A nylon modification process parameter adaptive control system and method
By constructing a parameter correlation matrix and dynamic modeling through adaptive control methods, the process parameters for nylon modification are adjusted in real time, solving the production instability problem caused by parameter instability in traditional processes, and achieving efficient and stable production process and product quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHANGJIAGANG OASIS NEW MATERIAL TECH CO LTD
- Filing Date
- 2026-04-14
- Publication Date
- 2026-06-12
AI Technical Summary
Traditional nylon modification processes cannot handle fluctuations in raw materials and changes in equipment status, leading to unstable production processes and affecting product quality consistency and efficiency.
By using adaptive control methods, a correlation matrix between parameters is constructed to analyze the direct impact paths and indirect chain reactions. Process parameters are adjusted in real time, and dynamic modeling and feedback optimization strategies are used to ensure production stability and quality consistency.
It enables real-time monitoring and adjustment of process parameters, improves the level of production automation, reduces quality fluctuations, and ensures high product quality and stable production process.
Smart Images
Figure CN122194682A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of nylon modification process technology, and in particular to an adaptive control system and method for nylon modification process parameters. Background Technology
[0002] In nylon modification processes, controlling process parameters is crucial for ensuring product quality and production efficiency. Traditional nylon modification processes largely rely on manually set empirical formulas to adjust process parameters such as screw speed, melt temperature, and residence time. However, these methods often cannot cope with the influence of raw material fluctuations and equipment status changes during actual production, leading to unstable process parameters and consequently affecting the consistency of final product quality and production efficiency.
[0003] Furthermore, existing process control technologies fail to adequately consider the coupling relationships between various parameters during nylon modification. In practical applications, parameters such as screw speed, melt temperature, and residence time interact, forming complex coupling effects. Traditional technologies often cannot accurately identify and model these interactions, thus failing to precisely predict the combined impact of changes in one parameter on other process parameters and the final product performance. This deficiency leads to instability in the production process and even unnecessary quality fluctuations.
[0004] To address the aforementioned issues, this invention proposes an adaptive control method for nylon modification process parameters. This method extracts coupling features from process data, analyzes direct influence paths and indirect chain reactions based on these features, and utilizes dynamic modeling and real-time adjustment to control process parameters. This approach allows for real-time monitoring and adjustment of various process parameters, reducing the impact of fluctuations on the production process and product quality. This significantly improves the intelligence and automation level of production, ensuring the stability of the production process and the high quality of the products. Summary of the Invention
[0005] This application provides an adaptive control system and method for nylon modification process parameters, which improves the accuracy and stability of nylon modification process control.
[0006] In a first aspect, this application provides an adaptive control method for nylon modification process parameters, the method comprising: S1. Extract relevant data streams including screw speed changes, melt temperature fluctuations and residence time differences from nylon modification process data, construct an initial correlation matrix of the relationship between parameters, store the initial correlation matrix in the digital mapping model, and obtain the coupling characteristics between parameters; S2. Based on the coupling characteristics, analyze the direct influence path and indirect chain reaction, obtain the degree of interference of the system viscosity, and calculate the effect weight according to the degree of interference. S3. Based on the weight of action, screen out potential fluctuation source parameters. For the potential fluctuation source parameters, extract the historical change data of their corresponding process parameters. Based on the historical change data, construct a dynamic model and simulate the impact of parameter changes on the key outputs of the system, and record the impact fluctuation values. S4. Determine whether the value of the fluctuation exceeds the preset safety threshold, detect the risk range based on the judgment result and determine the severity level of the abnormal fluctuation, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment. S5. Extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream. S6. Calculate the estimated viscosity and standard deviation of the melt temperature of the system based on the smoothed data stream. Compare the estimated value and standard deviation with the corresponding target value to obtain the comparison deviation. Based on the comparison deviation, determine whether the adjustment command is effective and output a stable control scheme or start a feedback optimization strategy.
[0007] Secondly, this application provides an adaptive control system for nylon modification process parameters, the system comprising: The data processing module is used to extract relevant data streams, including screw speed changes, melt temperature fluctuations and residence time differences, from nylon modification process data, construct an initial correlation matrix of the relationship between parameters, store the initial correlation matrix into a digital mapping model, and obtain the coupling characteristics between parameters. The weight calculation module is used to analyze the direct influence path and indirect chain reaction based on coupling characteristics, obtain the degree of interference of the system viscosity, and calculate the effect weight based on the degree of interference. The dynamic simulation module is used to filter out potential fluctuation source parameters based on their weights, extract historical change data of their corresponding process parameters for the potential fluctuation source parameters, build a dynamic model based on the historical change data, simulate the impact of parameter changes on the key outputs of the system, and record the impact fluctuation values. The risk assessment module is used to determine whether the value of the fluctuation exceeds the preset safety threshold, detect the risk range based on the judgment result, determine the severity level of the abnormal fluctuation, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment. The real-time adjustment module is used to extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream. The feedback optimization module is used to calculate the estimated value of the system viscosity and the standard deviation of the melt temperature based on the smoothed data stream. It compares the estimated value and the standard deviation with the corresponding target value to obtain the comparison deviation. Based on the comparison deviation, it judges whether the adjustment command is effective and outputs a stable control scheme or initiates the feedback optimization strategy.
[0008] Thirdly, this application provides a computer device, wherein the memory stores machine-readable instructions executable by the processor. When the computer device is running, the processor communicates with the memory via a bus, and when the machine-readable instructions are executed by the processor, the steps of the aforementioned adaptive control method for nylon modification process parameters are performed.
[0009] Compared with the prior art, the beneficial effects of the present invention are at least as follows: 1. Through adaptive control methods, various parameters in the nylon modification process, such as screw speed, melt temperature and residence time, can be monitored and adjusted in real time, effectively reducing manual intervention in traditional control methods and improving the automation and accuracy of the production process.
