A pigment dispersibility monitoring control system for waterless printing fabric

Abstract

The present application relates to the technical field of pigment dispersity monitoring, and particularly relates to a pigment dispersity monitoring control system for water-free printing fabric; initial parameters of a sample are obtained, a preset disturbance signal is applied to the sample, and disturbance parameters of the sample are collected; a stability index is calculated according to the initial parameters and the disturbance parameters, baseline calibration is performed, and calibrated sample state parameters are output; a sample flow process is measured, dynamic changes of the measurement signal are analyzed, and process characteristic parameters of sample particles are obtained; the sample state parameters and the process characteristic parameters are identified, and a process offset coefficient is obtained; the sample is compensated based on the process offset coefficient, and compensated process conveying parameters are obtained; running state parameters of a production device are obtained, and quality characteristic parameters of a processing object are read based on an optical detection device; process abnormalities are identified based on the running state parameters, the quality characteristic parameters and the process conveying parameters, and early warning control is performed.

Description

Technical Field

[0001] This invention relates to the field of pigment dispersion monitoring technology, specifically to a pigment dispersion monitoring and control system for waterless printed fabrics. Background Technology

[0002] In the field of technology that relies on high-precision fluid materials, the stability of the internal physical state of a dispersion system composed of micro-particles is a prerequisite for determining the final functional performance of the material. For example, the performance of pigments used in waterless printed fabrics is highly dependent on the size and distribution of the internal particles and the interaction forces between the particles. These factors together determine the overall physical properties of the system.

[0003] For this type of material, its performance under dynamic physical processes such as flow, shearing, and phase transformation is crucial. Any subtle changes in the microstructure of the system, such as irreversible agglomeration or flocculation of particles, will directly impair the uniformity and stability of its macroscopic physical properties. Therefore, accurately quantifying the physical properties of the dispersed system under dynamic conditions is a core technical challenge for achieving its quality control and reliability assurance.

[0004] However, existing physical property analysis methods have significant limitations in this regard. Traditional measurement techniques, such as offline particle size analysis or rheological measurement, typically evaluate samples under static or quasi-static conditions. These measurements cannot fully reflect the instantaneous state and evolution of materials under real, complex dynamic conditions, leading to discrepancies between analytical data and actual application performance. Furthermore, relying on macroscopic performance testing of the final application form is a lagging evaluation method that cannot provide preventative insights and is difficult to use for real-time control of material dynamic processes. Summary of the Invention

[0005] The purpose of this invention is to provide a pigment dispersion monitoring and control system for waterless printed fabrics, which identifies process anomalies based on operating status parameters, quality characteristic parameters, and process transport parameters, and performs early warning control.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A pigment dispersion monitoring and control system for waterless printed fabrics includes:

[0008] The characterization unit identifies the dispersion characteristics, stability characteristics, and rheological characteristics of the sample to obtain the initial parameters of the sample; applies a preset perturbation signal to the sample and collects the sample perturbation parameters; calculates the stability index based on the initial parameters and sample perturbation parameters, performs baseline calibration, and outputs the calibrated sample state parameters; the sample is a pigment.

[0009] The dynamic analysis unit measures the sample flow process using an acoustic detection device, analyzes the dynamic changes of the measurement signal, and obtains the process characteristic parameters of the sample particles during the flow process; it constructs a process state analysis model to identify the sample state parameters and process characteristic parameters, and obtains the process offset coefficient; based on the process offset coefficient, it performs state compensation on the sample to obtain the compensated process transport parameters.

[0010] The quality inspection unit acquires the operating status parameters of the production equipment and reads the quality characteristic parameters of the processed object based on the optical inspection device; it identifies process abnormalities based on the operating status parameters, quality characteristic parameters, and process conveying parameters, and performs early warning control.

[0011] The dispersion characteristics include the average particle size, polydispersity index, and particle size distribution curve of the sample particles, which are obtained by scanning the upper, middle, and lower layers of the sample in the storage space. The stability characteristics include the electrostatic repulsion between sample particles. The rheological characteristics include the sample viscosity.

[0012] The specific process of applying a preset disturbance signal and calculating the stability index includes:

[0013] By applying standardized temperature changes and high-frequency vibrations to the sample through a miniature heating and piezoelectric vibration unit within an integrated sensor, the dispersion characteristics, stability characteristics, and rheological characteristics after disturbance are identified, and the sample disturbance parameters are obtained.

[0014] Based on the changes in the sample perturbation parameters and the initial sample parameters, the perturbation change parameters are obtained, specifically including the average particle size difference, polydispersity index difference, particle size distribution curve shape difference, electrostatic repulsion difference, and sample viscosity difference.

[0015] The stability index is determined based on the degree of difference between the perturbation variation parameters and the standard variation parameters; when the stability index exceeds the stability threshold, baseline calibration is initiated.

[0016] The process of obtaining the process characteristic parameters includes:

[0017] During the sample flow process, an acoustic detection device emits ultrasonic waves into the sample flowing in the sample delivery pipe and acquires the reflected echoes caused by the movement of sample particles. The measurement signal is obtained by combining the emitted ultrasonic waves and the reflected echoes.

[0018] The Doppler frequency shift of the reflected echo in the real-time analysis and measurement signal is used to calculate the velocity distribution of sample particles in the cross-section of the conveying pipeline, which serves as a process characteristic parameter; the acoustic detection device includes an ultrasonic detection device.

