Polyurethane material production parameter monitoring and adjusting method and system

Abstract

The invention relates to the technical field of polyurethane material production, and particularly discloses a method and a system for monitoring and adjusting production parameters of a polyurethane material. According to the system, temperature, flow, rotating speed and pressure data are acquired in real time through a process parameter acquisition module, a performance index prediction module synchronously calculates predicted values of material hardness, elongation and wear resistance based on a deep learning model, and a multi-objective optimization decision module dynamically generates a collaborative optimization parameter set according to performance deviation and priority. And the actuator driving module converts the signal into a control signal to adjust production equipment. According to the invention, precise closed-loop control of key performance indexes of the polyurethane material is realized, and the product quality consistency and the intelligent production level are improved.

Description

Technical Field

[0001] This invention belongs to the field of polyurethane material production technology, specifically relating to a method and system for monitoring and adjusting polyurethane material production parameters. Background Technology

[0002] Polyurethane materials, as an important class of synthetic polymers, are widely used in various industrial products such as rubber wheels, screens, and liners due to their excellent mechanical properties, wear resistance, and chemical stability. The production process of these materials involves the polymerization reaction of raw materials such as polyols and isocyanates, and the precise control of its process parameters has a decisive impact on the performance of the final product.

[0003] Monitoring and adjusting the production parameters of polyurethane materials is a key technical aspect to ensure stable product quality. This technology aims to achieve dynamic monitoring and process optimization of the material synthesis state by real-time acquisition and analysis of key parameters such as temperature, pressure, and proportions during the production process.

[0004] Existing technologies typically employ single-parameter control strategies, which are ill-suited to the diverse performance requirements of different polyurethane products, such as hardness, elongation, and abrasion resistance. This control method fails to achieve synergistic optimization of multiple performance indicators, leading to a mismatch between actual product performance and design goals, severely impacting product lifespan and application effectiveness.

[0005] Therefore, how to achieve precise control of multiple performance indicators of polyurethane materials under complex and ever-changing production conditions has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The technical problem this invention aims to solve is to overcome the shortcomings of existing single-parameter control strategies in polyurethane material production processes, which are ill-suited to the synergistic optimization needs of multiple performance indicators. This invention provides a method and system for monitoring and adjusting polyurethane material production parameters. Its purpose is to construct a complete technical system capable of deeply integrating real-time process data and material performance indicators to achieve multi-objective dynamic optimization and precise control.

[0007] This invention provides a system for monitoring and adjusting production parameters of polyurethane materials. The system includes a process parameter acquisition module, a performance index prediction module, a multi-objective optimization decision-making module, and an actuator drive module.

[0008] The process parameter acquisition module is responsible for acquiring real-time temperature data inside the polyurethane synthesis reactor, instantaneous flow rate and cumulative ratio data of polyol and isocyanate in the reactor feed pipeline, speed and power data of the stirring motor, and pressure data inside the reaction system.

[0009] The performance index prediction module is based on the real-time data stream provided by the process parameter acquisition module. Through the built-in performance prediction model trained with massive historical production data, it simultaneously calculates the predicted hardness, predicted elongation, and predicted abrasion resistance index of the polyurethane material under the current process state.

[0010] The multi-objective optimization decision module receives multiple predicted performance indicators output by the performance indicator prediction module and compares and analyzes them with the target hardness range, target elongation lower limit, and target wear resistance threshold preset in the system database.

[0011] This module further dynamically generates a set of collaborative optimization parameters, including temperature adjustment amount, ratio fine-tuning coefficient and stirring speed correction value, based on preset priority rules and performance deviation tolerance.

[0012] The actuator drive module is connected to the temperature control valve, metering pump speed controller, and stirring motor frequency converter at the production site. It receives the collaborative optimization parameter set issued by the multi-objective optimization decision module and converts it into corresponding control signals to drive the aforementioned actuators to perform precise actions, thereby realizing closed-loop regulation of the polyurethane material production process.

