Textile fabric dyeing control system for spinning

Abstract

The invention relates to the technical field of textile printing and dyeing, and particularly discloses a textile fabric dyeing control system for textile, which comprises a collaborative acquisition module, a reference matching module, a deviation compensation module and an evaluation evolution module, and is characterized in that a multi-sensing collaborative acquisition unit is constructed, key material fingerprints and key process variables are acquired, and the evaluation evolution module is used for evaluating the key material fingerprints based on the key material fingerprints; matching a similar optimal batch in a historical database to set a reference process track for a current dyeing batch, constructing and adopting a lightweight real-time digital twin, performing simulation prediction based on an intercepted current actually measured track to form a dyeing real-time state track, and performing track synchronous comparison and systematic deviation diagnosis with the reference process track to obtain a real-time dyeing state track. If systematic deviation is judged to occur, a dynamic differential compensation process is triggered to generate an optimal control strategy, if dyeing batches are completed, dyeing quality is evaluated, if the dyeing batches are qualified, the dyeing batches are classified into a historical database, and a contribution degree analysis process is started based on the optimal control strategy to update the historical database.

Description

Technical Field

[0001] This invention relates to the field of textile printing and dyeing technology, and specifically to a textile dyeing control system. Background Technology

[0002] In the textile printing and dyeing industry, the dyeing process is the core link that determines the final appearance quality and performance of textiles. Its process control level directly affects key quality indicators such as color difference, uniformity, and color fastness. Traditional dyeing control relies heavily on operator experience, adjusting by setting fixed temperature and time curves and supplementing with manual sampling and comparison. This method is slow to respond and inefficient in the face of interference from differences in fabric raw materials, batch fluctuations in dyes, and changes in equipment status. This easily leads to quality problems such as batch-to-batch color differences and dyeing variations, resulting in rework and waste of energy and raw materials.

[0003] To address these issues, existing technologies typically introduce online monitoring and automated control systems. These systems utilize sensors for temperature, pH, flow rate, and other parameters to achieve real-time acquisition of process parameters and automatic control of certain loops. However, such systems are often limited to monitoring and adjusting single or a few isolated variables, lacking a holistic understanding and collaborative optimization of the inherent multi-variable, strongly coupled, and nonlinear dynamic characteristics of the staining process. When multiple parameters deviate simultaneously, the system often only issues alarms, failing to automatically diagnose the root cause and provide comprehensive adjustment strategies.

[0004] Further analysis reveals that existing technical solutions generally suffer from a disconnect between perception and decision-making. The collected data is not deeply correlated with the essential factors determining the dyeing results, and the setting of control benchmarks is still based on fixed formulas or coarse classifications, failing to provide personalized and precise initial settings according to the actual state of each dye batch and the specific characteristics of the fabric. On the other hand, control and prediction are disconnected, lacking the ability to make advanced predictions and rolling optimizations based on process dynamic models. This prevents proactive intervention before quality deviations actually occur, and the solidification of experience and knowledge hinders the iterative evolution of system intelligence.

[0005] To address the above problems, this invention proposes a textile dyeing control system. Summary of the Invention

[0006] The purpose of this invention is to provide a textile dyeing control system to solve the aforementioned background problems.