[0019] The identification process of the process state analysis model includes:

[0020] Obtain the sample state parameters output by the characteristic characterization unit; based on the sample state parameters, establish a flow velocity distribution benchmark model for the current batch of samples;

[0021] The flow velocity distribution in the process characteristic parameters is continuously compared and analyzed with the flow velocity distribution benchmark model to identify the degree of deviation of the real-time flow velocity distribution from the flow velocity distribution benchmark model, and the process deviation coefficient is quantified.

[0022] The process offset coefficient is compared with a preset process offset threshold. When the process offset coefficient is greater than the process offset threshold, compensation control for the sample at the end of the delivery is triggered.

[0023] By using a pigment processing anomaly model to identify operational status parameters, quality characteristic parameters, and process transport parameters, processing anomalies are determined. Specifically, this includes:

[0024] The pigment processing anomaly model performs time alignment processing on the operating status parameters, quality characteristic parameters, and process transport parameters to obtain pigment processing time series data.

[0025] By using a long short-term memory network algorithm, we learn the dynamic correlation patterns between pigment processing time-series data and processing anomalies that occurred in historical production.

[0026] In real-time production, the probability of a process anomaly is identified based on the dynamic correlation pattern. When the probability of an anomaly exceeds a preset process anomaly threshold, it is determined to be a process anomaly, and an early warning control is executed.

[0027] The method for obtaining the standard variation parameters is as follows:

[0028] The system also includes a final quality acquisition unit and a dynamic benchmark optimization unit;

[0029] The final quality acquisition unit is used to collect the final quality parameters of the printed fabric after baking and curing.

[0030] The dynamic benchmark optimization unit deeply correlates historical process data with corresponding final quality parameters, reverse-optimizes and generates a dynamic process benchmark that ensures the best finished product quality; the dynamic process benchmark includes standard variation parameters.

[0031] The final quality parameters are multidimensional quality vectors, including: acoustic fastness factor, dynamic flexibility index, and Raman crosslinking characteristic peak shift;

[0032] The acoustic fastness factor is used to characterize color fastness and is quantified by applying standardized micromechanical perturbations to the cured film layer and analyzing the spectral characteristics of the acoustic response signal.

[0033] The dynamic flexibility index is used to characterize the softness of the fabric. It is quantified by measuring the rate of change of the time autocorrelation function of the laser speckle pattern on the surface when micro-vibration is applied to the fabric.

[0034] The Raman crosslinking characteristic peak shift is used to characterize the quality of the cured film. It is obtained by monitoring the Raman characteristic peaks of the adhesive chemical bonds in the cured film with an online Raman spectrometer and calculating the shift relative to the standard peak position in the uncrosslinked state. The shift is directly related to the degree of crosslinking of the cured film.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. This invention uses a micro-heating and piezoelectric vibration unit to simulate the environmental pressures such as temperature fluctuations and mechanical shearing that pigments may encounter during actual production and transportation. By accurately comparing the changes in multi-dimensional parameters before and after the disturbance, and quantitatively comparing them with standard change parameters derived from best production practices, a comprehensive stability index is obtained. This effectively identifies batches with insufficient stability before materials are put into the production line, eliminating production anomalies caused by pigment agglomeration, sedimentation, and other problems at the source, significantly improving production stability and yield.

[0037] 2. This invention refines the internal logic of the process state analysis model and establishes a batch-specific flow velocity distribution benchmark model; it continuously quantifies complex flow velocity distribution differences into a single, intuitive process deviation coefficient, realizing standardized and real-time assessment of the degree of process deviation, thereby triggering compensation control; it can autonomously and quickly correct the deterioration trend of the flow state, eliminate potential process fluctuations in their infancy, and ensure a high degree of consistency and stability in the production process.

[0038] 3. This invention collects the final quality parameters and performs in-depth correlation analysis between the final quality parameters and massive process data in the production process; then, through the dynamic benchmark optimization unit, it reverse-calculates the ideal process parameters and dynamic benchmarks that should be followed to achieve the best quality, including standard variation parameters used for stability testing; so that the system's control target is no longer a rigid fixed value, but an optimal solution that can continuously iterate and dynamically optimize based on historical data and quality feedback, leading the production process to continuously and automatically evolve towards higher quality and higher efficiency.

[0039] 4. This invention further defines the final quality parameter as a multidimensional quality vector composed of acoustic fastness factor, dynamic flexibility index and Raman crosslinking characteristic peak shift; it provides a comprehensive and in-depth analysis of product quality from three key dimensions: physical and mechanical properties, fabric sensory properties and chemical curing degree, and realizes real-time, quantitative and non-destructive online characterization of the core quality indicators of printed fabrics; it provides a data foundation for the dynamic benchmark optimization unit to perform self-learning and reverse optimization. Attached Figure Description

[0040] Figure 1 This is a logical schematic diagram of the characteristic characterization unit of the present invention;

[0041] Figure 2 This is a logical schematic diagram of the dynamic analysis unit of the present invention;

[0042] Figure 3 This is a logical diagram of the quality inspection unit of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example 1:

[0045] This invention proposes a pigment dispersion monitoring and control system for waterless printed fabrics, including a characteristic characterization unit, a dynamic analysis unit, and a quality detection unit.

[0046] The characterization unit identifies the dispersion, stability, and rheological properties of the sample to obtain initial sample parameters; it applies a preset perturbation signal to the sample and acquires the perturbation parameters; it calculates the stability index based on the initial and perturbation parameters, performs baseline calibration, and outputs the calibrated sample state parameters; the sample is a pigment; the logic of the characterization unit is as follows: Figure 1 As shown.