[0010] 2. Coupled feature modeling based on process data analysis can accurately identify the complex coupling relationships between process parameters, and then predict and adjust the impact of parameter changes on system output, overcoming the shortcomings of existing technologies that cannot accurately model and adjust the interaction between parameters.
[0011] 3. Dynamic modeling and simulation based on historical data can predict and intervene in process fluctuation sources in advance, avoiding quality fluctuations and production failures caused by the inability to identify potential fluctuation sources in real time in traditional methods, thus improving production stability and product quality consistency.
[0012] 4. Through an intelligent feedback optimization mechanism, the effectiveness of the adjustment command can be judged by comparing the deviation between the target value and the actual value after parameter adjustment, and optimization can be carried out to ensure the efficiency and stability of the adjustment process and further improve the accuracy of process control. Attached Figure Description
[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 This is a flowchart of an adaptive control method for nylon modification process parameters according to this application; Figure 2 This is a multi-level threshold pressure monitoring diagram for this application; Figure 3 This is a severity rating chart for this application; Figure 4 The execution flow of the adaptive feedback optimization strategy in this application; Figure 5This is a schematic diagram of the structure of an adaptive control system for nylon modification process parameters according to this application; Figure 6 This is a schematic block diagram of an adaptive control device for nylon modification process parameters according to this application. Detailed Implementation
[0015] This application provides an adaptive control system and method for nylon modification process parameters. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0016] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of an adaptive control method for nylon modification process parameters in this application includes: Step S1: Extract relevant data streams, including screw speed changes, melt temperature fluctuations, and residence time differences, from the nylon modification process data; construct an initial correlation matrix of the relationships between parameters; store the initial correlation matrix in a digital mapping model; and obtain the coupling characteristics between parameters.
[0017] In one specific embodiment, the process of performing step S1 may specifically include the following steps: The relevant data streams are cleaned to obtain the initial data stream; Based on the initial data stream, an initial correlation matrix is constructed to represent the relationship between the parameters, and the initial correlation matrix is stored in the digital mapping model; Based on the digital mapping model, the initial correlation matrix is analyzed using time series analysis methods to obtain the coupling characteristics between parameters.
[0018] Specifically, in the data cleaning stage, a method based on... The sliding window method of the criterion is combined with a physical threshold for filtering. Taking screw speed variation data as an example, the system acquires signals at a frequency of 10Hz and sets a sliding window with a length of 100 data points. The mean μ and standard deviation of the data within the window are calculated. It will exceed Instantaneous values within a certain range are identified as noise points and removed. For example, during a steady-state production period, if the average rotational speed within the window is 300 rpm and the standard deviation is 2.5 rpm, then data points with instantaneous rotational speed values below 292.5 rpm or above 307.5 rpm are considered abnormal and removed. Simultaneously, for melt temperature fluctuation data, a process safety boundary of 220 is set. Up to 320 Sampled values outside this range will be marked as outliers and replaced with valid data from before and after using linear interpolation to obtain the initial data stream after denoising.
[0019] Based on this initial data stream, an initial correlation matrix is constructed by calculating the Pearson correlation coefficient among screw speed variation, melt temperature fluctuation, and residence time difference. The calculation formula is: in, and Let be the values of the two parameters at the i-th sampling point. and This is the average value over n sampling points. For example, analysis using 12 consecutive hours of production data shows that the correlation coefficient between screw speed variation and melt temperature fluctuation is +0.79, indicating a significant positive correlation; while the correlation coefficient between residence time difference and melt temperature fluctuation is -0.61, reflecting a negative correlation. This correlation coefficient matrix is stored as the initial correlation matrix in the database of the digital mapping model. The digital mapping model is a hierarchical computational framework used to characterize and simulate the dynamic coupling relationships of multiple parameters in the nylon modification process. Its basic layer consists of the initial correlation matrix, used to describe the static correlation between parameters. Based on this, the model's mechanism analysis layer uses time series analysis methods (such as vector autoregression models) to deeply mine time series data, extracting dynamic coupling features with time delay and feedback mechanisms, such as impulse response functions and variance decomposition contribution. Simultaneously, the model's intelligent prediction layer uses a long short-term memory network to construct a dynamic model, using the aforementioned dynamic features and real-time process data as input, to learn the complex nonlinear mapping relationship between parameters and key system outputs, achieving accurate prediction and simulation of process risks. Through a feedback optimization strategy, the model can be continuously updated: the intelligent prediction layer receives incremental learning training to optimize its predictive capabilities, and the parameters and features of the mechanism analysis layer are also iteratively updated accordingly. Thus, the digital mapping model evolves from a static relational performance into a digital mirror capable of adaptively characterizing and predicting the dynamic behavior of the process system, providing core model support for the entire adaptive control method.
[0020] Based on this digital mapping model, time series analysis methods are used to deeply mine the temporal relationships implied in the initial correlation matrix to extract dynamic coupling characteristics. A vector autoregression model is then used to jointly model the time series of the three parameters. The model expression is as follows: in, It is a 3D column vector containing the screw speed change at time t, the melt temperature fluctuation, and the residence time difference. It is a constant vector; Let be a 3×3 coefficient matrix of the k-th lag term, whose elements Characterizes the strength of the influence of the j-th variable on the i-th variable when lagged by k periods; The optimal lag order of the model is determined according to the AIC criterion; This is a white noise vector. Model estimation can quantitatively reveal the dynamic interactions between parameters. For example, coefficients... This indicates that for every 1 unit increase in screw speed at the previous moment, the average melt temperature fluctuation at the current moment increases by 0.38 units, directly characterizing the driving effect of speed on temperature. Furthermore, based on the estimation model, the impulse response function is calculated, allowing for the systematic simulation of the response trajectory of all parameters over time after any parameter is subjected to an impact. This fully reveals the dynamic coupling characteristics of time delay, feedback, and amplification between parameters, laying a quantitative analytical foundation for risk identification and precise control.