[0007] The objective of this invention can be achieved through the following technical solution: a textile dyeing control system, comprising the following modules: Collaborative acquisition module: Constructs a multi-sensor collaborative acquisition unit to collect key material fingerprints and key process variables; Benchmark matching module: Based on the fingerprint of key materials, it matches similar optimal batches in the historical database to set the benchmark process trajectory for the current dyeing batch; Deviation Compensation Module: Constructs and adopts a lightweight real-time digital twin, performs simulation prediction based on the captured current measured trajectory to form the real-time dyeing state trajectory, and performs trajectory synchronization comparison and systematic deviation diagnosis with the benchmark process trajectory. If a systematic deviation is determined to occur, a dynamic differential compensation process is triggered to generate the optimal control strategy. Evaluation and evolution module: If the staining batch is completed, the staining quality is evaluated. If it is qualified, it is added to the historical database, and the contribution analysis process is initiated based on the optimal control strategy to update the historical database. Furthermore, the baseline process trajectory is obtained as follows: The historical database stores dyeing data records for dyeing batches that have passed quality assessment, including key material fingerprints, process variable trajectories, and dyeing quality assessment indicators for each batch. The weighted Mahalanobis distance algorithm is used to calculate the similarity score between the current key material fingerprint and each key material fingerprint in the historical database. Dyeing batches with similarity scores that reach or exceed the preset similarity threshold are listed as high similarity candidate sets. Among the dyeing quality evaluation indicators of each dyeing batch in the high similarity candidate set, the dyeing batch with the smallest final color difference ΔE value is marked as the best similarity batch. The corresponding process variable trajectory is set as the benchmark process trajectory of the current dyeing batch. Furthermore, the similarity score is calculated as follows: Z-score standardization preprocessing is performed on the key material fingerprints of the current dyeing batch and the key material fingerprints of all dyeing data records in the historical database. Based on the weighted Mahalanobis distance algorithm, combined with the inverse matrix of the covariance matrix calculated from all key material fingerprints in the historical database, the Mahalanobis distance between the current key material fingerprint and each key material fingerprint in the historical database is calculated and mapped to a similarity score between 0 and 1 through the Gaussian kernel function. Furthermore, the method for obtaining the real-time coloring status trajectory is as follows: The key process variables with unified timestamps collected within a fixed time window, starting from the current moment, are used as the current measured trajectory. The current measured trajectory is continuously input into a lightweight real-time digital twin with an embedded reduced-order dynamics model as the core of the calculation. The reduced-order dynamics model is a set of transfer functions trained using a system identification method based on the process variable trajectories in the historical database. The model performs forward rolling simulation prediction with the current measured trajectory as the initial condition, and outputs the predicted evolution trajectory of each key process variable in a short time domain in the future. The current measured trajectory and the predicted evolution trajectory together constitute the coloring real-time state trajectory. Furthermore, the diagnostic methods for systematic deviations are as follows: A dynamic time warping algorithm is used to nonlinearly align the real-time dyeing state trajectory with the baseline process trajectory. For each key process variable, the instantaneous deviation of its corresponding value on the real-time dyeing state trajectory and the baseline process trajectory, the cumulative deviation integral and the trend of change of the instantaneous deviation within the sliding time window on the real-time dyeing state trajectory are calculated. When any key process variable meets the preset trend deviation condition rule, it is determined that the key process variable has a trend deviation. If two or more key process variables that reveal a strong coupling relationship based on the reduced-order dynamic model are simultaneously deviated from the trend, and the direction of deviation is consistent with the logic of the revealed strong coupling relationship, it is determined that a systematic deviation has occurred. Furthermore, the trend deviation condition rules include amplitude conditions and trend conditions: The amplitude condition is that the absolute value of the instantaneous deviation continuously exceeds the corresponding preset static allowable deviation for a preset deviation duration. The trend condition is that the sign of the changing trend is consistent over a consecutive preset number of sliding time windows, and the absolute value of the cumulative deviation integral exceeds the preset dynamic deviation threshold. Furthermore, the optimal control strategy is generated as follows: If a systematic deviation occurs, the systematic deviation state is recorded as the initial condition. Under multiple future tentative control strategies for the main control equipment generated by a quadratic programming algorithm, forward rolling simulation is performed in a lightweight real-time digital twin to predict the evolution path of each key process variable under different control strategies and its comprehensive deviation value from the baseline process trajectory. A multi-objective optimization problem is constructed. Under the condition of satisfying the physical limits and process safety constraints of each main control equipment, the quadratic programming algorithm is used to solve the multi-objective optimization problem to obtain the optimal control strategy in the optimization time domain. Furthermore, the multi-objective optimization problem includes a primary objective function and a secondary objective function. The primary objective function aims to minimize the overall prediction deviation, while the secondary objective function aims to constrain the amplitude and frequency of the main control equipment's actions. Furthermore, the comprehensive deviation value is calculated as follows: For each key process variable, the difference between its predicted value at each moment in the optimization time domain and its corresponding value on the baseline process trajectory is calculated, and normalization is performed based on a preset normalization factor to obtain the normalized deviation. The analytic hierarchy process is used to assign a quality weight coefficient to the normalized deviation of each key process variable to calculate the weighted deviation square. Along the time axis of the optimization time domain, a time discount factor is introduced to accumulate and sum the weighted deviation squares of each key process variable at each moment to obtain the comprehensive deviation value. Furthermore, the historical database is updated in the following ways: The key material fingerprints, process variable trajectories, and dyeing quality assessment indicators of the qualified current dyeing batch are constructed into dyeing data records and stored in the historical database. The SHAP value analysis algorithm is used to compare the process variable trajectory under the intervention of the optimal control strategy for the current dyeing batch with the benchmark process trajectory, quantify the contribution score of the control actions of each main control device to the dyeing quality assessment indicators, associate the calculated contribution scores with the corresponding control actions, and update the metadata fields of the corresponding dyeing data records in the historical database.