[0047] The dynamic analysis unit measures the sample flow process using an acoustic detection device, analyzes the dynamic changes of the measurement signal, and obtains the process characteristic parameters of the sample particles during the flow process; it constructs a process state analysis model, identifies the sample state parameters and process characteristic parameters, and obtains the process offset coefficient; based on the process offset coefficient, it performs state compensation on the sample to obtain the compensated process transport parameters; the logic of the dynamic analysis unit is as follows: Figure 2 As shown.

[0048] The quality inspection unit acquires the operating status parameters of the production equipment and reads the quality characteristic parameters of the processed object using an optical inspection device. Based on the operating status parameters, quality characteristic parameters, and process transport parameters, it identifies process anomalies and performs early warning control. The logic of the quality inspection unit is as follows: Figure 3 As shown.

[0049] Preferably, the dispersion characteristics include the average particle size, polydispersity index, and particle size distribution curve of the sample particles, which are obtained by scanning the upper, middle, and lower layers of the sample in the storage space; the stability characteristics include the electrostatic repulsion between sample particles; and the rheological characteristics include the sample viscosity.

[0050] Average particle size: refers to the average size of pigment particles; the smaller the particle size, the larger the specific surface area of ​​the pigment, the stronger the tinting strength, and the higher the gloss. It is obtained by multi-level scanning of the upper, middle, and lower layers of the sample within the storage container. This avoids sampling bias caused by gravity settling and ensures data accuracy.

[0051] Polydispersity index (PDI): This is a dimensionless parameter used to characterize the width of the particle size distribution. A PPI value closer to 0 indicates more uniform particle size, a more stable dispersion system, and less susceptibility to aggregation; conversely, a higher PPI value indicates a wider particle size distribution range and relatively poorer system stability. It is obtained in the same way as the average particle size, measured simultaneously in multi-level sigma analysis.

[0052] Particle size distribution curve: This curve visually displays the percentage of particles in different size ranges. It provides more complete information than average particle size and polydispersity index, revealing the presence of small amounts of large particles that could cause printhead clogging or printing defects. It is an important basis for precise quality control.

[0053] Electrostatic repulsion between sample particles refers to the mutual repulsion force generated by the like charges on the surfaces of pigment particles. In other words, electrostatic repulsion is a key force in maintaining the stable suspension of particles in the dispersion medium, counteracting van der Waals forces, and thus preventing aggregation or flocculation. The greater the repulsion, the better the dispersion stability of the pigment and the longer its shelf life.

[0054] Sample viscosity: refers to the fluid's ability to resist flow. Its meaning is directly related to the processing performance of pigments. If the viscosity is too high, it will lead to pumping difficulties, increased energy consumption, and difficulty in uniformly penetrating into the fabric fibers; if the viscosity is too low, it may cause problems such as blurred edges of printed patterns and bleeding.

[0055] Among them, a dynamic light scattering particle size analyzer is used to perform multi-layer scanning of the upper, middle and lower layers of the sample in the storage space to obtain the average particle size, polydispersity index and particle size distribution curve of the sample particles; a Zeta potential analyzer is used to measure the electrostatic repulsion between sample particles; and a Brookfield rotational viscometer is used to measure the viscosity of the sample.

[0056] This invention decomposes sample characteristics into a series of precise and quantifiable physical indicators such as average particle size, polydispersity index, electrostatic repulsion, and viscosity. By comprehensively scanning and analyzing different layers, it effectively overcomes measurement errors caused by sample inhomogeneity, ensuring the representativeness and accuracy of initial parameters. This greatly improves the reliability and accuracy of subsequent baseline calibration, model building, and anomaly diagnosis, providing a crucial and unbiased data source for achieving high-precision process control.

[0057] Preferably, the specific process of applying a preset disturbance signal and calculating the stability index includes:

[0058] By applying standardized temperature changes and high-frequency vibrations to the sample through a miniature heating and piezoelectric vibration unit within an integrated sensor, the dispersion characteristics, stability characteristics, and rheological characteristics after disturbance are identified, and the sample disturbance parameters are obtained.

[0059] Based on the changes in the sample perturbation parameters and the initial sample parameters, the perturbation change parameters are obtained, specifically including the average particle size difference, polydispersity index difference, particle size distribution curve shape difference, electrostatic repulsion difference, and sample viscosity difference.

[0060] The stability index is determined based on the degree of difference between the perturbation variation parameters and the standard variation parameters; when the stability index exceeds the stability threshold, baseline calibration is initiated.

[0061] By actively applying standardized physical perturbations to samples and quantifying the changes in their state parameters before and after perturbation, the stability of samples in complex production environments is prospectively assessed. The measured perturbation parameters are then mathematically compared with built-in standard parameters, such as by calculating Euclidean distance, weighted difference, or Mahalanobis distance, ultimately yielding a quantified stability index. The stability index signifies that a smaller index indicates the sample's actual changes are closer to the ideal state, indicating higher stability; an index exceeding a threshold indicates poor sample stability and a high risk of aggregation and sedimentation in actual production.

[0062] Among them, the standard variation parameters are obtained by reverse optimization through the dynamic benchmark optimization unit. They represent the ideal range of changes in the various physical parameters of the pigment that can produce the best quality printed fabric after being subjected to standardization perturbation.