[0013] As one embodiment of the present invention, the process parameter acquisition module specifically includes a temperature sensing unit, a flow metering unit, a speed monitoring unit, and a pressure sensing unit.

[0014] The temperature sensing unit uses multiple platinum resistance temperature sensors distributed on the inner wall of the reactor and in the core area of ​​the material. Its measurement accuracy is greater than ±0.5℃ and the sampling frequency is 2 times / s.

[0015] The flow metering unit uses a high-precision electromagnetic flow meter to monitor the instantaneous flow of the polyol and isocyanate feed lines, and integrates a pulse counter to calculate the cumulative flow, ensuring the dynamic accuracy of the ratio calculation.

[0016] The speed monitoring unit captures the speed signal in real time through a photoelectric encoder installed on the main shaft of the stirring motor.

[0017] The pressure sensing unit uses a corrosion-resistant piezoresistive sensor, which is directly immersed in the reactants to monitor changes in internal pressure of the system.

[0018] Furthermore, the performance prediction model relied upon by the performance indicator prediction module is a deep neural network model.

[0019] The input layer nodes of this model correspond to all real-time parameter dimensions acquired by the process parameter acquisition module.

[0020] The model contains three hidden layers, with 128, 64, and 32 neurons in each layer, respectively, and the ReLU activation function is used.

[0021] The output layer contains three nodes, which correspond to the predicted values ​​of three performance indicators: hardness, elongation, and abrasion resistance index.

[0022] The training process of this model is based on data samples of more than 10,000 historical production batches. The adaptive moment estimation algorithm is used for weight optimization, and the loss function is the mean square error function.

[0023] As one embodiment of the present invention, the internal decision-making logic of the multi-objective optimization decision module follows a multi-objective optimization algorithm based on weighted compromise programming.

[0024] The algorithm assigns appropriate weight coefficients to the three optimization objectives: hardness deviation, insufficient elongation, and negative shift of wear resistance index.

[0025] The weighting coefficients are assigned based on the specific product type of polyurethane material. For rubber wheel products, the abrasion resistance index has the highest weight; for screen products, the elongation has the highest weight.

[0026] The module determines which set of collaborative optimization parameters to use by finding the solution that minimizes the weighted objective function value.

[0027] Furthermore, the multi-objective optimization decision-making module also integrates a rule engine. This rule engine predefines several emergency situation handling rules.

[0028] When any process parameter collected in real time, such as reaction temperature, exceeds its preset safety threshold upper or lower limit, the rule engine will immediately trigger and override the conventional multi-objective optimization algorithm, and prioritize sending an emergency adjustment command to the actuator driver module to ensure production safety.

[0029] As one embodiment of the present invention, the actuator driving module includes a signal conversion submodule and an execution verification submodule.

[0030] The signal conversion submodule is responsible for converting the digital instructions in the collaborative optimization parameter set into analog current signals of 4mA to 20mA or pulse width modulation signals of a specific frequency, according to the pre-calibrated control curve.

[0031] After the drive command is issued, the execution verification submodule will continuously monitor the actual feedback status of the target actuator, such as the opening feedback signal of the temperature control valve, and compare it with the expected status.

[0032] If the deviation persists for more than 5 seconds and does not converge, a fault alarm will be sent to the system's main controller.

[0033] This invention also provides a method for monitoring and adjusting polyurethane material production parameters based on the above system. The method includes the following steps: First, a sensor array deployed at the production site continuously collects temperature, flow rate, rotation speed, and pressure parameters during the polyurethane synthesis reaction process.

[0034] Next, the collected real-time process parameter sequence is input into the pre-trained performance prediction model, and the predicted hardness, predicted elongation and predicted abrasion resistance index of the polyurethane material under the current production batch are calculated and output in real time.

[0035] Next, the obtained multiple predicted performance indicators are compared with the preset target performance indicators. Using a multi-objective optimization algorithm, considering the priority of each performance indicator and process constraints, a set of production parameter adjustment schemes that minimize the overall performance deviation is calculated, namely the collaborative optimization parameter set.

