CN-122018471-A - Intelligent polyurethane production control method and system based on digital twin

CN122018471ACN 122018471 ACN122018471 ACN 122018471ACN-122018471-A

Abstract

The invention relates to the technical field of intelligent control, in particular to an intelligent control method and system for polyurethane production based on digital twinning. The method comprises the steps of constructing a digital twin simulation model corresponding to a full process flow of polyurethane production, synchronously accessing a real-time sensor data stream corresponding to a production line, dynamically assigning values based on the real-time sensor data stream to generate a digital twin model instance, performing prospective process simulation operation based on the digital twin model instance to generate multi-physical-field prediction data, extracting a prediction time sequence curve set of key process state parameters from the multi-physical-field prediction data, dynamically comparing the multi-physical-field prediction data at high frequency to generate a deviation attribution analysis report, generating a real-time optimization control instruction set based on the deviation attribution analysis report, and acquiring real response data of the production line after execution as real-time sensor data stream execution feedback of a new round to form a closed-loop intelligent control loop. The invention can comprehensively improve the polyurethane production regulation and control efficiency.

Inventors

SUN ZHENFEI
LI TIEHU
GE JINXING
Yin Dianxue
MA ZHENCHAO

Assignees

青岛仁成海绵制品有限公司

Dates

Publication Date: 20260512
Application Date: 20260326

Claims (10)