[0063] Furthermore, based on the degree of difference between the disturbance variation parameters and the standard variation parameters, the stability index is specifically determined to include:

[0064] The shape difference of the particle size distribution curve is not a single numerical value, but is decomposed into a shape feature vector, which includes three dimensions: kurtosis, skewness, and multimodal index, in order to capture early subtle signs of aggregation or flocculation.

[0065] All perturbation parameters, including normalized average particle size difference, polydispersity index difference, shape eigenvector, electrostatic repulsion difference, and sample viscosity difference, are combined to form a multidimensional variation vector.

[0066] A benchmark covariance matrix is ​​established by reverse optimization using a dynamic benchmark optimization unit; based on the multidimensional change vector and the benchmark covariance matrix, the Mahalanobis distance of the change vector of the current batch of samples relative to the standard change parameter is calculated as a stability index.

[0067] The aforementioned benchmark covariance matrix system utilizes historical batch data obtained through back-optimization by a dynamic benchmark optimization unit, which yields the best quality finished product. It calculates the covariance matrix of the variation vector of the optimal sample after undergoing standardized perturbation. This matrix not only includes the normal fluctuation range of each parameter (diagonal elements) but, more importantly, the strength and direction of the interrelationships between parameters (off-diagonal elements). For example, it describes the extent to which the viscosity should change accordingly when the particle size slightly increases for a stable sample.

[0068] This invention decomposes the shape differences of particle size distribution curves into multi-dimensional feature vectors such as kurtosis, skewness, and multi-peak index. The system can capture extremely subtle early signs of particle agglomeration that cannot be detected by traditional average particle size or multi-dispersion index. By using Mahalanobis distance instead of traditional weighted algorithms, it can comprehensively consider the intrinsic correlation between various parameters and intelligently distinguish between harmless normal fluctuations under the synergy of multiple parameters and abnormal change patterns that truly indicate the risk of instability. This avoids false alarms caused by a single parameter exceeding the limit and can more realistically reflect the overall health status of the pigment dispersion system.

[0069] This invention uses a micro-heating and piezoelectric vibration unit to simulate the environmental pressures such as temperature fluctuations and mechanical shearing that pigments may encounter during actual production and transportation. By accurately comparing the changes in multi-dimensional parameters before and after the disturbance, and quantitatively comparing them with standard change parameters derived from best production practices, a comprehensive stability index is obtained. This effectively identifies batches with insufficient stability before materials are put into the production line, eliminating production anomalies caused by pigment agglomeration, sedimentation, and other problems at the source, significantly improving production stability and yield.

[0070] Preferably, the process for obtaining the process characteristic parameters includes:

[0071] During the sample flow process, an acoustic detection device emits ultrasonic waves into the sample flowing in the sample delivery pipe and acquires the reflected echoes caused by the movement of sample particles. The measurement signal is obtained by combining the emitted ultrasonic waves and the reflected echoes.

[0072] The Doppler frequency shift of the reflected echo in the real-time analysis and measurement signal is used to calculate the velocity distribution of sample particles in the cross-section of the conveying pipeline, which serves as a process characteristic parameter; the acoustic detection device includes an ultrasonic detection device.

[0073] An acoustic detection device emits a beam of ultrasonic waves of a known frequency into the pipe. When the ultrasonic waves encounter moving pigment particles, they are reflected. Due to the Doppler effect, the frequency of the reflected echo changes (frequency shift). The magnitude of the frequency shift is proportional to the velocity of the particles. By measuring the magnitude of the frequency shift at different locations, the system can calculate the precise velocity distribution of the sample particles throughout the entire pipe cross-section in real time and use it as a process characteristic parameter.

[0074] By utilizing the ultrasonic Doppler effect, the flow velocity at various points on the cross-section of the pipe is measured in real time in a non-invasive manner, and the results are compiled into a complete flow velocity distribution, thereby dynamically monitoring the flow behavior of the pigment.

[0075] The velocity distribution describes the magnitude of the pigment flow velocity at various locations from the pipe wall to the center in the cross-section of the pipe; under ideal laminar flow conditions, this distribution should be parabolic; if the pigment agglomerates to form large particles, or if the viscosity changes, the velocity distribution profile will deviate from the ideal shape.

[0076] This invention uses an acoustic detection device to perform ultrasonic testing on the sample flowing inside the sample delivery pipe, non-invasively measuring the flow velocity at each point on the pipe cross-section in real time, thereby dynamically monitoring the flow behavior of the pigment; it can calculate the precise flow velocity distribution of sample particles throughout the entire pipe cross-section in real time.

[0077] Preferably, the identification process of the process state analysis model includes:

[0078] Obtain the sample state parameters output by the characteristic characterization unit; based on the sample state parameters, establish a flow velocity distribution benchmark model for the current batch of samples;

[0079] The flow velocity distribution in the process characteristic parameters is continuously compared and analyzed with the flow velocity distribution benchmark model to identify the degree of deviation of the real-time flow velocity distribution from the flow velocity distribution benchmark model, and the process deviation coefficient is quantified.

[0080] The process offset coefficient is compared with a preset process offset threshold. When the process offset coefficient is greater than the process offset threshold, compensation control for the sample at the end of the delivery is triggered.

[0081] A theoretically ideal flow model is established for the current batch of pigments and compared with the flow state measured in real time. If the deviation exceeds the allowable range, compensation control is immediately implemented.