[0036] Finally, the collaborative optimization parameter set is converted into corresponding control signals to drive actuators such as temperature control valves, metering pumps, and stirring motors to make precise adjustments, thereby achieving precise control of key performance indicators of polyurethane materials.

[0037] Furthermore, the update mechanism of the performance prediction model is as follows: every 24 hours, the system automatically compares the actual final performance test data of the production batch completed that day with the corresponding predicted value of the model, and calculates the model prediction error.

[0038] When the average prediction error over three consecutive days exceeds the preset model update threshold, such as 5%, the system automatically initiates the incremental learning process of the model, using new production data to fine-tune the model weights in order to maintain prediction accuracy.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention fundamentally changes the limitations of traditional single-parameter control by constructing a complete closed-loop system that integrates high-precision data acquisition, deep learning-based performance prediction, multi-objective dynamic optimization, and precise execution drive.

[0040] The system can observe the complex nonlinear mapping relationship between process parameter fluctuations and multiple performance indicators of the final material in real time, and make collaborative optimization decisions based on this, ensuring that polyurethane products meet design requirements in key properties such as hardness, elongation, and abrasion resistance.

[0041] This method and system improve the level of intelligence in polyurethane production and the consistency of product quality, effectively extend the service life of products, and broaden their application scope. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the overall technical architecture of the polyurethane material production parameter monitoring and adjustment system proposed in this invention; Figure 2This is a schematic diagram of the core principle framework of the performance index prediction module in this invention; Figure 3 This is a logical flowchart of the multi-objective optimization decision-making module in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between process parameter acquisition and actuator driving in this invention; Detailed Implementation

[0043] Please refer to the attached document. Figure 1 This embodiment details a specific implementation of a polyurethane material production parameter monitoring and adjustment system. This system is a complete technical system that deeply integrates real-time process data and material performance indicators to achieve multi-objective dynamic optimization and precise control. The core of the system consists of a process parameter acquisition module, a performance indicator prediction module, a multi-objective optimization decision-making module, and an actuator drive module. The modules interact with each other via industrial Ethernet and fieldbus, forming a high-speed and reliable data closed loop.

[0044] The process parameter acquisition module is the foundation for the system to perceive the status of the production site. This module specifically includes a temperature sensing unit, a flow metering unit, a speed monitoring unit, and a pressure sensing unit.

[0045] The temperature sensing unit adopts a distributed arrangement scheme, with a total of eight platinum resistance temperature sensors installed at different heights on the inner wall of the polyurethane synthesis reactor and in the core area of ​​the material.

[0046] These sensors are designed to achieve a measurement accuracy greater than ±0.5℃, with a uniform sampling frequency of 2 times / s. All temperature sensors are connected to an intrinsically safe signal conditioner via shielded twisted-pair cables. The conditioner amplifies and filters the weak resistance signals collected, then converts them into digital temperature values ​​via an analog-to-digital converter. These values ​​are periodically uploaded to the system data buffer via the industrial Ethernet protocol.

[0047] The flow metering unit is responsible for monitoring the material flow of polyols and isocyanates in the reactor feed line. High-precision electromagnetic flow meters are installed on two independent feed lines, with a measurement accuracy of 0.5% of the indicated value.

[0048] The electromagnetic flowmeter outputs a 4mA to 20mA analog current signal that is proportional to the instantaneous flow rate in real time.

[0049] Meanwhile, the pulse counter integrated inside the unit accumulates the pulses emitted by the flow meter to calculate the cumulative flow of each batch of production materials, thereby dynamically calculating and verifying the actual ratio of polyol to isocyanate.

[0050] The speed monitoring unit utilizes a photoelectric encoder directly mounted on the rear end of the mixing motor's main shaft. This encoder generates 1024 pulses per revolution, and by measuring the number of pulses per unit time, the real-time speed of the mixing motor is accurately calculated in r / min.

[0051] The pressure sensing unit uses a specially packaged piezoresistive sensor that is resistant to polyurethane corrosion, and its sensing diaphragm head is directly immersed in the material inside the reactor.