[0021] Step S2: Based on the coupling characteristics, analyze the direct influence path and indirect chain reaction, obtain the interference degree of the system viscosity, and calculate the effect weight according to the interference degree.
[0022] In one specific embodiment, the process of performing step S2 may specifically include the following steps: Based on the coupling characteristics, the direct influence path and indirect chain reaction between screw speed, melt temperature and residence time are analyzed. Based on the direct influence path and indirect chain reaction, the degree of interference of each parameter change on the system viscosity is quantified; Based on the degree of interference, the weights of the effects of screw speed variation, melt temperature fluctuation and residence time difference on the process state are determined.
[0023] Specifically, based on the dynamic coupling characteristics, this study uses impulse response functions and variance decomposition techniques from time series analysis to analyze the direct influence paths and indirect chain reactions among screw speed, melt temperature, and residence time. The impulse response function describes the dynamic response trajectory and direction of other parameters over multiple future periods after a single parameter is subjected to a standard unit impact. For example, the impulse response extracted from the vector autoregression model of the digital mapping model shows that a positive impact on screw speed directly leads to a sustained increase in melt temperature after a lag of 2-4 sampling periods, constituting the direct influence path of screw speed on melt temperature. Furthermore, this temperature increase impulse response causes a shortening of residence time in subsequent periods, thus constituting the indirect chain reaction from screw speed to melt temperature to residence time.
[0024] To map the coupling relationship between the above parameters to the final process target, based on the real-time acquired screw speed, melt temperature and die pressure data, a pre-set material rheology model is used. The material rheology model is a mathematical model pre-established based on the rheological properties of a specific nylon modified material. The core of this model usually adopts a modified power law model, which expresses the relationship between melt viscosity and shear rate and temperature. Then, the estimated value sequence of system viscosity is calculated online, thereby obtaining continuous viscosity time series data synchronized with the process parameters.
[0025] Based on variance decomposition technology and combined with viscosity time series data, this study quantifies the degree of interference of various parameter changes on the system viscosity. Specifically, an extended vector autoregressive model is constructed, comprising four variables: screw speed variation, melt temperature fluctuation, residence time difference, and the estimated system viscosity. Variance decomposition is then performed on this model to quantify the proportion of the prediction error variance of the system viscosity estimate explained by the impact of each process parameter within a specific prediction step (e.g., the next 30 seconds).
[0026] This percentage directly reflects the contribution of the parameter's impact to viscosity fluctuations, i.e., the degree of interference. For example, in a specific embodiment, variance decomposition of the above extended vector autoregressive model shows that, in a prediction time domain of 30 seconds, the contribution of screw speed variation to viscosity fluctuations is approximately 41%, melt temperature fluctuation is approximately 33%, residence time difference is approximately 18%, and the remainder are interaction terms and unexplained components. These contribution percentages represent the degree of interference of each parameter on viscosity.
[0027] The interference levels are normalized to determine the weights of screw speed variation, melt temperature fluctuation, and residence time difference on the process state. Specifically, the interference level value of each parameter is divided by the sum of the interference levels of all parameters. Using the previous example, the sum of interference levels is 41% + 33% + 18% = 92%. Therefore, the normalized weights are: screw speed variation weight = 41% / 92% ≈ 0.446, melt temperature fluctuation weight = 33% / 92% ≈ 0.359, and residence time difference weight = 18% / 92% ≈ 0.196. This weight vector accurately quantifies the relative importance of each parameter's impact on the current process state, providing a direct and quantitative decision-making basis for subsequent risk identification and control priority ranking.
[0028] Step S3: Based on the impact weight, screen out potential fluctuation source parameters. For the potential fluctuation source parameters, extract the historical change data of their corresponding process parameters. Based on the historical change data, construct a dynamic model and simulate the impact of parameter changes on the key outputs of the system, and record the impact fluctuation values.
[0029] In one specific embodiment, the process of performing step S3 may specifically include the following steps: Parameters whose weight exceeds a preset weight threshold are marked as potential sources of fluctuation. For potential source parameters of fluctuation, extract historical change data of their corresponding process parameters; Based on historical change data, a dynamic model is constructed to simulate the parameter coupling effect; The impact of expected changes in the parameters of potential fluctuation sources on key output indicators of the system is simulated using a dynamic model, and the fluctuation values are recorded.
[0030] Specifically, the calculated impact weights are compared with preset weight thresholds to filter out potential fluctuation source parameters. The weight thresholds can be set according to process stability requirements, for example, 0.25. If the impact weight of screw speed change is 0.446 and the impact weight of melt temperature fluctuation is 0.359, both exceeding the threshold, then these two parameters are marked as potential fluctuation source parameters. Historical change data is a time series; for example, complete data of screw speed and melt temperature sampled at a frequency of 1Hz for the most recent week is extracted.
[0031] The dynamic model is constructed using a Long Short-Term Memory (LSTM) network. To highlight the key technical aspects and ensure the model focuses on learning coupled fluctuations, the following unique method is employed when preparing training and testing data: historical data is divided into independent data blocks based on continuous production batches or fixed long-term windows (e.g., every 8 hours). The first 80% of these blocks are allocated as the training dataset, and the last 20% as the testing dataset, simulating real-world scenarios of predicting future operating conditions using past data. Secondly, when constructing the model's input, differentiated historical review window lengths are set as input features to account for the differences in the physical response characteristics of parameters from different potential fluctuation sources. For example, the input window length for melt temperature fluctuations is set. The length of the input window for screw speed variation in seconds (60 time steps). Seconds (120 time steps). This multi-scale input design allows the model to capture the time-scale differences in the dynamic characteristics of different parameters more precisely. The most critical technical point lies in the construction of the model output and training labels: the model's learning objective is not the absolute value of the system's key output indicators (such as head pressure), but rather the numerical fluctuation of its influence due to parameter coupling. Specifically, for each sample in the training data, its label... For the future Time (e.g.) Actual pressure value (seconds) , and a dynamic baseline pressure prediction value at that moment The percentage of relative deviation between them. The calculation formula is: in, The future is derived from historical change data. Real-time, accurate mold head pressure value. It is a prediction of the pressure at the same moment through a robust benchmark model (e.g., a simple linear regression model based only on the current melt temperature), which characterizes the expected pressure under the assumption of no anomalous coupling disturbances. This represents the actual impact fluctuation value indicated by the sample, expressed as a percentage. By training with this percentage deviation as a supervisory signal, the dynamic model is forced to learn how coupled changes in the parameters of potential fluctuation sources cause the system output to deviate from its normal baseline, thus making it extremely sensitive to abnormal fluctuations during risk warnings.