[0008] The beneficial effects of this invention are as follows: 1. This invention, by constructing a multi-sensor collaborative acquisition unit and a lightweight real-time digital twin, achieves a leap from experience-driven to data and model-driven collaborative dyeing processes. It can intelligently match and set personalized optimal baseline process trajectories for the current batch based on real-time acquired key material fingerprints, thereby improving the accuracy and adaptability of control at the source. In actual operation, it can proactively diagnose systematic process deviations. Once a trend deviation is detected, the system automatically triggers dynamic differential compensation based on rolling time-domain optimization, generating and executing a collaborative control strategy to drive process parameters back to the correct track quickly and smoothly. This effectively solves the problems of lag and coarse adjustment in traditional methods, significantly improving the first-pass success rate, batch-to-batch consistency, and uniform dyeing.

[0009] 2. This invention further establishes a knowledge loop for continuous optimization, endowing the control system with self-evolution capabilities. The complete data of each qualified batch is structured and stored in a historical database. Through SHAP contribution analysis, the marginal impact of the control actions of different devices in the optimal control strategy on the final dyeing quality is quantified and archived. This makes the historical database not only a repository of cases, but also a knowledge base containing causal relationships. In subsequent production, this knowledge can be used to improve the accuracy of matching similar batches and enhance the reliability of digital twin simulation prediction. Through this continuous iteration of production-learning-application, the system can continuously accumulate and precipitate process knowledge, reduce dependence on the experience of core personnel, and ultimately achieve a step-by-step improvement in the overall control level and stability of the dyeing process. Attached Figure Description

[0010] The invention will now be further described with reference to the accompanying drawings.

[0011] Figure 1 This is a module architecture diagram of a textile dyeing control system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the specific steps of a textile dyeing control system according to an embodiment of the present invention. Detailed Implementation