[0052] The sensor converts the sensed pressure signal into a millivolt-level voltage signal, which is then processed by a dedicated amplifier circuit to output a standard 4mA to 20mA current signal, corresponding to a pressure measurement range of 0MPa to 2.5MPa.

[0053] Please refer to the attached document. Figure 2 The performance index prediction module is the intelligent core of the system. Based on the real-time, multi-dimensional data stream provided by the process parameter acquisition module, it makes advanced predictions on the performance index of the final polyurethane material product.

[0054] The performance prediction model used in this module is a deep neural network model.

[0055] The input layer of the model has 10 nodes, which correspond to eight temperature measurement points, two instantaneous flow values, one rotation speed value and one pressure value obtained by the process parameter acquisition module. A total of 12 real-time parameter dimensions are normalized to form feature vectors.

[0056] The model contains three fully connected hidden layers: the first hidden layer has 128 neurons, the second hidden layer has 64 neurons, and the third hidden layer has 32 neurons.

[0057] Each neuron in the model employs a linear rectified function as its activation function. The output layer contains three nodes, corresponding to the predicted key performance indicators of the finished polyurethane material: predicted hardness (in HA); predicted elongation (in %); and predicted abrasion resistance index, a dimensionless indicator. The training process for this deep neural network model is based on a massive dataset containing over 10,000 historical production batch data samples.

[0058] Each data sample includes a complete sequence of production process parameters, as well as the actual hardness, elongation, and abrasion resistance index of that batch of products measured in the laboratory.

[0059] The model training employs an adaptive moment estimation algorithm to optimize the weights and biases. The loss function for the entire training process is defined as the mean squared error function, and its mathematical expression is: ; in, Represents the total number of training batches. Representing the The actual measured values ​​of the performance of each batch of products Representative model for the first Predicted performance values ​​for each batch of products.

[0060] This loss function measures the average deviation between the model's predicted values ​​and the true values, and guides the updating of model parameters through the backpropagation algorithm until the model's prediction accuracy on the validation set reaches a stable state.

[0061] In actual operation, the performance index prediction module calls the trained model at a frequency of 1 time / s to perform forward propagation calculations on the real-time collected and preprocessed process parameters, thereby continuously outputting three predicted performance indicators of the polyurethane material product under the current process state.

[0062] Please refer to the attached document. Figure 3 The multi-objective optimization decision module receives three predicted performance indicators from the performance indicator prediction module and compares and analyzes them with the target performance indicators preset in the system's central database.

[0063] The target performance indicators are preset according to the specific product type of polyurethane material, including the target hardness range, such as 80HA to 85HA; the target lower limit of elongation, such as 350%; and the target abrasion resistance threshold, such as 120.

[0064] The core decision-making logic of this module follows a multi-objective optimization algorithm based on weighted compromise programming.

[0065] The algorithm assigns appropriate weighting coefficients to three optimization objectives: hardness deviation, insufficient elongation, and negative shift in the abrasion resistance index. The weighting coefficient assignment strategy is strictly determined based on the specific product type of the polyurethane material.

[0066] For rubber wheel products, the wear resistance index has the highest weight, usually set at 0.5; the weight of hardness deviation is set at 0.3; and the weight of insufficient elongation is set at 0.2.

[0067] For screen products, elongation has the highest weight, usually set at 0.5; hardness deviation has a weight of 0.3; and abrasion resistance index has a weight of 0.2.

[0068] The module determines an optimal set of co-optimization parameters by constructing a weighted objective function and solving for the solution that minimizes the value of the function.

[0069] This parameter set typically includes adjustments to the reactor temperature in °C; a fine-tuning factor for the polyol to isocyanate ratio, which is a dimensionless multiplier; and a correction value for the stirring speed in r / min. The multi-objective optimization decision module also integrates an independent rule engine.

[0070] The rules engine predefines several rules for handling emergency situations.