[0032] Using the training dataset, with the aforementioned multi-scale window data as input, the calculated percentage of deviation is... To achieve the desired output, the LSTM network is trained under supervised instruction. The Adam optimizer is used during training, with an initial learning rate set to [value missing]. The Huber loss function was used, and early stopping was employed to prevent overfitting. After training, the model performance was evaluated using a test dataset.
[0033] In practice, the actual sequence of potential fluctuation source parameters (organized according to the aforementioned multi-scale rules) within the current and most recent historical window is input into the model, and the model directly outputs a prediction of the future. The percentage deviation of the pressure at any given moment from its dynamic baseline pressure is predicted and recorded as the impact fluctuation value for this simulation. For example, if the temperature sequence of the previous 60 seconds and the rotational speed sequence of the previous 120 seconds are input into the model, and the model predicts that the pressure will be +15.2% higher than its baseline value in the next 30 seconds, this "+15.2%" is quantified as the impact fluctuation value for this simulation and is used to assess potential risks.
[0034] Step S4: Determine whether the value of the fluctuation exceeds the preset safety threshold, detect the risk range based on the judgment result and determine the severity level of the abnormal fluctuation, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment.
[0035] In one specific embodiment, the process of performing step S4 may specifically include the following steps: Determine whether the fluctuation value exceeds the preset safety threshold; if so, determine that there is a risk range. Detect risk ranges, obtain real-time pressure data corresponding to the risk ranges, and compare them with preset pressure thresholds to determine the severity level of abnormal fluctuations; Based on the severity level, the corresponding key intervention nodes are matched and selected from the pre-set adjustment knowledge base; Calculate the priority sequence for parameter adjustments based on key intervention points and severity levels.
[0036] Specifically, the fluctuation value is compared with a preset safety threshold in real time. The preset safety threshold is set in advance based on the safe operating limit of the process equipment and product quality requirements. For example, the safe upper limit for die head pressure fluctuation is set to ±15% of the baseline pressure. If the fluctuation value exceeds this threshold, a risk range is immediately determined to exist in the present or near future.
[0037] After identifying a risk zone, the system monitors the actual process status corresponding to that risk zone in real time. Specifically, it acquires real-time pressure data near the predicted time point of the risk zone and calculates its mean. Average this actual pressure The severity level of abnormal fluctuations is determined by comparing the data to a more stringent preset pressure threshold, which can be divided into multiple levels, such as: the first threshold. (e.g., 90% of the baseline value), second threshold (110% of the baseline value), third threshold (120% of the baseline value) and the fourth threshold (135% of the baseline value). The comparison rule is: if Between and If it falls between these two levels, it is classified as mildly severe; if it falls between these two levels... and Between these ranges is considered moderate; if it exceeds this range... If the baseline pressure is 80 MPa, then it is considered severe. For example, if the actual measured pressure is... If the pressure is 98 MPa (i.e., 122.5% of the baseline), it falls into the moderately severe category.
[0038] Based on the determined severity level, corresponding key intervention nodes are matched and selected from a pre-set regulation knowledge base. The regulation knowledge base is a library of rules and cases built upon historical successful regulation experience. The matching process is as follows: using the current severity level and the identified potential fluctuation source parameter combination as query conditions, all matching historical records are retrieved from the knowledge base. The intervention nodes contained in these records are summarized and their frequencies are statistically analyzed. The top few nodes with the highest frequencies are selected as the key intervention nodes for this regulation. For example, for the condition "moderate risk caused by the coupling of screw speed and melt temperature," the knowledge base matching results show that "reducing screw speed" and "stabilizing melt temperature" are the most commonly used effective measures; therefore, these two are selected as the key intervention nodes for this regulation.
[0039] Finally, based on the selected key intervention nodes and their current severity levels, a priority sequence for parameter adjustments is calculated. The calculation considers not only the matching frequency of nodes in the knowledge base but also the immediate severity of the current risk and the latest coupling strength reflected in the process model. Its priority score is then used. The calculation formula is: in, Priority scores representing key intervention nodes; The frequency weight represents the normalized frequency of the node's appearance in historical matching records. This represents the severity coefficient, which is assigned a value according to the severity level. For example, mild is 1.0, moderate is 1.5, and severe is 2.0. Represents the real-time coupling strength, which is divided into several levels based on the coupling strength (e.g., weak = 0.5, medium = 1.0, strong = 1.5). These are adjustable weighting coefficients used to balance the contributions of each component. A priority score is calculated for each key intervention node. For example, the calculation shows that the "reducing screw speed" node... The value is 2.1, for the "Stable Melt Temperature" node. The value is 1.8. Sorting the scores from highest to lowest, the priority sequence for parameter adjustments is: first, reduce screw speed; second, stabilize melt temperature. This sequence provides the core decision-making basis for generating precise and orderly adjustment commands. (Reference) Figure 2 and Figure 3 The figure shows a multi-level threshold pressure monitoring chart and a severity level determination chart.
[0040] Step S5: Extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream.