[0082] Model building: When a new batch of pigments begins to be delivered, the system will establish a dedicated flow velocity distribution benchmark model for the specific sample based on its sample state parameters, such as viscosity and particle size distribution, using fluid dynamics algorithms. This model represents the most ideal and efficient flow velocity state of the pigment under the current pipeline and delivery settings.

[0083] Continuous Comparison: The system continuously compares the velocity distribution measured in real time by the dynamic analysis unit with a pre-established benchmark model. Through algorithms such as integral calculations or vector comparisons, the difference between the two flow data points is quantified into a single numerical process offset coefficient.

[0084] Trigger control: The coefficient is compared with a preset process offset threshold; if the coefficient is less than the threshold, the flow condition is normal. Once it exceeds the threshold, it means that the flow condition has deteriorated significantly, such as the appearance of signs of blockage. The system will immediately trigger compensatory control on the end-of-line equipment, such as adjusting the pump speed or pressure, to correct the flow condition.

[0085] The compensation control for the sample at the end of the delivery process is an adaptive hybrid control method based on a combination of fuzzy logic and PID (proportional-integral-derivative) logic.

[0086] Fuzzy logic controller: Taking the process offset coefficient and its rate of change as input, it quickly outputs a basic adjustment amount of pump speed or pressure according to the preset fuzzy rule base to deal with nonlinear flow state deviations.

[0087] PID controller: With the process offset coefficient as input and a target value of zero, it is used to finely adjust the residual steady-state error after fuzzy control, ensuring that the flow state can accurately and smoothly return to the flow velocity distribution reference model.

[0088] Adaptive adjustment: The membership function of the fuzzy logic controller and the parameters of the PID controller can learn and fine-tune online based on the historical best control data output by the dynamic benchmark optimization unit to adapt to the rheological characteristics of different batches of pigments.

[0089] This invention achieves a balance of speed, accuracy, and robustness in compensating for pigment flow state through complementary advantages;

[0090] Fuzzy logic controllers, relying on empirical rules, can react quickly and coarsely to large, nonlinear deviations in flow state, preventing significant system fluctuations. Subsequently, a PID controller intervenes to fine-tune the small residual errors after fuzzy control, ensuring that the flow state smoothly and accurately returns to the baseline model without overshoot. This overcomes the limitations of a single controller and improves the system's dynamic performance and stability.

[0091] It possesses strong adaptability and generalization capabilities: Due to the differences in the viscosity and other rheological properties of different batches of pigments, controllers with fixed parameters are difficult to apply universally; this solution, through linkage with the dynamic benchmark optimization unit, uses historical optimal data to perform self-learning and online fine-tuning of the controller parameters, enabling it to autonomously adapt to the characteristics of the current material; it enhances the robustness of the system, ensuring that it can always maintain the optimal control effect when facing different batches of pigments, and guaranteeing a high degree of consistency in the production process.

[0092] This invention refines the internal logic of the process state analysis model and establishes a batch-specific flow velocity distribution benchmark model. It continuously quantifies complex flow velocity distribution differences into a single, intuitive process deviation coefficient, realizing standardized and real-time assessment of the degree of process deviation, thereby triggering compensation control. It can autonomously and quickly correct the deterioration trend of the flow state, eliminate potential process fluctuations in their infancy, and ensure a high degree of consistency and stability in the production process.

[0093] Preferably, processing anomalies are identified by using a pigment processing anomaly model to determine operating status parameters, quality characteristic parameters, and process transport parameters. Specifically, this includes:

[0094] The pigment processing anomaly model performs time alignment processing on the operating status parameters, quality characteristic parameters, and process transport parameters to obtain pigment processing time series data.

[0095] By using a long short-term memory network algorithm, we learn the dynamic correlation patterns between pigment processing time-series data and processing anomalies that occurred in historical production.

[0096] In real-time production, the probability of a process anomaly is identified based on the dynamic correlation pattern. When the probability of an anomaly exceeds a preset process anomaly threshold, it is determined to be a process anomaly, and an early warning control is executed.

[0097] By using the Long Short-Term Memory (LSTM) network algorithm, we can learn the complex correlation between the time series changes of various process parameters in massive historical production data and the final process anomalies, thereby enabling prediction before anomalies occur.

[0098] The core of the pigment processing anomaly model is a Long Short-Term Memory (LSTM) network, a type of recurrent neural network suitable for processing and predicting time-series data. The input layer of the model receives a multi-dimensional data stream after time alignment, including operating status parameters of the production equipment, quality characteristic parameters read by the optical inspection device, and process transport parameters. The output layer of the model outputs the risk probability of one or more process anomalies.

[0099] Training process: The pigment processing anomaly model is trained using a large amount of historical production data. The training data is organized into "input sequence-output label" pairs; for example, the time series of all relevant parameters within a period of time before a production accident (such as "severe color difference") occurs is used as input, and "severe color difference" is used as the label. By repeatedly learning from this data, the LSTM network can capture the parameter evolution patterns and long-term dependencies that lead to specific anomalies.

[0100] The quality characteristic parameters include color difference before baking and color difference after baking obtained by optical detection equipment; furthermore, the quality characteristic parameters may also include acoustic fastness factor, dynamic flexibility index, and Raman crosslinking characteristic peak shift in the final quality parameters.

[0101] This invention elevates the system's anomaly diagnosis capabilities from traditional threshold-based simple judgments to intelligent prediction and risk assessment based on deep learning. Through training on massive amounts of historical data, the model can deeply understand the complex causal chain from process parameters to final quality anomalies. In real-time production, it can comprehensively analyze the dynamic changes of all relevant parameters, identify subtle signs that may indicate future failures, and provide accurate early warnings in the form of risk probabilities.