[0071] For example, when the temperature of the core area of ​​the reactor, which is collected in real time, exceeds its preset safety threshold upper limit, such as 120°C, or is lower than its safety threshold lower limit, such as 70°C, the rule engine will be triggered immediately.

[0072] Once triggered, the rules engine will override the conventional multi-objective optimization algorithm and prioritize sending emergency adjustment commands to the actuator driver module, such as forcibly opening the cooling water valve or stopping the feeding, to ensure the absolute safety of the production process.

[0073] Please refer to the attached document. Figure 4 The actuator drive module is the final execution link of the system instructions, and it is directly connected to the key actuators on the production site.

[0074] This module contains a signal conversion submodule and an execution verification submodule.

[0075] The signal conversion submodule is responsible for converting the digital instructions in the collaborative optimization parameter set issued by the multi-objective optimization decision module into standard industrial control signals that the actuator can recognize, based on the nonlinear control curves that have been calibrated in the laboratory in advance.

[0076] For the temperature control valve of the reactor jacket, it is converted into an analog current signal of 4mA to 29mA, where 4mA corresponds to the valve being fully closed and 20mA corresponds to the valve being fully open.

[0077] For the speed controller of the metering pump, it is converted into a pulse width modulation signal of a specific frequency, ranging from 0Hz to 10kHz, corresponding to the pump speed from zero to the rated speed. For the frequency converter of the mixing motor, it is converted into an analog voltage signal from 0V to 10V, corresponding to the motor speed from 0 to the maximum allowable speed.

[0078] After the driver command is issued, the execution verification submodule will continuously monitor the actual feedback status of the target actuator.

[0079] For example, for a temperature control valve, it reads the actual opening signal fed back by the valve positioner; for a metering pump, it reads the actual pump speed feedback.

[0080] The execution verification submodule compares the actual feedback value with the expected state value in real time.

[0081] If the deviation between the two is found to be greater than 5 seconds and does not show a convergence trend, for example, the temperature control valve command opening degree is 50%, but the actual feedback opening degree always hovers around 35%, the execution verification submodule will immediately send a level 3 fault alarm containing the actuator identifier and deviation details to the system main controller to prompt the operation and maintenance personnel to intervene.

[0082] The system also has an important adaptive learning mechanism, namely the online update mechanism for the performance prediction model.

[0083] The system's main controller has a built-in timed task that automatically performs a model accuracy assessment every 24 hours.

[0084] This task compares the actual final performance laboratory test data of all polyurethane materials produced on that day with the corresponding predicted values ​​given by the performance index prediction module during the production process, and calculates the average prediction error of the model under the current production conditions.

[0085] When the system detects that the average prediction error over three consecutive days is greater than the preset model update threshold, such as 5%, the system will automatically start the incremental learning process of the model.

[0086] This process utilizes all new, validated production data samples from the past 72 hours to fine-tune the weights and bias parameters of the existing deep neural network model.

[0087] Incremental learning uses a small learning rate, such as 0.001, to ensure that new knowledge is incorporated into the model without forgetting the original generalization ability.

[0088] Through this mechanism, the system can continuously adapt to dynamic changes in the process caused by fluctuations in raw material properties and slow aging of equipment, and maintain high-precision prediction performance over the long term.

[0089] The entire system begins with the continuous capture of production site data by the process parameter acquisition module.

[0090] Temperature, flow rate, rotation speed, and pressure data are collected, preprocessed, and packaged into data frames several times per second, and then sent to the system data server via the industrial network.

[0091] After the data server performs timestamp alignment and validity verification on the data, it pushes it to the performance metric prediction module.

[0092] The performance index prediction module calls its internal deep neural network model to perform a forward calculation and outputs predicted hardness, predicted elongation and predicted wear resistance index.

[0093] These predicted values ​​are then fed into the multi-objective optimization decision-making module.

[0094] The multi-objective optimization decision module first calls the rule engine to perform a safety scan on the input process parameters and prediction performance.

[0095] If everything is normal, the weighted compromise programming algorithm is started to calculate a set of co-optimization parameters that minimizes the overall performance deviation.

