient for injection into the humidity compensation sub-model. The optimized parameters are synchronized to the multi-process collaborative execution unit to construct a collaborative control digital twin model. A dynamic time warping algorithm aligns the spray flow correction signal with the drying temperature control command timing. An incremental fine-tuning algorithm updates the temperature-speed correlation model parameters, generating a collaborative control sequence. A priority scheduling algorithm parses the command sequence and dynamically adjusts the sprayer reciprocating frequency, drying fan speed, and conveyor acceleration to prevent uncured boards from prematurely entering the conveyor section. Under abnormal operating conditions, the collaborative control digital twin model triggers a parameter rollback mechanism, calls historical stable parameter snapshots and activates the genetic algorithm compensation coefficients for temporary correction. After the infrared spectrometer detection data returns to the threshold range, the incremental learning process is reset and the neuron parameters that are fixed by elastic weights are reloaded, thereby achieving a dynamic balance between production cycle and process quality and the system's self-healing capability.

[0155] This invention utilizes a real-time data acquisition module to deploy speed sensors, temperature and humidity sensors, and a multispectral imaging unit at the feeding conveyor section and the spraying machine inlet, simultaneously acquiring data on the material's travel speed, ambient temperature and humidity, and substrate porosity distribution. A dynamic flow adaptation module generates an initial flow baseline value based on the speed data and a preset spray coverage model. It then calculates the atomization efficiency attenuation coefficient using a humidity compensation sub-model, weights and corrects the baseline value, and integrates a substrate porosity weighting factor to generate a dynamic flow instruction set. This instruction set is converted into a continuous flow distribution map using a spatial interpolation algorithm, driving a multi-nozzle collaborative control unit to adjust the atomization pressure and opening / closing sequence according to regional characteristics, achieving precise adaptation of spraying parameters to substrate characteristics.

[0156] The curing quality closed-loop control unit deploys an infrared spectrometer at the drying oven outlet to detect the coating curing degree and thickness uniformity, generating quality deviation data and using a PID algorithm to reversely correct the spraying flow parameters. The adaptive compensation algorithm module calls upon environmental variable and quality deviation correlation data from the historical process database to construct a multi-dimensional parameter space. It uses a genetic algorithm for crossover and mutation to generate a candidate set of compensation coefficients, and a sliding time window mechanism selects the compensation coefficient with the highest fitness score, injecting it into the humidity compensation sub-model for dynamic correction of control parameters. The optimized compensation coefficients are synchronously updated to the historical database, forming a closed-loop iterative accumulation of process knowledge.

[0157] A digital twin model is constructed using a multi-process collaborative execution unit. The spraying flow correction signal and the drying oven temperature control command are time-aligned using a dynamic time warping algorithm, and the temperature-speed correlation model parameters are updated using an incremental fine-tuning algorithm. Based on the collaborative control sequence output by the model, the reciprocating frequency of the spraying machine, the speed of the drying fan, and the conveyor acceleration are synchronously adjusted through a priority scheduling algorithm to dynamically balance the coating curing quality and production cycle time. Under abnormal operating conditions, a parameter rollback mechanism is triggered, which calls a snapshot of historical stable parameters and activates the genetic algorithm compensation coefficient for temporary correction. After the system returns to steady state, the learning process is reset, achieving collaborative optimization of cross-process control robustness and adaptability.

Claims

1. A functional gypsum board processing control system, characterized in that, It includes a real-time data acquisition module, a dynamic flow adaptation module, a multi-nozzle collaborative control unit, a curing quality closed-loop control unit, an adaptive compensation algorithm module, and a multi-process collaborative execution unit; The real-time data acquisition module is deployed at the feeding conveyor section and the spraying machine inlet. It is used to collect data on the board's travel speed, ambient temperature and humidity, as well as the surface porosity and adsorption characteristics of the substrate, and transmit them to the dynamic flow adaptation module. The dynamic flow adaptation module generates an initial flow reference value for each nozzle based on the received plate speed data and a preset spray coverage model. It then performs humidity compensation correction on the initial flow reference value by combining the ambient temperature and humidity data and outputs the compensated dynamic flow command to the multi-nozzle collaborative control unit. The multi-nozzle collaborative control unit generates a dynamic flow distribution map based on the substrate porosity distribution data, controls the nozzles to adjust the atomization pressure and opening and closing time differently during the movement, and feeds the execution parameters back to the curing quality closed-loop control unit. The curing quality closed-loop control unit generates quality deviation data by detecting the curing degree and thickness uniformity of the coating, uses the PID algorithm to reverse correct the spraying flow parameters and triggers the adaptive compensation algorithm module. The adaptive compensation algorithm module calls the environmental variable and quality deviation correlation data in the historical process database, optimizes the compensation coefficient matrix through a genetic algorithm to dynamically correct the control parameters, and synchronizes the optimization results to the multi-process collaborative execution unit. The multi-process collaborative execution unit coordinates the timing of the spraying machine, drying oven, and conveying equipment based on optimized control parameters.