[0041] In one specific embodiment, the process of performing step S5 may specifically include the following steps: Based on the priority sequence, adjustment instructions are extracted from key intervention nodes. These instructions are used to determine the adjustment range for screw speed changes and residence time differences. The adjustment command is transmitted to the control module, which drives the control module to adjust the screw speed or residence time in real time and obtain the adjusted process data stream; Based on the adjusted process data stream, a smoothed data stream is generated to monitor changes in process status.
[0042] Specifically, based on a priority sequence, the system automatically extracts the corresponding adjustment command from the top-priority critical intervention node. The adjustment command is a structured control instruction, whose core content includes the target parameter, adjustment direction, specific adjustment range, maximum allowable rate of change, and expected effective time window. The adjustment range is calculated by querying a pre-set adjustment response model library and combining it with the current operating conditions. For example, for the node "reduce screw speed," the system calculates the specific command based on the current speed, target pressure deviation, and historical records of similar adjustments.
[0043] The generated adjustment commands are transmitted to the control module via a real-time industrial Ethernet protocol. The control module consists of a high-speed programmable logic controller (PLC) and an intelligent servo driver. If the command is to adjust the screw speed, the PLC sends a new speed setpoint and slope limit to the main drive servo driver. The driver precisely adjusts the motor through closed-loop control, causing the screw to change its speed according to a preset smooth curve. If the command is to adjust the residence time, this is achieved by adjusting the speed of the feeder's servo motor, thereby changing the filling and conveying conditions of the material in the screw barrel and indirectly controlling the residence time. Throughout the real-time adjustment process, the control module continuously feeds back the actual status data of the actuators to ensure that the adjustment actions are executed accurately and safely.
[0044] While the control module drives the actuator to make real-time adjustments, and subsequently, a sensor network deployed on the production line synchronously collects high-frequency data to obtain the adjusted process data stream. This data includes the actual screw speed curve measured by a photoelectric encoder, melt temperature fluctuations monitored by a temperature sensor, melt pressure changes captured by a pressure sensor, and estimated residence time differences indirectly derived through calculation. All data is synchronized and recorded using a unified time base, forming a multi-dimensional, high-frequency, time-aligned adjusted process data stream that fully covers the evolution of the process state before and after adjustment.
[0045] Based on the adjusted process data stream, a smoothed data stream is generated using signal processing algorithms. The core processing employs a moving average filtering method, optimized for the characteristics of the process data. This method calculates a weighted average of multiple data points within a continuous sliding time window, assigning higher weights to more recent data points at the center of the window to better preserve the true trend of parameter changes while suppressing random noise and instantaneous spikes. For example, applying this filter to the die pressure sequence effectively smooths short-term disturbances, while clearly revealing the slow pressure decrease or recovery process caused by speed adjustments. The resulting smoothed data stream exhibits significantly reduced short-term fluctuations, more stably and reliably reflecting the overall trend of the process system after adjustment. This provides a high-quality, interference-resistant input signal for subsequent accurate evaluation of the adjustment effect and triggering of feedback mechanisms.
[0046] Step S6: Calculate the estimated viscosity and standard deviation of the melt temperature of the system based on the smoothed data stream. Compare the estimated value and standard deviation with the corresponding target value to obtain the comparison deviation. Based on the comparison deviation, determine whether the adjustment command is effective and output a stable control scheme or start a feedback optimization strategy.
[0047] In one specific embodiment, the process of performing step S6 may specifically include the following steps: Based on the smoothed data stream, sliding window statistical calculations are performed at fixed time intervals to obtain the estimated viscosity of the system and the standard deviation of the melt temperature. The estimated viscosity and standard deviation of the melt temperature of the system are compared with the preset viscosity target value and the preset temperature stability target value, respectively, to obtain the viscosity deviation and temperature stability deviation of the system. Determine whether the absolute value of the system viscosity deviation is less than the first deviation threshold and the temperature stability deviation is less than the second deviation threshold; If so, the adjustment command is deemed valid, and a stable control scheme for the nylon modification process is output. If not, the adjustment instruction is determined to need optimization, and a feedback optimization strategy is initiated based on the adjustment instruction until the adjustment instruction is determined to be valid.
[0048] Specifically, the smoothed data stream is continuously monitored, and a process status assessment is triggered at fixed time intervals (e.g., every 30 seconds). Within each assessment cycle, the following steps are performed: A smoothed data stream of a preset time length preceding the current evaluation moment is extracted as an analysis window. This window length is typically set to reflect the minimum stable period of the process dynamics, such as 60 to 180 seconds. Based on the data within this window, two key state indicators are calculated in parallel: 1. Estimated system viscosity: This value is not directly measured, but estimated online using a mechanistic-empirical fusion model based on the relationship between process parameters and viscosity. This model uses a smoothed screw speed within a window. Melt temperature and length of stay The average or eigenvalues are inputs. A typical simplified estimation formula is as follows: in, This is an estimate of the system viscosity; These are material constants; It is the activation energy of the flow. It is the gas constant; This is the weighted average of the melt temperatures within the window; Based on screw speed The effective shear rate is calculated based on the screw's geometric parameters; The non-Newtonian exponents of the material. Model parameters ( The model has been pre-calibrated for specific nylon-modified formulations. It can be used to calculate, in real-time, an estimate of the system viscosity under current process conditions.
[0049] 2. Standard deviation of melt temperature: This indicator is directly used to quantify the severity of temperature fluctuations. It calculates the statistical standard deviation of the smoothed melt temperature data series within the analysis window. This standard deviation effectively eliminates the influence of random noise and truly reflects the dispersion of melt temperature over a short period of time. It is a direct quantitative indicator for assessing whether thermal stability meets the standards.
[0050] After completing the index calculations, the comparison and judgment phase begins. The estimated viscosity values of the calculated system are then compared and judged. Compared with the preset viscosity target value Compare the viscosity target values. This is the optimal value or the median of the optimal range set by process experts based on the product grade, screw design, and desired extrusion stability. Calculate the system viscosity deviation. This is usually expressed as a percentage. Simultaneously, the standard deviation of the calculated melt temperature will be... Compared with the preset temperature stability target value Compare the target values for temperature stability. It is set based on the maximum allowable temperature fluctuation level without affecting the performance of the final product (such as dimensional and mechanical property uniformity). Calculate the temperature stability deviation. , which is a ratio.