1. The intelligent control method for polyurethane production based on digital twinning is characterized by comprising the following steps: Step S1, constructing a digital twin simulation model corresponding to the whole process flow of polyurethane production, and synchronously accessing a real-time sensor data stream corresponding to a production line, wherein the real-time sensor data stream comprises the real-time temperature, the real-time pressure, the material viscosity, the isocyanate index and the polyol feeding rate in a reaction kettle; S2, injecting preset production formula parameters and a process set point sequence based on a digital twin model example to perform prospective process simulation operation to generate multi-physical-field prediction data covering a full reaction period; S3, carrying out high-frequency dynamic comparison on the predicted time sequence curve set and an actual process state parameter time sequence acquired from a production line in real time, and calculating a state deviation amount set of each parameter between a digital twin virtual space and a physical entity space at each comparison time; Step S4, a real-time optimal control instruction set aiming at a physical production line is generated based on that the deviation attribution analysis report matches with a current deviation attribution scene, the real-time optimal control instruction set is issued to a distributed control system of the polyurethane production line to be executed, physical production line equipment is driven to be dynamically adjusted, and meanwhile, the actual response data of the production line after being executed is fed back to the step S1 as a new round of real-time sensor data stream to form a closed-loop intelligent control loop.
2. The intelligent control method for producing polyurethane based on digital twinning according to claim 1, wherein the step S1 comprises the steps of: Step S11, a core unit of the whole process flow of polyurethane carding production comprises a raw material pretreatment unit, a reaction synthesis unit, a curing modification unit, a molding cutting unit and a quality detection unit, equipment geometric parameters, material properties, process constraint conditions and historical production operation data of all units are collected, an infrastructure of a digital twin simulation model is constructed, and a material transmission path, an energy exchange relationship and a time sequence linkage logic among all units are defined; Step S12, deploying a distributed sensor network, synchronously accessing real-time sensor data streams of all core units of a production line, performing real-time denoising processing, eliminating abnormal pulse data and timing missing data, and generating a purified real-time sensor data stream; s13, extracting real-time running state parameters of each core unit based on the purified real-time sensor data stream, establishing a parameter mapping relation, correspondingly assigning the real-time state parameters of the physical production line to each corresponding node of the digital twin simulation model, and generating an initial digital twin model instance; And S14, acquiring running state data of the initial digital twin model instance in real time, performing preliminary comparison with the purified real-time sensor data stream, calculating an assigned deviation of the initial state of the model, dynamically correcting initial state parameters of the initial digital twin model instance based on the assigned deviation, and generating a corrected digital twin model instance.
3. The intelligent control method for producing polyurethane based on digital twinning according to claim 1, wherein the step S2 comprises the steps of: S21, determining production formula parameters and time sequence process set point sequences based on preset specification requirements of polyurethane products, defining parameter control ranges, time sequence nodes and switching conditions of each process stage, and carrying out structural coding on the production formula parameters and the time sequence process set point sequences to generate process control instructions; S22, injecting a process control instruction into a digital twin model example, starting prospective process simulation operation, simulating a material reaction process, an energy transfer process and a device operation process of a full reaction period in polyurethane production, and generating multi-physical-field prediction data covering the full reaction period, wherein the multi-physical-field prediction data comprises time sequence change data of a temperature field, a pressure field, a material concentration field and a viscosity field in a reaction kettle; S23, screening out key process state parameters from the multi-physical-field predicted data, wherein the key process state parameters comprise a real-time temperature predicted value, a real-time pressure predicted value, a material viscosity predicted value, an isocyanate index predicted value, a polyol feeding rate predicted value and a reaction conversion rate predicted value in a reaction kettle, and extracting time sequence data of each key process state parameter in a full reaction period; step S24, based on the time sequence data of each key process state parameter, drawing a corresponding prediction time sequence curve, marking time sequence nodes, parameter change trend and key characteristic points of each curve, and generating a prediction time sequence curve set of the key process state parameter; and S25, performing time sequence consistency check on the predicted time sequence curve set, calculating time sequence synchronization errors among the curves, correcting the predicted time sequence curve set based on the time sequence synchronization errors, and ensuring that the predicted time sequence of each key process state parameter is consistent with the actual production time sequence.
4. A digital twin based polyurethane production intelligent control method according to claim 3, wherein step S23 comprises the steps of: Determining the type of a key process state parameter affecting the quality and the production stability of a product based on a polyurethane production reaction mechanism, defining the detection standard and the time sequence acquisition requirement of each key parameter, and establishing a key parameter screening criterion; Traversing the multi-physical field prediction data, extracting time sequence data corresponding to each key process state parameter according to a key parameter screening criterion, removing invalid prediction data and abnormal prediction data exceeding a process control range to generate a key parameter original time sequence data set; Calculating the change rate and time sequence fluctuation amplitude of each key process state parameter in different reaction stages based on the continuous key parameter time sequence data set, extracting the change characteristics of each parameter, and generating a key parameter change characteristic data set; according to the key parameter change characteristic data set, the predicted time sequence data of each key process state parameter is matched, and the time sequence data of each key process state parameter in the full reaction period is extracted.
5. The intelligent control method for producing polyurethane based on digital twin according to claim 4, wherein calculating the change rate and the time sequence fluctuation amplitude of each key process state parameter in different reaction phases based on the continuous key parameter time sequence data set, and extracting the change characteristics of each parameter comprises the following steps: Dividing the continuous key parameter time sequence data set according to reaction stages of polyurethane production, and defining the time sequence interval and the time sequence length of each reaction stage to generate a staged key parameter time sequence data set; calculating the time sequence change rate of each key process state parameter in each reaction phase based on the time sequence data set of the key parameters in stages, obtaining the parameter change rate value at each time through a time sequence difference algorithm, and generating the time sequence change rate data set of the key parameters; Calculating the fluctuation amplitude of the change rate of each key process state parameter in each reaction phase according to the key parameter staged change rate data set, determining the stability characteristics of each parameter change, and generating a key parameter fluctuation characteristic data set; Extracting fluctuation peak values, fluctuation frequency and time distribution rules of state parameters of each key process in different reaction stages based on the key parameter fluctuation characteristic data set, and generating key parameter fluctuation characteristic indexes; And integrating the key parameter staged change rate data set, the key parameter fluctuation feature data set and the key parameter fluctuation feature index to generate a key parameter change feature data set, and determining the time sequence change rule and the fluctuation characteristic of each key parameter.