[0102] Preferably, the method for obtaining the standard variation parameter is as follows:

[0103] The system also includes a final quality acquisition unit and a dynamic benchmark optimization unit; the final quality acquisition unit is used to acquire the final quality parameters of the printed fabric after baking and curing; the dynamic benchmark optimization unit deeply correlates historical process data with the corresponding final quality parameters, reversely optimizes and generates a dynamic process benchmark to ensure the best finished product quality; the dynamic process benchmark includes standard variation parameters.

[0104] The dynamic baseline optimization unit employs a generative adversarial network (GAN) for backward optimization:

[0105] Generator: Its input is a vector of target final quality parameters, and its output is a set of dynamic process benchmarks that predict the achievement of that quality, including standard variation parameters.

[0106] Discriminator: Its input is a pair of real process benchmarks and final quality parameters from historical production, used to learn the real physical relationship between the two.

[0107] The optimization process involves the generator continuously generating new process baselines and combining them with a forward prediction model built on a long short-term memory network. The forward prediction model can predict the final quality based on the input process baselines. The discriminator then judges whether the "process-quality" output by the generator is close to the actual production data. Through adversarial training, the generator can eventually accurately reverse-engineer a set of optimal dynamic process baselines for any set optimal quality target.

[0108] This invention employs generative adversarial networks (GANs) for reverse optimization, achieving a leap from empirical fitting to deep learning in process parameter optimization, resulting in higher accuracy, efficiency, and innovation. Compared to methods such as response surface methodology that rely on pre-defined models, GANs can autonomously learn extremely complex, high-dimensional nonlinear relationships from massive amounts of historical data; they also gain a deeper understanding of the true physical mapping relationship between process parameters and final quality parameters, thereby generating more accurate and reliable optimal process benchmarks.

[0109] This invention collects the final quality parameters and performs in-depth correlation analysis between the final quality parameters and massive process data in the production process; then, through the dynamic benchmark optimization unit, it reverse-calculates the ideal process parameters and dynamic benchmarks that should be followed to achieve the best quality, including standard variation parameters for stability testing; so that the system's control target is no longer a rigid fixed value, but an optimal solution that can continuously iterate and dynamically optimize based on historical data and quality feedback, leading the production process to continuously and automatically evolve towards higher quality and higher efficiency.

[0110] Preferably, the final quality parameter is a multidimensional quality vector, including: acoustic fastness factor, dynamic flexibility index, and Raman crosslinking characteristic peak shift;

[0111] The acoustic fastness factor is used to characterize color fastness and is quantified by applying standardized micromechanical perturbations to the cured film layer and analyzing the spectral characteristics of the acoustic response signal. It means that a strong and well-adhesive film layer will exhibit a specific pattern in its acoustic response spectrum, based on which the fastness factor can be calculated.

[0112] By exciting the cured film layer through micromechanical perturbation (such as sound waves or vibration) and analyzing its acoustic response (such as resonant frequency, damping coefficient, etc.), the physical / mechanical properties of the film layer, such as adhesion, hardness, Young's modulus and integrity, can be effectively evaluated; thus effectively characterizing the fastness related to physical wear resistance.

[0113] The dynamic flexibility index is used to characterize the softness of the fabric. It is quantified by measuring the rate of change of the time autocorrelation function of the laser speckle pattern on the surface when micro-vibration is applied to the fabric. Soft fabrics have more free movement of surface fibers under vibration, resulting in faster speckle changes and thus a higher flexibility index.

[0114] The Raman crosslinking characteristic peak shift is used to characterize the quality of the cured film. It is obtained by monitoring the Raman characteristic peaks of the adhesive chemical bonds in the cured film using an online Raman spectrometer and calculating the shift of the standard peak position relative to the uncrosslinked state. The shift is directly related to the degree of crosslinking of the cured film. When the adhesive undergoes crosslinking and curing, its chemical bond state changes, causing the Raman peak position to shift compared to the uncrosslinked state. The magnitude of the shift is directly related to the degree of crosslinking of the cured film, i.e., the quality of curing.

[0115] This invention further defines a multidimensional quality vector consisting of acoustic fastness factor, dynamic flexibility index, and Raman crosslinking characteristic peak shift as the final quality parameter; it provides a comprehensive and in-depth analysis of product quality from three key dimensions: physical and mechanical properties, fabric sensory properties, and chemical curing degree, realizing real-time, quantitative, and non-destructive online characterization of the core quality indicators of printed fabrics; and it provides a data foundation for the dynamic benchmark optimization unit to perform self-learning and reverse optimization.

[0116] Example 2:

[0117] This invention proposes a pigment dispersion monitoring and control system for waterless printed fabrics, including a characteristic characterization unit, a dynamic analysis unit, and a quality detection unit.

[0118] The characterization unit identifies the dispersion characteristics, stability characteristics, and rheological characteristics of the sample to obtain the initial parameters of the sample; applies a preset perturbation signal to the sample and collects the sample perturbation parameters; calculates the stability index based on the initial parameters and sample perturbation parameters, performs baseline calibration, and outputs the calibrated sample state parameters; the sample is a pigment.

[0119] The dispersion characteristics include the average particle size, polydispersity index, and particle size distribution curve of the sample particles, which are obtained by scanning the upper, middle, and lower layers of the sample in the storage space. The stability characteristics include the electrostatic repulsion between sample particles. The rheological characteristics include the sample viscosity.