[0012] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0013] Example 1 Please see Figure 1 and Figure 2 As shown in the embodiment of the present invention, a textile dyeing control system aims to solve the problem that existing dyeing processes rely on manual experience and are difficult to control in real time to ensure batch consistency. It constructs a multi-sensor collaborative acquisition unit to collect key material fingerprints and key process variables in real time, and matches similar optimal batches in a historical database to set a benchmark process trajectory. Then, it constructs and runs a lightweight real-time digital twin, continuously simulating and generating a real-time dyeing status trajectory. This trajectory is then compared synchronously with the benchmark trajectory for systematic deviation diagnosis. If a deviation is detected, a dynamic differential compensation process is triggered to generate and distribute the optimal control strategy for the main control equipment. After the dyed batch is qualified, a contribution analysis process is initiated based on the optimal control strategy to update the historical database. This achieves real-time perception, intelligent decision-making, precise control, and continuous knowledge evolution in the dyeing process, improving dyeing quality and production stability. Specifically, it includes the following modules: Collaborative acquisition module: Constructs a multi-sensor collaborative acquisition unit to collect key material fingerprints and key process variables; Specifically, a multi-sensor collaborative acquisition unit is constructed to collect the key material fingerprints and key process variables of the current dyeing batch. The multi-sensor collaborative acquisition unit includes a key material fingerprint acquisition subunit and a key process variable acquisition subunit. Specifically, the key material fingerprint acquisition subunit uses an online UV-Vis spectrometer installed in the bypass flow tank of the main circulation pipeline of the dyeing vat to perform real-time, non-contact transmission spectral scanning of the circulating dye liquor. It acquires characteristic absorption spectra containing information on dye type, concentration, and compatibility. Through the API protocol, based on the Manufacturing Execution System (MES) data interface, it obtains the textile model information of the current dyeing batch in real time. The textile model information is a structured process data package, for example: {Order No.: PO20231001, Fabric Model: CM40S, Weight: 120g / m²}. 2 Pretreatment process code: PT-03}; The key process variable acquisition subunit measures the color value of the outflowing dye liquor in real time using an online colorimeter installed at the dyeing vat outlet, and calculates the real-time dyeing rate (unit: %) using the Kubelka-Munk relation as a built-in algorithm. Temperature values ​​(unit: °C) are collected in real time using an array of temperature sensors deployed in key temperature zones such as the heating zone, fabric accumulation core zone, and outlet zone of the dyeing vat. The main circulation flow rate (unit: m³ / s) is monitored in real time using an electromagnetic flowmeter and pressure transmitter installed on the outlet pipeline of the main circulation pump. 3The pH value of the dye solution is monitored in real time by immersing a composite pH electrode with automatic cleaning function into the main circulation of the dyeing tank and the system pressure (unit: bar). The data collected by the multi-sensor collaborative acquisition unit is validated for rationality, and a first-order low-pass digital filter is used to smooth transient noise to obtain stable measurement values. Among them, the key material fingerprint collection subunit is triggered in the initialization stage of the current dyeing batch. It extracts the characteristic peak position, peak area and specific band absorbance ratio of the characteristic absorption spectrum collected in the initialization stage as spectral feature vectors. It performs feature-level fusion with the model feature vector composed of fabric model information provided by MES to generate key material fingerprints. The key material fingerprint is a structured vector. For example, the fingerprint of a key material is expressed as: FP_Batch_A={spectral characteristics:(λ_max:610nm, peak area:18500, A450 / A600:1.25), fabric characteristics:(model ID:CM40S, preprocessing code:PT-03)}; The key process variable acquisition subunit is synchronously triggered at a preset acquisition frequency during the actual dyeing stage of the current dyeing batch. It reads the stable measurement values ​​of each sensor in the key process variable acquisition subunit to obtain key process variables, including real-time dyeing rate, temperature value, main circulation flow rate, system pressure and pH value, and assigns a unified timestamp to the key process variables. Benchmark matching module: Based on the fingerprint of key materials, it matches similar optimal batches in the historical database to set the benchmark process trajectory for the current dyeing batch; Specifically, the historical database stores dyeing data records for dyeing batches that have passed quality assessment. The dyeing data records include the key material fingerprints, process variable trajectories, and dyeing quality assessment indicators for the dyeing batches. The process variable trajectories represent the time sequence of key process variables during the actual dyeing stage of the dyeing batches. The dyeing quality assessment indicators include the final color difference ΔE value measured by a laboratory colorimeter, the dyeing uniformity grade, and the physicochemical indicators required by the order. Z-score standardization preprocessing is performed on the key material fingerprints of the current dyeing batch and the key material fingerprints of all dyeing data records in the historical database. The weighted Mahalanobis distance algorithm is used as the core matching algorithm. Combined with the inverse matrix of the covariance matrix calculated from all key material fingerprints in the historical database, the Mahalanobis distance between the key material fingerprint of the current dyeing batch and each key material fingerprint in the historical database is calculated. The Gaussian kernel function is used to map the calculated Mahalanobis distance into a similarity score between 0 and 1. It should be noted that the weighted Mahalanobis distance algorithm considers the correlation and dimensional differences between the elements of the key material fingerprint through the covariance matrix, ensuring scale independence in the calculation process. The calculated similarity score is compared with the preset similarity threshold. All dyeing batches with