6. The intelligent control method for producing polyurethane based on digital twinning according to claim 1, wherein the step S3 comprises the steps of: Step S31, establishing a time synchronization mapping channel between each prediction time sequence curve and an actual process state parameter time sequence acquired from a production line in real time, carrying out point-by-point difference value operation on each prediction time sequence curve and a corresponding actual process state parameter time sequence on the basis of the time synchronization mapping channel, calculating instantaneous absolute deviation and relative deviation between a predicted value and an actual measured value, and carrying out time sequence integration to generate a state deviation value set; S32, carrying out time-frequency analysis on the state deviation amount set, extracting fluctuation modes, trend components and mutation features of each parameter deviation signal on different time scales, and clustering deviation signals with similar fluctuation modes and trend features to generate a deviation mode feature vector set; Step S33, a reverse mapping network corresponding to a drift factor is included through pre-training, and a complex nonlinear mapping relation between a typical deviation mode and a series of potential reasons is learned, wherein the disturbance source at least comprises raw material component fluctuation, catalyst deactivation and temperature control loop disturbance, and the drift factor at least comprises a stirring efficiency attenuation factor and a heat transfer coefficient drift factor; And step S34, inputting the characteristic vector set of the deviation modes into a reverse mapping network to calculate and output the probability that the potential process disturbance source corresponding to each deviation mode exists and the equipment performance drift factor and the quantized contribution degree estimated value of the potential process disturbance source to the current overall deviation, and integrating all the probability and the quantized contribution degree estimated value to generate a deviation attribution analysis report.
7. The intelligent control method for producing polyurethane based on digital twinning according to claim 1, wherein the step S4 comprises the steps of: Step 41, establishing a bias attribution scene matching library, covering bias scenes and corresponding control strategies corresponding to various process disturbance and equipment drift, comparing attribution results in a bias attribution analysis report with the bias attribution scene matching library, and determining attribution scenes and control directions corresponding to the current bias; Step S42, based on the current deviation attribution scene and the control direction, combining the real-time running state of the digital twin model example, generating real-time optimization control instructions aiming at each core unit of the physical production line, wherein the real-time optimization control instructions comprise a raw material feeding rate adjustment instruction, a reaction kettle temperature and pressure regulation instruction and an equipment running parameter correction instruction, and integrating all control instructions to generate a real-time optimization control instruction set; S43, inputting a real-time optimized control instruction set into a digital twin model instance for virtual simulation verification, simulating the line response effect after the control instruction is executed, and calculating the predicted deviation amount after the control instruction is executed; if the predicted deviation amount does not meet the preset control requirement, correcting the real-time optimized control instruction set based on the predicted deviation amount, and repeating the virtual simulation verification step until the predicted deviation amount meets the requirement; And S44, acquiring actual response data of the production line after the execution of the corresponding control instruction meeting the requirements in real time, feeding back the actual response data to the step S1 as a new round of real-time sensor data flow, updating the state parameters of the digital twin model instance, and starting the simulation, comparison, attribution and control flow of the next round to form a closed-loop intelligent control loop.
8. The intelligent control method for producing polyurethane based on digital twinning according to claim 7, wherein the step S43 comprises the steps of: extracting all control instruction parameters in a real-time optimized control instruction set, defining an execution object, an execution time sequence, an adjustment amplitude and an execution time length of all control instructions, carrying out structural analysis on the control instruction parameters, and generating a virtual control instruction capable of being injected into a digital twin model instance; injecting a virtual control instruction into the current digital twin model instance, keeping other parameters of the model consistent with the real-time state of the physical production line, starting virtual simulation verification, simulating the state change of the digital twin model instance in the execution process of the control instruction, and generating virtual response data; The method comprises the steps of extracting time sequence change data of each key process state parameter from virtual response data, drawing a virtual response time sequence curve, comparing the curve with a preset process standard time sequence curve, calculating a predicted deviation amount after virtual response control instruction execution, analyzing distribution characteristics and time sequence change trend of the predicted deviation amount, judging whether the predicted deviation amount is within a preset control requirement range, if the predicted deviation amount meets the preset control requirement, issuing a real-time optimized control instruction set to a distributed control system of a polyurethane production line, driving each device of the physical production line to dynamically adjust according to the control instruction, if the predicted deviation amount exceeds the range, marking abnormal deviation amount exceeding the control requirement, time sequence nodes and corresponding virtual control instructions, correcting the deviation nodes and the corresponding virtual control instructions based on the predicted deviation amount, and repeating virtual simulation verification steps until the predicted deviation amount meets the requirement.
9. The intelligent control method for polyurethane production based on digital twin according to claim 8, wherein if the range is out of range, marking the abnormal deviation amount and the time sequence node which are out of control requirements and the corresponding virtual control instruction comprises the following steps: Determining a deviation threshold range corresponding to a preset control requirement, defining an upper limit and a lower limit of the allowable deviation of each key process state parameter, and establishing a deviation judgment standard; Analyzing a time sequence node and a key parameter type corresponding to the abnormal deviation amount, determining a target control instruction which causes the abnormal deviation amount to be generated by combining the execution time sequence of the virtual control instruction, and determining the association relation among the adjustment amplitude, the execution time sequence and the abnormal deviation amount of the target control instruction, wherein the method specifically comprises the following steps: Determining a virtual control instruction execution stage corresponding to an abnormal deviation amount based on a time sequence node corresponding to the time sequence node, determining all control instructions being executed in the stage and the execution state of each instruction, analyzing the change rule of key process state parameters in the execution process of each control instruction based on virtual response data of a digital twin model instance, establishing a correlation model between the control instructions and parameter changes, inputting the virtual control instruction execution stage corresponding to the time sequence node into the correlation model, inverting the control instruction causing the abnormal change of the parameter to determine a target control instruction, extracting the adjustment amplitude and the execution time sequence of the target control instruction, analyzing the quantization relation between the adjustment amplitude and the execution time sequence of the target control instruction and the abnormal deviation amount, calculating the influence coefficient of the parameter change of the control instruction on the abnormal deviation amount, and determining the correlation relation among the adjustment amplitude and the execution time sequence of the target control instruction and the abnormal deviation amount based on the influence coefficient; calculating the difference between the abnormal deviation amount and the deviation threshold value, determining the deviation exceeding degree, generating abnormal deviation analysis data, and determining the severity level and the time sequence propagation range of the abnormal deviation; Based on the abnormal deviation analysis data, judging whether the current real-time optimal control instruction set needs to be corrected, and if so, marking the control instruction and the correction direction which need to be corrected.
10. A digital twin based intelligent control system for polyurethane production, comprising a processor, a memory and a computer program stored on the memory and executable on the processor for performing the digital twin based intelligent control method for polyurethane production as claimed in any one of claims 1 to 9.