[0051] Two preset deviation thresholds: the first deviation threshold (Regarding viscosity) and second deviation threshold (Regarding temperature stability). For example, It can be set to 5% (i.e., the viscosity estimate should be within ±5% of the target value). It can be set to 1.0 or 1.2 (meaning the actual temperature standard deviation must not exceed 1.0 or 1.2 times the target value).
[0052] The judgment logic is as follows: If both conditions are met and If the previously executed adjustment commands are deemed effective, the process state has been successfully guided and stabilized within the expected target range. At this point, a stable control scheme for the nylon modification process is output. This scheme is a structured file or data package containing not only the steady-state values of key process parameters (screw speed, temperature setpoints for each zone, feed rate, etc.) at the point of final stabilization, but also a complete record of the priority sequence executed to achieve this stable state, the specific adjustment commands (including adjustment magnitude and rate), and key process data from fluctuation to stability. This stable control scheme can be stored and used to guide the production of subsequent products with the same formulation and specifications, or as an accumulation of advanced process knowledge.
[0053] If any of the above conditions are not met (i.e.) or If the current adjustment command fails to achieve the expected optimization effect, further adjustments are required. Immediately initiate a feedback optimization strategy based on the adjustment command. The core of this strategy is to treat this adjustment as an "experiment" and use the experimental results, i.e., the current process state (as...). and Characterization), the content of the executed adjustment instructions, and the calculated deviation ( and This data is packaged into optimization requirement data. This data package is sent to the feedback optimization loop, triggering a new round of analysis. Based on the new process state, the parameter weights are reassessed, or the digital mapping model is fine-tuned. It may also incorporate experience from unsuccessful adjustments (e.g., insufficient or excessive adjustment) to recalculate the priority sequence of parameter adjustments, generating new, optimized adjustment commands. This feedback optimization strategy is a closed-loop iterative process that continues until, after the new command is executed, the deviation calculated in the new evaluation cycle meets the validity conditions, ultimately deriving an effective and stable control scheme. This mechanism ensures that the control system possesses self-learning and adaptive capabilities, enabling it to cope with uncertainties such as raw material fluctuations and equipment status changes, gradually approaching and maintaining the optimal process state.
[0054] In one specific embodiment, executing the feedback optimization strategy in step S6 may specifically include the following steps: Collect current process status data, executed adjustment commands, and adjustment effect data as optimization requirement data; Based on the optimized demand data, the digital mapping model is incrementally trained to obtain an updated digital mapping model. Based on the updated digital mapping model, the interaction weights between screw speed, melt temperature and residence time differences are recalculated. Using the action weight as input and combining it with the current process status, the optimized adjustment command is iteratively calculated through a dynamic model.
[0055] Specifically, when it is determined that an adjustment command needs optimization, a feedback optimization strategy is immediately initiated. This strategy first performs a data acquisition step: acquiring current process status data, including the latest smoothing data stream; acquiring a complete record of executed adjustment commands; and acquiring adjustment effect data. These three types of data are structured into a single data package, serving as the optimization requirement data for this iteration.
[0056] Based on this optimized data, the digital mapping model is updated collaboratively. This is mainly reflected in targeted optimizations of its two core layers: 1. Parameter Updates for the Mechanism Analysis Layer: New data is used to re-estimate the parameters of the model's mechanism analysis layer (whose core is time series analysis tools such as the vector autoregression model). For example, after re-estimating the vector autoregression model based on the latest 2-hour data, the current influence coefficient of screw speed on melt temperature is updated from 0.38 to 0.42, more accurately reflecting the current material's stronger shear heat generation characteristics; variance decomposition results show that the contribution of melt temperature fluctuations to pressure variance increases from 33% to 38%. These updated dynamic coupling features are synchronized to the feature library. This layer update provides a more accurate "physical understanding" foundation reflecting the current operating conditions for subsequent intelligent predictions.
[0057] 2. Incremental Training of the Intelligent Prediction Layer: For the risk prediction LSTM model in the model, i.e., the intelligent prediction layer, the dynamic coupling features extracted from the updated mechanism analysis layer are used as one of the important input features. Combined with the latest process status data and regulation effect data, incremental learning training is performed. This process employs an elastic weight solidification strategy to prevent knowledge forgetting. For example, in an incremental learning exercise targeting a new fluctuation pattern caused by changes in raw material batches, training uses a mixed dataset containing 50 new samples (whose input features already include the new coupling features calculated by the updated VAR model) and 200 historical samples, with only 5 rounds of training. This sequential update method ensures that the LSTM model learns prediction patterns based on the latest physical understanding. Results show that the root mean square error of the model's prediction of pressure peaks under similar new fluctuations decreased from 12% before the update to 7%, while the prediction accuracy for historical steady-state conditions remained stable.
[0058] After completing the above updates, the updated digital mapping model was obtained. Next, the analysis was re-executed based on this updated model: First, using its updated mechanistic analysis layer, the influence weights between screw speed, melt temperature, and residence time differences were recalculated. For example, due to changes in contribution, the new influence weights were adjusted from [0.61, 0.27, 0.12] to [0.55, 0.35, 0.10], indicating that the importance of temperature fluctuations has relatively increased, providing more accurate guidance for subsequent control. Then, these new influence weights, along with the current process status data, were input into the incrementally trained risk prediction LSTM model. (Reference) Figure 4 The figure illustrates the execution flow of the adaptive feedback optimization strategy.