similarity scores that reach or exceed the threshold are listed as high similarity candidate sets. Within the high similarity candidate set, an optimization sorting process based on quality constraints is initiated. The dyeing quality evaluation index associated with each dyeing batch in the high similarity candidate set is read. According to the preset sorting rules, the batches are sorted in ascending order based on the final color difference ΔE value. Finally, the single dyeing batch with the smallest final color difference ΔE value is selected from the high similarity candidate set and marked as the best similarity batch. The process variable trajectory of the best similarity batch is set as the baseline process trajectory of the current dyeing batch. It should be noted that the function of this module is to standardize and calculate the similarity between the key material fingerprint of the current batch and the key material fingerprint of each batch in the historical database, screen out the highly similar candidate set, and select the most similar batch based on its historical dyeing quality evaluation index. The process variable trajectory of the batch is set as the benchmark process trajectory of the current batch, which ensures that the set benchmark process trajectory is not only similar in working conditions, but also has the best results, thus improving the practical guiding value of the matching results. Deviation Compensation Module: Constructs and adopts a lightweight real-time digital twin, performs simulation prediction based on the captured current measured trajectory to form the real-time dyeing state trajectory, and performs trajectory synchronization comparison and systematic deviation diagnosis with the benchmark process trajectory. If a systematic deviation is determined to occur, a dynamic differential compensation process is triggered to generate the optimal control strategy. Specifically, a lightweight real-time digital twin for textile dyeing is constructed and continuously run. The lightweight real-time digital twin uses an embedded reduced-order dynamics model as its computational core. The reduced-order dynamics model is a set of transfer functions trained using a system identification method based on the process variable trajectories in the historical database. It simulates the dynamic coupling relationship and evolution law between key process variables during the dyeing process. In the actual dyeing stage of the current dyeing batch, the key process variables with a unified timestamp collected within a fixed time window with the current time as the endpoint are used as the current measured trajectory. The current measured trajectory is continuously input into the lightweight real-time digital twin. The reduced-order dynamic model performs forward rolling simulation prediction with the current measured trajectory as the initial condition, and outputs the predicted evolution trajectory of each key process variable in a short time domain in the future. The current measured trajectory and the predicted evolution trajectory together constitute the real-time dyeing state trajectory. Based on the real-time dyeing status trajectory and the benchmark process trajectory, the trajectory is synchronously compared and systematic deviation is diagnosed; Specifically, a dynamic time warping algorithm is used to nonlinearly align the real-time dyeing status trajectory with the baseline process trajectory to eliminate the small time phase difference that may exist in the initialization stage of the dyeing batch. For each key process variable i, the instantaneous deviation e_i(t) of the key process variable on the real-time dyeing status trajectory and the corresponding value on the baseline process trajectory is calculated, and the cumulative deviation integral ∫e_i(t)dt and the trend of change de_i(t) / dt of the instantaneous deviation within a sliding time window on the real-time dyeing status trajectory are calculated, where t represents time. The rules for defining trend deviation include amplitude conditions and trend conditions. The amplitude condition is that the absolute value of the instantaneous deviation e_i(t) continuously exceeds the corresponding preset static allowable deviation for a preset deviation duration. The trend condition is that the sign of the changing trend is consistent over a consecutive preset number of sliding time windows, and the absolute value of the cumulative deviation integral exceeds the preset dynamic deviation threshold. If any key process variable simultaneously satisfies the trend deviation condition rule, it is determined that the corresponding key process variable has a trend deviation. If two or more key process variables that reveal a strong coupling relationship based on the reduced-order dynamics model simultaneously exhibit trend deviations, and the direction of the trend deviation is consistent with the logic of the strong coupling relationship revealed by the reduced-order dynamics model, it is determined that a systematic deviation has occurred, the systematic deviation status is recorded, and the dynamic differential compensation process is triggered. If the dynamic differential compensation process is triggered, a lightweight real-time digital twin is used as the simulation engine. The currently diagnosed systematic deviation state is used as the initial condition. Forward rolling simulation is performed under multiple future tentative control strategies for the main control equipment generated by the quadratic programming algorithm. The simulation predicts the evolution path of each key process variable under different future tentative control strategies and its comprehensive deviation from the baseline process trajectory in a future optimization time domain. Specifically, the calculation method for the comprehensive deviation value is as follows: For each key process variable, the difference between the predicted value of the key process variable at each moment in the optimization time domain and the corresponding value in the baseline process trajectory is calculated, and normalization is performed based on a preset normalization factor to obtain the normalized deviation. The analytic hierarchy process is used to assign a quality weight coefficient to the normalized deviation of each key process variable to calculate the weighted deviation square. Along the time axis of the optimization time domain, a time discount factor is introduced to accumulate and sum the weighted deviation squares of each key process variable at each moment to obtain the comprehensive deviation value. The main control equipment includes a main circulation pump frequency converter, a main heating and cooling power regulator, and a central dye and auxiliary metering pump. A multi-objective optimization problem is constructed, in