Description

Intelligent polyurethane production control method and system based on digital twin Technical Field The invention relates to the technical field of intelligent control, in particular to an intelligent control method and system for polyurethane production based on digital twinning. Background In recent years, with the wide application of polyurethane materials in a plurality of fields such as buildings, automobiles, household appliances and the like, the market has higher requirements on the performance consistency, the production efficiency and the green low-carbon level of polyurethane products, and meanwhile, the polyurethane production process has the characteristics of strong nonlinearity, multi-parameter coupling and complex reaction mechanism, and relates to the cooperative regulation and control of a plurality of key parameters such as the temperature, the pressure, the material viscosity, the isocyanate index, the polyol feeding rate and the like of a reaction kettle. At present, some polyurethane production control methods based on digital technology are proposed, and most of the methods collect real-time operation data by deploying various sensors on a production line, combine a fixed parameter regulation and control mode and an experience feedback mode, start and stop key equipment or fine adjustment parameters in the production process, and then perform preliminary optimization on a production process according to historical production data and an experience model, however, the existing methods have insufficient global simulation and dynamic mapping capability on the whole production process, are influenced by factors such as raw material batch fluctuation, multiple physical field coupling in a reaction kettle, equipment performance drift and the like, and easily cause deviation between sensor collection data and actual reaction state, so that process parameter regulation and control are lagged. Disclosure of Invention Based on the above, the present invention is needed to provide an intelligent control method and system for polyurethane production based on digital twinning, so as to solve at least one of the above technical problems. In order to achieve the above purpose, the intelligent control method for polyurethane production based on digital twin comprises the following steps: Step S1, constructing a digital twin simulation model corresponding to the whole process flow of polyurethane production, and synchronously accessing a real-time sensor data stream corresponding to a production line, wherein the real-time sensor data stream comprises the real-time temperature, the real-time pressure, the material viscosity, the isocyanate index and the polyol feeding rate in a reaction kettle; S2, injecting preset production formula parameters and a process set point sequence based on a digital twin model example to perform prospective process simulation operation to generate multi-physical-field prediction data covering a full reaction period; S3, carrying out high-frequency dynamic comparison on the predicted time sequence curve set and an actual process state parameter time sequence acquired from a production line in real time, and calculating a state deviation amount set of each parameter between a digital twin virtual space and a physical entity space at each comparison time; Step S4, a real-time optimal control instruction set aiming at a physical production line is generated based on that the deviation attribution analysis report matches with a current deviation attribution scene, the real-time optimal control instruction set is issued to a distributed control system of the polyurethane production line to be executed, physical production line equipment is driven to be dynamically adjusted, and meanwhile, the actual response data of the production line after being executed is fed back to the step S1 as a new round of real-time sensor data stream to form a closed-loop intelligent control loop. Further, step S1 includes the steps of: Step S11, a core unit of the whole process flow of polyurethane carding production comprises a raw material pretreatment unit, a reaction synthesis unit, a curing modification unit, a molding cutting unit and a quality detection unit, equipment geometric parameters, material properties, process constraint conditions and historical production operation data of all units are collected, an infrastructure of a digital twin simulation model is constructed, and a material transmission path, an energy exchange relationship and a time sequence linkage logic among all units are defined; Step S12, deploying a distributed sensor network, synchronously accessing real-time sensor data streams of all core units of a production line, performing real-time denoising processing, eliminating abnormal pulse data and timing missing data, and generating a purified real-time sensor data stream; s13, extracting real-time running state parameters of each core unit based on the purified real-