[0059] With the goal of minimizing prediction risk, an optimization algorithm is used to simulate and iteratively calculate the optimized adjustment command within the model. For example, guided by the new weights, the simulation prioritizes testing "first reducing the melt temperature setpoint by 3". Instead of the previous strategy of "simply drastically reducing the speed," the new strategy combines "adjusting the screw speed" with "further fine-tuning the screw speed." Simulations predict that this new strategy can smoothly reduce pressure to a safe range within 45 seconds, reducing overshoot risk by 20%. This optimization command is transmitted and executed, completing a full feedback optimization cycle. This process can be iterated repeatedly until the process reaches a stable state. In this way, the system achieves a shift from "experience-based trial and error" to "data-driven precision optimization," reducing the average time to process stability by approximately 42.3% and decreasing the incidence of secondary fluctuations caused by improper adjustments by more than 23%.
[0060] It is understood that the executing entity of this application can be an adaptive control system for nylon modification process parameters, or it can be a terminal or a server; no specific limitation is made here. This application's embodiments use a server as an example for illustration.
[0061] The above describes an adaptive control method for nylon modification process parameters in an embodiment of this application. The following describes an adaptive control system for nylon modification process parameters in an embodiment of this application. Please refer to [link to relevant documentation]. Figure 5 One embodiment of an adaptive control system for nylon modification process parameters in this application includes: The data processing module is used to extract relevant data streams, including screw speed changes, melt temperature fluctuations and residence time differences, from nylon modification process data, construct an initial correlation matrix of the relationship between parameters, store the initial correlation matrix into a digital mapping model, and obtain the coupling characteristics between parameters. The weight calculation module is used to analyze the direct influence path and indirect chain reaction based on coupling characteristics, obtain the degree of interference of the system viscosity, and calculate the effect weight based on the degree of interference. The dynamic simulation module is used to filter out potential fluctuation source parameters based on their weights, extract historical change data of their corresponding process parameters for the potential fluctuation source parameters, build a dynamic model based on the historical change data, simulate the impact of parameter changes on the key outputs of the system, and record the impact fluctuation values. The risk assessment module is used to determine whether the value of the fluctuation exceeds the preset safety threshold, detect the risk range based on the judgment result, determine the severity level of the abnormal fluctuation, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment. The real-time adjustment module is used to extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream. The feedback optimization module is used to calculate the estimated value of the system viscosity and the standard deviation of the melt temperature based on the smoothed data stream. It compares the estimated value and the standard deviation with the corresponding target value to obtain the comparison deviation. Based on the comparison deviation, it judges whether the adjustment command is effective and outputs a stable control scheme or initiates the feedback optimization strategy.
[0062] Through the collaborative efforts of the aforementioned components, a closed-loop intelligent control system capable of self-perception, dynamic analysis, accurate prediction, proactive intervention, and continuous optimization is constructed, fundamentally changing the traditional nylon modification process's reliance on manual experience and isolated parameter control. The core advantage of this system lies in its holistic and adaptive nature. The process from data perception to feature extraction forms the foundation of the system's cognition. The data processing module, acting as the system's "sensory organs," continuously collects high-dimensional raw process data and extracts coupling features representing the interdependencies between parameters. This provides a structured knowledge base for subsequent analysis, enabling the system to "understand" rather than merely "see" the process state. The chain from feature analysis to risk prediction achieves a leap from cognition to prediction. The weight calculation module and the dynamic simulation module jointly serve as the analysis and deduction center of the system's "brain." The former, through quantitative analysis, accurately locates the key parameters (i.e., potential sources of fluctuation) that have the greatest impact on system stability; the latter, based on a dynamic model built from historical data, simulates future changes in these key parameters, "pre-simulating" potential chain reactions and risk consequences. This forward-looking predictive capability enables the system to identify risk zones before abnormal fluctuations actually occur. The closed loop from decision execution to feedback optimization ensures precise control and system evolution. The risk assessment module and the real-time adjustment module constitute the system's "decision-making and execution mechanism." The former assesses risks and prioritizes them based on prediction results, generating the optimal intervention strategy; the latter precisely executes adjustment commands, achieving real-time and smooth adjustments to process parameters. Crucially, the feedback optimization module acts as the system's "learning and evolution hub," continuously monitoring the adjustment effects and feeding the results back to the upstream digital mapping model and dynamic model, driving incremental learning and iterative updates. This transforms the entire system from a static program into an adaptive system that continuously improves itself with the accumulation of production data, becoming increasingly intelligent with use.
[0063] above Figure 5 The adaptive control system for nylon modification process parameters in this embodiment of the invention is described in detail from the perspective of modular functional entities. The adaptive control device for nylon modification process parameters in this embodiment of the invention is described in detail from the perspective of hardware processing.
[0064] Reference Figure 6 This invention also provides an adaptive control device for nylon modification process parameters. This adaptive control device for nylon modification process parameters can be a server, and its internal structure can be as follows: Figure 6As shown, this adaptive control device for nylon modification process parameters includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor, designed as a computer, provides computational and control capabilities. The memory of this adaptive control device for nylon modification process parameters includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of this adaptive control device for nylon modification process parameters stores the data corresponding to this embodiment. The network interface of this adaptive control device for nylon modification process parameters is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it can implement the above-described method.
[0065] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the adaptive control device for nylon modification process parameters applied thereto.
[0066] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the adaptive control method for nylon modification process parameters.
[0067] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0068] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0069] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. An adaptive control method for nylon modification process parameters, characterized in that, The method includes: S1. Extract relevant data streams including screw speed changes, melt temperature fluctuations and residence time differences from nylon modification process data, construct an initial correlation matrix of the relationship between parameters, store the initial correlation matrix into a digital mapping model, and obtain the coupling characteristics between parameters; S2. Based on the coupling characteristics, analyze the direct influence path and indirect chain reaction, obtain the degree of interference of the system viscosity, and calculate the effect weight according to the degree of interference; S3. Based on the said weight, screen out potential fluctuation source parameters, extract historical change data of their corresponding process parameters for the potential fluctuation source parameters, construct a dynamic model based on the historical change data, simulate the impact of parameter changes on key system outputs, and record the impact fluctuation values. S4. Determine whether the value of the fluctuation exceeds the preset safety threshold, detect the risk range and determine the severity level of the abnormal fluctuation based on the judgment result, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment. S5. Extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream. S6. Calculate the estimated viscosity and standard deviation of the melt temperature of the system based on the smoothed data stream, compare the estimated value and the standard deviation with the corresponding target value, obtain the comparison deviation, determine the effectiveness of the adjustment command based on the comparison deviation, and output a stable control scheme or start a feedback optimization strategy.