[0096] This set of parameters is sent to the actuator driver module via the internal data bus.

[0097] The signal conversion submodule of the actuator drive module converts it into corresponding analog or digital control signals, drives the temperature control valve to change the opening degree to adjust the reaction temperature, drives the metering pump speed controller to fine-tune the flow rate of polyol or isocyanate to change the ratio, and drives the stirring motor frequency converter to adjust the speed to change the mixing intensity.

[0098] The execution verification submodule monitors the action feedback of these actuators in parallel to ensure that control commands are executed accurately.

[0099] This cycle repeats continuously, forming a real-time, closed-loop intelligent control system aimed at the final performance of the material, ensuring that the key indicators such as hardness, elongation, and abrasion resistance of each batch of polyurethane products can stably fall within the design window.

[0100] This embodiment provides an alternative implementation scheme for a polyurethane material production parameter monitoring and adjustment system. Its core feature is that the multi-objective optimization decision module adopts a different decision logic architecture and optimizes the input features of the performance index prediction model for specific types of polyurethane products, such as high elastomers.

[0101] In terms of the multi-objective optimization decision-making module, this embodiment adopts a decision-making mechanism based on a combination of fuzzy reasoning and expert system to replace the weighted compromise planning algorithm.

[0102] This module has a pre-defined fuzzy rule base. Each rule in the rule base consists of a premise and a conclusion.

[0103] The premise is a vague description of the deviation between the predicted performance index and the target value, such as a large positive deviation in hardness, a small negative deviation in elongation, and a moderate positive deviation in abrasion resistance.

[0104] The conclusion section provides vague suggestions regarding temperature adjustment, ratio fine-tuning coefficient, and stirring speed correction, such as slightly increasing the temperature, slightly decreasing the ratio, and maintaining the stirring speed.

[0105] The module's input interface converts the precise deviations of the three received predictive performance indicators from their target values ​​into corresponding fuzzy set membership degrees through three independent fuzzification interfaces.

[0106] The fuzzy inference engine then performs fuzzy logic inference based on these membership degrees and a pre-set fuzzy rule base, such as if the hardness has a large positive deviation and the elongation has a small negative deviation, then the temperature increases slightly and the ratio decreases slightly.

[0107] The reasoning process employs the Madani minima reasoning method, and finally defuzzifies the data using the centroid method, outputting a precise set of co-optimized parameters.

[0108] The advantage of this method is that it can better handle the nonlinear and coupling relationships between performance indicators, and the decision-making process is closer to the experience judgment of human experts. It is especially suitable for polyurethane synthesis processes with complex process mechanisms that are difficult to describe with precise mathematical models.

[0109] For polyurethane products, such as high elastomers, which have extremely high requirements for elongation, this embodiment optimizes the deep neural network model used in the performance index prediction module through feature engineering.

[0110] In addition to the 12 basic process parameters, the model input layer introduces two additional derived feature parameters.

[0111] The first derived characteristic is the average rate of temperature rise of the reaction system within a specific temperature range, such as 80°C to 100°C. This parameter is obtained by calculating the rate of change of temperature sensor readings over time within this temperature range.

[0112] The second derived feature is the variance of the stirring power fluctuation, which is obtained by calculating the variance of the stirring motor power readings over the most recent 30 seconds.

[0113] These two derived characteristics are believed to be strongly correlated with the linearity and crosslinking density of polyurethane molecular chains, and the molecular chain structure directly determines the elongation performance of the final product.

[0114] Therefore, incorporating these two features into the model input can improve the model's sensitivity and accuracy in predicting the elongation of highly elastic materials.

[0115] The network structure of the model was also adjusted accordingly, with the number of input layer nodes increased to 14 and the number of neurons in the first hidden layer increased to 154, in order to accommodate the newly added feature information and extract its deep patterns.

[0116] In the model's training dataset, the proportion of data samples from high-elasticity production batches has been increased to 40%, ensuring that the model has specialized predictive capabilities for this type of product.