which the primary objective function aims to minimize the comprehensive deviation of the prediction, while the secondary objective function constrains the action amplitude and frequency of the main control equipment to ensure control stability. Under the condition of satisfying the physical limits and process safety constraints of each main control equipment, the optimal control strategy in the optimization time domain is obtained by using a quadratic programming algorithm. The optimal control strategy is immediately calculated into a specific set of coordinated control instructions and sent to the main control equipment to implement comprehensive regulation through coordinated actions, driving key process variables back to the baseline process trajectory. It should be noted that the function of this module is to construct a lightweight real-time digital twin with an embedded reduced-order dynamic model as its core. Based on the current measured trajectory, it performs rolling simulation prediction to form a real-time dyeing state trajectory. It performs dynamic time warping comparison and systematic deviation diagnosis with the benchmark process trajectory. If a systematic deviation is determined, it triggers a dynamic differential compensation process. Using the digital twin as a simulation engine, it solves the optimal control strategy for the future through a rolling time-domain optimization algorithm and sends it to the main control equipment for execution. This realizes dynamic monitoring and active intervention of complex dyeing processes, can correct systematic deviations in a timely manner, ensure that the process operation closely follows the high-quality benchmark, and greatly improve the accuracy and stability of control. Evaluation and evolution module: If the staining batch is completed, the staining quality is evaluated. If it is qualified, it is added to the historical database, and the contribution analysis process is initiated based on the optimal control strategy to update the historical database. Specifically, when the current dyeing batch is completed, the dyeing quality assessment process is triggered to conduct laboratory standard measurements on the dyed textiles and obtain dyeing quality assessment indicators. The dyeing quality assessment process involves using a laboratory colorimeter to measure multiple points on dyed textile samples, calculating the final color difference ΔE value and dyeing uniformity grade, and obtaining the physicochemical indicators required by the order through physicochemical testing. Examples include physicochemical properties such as friction fastness and wash fastness; The dyeing quality assessment indicators are judged to be qualified. If all indicators meet the preset quality standards, the current dyeing batch is judged to be qualified. If it is not qualified, a prompt is sent to the administrator terminal. If the current dyeing batch is qualified, the key material fingerprints, process variable trajectories, and dyeing quality evaluation indicators of the current dyeing batch will be constructed as dyeing data records and stored in the historical database in a structured manner through the database API protocol. For example, the dyeing data record is expressed as: Record_Batch_A={Key material fingerprint:FP_Batch_A, Process variable trajectory:{Timestamp sequence:[t1,t2,...], Dyeing rate:[% sequence], Temperature:[℃ sequence], Main circulation flow rate:[m 3[h sequence], System pressure: [bar sequence], pH value: [sequence]}, Staining quality assessment index: {ΔE: 1.2, Uniformity grade: 4, Fastness index: {Friction: grade 4, Washing: grade 3-4}}}; Based on the optimal control strategy generated by triggering the dynamic differential compensation process in the current staining batch, the contribution analysis process is initiated to update the historical database. Specifically, the contribution analysis process uses the SHAP value analysis algorithm as the core analysis method. It takes the process variable trajectory of similar best batches in the historical database as the benchmark process trajectory, compares the process variable trajectory under the intervention of the optimal control strategy of the current dyeing batch with the benchmark process trajectory, and quantifies the contribution score of the control actions of each main control equipment to the dyeing quality evaluation index. It should be noted that the SHAP value analysis algorithm is based on game theory, assigning a contribution score to each control action to explain its marginal impact on quality indicators. The calculated contribution score is associated with the corresponding control action and updated to the metadata field of the corresponding colored data record in the historical database; For example, the updated dyeing data record adds contribution metadata: {Main circulation pump inverter action contribution: 0.35, main heater regulator action contribution: 0.45, central dye metering pump action contribution: 0.20}}; By regularly executing the contribution analysis process, the historical staining data records continuously accumulate contribution information, thereby improving the accuracy of subsequent staining batches in matching similar optimal batches, as well as the reliability of lightweight real-time digital twin simulation prediction. It should be noted that the role of this module is to introduce the SHAP value analysis algorithm to analyze the contribution of control actions, transforming the difficult-to-quantify expert experience into explicit knowledge that can be calculated, stored, and reused, driving the entire system to continuously evolve from case-based to causal knowledge-based, ensuring basic compensation while achieving long-term adaptive optimization. The technical solution of this invention is as follows: a multi-sensor collaborative acquisition unit is constructed to collect key material fingerprints and key process variables. Based on the key material fingerprints, similar optimal batches are matched in the historical database to set the benchmark process trajectory for the current dyeing batch. A lightweight real-time digital twin is constructed and adopted. Based on the captured current measured trajectory, simulation prediction is performed to form the real-time dyeing state trajectory. The trajectory is synchronously compared with the benchmark process trajectory and systematic deviation diagnosis is performed. If a systematic deviation is determined to occur, a dynamic differential compensation process is triggered to generate the optimal control strategy. If the dyeing batch is completed, the dyeing quality is evaluated. If it is qualified, it is added to the historical database, and a contribution analysis process is initiated based on the optimal control strategy to update the historical database.