2. The method according to claim 1, characterized in that, S1 includes: The relevant data streams are cleaned to obtain an initial data stream; Based on the initial data stream, an initial correlation matrix is constructed to represent the relationship between the parameters, and the initial correlation matrix is stored in the digital mapping model; Based on the digital mapping model, the initial correlation matrix is analyzed using time series analysis methods to obtain the coupling characteristics between parameters.
3. The method according to claim 1, characterized in that, S2 includes: Based on the aforementioned coupling characteristics, the direct influence path and indirect chain reaction between screw speed, melt temperature and residence time are analyzed. Based on the direct influence path and the indirect chain reaction, the degree of interference of each parameter change on the system viscosity is quantified; Based on the degree of interference, the weights of the effects of screw speed variation, melt temperature fluctuation and residence time difference on the process state are determined.
4. The method according to claim 1, characterized in that, S3 includes: Parameters whose influence weight exceeds a preset weight threshold are marked as potential fluctuation source parameters; For the potential source parameters of fluctuation, extract the historical change data of their corresponding process parameters; Based on the historical change data, a dynamic model is constructed to simulate the parameter coupling effect; The dynamic model simulates the impact of expected changes in the parameters of the potential fluctuation source on the key output indicators of the system, and records the fluctuation values.
5. The method according to claim 1, characterized in that, S4 includes: Determine whether the fluctuation value exceeds a preset safety threshold; if so, determine that a risk range exists. The risk range is detected, the real-time pressure data corresponding to the risk range is obtained, and it is compared with the preset pressure threshold to determine the severity level of the abnormal fluctuation. Based on the severity level, the corresponding key intervention nodes are matched and filtered from the preset adjustment knowledge base; Based on the key intervention nodes and the severity level, calculate the priority sequence for parameter adjustments.
6. The method according to claim 1, characterized in that, S5 includes: According to the priority sequence, adjustment instructions are extracted from the key intervention nodes, and the adjustment instructions are used to determine the adjustment range of the screw speed change and the residence time difference; The adjustment command is transmitted to the control module, which drives the control module to adjust the screw speed or residence time in real time and obtain the adjusted process data stream; Based on the adjusted process data stream, a smoothed data stream is generated for monitoring changes in process status.
7. The method according to claim 1, characterized in that, S6 includes: Based on the smoothed data stream, sliding window statistical calculations are performed at fixed time intervals to obtain the estimated value of the system viscosity and the standard deviation of the melt temperature. The estimated viscosity of the system and the standard deviation of the melt temperature are compared with the preset viscosity target value and the preset temperature stability target value, respectively, to obtain the system viscosity deviation and temperature stability deviation. Determine whether the absolute value of the viscosity deviation of the system is less than a first deviation threshold and whether the temperature stability deviation is less than a second deviation threshold; If so, the adjustment command is deemed valid, and a stable control scheme for the nylon modification process is output. If not, the adjustment instruction is determined to need optimization, and a feedback optimization strategy is initiated based on the adjustment instruction until the adjustment instruction is determined to be valid.
8. The method according to claim 7, characterized in that, The feedback optimization strategy includes: Collect current process status data, executed adjustment commands, and adjustment effect data as optimization requirement data; Based on the optimization requirement data, the digital mapping model is incrementally trained to obtain an updated digital mapping model. Based on the updated digital mapping model, the interaction weights between screw speed, melt temperature and residence time differences are recalculated. Using the aforementioned action weights as input, and combining them with the current process status, the optimized adjustment instructions are iteratively calculated through the dynamic model.
9. An adaptive control system for nylon modification process parameters, used to implement the method as described in any one of claims 1-8, characterized in that, The adaptive control system for nylon modification process parameters includes: The data processing module is used to extract relevant data streams, including screw speed changes, melt temperature fluctuations and residence time differences, from nylon modification process data, construct an initial correlation matrix of the relationship between parameters, store the initial correlation matrix into a digital mapping model, and obtain the coupling characteristics between parameters. The weight calculation module is used to analyze the direct influence path and indirect chain reaction based on the coupling characteristics, obtain the degree of interference of the system viscosity, and calculate the effect weight according to the degree of interference. The dynamic simulation module is used to filter out potential fluctuation source parameters based on the said weight, extract historical change data of the corresponding process parameters for the potential fluctuation source parameters, construct a dynamic model based on the historical change data, simulate the impact of parameter changes on the key outputs of the system, and record the impact fluctuation values. The risk assessment module is used to determine whether the fluctuation value exceeds a preset safety threshold, detect the risk range based on the judgment result and determine the severity level of the abnormal fluctuation, match the corresponding key intervention nodes, and calculate the priority sequence of parameter adjustment. The real-time adjustment module is used to extract adjustment instructions according to the priority sequence, determine the adjustment range, adjust the process parameters in real time, obtain the adjusted process data stream, and form a smoothed data stream. The feedback optimization module is used to calculate the estimated value of the system viscosity and the standard deviation of the melt temperature based on the smoothed data stream, compare the estimated value and the standard deviation with the corresponding target value, obtain the comparison deviation, determine the effectiveness of the adjustment command based on the comparison deviation, and output a stable control scheme or start the feedback optimization strategy.
10. An adaptive control device for nylon modification process parameters, characterized in that, The device includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the adaptive control device for nylon modification process parameters to execute an adaptive control method for nylon modification process parameters as described in any one of claims 1-8.