[0117] Regarding the actuator drive module, this embodiment enhances the accuracy and response speed of its drive capability.

[0118] For the speed control of the metering pump, the signal conversion submodule uses a higher resolution digital-to-analog converter. When the digital command of the ratio fine-tuning coefficient is converted into a pulse width modulation signal, the duty cycle adjustment accuracy is improved from 1% to 1‰.

[0119] This allows for precise adjustments to the ratio of polyols to isocyanates down to the part-in-a-million level, which is crucial for synthesizing high-performance polyurethane elastomers with narrow molecular weight distributions.

[0120] Meanwhile, the monitoring frequency of the verification submodule was increased from 2 times / s to 10 times / s, and a predictive verification mechanism was introduced.

[0121] This mechanism not only compares the current values ​​of the command and feedback, but also predicts the state within the next 2 seconds by analyzing the changing trend of the actuator feedback signal.

[0122] If the predicted state will exceed the allowable deviation range, the execution verification submodule will send a feedforward compensation signal to the signal conversion submodule in advance to fine-tune its output, thereby actively suppressing the generation of deviation and achieving more stable and accurate control.

[0123] The system's model update mechanism is also more refined in this embodiment.

[0124] In addition to the triggering condition based on overall prediction error, this embodiment adds a triggering condition based on feature importance analysis. Every 24 hours, the system uses the Shapleyga and interpretation methods to rank the importance of the input features of the performance prediction model.

[0125] If one or more key features are found to have a continuously declining importance ranking for three consecutive days, such as the average temperature rise rate of a high-elasticity material, this may indicate a shift in the basic properties of the production process or raw materials. Even if the overall prediction error has not exceeded the standard, the system will initiate the incremental learning process of the model in advance.

[0126] This feature-based early warning mechanism enables the system to detect potential changes in the production environment earlier and make adaptive adjustments in advance, further ensuring the robustness and predictive reliability of long-term operation.

[0127] In summary, this embodiment, by introducing a fuzzy expert system into the multi-objective optimization decision-making module, enhanced feature engineering for specific products into the performance prediction module, high-precision and predictive control into the actuator drive module, and feature importance early warning into the model update mechanism, together constitute a more refined and adaptive monitoring and adjustment system implementation plan for the production of high-performance, special-requirement polyurethane materials.

[0128] This system is particularly adept at handling production scenarios with strong process nonlinearity and stringent performance requirements, and can achieve better control quality and product consistency than the system described in Example 1.

[0129] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0130] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A polyurethane material production parameter monitoring and adjustment system, characterized in that, include: The process parameter acquisition module is used to acquire in real time the temperature data inside the polyurethane synthesis reactor, the instantaneous flow rate and cumulative ratio data of polyol and isocyanate in the reactor feed pipeline, the speed and power data of the stirring motor, and the pressure data inside the reaction system. The performance index prediction module is used to simultaneously calculate the predicted hardness, predicted elongation and predicted abrasion resistance index of the polyurethane material under the current process state based on the real-time data stream provided by the process parameter acquisition module and the performance prediction model trained by the built-in historical production data. The multi-objective optimization decision module receives multiple predicted performance indicators output by the performance indicator prediction module and compares and analyzes them with the target hardness range, target elongation lower limit and target wear resistance threshold preset in the system database. Based on the preset priority rules and performance deviation tolerance, it dynamically generates a set of collaborative optimization parameters including temperature adjustment amount, ratio fine adjustment coefficient and stirring speed correction value. The actuator drive module is connected to the temperature control valve, metering pump speed controller, and stirring motor frequency converter at the production site. It is used to receive the collaborative optimization parameter set issued by the multi-objective optimization decision module and convert it into corresponding control signals to drive the aforementioned actuators to perform precise actions.