[0014] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A textile dyeing control system, characterized in that: Includes the following modules: Collaborative acquisition module: Constructs a multi-sensor collaborative acquisition unit to collect key material fingerprints and key process variables; Benchmark matching module: Based on the fingerprint of key materials, it matches similar optimal batches in the historical database to set the benchmark process trajectory for the current dyeing batch; Deviation Compensation Module: Constructs and adopts a lightweight real-time digital twin, performs simulation prediction based on the captured current measured trajectory to form the real-time dyeing state trajectory, and performs trajectory synchronization comparison and systematic deviation diagnosis with the benchmark process trajectory. If a systematic deviation is determined to occur, a dynamic differential compensation process is triggered to generate the optimal control strategy. Evaluation and evolution module: Once the staining batch is completed, the staining quality is evaluated. If it is qualified, it is added to the historical database, and the contribution analysis process is initiated based on the optimal control strategy to update the historical database.

2. The textile dyeing control system according to claim 1, characterized in that: The baseline process trajectory is obtained as follows: The historical database stores dyeing data records for dyeing batches that have passed quality assessment, including key material fingerprints, process variable trajectories, and dyeing quality assessment indicators for each batch. The weighted Mahalanobis distance algorithm is used to calculate the similarity score between the current key material fingerprint and each key material fingerprint in the historical database. Dyeing batches with similarity scores that reach or exceed the preset similarity threshold are listed as high similarity candidate sets. Among the dyeing quality evaluation indicators of each dyeing batch in the high similarity candidate set, the dyeing batch with the smallest final color difference ΔE value is marked as the most similar batch. The corresponding process variable trajectory is set as the benchmark process trajectory of the current dyeing batch.