[0120] The specific process of applying a preset disturbance signal and calculating the stability index includes:

[0121] The sample is subjected to standardized temperature changes and high-frequency vibrations via a miniature heating and piezoelectric vibration unit integrated within the sensor. The specific standardized parameters are as follows: Temperature change: The temperature is increased from room temperature (e.g., 25°C) to 45°C at a rate of 5°C / minute, and then held at this temperature for 120 seconds. High-frequency vibration: During the isothermal holding phase, ultrasonic vibrations with a frequency of 20 kHz and an amplitude of 5 micrometers are simultaneously applied for 60 seconds.

[0122] The dispersion characteristics, stability characteristics, and rheological characteristics after disturbance are identified to obtain the sample disturbance parameters;

[0123] Based on the changes in the sample perturbation parameters and the initial sample parameters, the perturbation change parameters are obtained, specifically including the average particle size difference, polydispersity index difference, particle size distribution curve shape difference, electrostatic repulsion difference, and sample viscosity difference.

[0124] Based on the parameter changes before and after the perturbation, the stability index is determined using a weighted Euclidean distance formula; this index aims to quantify the degree of deviation of the overall state of the sample under pressure; when the stability index exceeds the stability threshold, baseline calibration is initiated.

[0125] The dynamic analysis unit measures the sample flow process using an acoustic detection device, analyzes the dynamic changes of the measurement signal, and obtains the process characteristic parameters of the sample particles during the flow process; it constructs a process state analysis model, identifies the sample state parameters and process characteristic parameters, and obtains the process offset coefficient; based on the process offset coefficient, it performs state compensation on the sample and obtains the compensated process transport parameters.

[0126] The process of obtaining the process characteristic parameters includes:

[0127] During the sample flow process, an acoustic detection device emits ultrasonic waves into the sample flowing in the sample delivery pipe and acquires the reflected echoes caused by the movement of sample particles. The measurement signal is obtained by combining the emitted ultrasonic waves and the reflected echoes.

[0128] The Doppler frequency shift of the reflected echo in the real-time analysis and measurement signal is used to calculate the velocity distribution of sample particles in the cross-section of the conveying pipeline, which serves as a process characteristic parameter; the acoustic detection device includes an ultrasonic detection device.

[0129] The identification process of the process state analysis model includes:

[0130] Obtain the sample state parameters output by the characteristic characterization unit; based on the sample state parameters, establish a flow velocity distribution benchmark model for the current batch of samples;

[0131] The flow velocity distribution in the process characteristic parameters is continuously compared and analyzed with the flow velocity distribution benchmark model to identify the degree of deviation of the real-time flow velocity distribution from the flow velocity distribution benchmark model, and the process deviation coefficient is quantified.

[0132] The process offset coefficient is compared with a preset process offset threshold. When the process offset coefficient is greater than the process offset threshold, compensation control for the sample at the end of the delivery is triggered.

[0133] Considering that pigment slurries are generally non-Newtonian fluids, this embodiment uses a power-law fluid model to establish a baseline model for velocity distribution; this model can more accurately describe their shear-thinning characteristics.

[0134] First, the consistency coefficient and flow behavior index of the power-law model are obtained by fitting the initial rheological parameters of the sample measured by the characteristic characterization unit. Then, based on the consistency coefficient and flow behavior index, a baseline model of the flow velocity distribution of the current batch of samples under a specific pipe diameter and pressure difference is established. By comparing the real-time measured flow velocity distribution with the above-mentioned flow velocity baseline model, the process offset coefficient is quantified using the normalized root mean square error algorithm.

[0135] The quality inspection unit acquires the operating status parameters of the production equipment and reads the quality characteristic parameters of the processed object based on the optical inspection device; it identifies process abnormalities based on the operating status parameters, quality characteristic parameters, and process conveying parameters, and performs early warning control.

[0136] In this embodiment, the pigment processing anomaly model employs a network structure containing two stacked LSTM layers, each with 128 neurons, followed by a fully connected layer with 64 neurons and a sigmoid activation function layer for output probability. The model is trained using the Adam optimizer with a learning rate of 0.001 and a binary cross-entropy loss function. Training data is sampled every second in 30-minute time windows.

[0137] The system was trained using a sample set of historical failure data, with the Adam optimizer employed. The initial learning rate was set to 0.001. A binary cross-entropy loss function was used. The batch size was set to 64.

[0138] Input data: The model input is multidimensional time series data after time alignment; the series data contains 12 key features, the sampling frequency is 1Hz, and the time window is 600 seconds.

[0139] Key features include: Process delivery parameters: pump speed, pipeline pressure, process offset coefficient. Operating status parameters: oven temperature, fabric conveying speed, printing nozzle pressure, ambient humidity, ambient temperature. Quality characteristic parameters: online measured acoustic fastness factor, dynamic flexibility index, Raman crosslinking characteristic peak shift, and a color difference value acquired by an optical camera.