2. The polyurethane material production parameter monitoring and adjustment system according to claim 1, characterized in that, The process parameter acquisition module includes a temperature sensing unit, a flow metering unit, a speed monitoring unit, and a pressure sensing unit. The temperature sensing unit uses multiple platinum resistance temperature sensors distributed in the inner wall of the reactor and the core area of ​​the material. The flow metering unit uses a high-precision electromagnetic flow meter to monitor the instantaneous flow of the polyol and isocyanate feed lines, and integrates a pulse counter for cumulative flow calculation. The speed monitoring unit captures the speed signal in real time through a photoelectric encoder installed on the main shaft of the stirring motor; The pressure sensing unit uses a corrosion-resistant piezoresistive sensor, which is directly immersed in the reactants to monitor changes in internal pressure of the system.

3. The polyurethane material production parameter monitoring and adjustment system according to claim 1, characterized in that, The performance prediction model relied upon by the performance index prediction module is a deep neural network model. The input layer nodes of this model correspond to all real-time parameter dimensions acquired by the process parameter acquisition module; The model contains three hidden layers, and the activation function is ReLU. The output layer contains three nodes, which correspond to the predicted values ​​of three performance indicators: hardness, elongation, and abrasion resistance index.

4. The polyurethane material production parameter monitoring and adjustment system according to claim 3, characterized in that, The training process of the deep neural network model is based on data samples from historical production batches, and the adaptive moment estimation algorithm is used for weight optimization. The loss function is the mean squared error function.

5. The polyurethane material production parameter monitoring and adjustment system according to claim 1, characterized in that, The internal decision-making logic of the multi-objective optimization decision-making module follows a multi-objective optimization algorithm based on weighted compromise programming. The algorithm assigns appropriate weight coefficients to the three optimization objectives: hardness deviation, insufficient elongation, and negative shift of wear resistance index. The weighting coefficients are assigned based on the specific product type of the polyurethane material.

6. The polyurethane material production parameter monitoring and adjustment system according to claim 5, characterized in that, The multi-objective optimization decision module determines the collaborative optimization parameter set by solving for the solution that minimizes the weighted objective function value.

7. The polyurethane material production parameter monitoring and adjustment system according to claim 6, characterized in that, The multi-objective optimization decision-making module also integrates a rule engine; The rules engine predefines several emergency situation handling rules; When any process parameter collected in real time exceeds its preset safety threshold upper or lower limit, the rule engine immediately triggers and overrides the conventional multi-objective optimization algorithm, and prioritizes sending an emergency adjustment command to the actuator driver module.

8. The polyurethane material production parameter monitoring and adjustment system according to claim 1, characterized in that, The actuator driving module includes a signal conversion submodule and an execution verification submodule; The signal conversion submodule is responsible for converting the digital instructions in the co-optimization parameter set into analog current signals or pulse width modulation signals of a specific frequency, according to the pre-calibrated control curves. After the driver command is issued, the execution verification submodule continuously monitors the actual feedback status of the target actuator and compares it with the expected status.

9. A polyurethane material production parameter monitoring and adjustment system according to claim 8, characterized in that, When the execution verification submodule detects that the deviation lasts for more than 5 seconds and has not converged, it sends a fault alarm to the system main controller.

10. A method for monitoring and adjusting production parameters of polyurethane materials, characterized in that, Using the polyurethane material production parameter monitoring and adjustment system according to any one of claims 1 to 9 to achieve polyurethane material production parameter monitoring and adjustment includes the following steps: The temperature, flow rate, rotation speed, and pressure parameters during the polyurethane synthesis reaction are continuously collected by a sensor array deployed on the production site. The collected real-time process parameter sequence is input into the pre-trained performance prediction model, which calculates and outputs the predicted hardness, predicted elongation and predicted abrasion resistance index of polyurethane material in the current production batch in real time. The obtained predicted performance indicators are compared with the preset target performance indicators. Using a multi-objective optimization algorithm, considering the priority of each performance indicator and process constraints, a set of production parameter adjustment schemes that can minimize the overall performance deviation is calculated, namely the collaborative optimization parameter set. The collaborative optimization parameter set is converted into corresponding control signals to drive actuators such as temperature control valves, metering pumps, and stirring motors for precise adjustment.

Images