3. A textile dyeing control system according to claim 2, characterized in that: The similarity score is calculated as follows: Z-score normalization preprocessing is performed on the key material fingerprints of the current dyeing batch and the key material fingerprints of all dyeing data records in the historical database. Based on the weighted Mahalanobis distance algorithm, combined with the inverse matrix of the covariance matrix calculated from all key material fingerprints in the historical database, the Mahalanobis distance between the current key material fingerprint and each key material fingerprint in the historical database is calculated and mapped to a similarity score between 0 and 1 through the Gaussian kernel function.

4. A textile dyeing control system according to claim 1, characterized in that: The method for obtaining the real-time coloring status trajectory is as follows: The key process variables with uniform timestamps collected within a fixed time window, starting from the current moment, are used as the current measured trajectory. The current measured trajectory is continuously input into a lightweight real-time digital twin with an embedded reduced-order dynamics model as the core of the calculation. The reduced-order dynamics model is a set of transfer functions trained using a system identification method based on the process variable trajectories in the historical database. The model performs forward rolling simulation prediction with the current measured trajectory as the initial condition, and outputs the predicted evolution trajectory of each key process variable in a short time domain in the future. The current measured trajectory and the predicted evolution trajectory together constitute the coloring real-time state trajectory.

5. A textile dyeing control system according to claim 4, characterized in that: The methods for diagnosing systematic deviations are as follows: A dynamic time warping algorithm is used to nonlinearly align the real-time dyeing state trajectory with the baseline process trajectory. For each key process variable, the instantaneous deviation of its corresponding value on the real-time dyeing state trajectory and the baseline process trajectory, the cumulative deviation integral and the trend of change of the instantaneous deviation within the sliding time window on the real-time dyeing state trajectory are calculated. When any key process variable meets the preset trend deviation condition rule, it is determined that the key process variable has a trend deviation. If two or more key process variables that reveal a strong coupling relationship based on the reduced-order dynamic model are simultaneously deviated from the trend, and the direction of deviation is consistent with the logic of the revealed strong coupling relationship, it is determined that a systematic deviation has occurred.

6. A textile dyeing control system according to claim 5, characterized in that: Trend deviation condition rules include magnitude condition and trend condition: The amplitude condition is that the absolute value of the instantaneous deviation continuously exceeds the corresponding preset static allowable deviation for a preset deviation duration. The trend condition is that the sign of the changing trend is consistent over a consecutive preset number of sliding time windows, and the absolute value of the cumulative deviation integral exceeds the preset dynamic deviation threshold.

7. A textile dyeing control system according to claim 1, characterized in that: The optimal control strategy is generated as follows: If a systematic deviation occurs, the systematic deviation state is recorded as the initial condition. Under multiple future tentative control strategies for the main control equipment generated by a quadratic programming algorithm, forward rolling simulation is performed in a lightweight real-time digital twin to predict the evolution path of each key process variable under different control strategies and its comprehensive deviation value from the baseline process trajectory. A multi-objective optimization problem is constructed. Under the condition of satisfying the physical limits and process safety constraints of each main control equipment, the quadratic programming algorithm is used to solve the multi-objective optimization problem to obtain the optimal control strategy in the optimization time domain.

8. A textile dyeing control system according to claim 7, characterized in that: Multi-objective optimization problems include a primary objective function and a secondary objective function. The primary objective function aims to minimize the overall deviation of the prediction, while the secondary objective function aims to constrain the amplitude and frequency of the main control equipment's actions.

9. A textile dyeing control system according to claim 8, characterized in that: The calculation method for the overall deviation value is as follows: For each key process variable, the difference between its predicted value at each moment in the optimization time domain and its corresponding value on the baseline process trajectory is calculated, and normalization is performed based on a preset normalization factor to obtain the normalized deviation. The analytic hierarchy process (AHP) is used to assign a quality weight coefficient to the normalized deviation of each key process variable to calculate the weighted deviation square. Along the time axis of the optimization time domain, a time discount factor is introduced to accumulate and sum the weighted deviation squares of each key process variable at each moment to obtain the comprehensive deviation value.

10. A textile dyeing control system according to claim 1, characterized in that: The historical database is updated in the following ways: The key material fingerprints, process variable trajectories, and dyeing quality assessment indicators of the qualified current dyeing batch are constructed into dyeing data records and stored in the historical database. The SHAP value analysis algorithm is used to compare the process variable trajectory under the intervention of the optimal control strategy for the current dyeing batch with the benchmark process trajectory, quantify the contribution score of the control actions of each main control device to the dyeing quality assessment indicators, associate the calculated contribution scores with the corresponding control actions, and update the metadata fields of the corresponding dyeing data records in the historical database.

Images