[0140] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A pigment dispersion monitoring and control system for waterless printed fabrics, characterized in that, include: The characterization unit identifies the dispersion characteristics, stability characteristics, and rheological characteristics of the sample, and obtains the initial parameters of the sample. A preset perturbation signal is applied to the sample, and the sample perturbation parameters are collected; The stability index is calculated based on the initial parameters and perturbation parameters of the sample, baseline calibration is performed, and the calibrated sample state parameters are output. The sample is a pigment; the sample state parameters include viscosity and particle size distribution. The dispersion characteristics include the average particle size, polydispersity index, and particle size distribution curve of the sample particles, which are obtained by scanning the upper, middle, and lower layers of the sample in the storage space. The stability characteristics include the electrostatic repulsion between sample particles. The rheological characteristics include the sample viscosity. The specific process for calculating the stability index includes: By applying standardized temperature changes and high-frequency vibrations to the sample through a miniature heating and piezoelectric vibration unit within an integrated sensor, the dispersion characteristics, stability characteristics, and rheological characteristics after disturbance are identified, and the sample disturbance parameters are obtained. Based on the changes in the sample perturbation parameters and the initial sample parameters, the perturbation change parameters are obtained, specifically including the average particle size difference, polydispersity index difference, particle size distribution curve shape difference, electrostatic repulsion difference, and sample viscosity difference. The stability index is determined based on the degree of difference between the perturbation variation parameters and the standard variation parameters; when the stability index exceeds the stability threshold, baseline calibration is initiated. The dynamic analysis unit measures the sample flow process using an acoustic detection device, analyzes the dynamic changes of the measurement signal, and obtains the process characteristic parameters of the sample particles during the flow process; it constructs a process state analysis model to identify the sample state parameters and process characteristic parameters, and obtains the process offset coefficient; based on the process offset coefficient, it performs state compensation on the sample to obtain the compensated process transport parameters. The process of obtaining the process characteristic parameters includes: during the sample flow process, an acoustic detection device emits ultrasonic waves into the sample flowing in the sample delivery pipe and acquires the reflected echoes caused by the movement of sample particles; a measurement signal is obtained by combining the emitted ultrasonic waves and the reflected echoes; the Doppler frequency shift of the reflected echoes in the measurement signal is analyzed in real time, and the flow velocity distribution of sample particles in the cross-section of the delivery pipe is calculated as the process characteristic parameters; the acoustic detection device includes an ultrasonic detection device. The process delivery parameters include pump speed, pipeline pressure, and process offset coefficient; The identification process of the process state analysis model includes: Obtain the sample state parameters output by the characteristic characterization unit; based on the sample state parameters, establish a flow velocity distribution benchmark model for the current batch of samples; The flow velocity distribution in the process characteristic parameters is continuously compared and analyzed with the flow velocity distribution benchmark model to identify the degree of deviation of the real-time flow velocity distribution from the flow velocity distribution benchmark model, and the process deviation coefficient is quantified. The process offset coefficient is compared with a preset process offset threshold. When the process offset coefficient is greater than the process offset threshold, compensation control for the sample at the end of the delivery is triggered. The state compensation is based on an adaptive hybrid control that combines fuzzy logic and PID. The quality inspection unit acquires the operating status parameters of the production equipment and reads the quality characteristic parameters of the processed object based on the optical inspection device; it identifies process abnormalities based on the operating status parameters, quality characteristic parameters, and process conveying parameters, and performs early warning control. The quality characteristic parameters include an online measured acoustic fastness factor, a dynamic flexibility index, a Raman crosslinking characteristic peak shift, and a color difference value acquired by an optical camera; the acoustic fastness factor is used to characterize color fastness, the dynamic flexibility index is used to characterize the softness of the fabric, and the Raman crosslinking characteristic peak shift is used to characterize the quality of the cured film layer.

2. The pigment dispersion monitoring and control system for waterless printed fabrics according to claim 1, characterized in that, By using a pigment processing anomaly model to identify operational status parameters, quality characteristic parameters, and process transport parameters, processing anomalies are determined. Specifically, this includes: The pigment processing anomaly model performs time alignment processing on the operating status parameters, quality characteristic parameters, and process transport parameters to obtain pigment processing time series data. By using a long short-term memory network algorithm, we learn the dynamic correlation patterns between pigment processing time-series data and processing anomalies that occurred in historical production. In real-time production, the probability of a process anomaly is identified based on the dynamic correlation pattern. When the probability of an anomaly exceeds a preset process anomaly threshold, it is determined to be a process anomaly, and an early warning control is executed.

3. The pigment dispersion monitoring and control system for waterless printed fabrics according to claim 1, characterized in that, The method for obtaining the standard variation parameters is as follows: The system also includes a final quality acquisition unit and a dynamic benchmark optimization unit; The final quality acquisition unit is used to collect the final quality parameters of the printed fabric after baking and curing. The dynamic benchmark optimization unit deeply correlates historical process data with corresponding final quality parameters, reverse-optimizes and generates a dynamic process benchmark that ensures the best finished product quality; the dynamic process benchmark includes standard variation parameters.

4. The pigment dispersion monitoring and control system for waterless printed fabrics according to claim 3, characterized in that: The final quality parameters are multidimensional quality vectors, including: acoustic fastness factor, dynamic flexibility index, and Raman crosslinking characteristic peak shift; The acoustic fastness factor is quantified by applying standardized micromechanical perturbations to the cured film layer and analyzing the spectral characteristics of the acoustic response signal. The dynamic flexibility index is quantified by measuring the rate of change of the time autocorrelation function of the laser speckle pattern on the surface when micro-vibration is applied to the fabric. The shift of the Raman crosslinking characteristic peak is quantified by monitoring the Raman characteristic peaks of the adhesive chemical bonds in the cured film layer using an online Raman spectrometer and calculating the shift relative to the standard peak position in the uncrosslinked state. The shift is directly related to the degree of crosslinking of the cured film layer.

Images