2. The functional gypsum board processing control system according to claim 1, characterized in that, The dynamic traffic adaptation module includes: The temperature and humidity data transmitted by the real-time data acquisition module are input into the humidity compensation sub-model in the dynamic flow adaptation module to calculate the atomization efficiency attenuation coefficient. The initial flow rate reference value is weighted and corrected based on the atomization efficiency attenuation coefficient to generate a dynamic flow rate instruction set after humidity compensation. The dynamic flow instruction set and the substrate porosity distribution data obtained by the real-time data acquisition module are matrix superimposed in a spatial coordinate system to generate nozzle control parameters with porosity weighting factors, and then output to the multi-nozzle collaborative control unit.

3. The functional gypsum board processing control system according to claim 2, characterized in that, The multi-nozzle collaborative control unit includes: Receive nozzle control parameters with porosity weighting factor output by the dynamic flow adaptation module; convert the discrete nozzle control parameters into a continuous flow distribution map through a spatial interpolation algorithm; Based on the mechanical motion trajectory equation of the nozzle, the real-time atomization pressure gradient value on the movement path of each nozzle is calculated in conjunction with the continuous flow distribution diagram. Based on the real-time atomization pressure gradient value and the current position coordinates of the nozzle, the opening and closing timing pulse width of the corresponding nozzle is dynamically adjusted to generate area spraying control commands.

4. The functional gypsum board processing control system according to claim 1, characterized in that, The adaptive compensation algorithm module includes: The deviation vector is extracted from the quality deviation data sent by the solidification quality closed-loop control unit, and combined with the environmental variables collected by the real-time data acquisition module to construct a multi-dimensional parameter space; A candidate set of compensation coefficients is generated by crossover mutation, and the fitness score of each candidate set in the dynamic traffic adaptation module is calculated based on the sliding time window mechanism. The compensation coefficient with the highest fitness score is injected into the humidity compensation sub-model of the dynamic flow adaptation module, and the mapping relationship between environmental variables and compensation coefficients in the historical process database is updated.

5. The functional gypsum board processing control system according to claim 1, characterized in that, The multi-process collaborative execution unit includes: Receive the spray flow correction signal and equipment status parameters generated by the curing quality closed-loop control unit; A collaborative control digital twin model is constructed, and the spray flow correction signal and the drying oven temperature control command are time-aligned using a dynamic time warping algorithm. The parameters of the drying temperature and conveying speed correlation model in the digital twin model are updated using an incremental fine-tuning algorithm; Based on the updated model parameters, a collaborative control sequence for the equipment is generated, which synchronously adjusts the reciprocating frequency of the spraying machine, the speed of the drying fan, and the acceleration of the discharge conveyor.

6. The functional gypsum board processing control system according to claim 5, characterized in that, The incremental fine-tuning algorithm includes: The neuron weights associated with the historical process database in the collaborative control digital twin model are locked using an elastic weight solidification technique. The momentum gradient descent is fine-tuned on the newly added real-time spraying quality data feature layer to generate the fine-tuned model parameters. The fine-tuned model parameters are mapped to the temperature setpoint correction amount of each temperature zone of the drying oven and the discharge conveyor acceleration adjustment threshold.

7. The functional gypsum board processing control system according to claim 1, characterized in that, Also includes: When the input feature data exceeds the confidence interval of the historical process database, the parameter rollback mechanism of the collaborative control digital twin model is triggered, and the model parameter snapshot of the previous stable production cycle is called to replace the current parameters. Activate the compensation coefficients optimized by the genetic algorithm in the adaptive compensation algorithm module for temporary correction; After the system returns to a steady state, the learning process of the incremental fine-tuning algorithm is reset, and the neuron weights locked by the elastic weight solidification technology in the collaborative control digital twin model are reloaded.

Images