Search

CN-121095851-B - Lining cloth dyeing control method and system based on artificial intelligence

CN121095851BCN 121095851 BCN121095851 BCN 121095851BCN-121095851-B

Abstract

The invention discloses an artificial intelligence-based lining cloth dyeing control method and system, which are used for the control field, wherein the method comprises the following steps of monitoring multi-source data in a lining cloth dyeing vat in real time and capturing video streams in the lining cloth dyeing vat; the method comprises the steps of carrying out feature extraction on multi-source data by utilizing a Gaussian mixture variation self-encoder model to obtain multi-source feature data, obtaining variation time of fabric states in a video stream, extracting fabric state variation feature data according to fabric color variation and heterogeneity of dye distribution observed in the video stream, fusing the multi-source feature data and the fabric state variation feature data, introducing a model prediction control layer, adjusting dyeing parameters in real time according to a prediction result, and establishing an intelligent feedback mechanism. The invention trains and optimizes the Gaussian mixture variation self-encoder model through a small batch random gradient descent and back propagation algorithm, and learns and captures important characteristics from data.

Inventors

  • YANG CHENGUANG

Assignees

  • 南通摩瑞纺织有限公司

Dates

Publication Date
20260508
Application Date
20251107

Claims (6)

  1. 1. The lining cloth dyeing control method based on artificial intelligence is characterized by comprising the following steps of: s1, monitoring multi-source data in an lining cloth dyeing vat in real time, and capturing video streams in the lining cloth dyeing vat at the same time; S2, performing feature extraction on the multi-source data by utilizing a Gaussian mixture variation self-encoder model to obtain multi-source feature data; s3, acquiring change time of fabric state in the video stream, and extracting fabric state change characteristic data according to fabric color change and dye distribution heterogeneity observed in the video stream; S4, merging the multi-source characteristic data and the fabric state change characteristic data, creating a comprehensive data set, and analyzing the comprehensive data set by utilizing a depth deterministic strategy gradient algorithm and combining with an adaptive control theory to obtain an analysis result; s5, introducing a model prediction control layer, predicting the dyeing process state of the lining cloth based on the analysis result, and adjusting the dyeing parameters in real time according to the prediction result; s6, establishing an intelligent feedback mechanism, collecting deviation between the dyeing effect of the lining cloth and an expected target, feeding back in real time, and adjusting parameters of the lining cloth dyeing process according to the feedback result; the method for extracting the characteristics of the multi-source data by utilizing the Gaussian mixture variation self-encoder model comprises the following steps of: S21, constructing a Gaussian mixture variation self-encoder model, and setting network layers and activation functions of an encoder and a decoder of the Gaussian mixture variation self-encoder model; S22, training a Gaussian mixture variation self-encoder model by utilizing small batch random gradient descent, and calculating reconstruction errors and KL divergence; s23, mapping the multi-source data to a potential space by using an encoder; S24, classifying the mapped multi-source data into a set component according to Gaussian mixture distribution of the potential space to obtain multi-source characteristic data; the method for acquiring the change moment of the fabric state in the video stream and extracting the characteristic data of the fabric state change according to the observed fabric color change and the heterogeneity of dye distribution in the video stream comprises the following steps: S31, analyzing the captured video stream, and identifying the change of the color of the fabric and the distribution pattern of the dye on the fabric; s32, evaluating the degree of color change of the fabric and the heterogeneity of dye distribution by comparing the image data in the video stream with a standard color template; S33, according to the degree of the color change of the fabric and the heterogeneity of dye distribution, adjusting the feature extraction of the Gaussian mixture variation from the encoder model; s34, identifying and recording the time of the change of the fabric color and the dye distribution state observed in the video stream; s35, integrating the fabric color and dye distribution information extracted from the video stream into multi-source characteristic data; the fusion of the multi-source signature data and the fabric state change signature data creates a comprehensive dataset, and analyzing the comprehensive data set by combining a depth deterministic strategy gradient algorithm with an adaptive control theory, and acquiring an analysis result comprises the following steps: S41, fusing multi-source characteristic data and fabric state change characteristic data to create a comprehensive data set; S42, analyzing the comprehensive data set by using a depth deterministic strategy gradient algorithm; S43, combining the output of the depth deterministic strategy gradient algorithm with the self-adaptive control theory, and adjusting and optimizing the dyeing process; s44, adjusting dyeing parameters based on analysis results of a depth deterministic strategy gradient algorithm and an adaptive control theory; the analysis of the integrated data set using the depth deterministic strategy gradient algorithm comprises the steps of: s421, initializing a behavior network and a reviewer network; Wherein the actor network is used for predicting the optimal action under a given state; the reviewer network is operable to evaluate expected rewards for a given combination of states and actions; S422, preprocessing a comprehensive data set combining the multi-source characteristic data and the fabric state change moment, and dividing the comprehensive data set into a training set and a verification set; s423, training a depth deterministic strategy gradient algorithm model by using training set data; S424, the actor network predicts actions according to the current state, executes the predicted actions, and observes the results and rewards corresponding to the results; S425, the reviewer network evaluates the expected return of the action predicted by the actor network and updates the actor network by using feedback of the reviewer network; S426, storing the state, action, rewards and new state of each step in a playback buffer zone by adopting an experience playback mechanism, and periodically updating a actor network and a reviewer network; s427, extracting final dyeing parameters from the actor network, and adjusting the optimal dyeing parameters under a given dyeing process state.
  2. 2. The method for controlling the dyeing of the lining cloth based on the artificial intelligence according to claim 1, wherein the real-time monitoring of the multi-source data in the lining cloth dyeing vat and the capturing of the video stream in the lining cloth dyeing vat simultaneously comprise the following steps: s11, installing sensors at all positions of the lining cloth dyeing vat, and monitoring multi-source data in the lining cloth dyeing vat in real time; s12, installing a high-definition camera to align with the inside of the lining dyeing vat, and capturing video stream of fabric inside the lining dyeing vat; s13, synchronizing the time stamps of the multi-source data and the video stream, ensuring the consistency of data analysis, cleaning the data, and removing the erroneous or invalid data.
  3. 3. The artificial intelligence based lining cloth dyeing control method according to claim 1, wherein training the gaussian mixture variation self-encoder model by using small batch random gradient descent and calculating the reconstruction error and KL divergence comprises the following steps: S221, setting initial values for parameters of a Gaussian mixture variation self-encoder model; s222, randomly selecting partial data from the multi-source data as batch data; s223, inputting batch data into a Gaussian mixture variation self-encoder model, acquiring parameters of a potential space through an encoder, and reconstructing the input data through a decoder to obtain reconstructed data and potential representation; S224, calculating a reconstruction error between output and input data of the Gaussian mixture variation self-encoder model by using a loss function according to the reconstruction data and the potential representation; S225, calculating KL divergence of the potential space parameters and the Gaussian mixture variation self-encoder model; S226, adding the reconstruction error and the KL divergence to obtain the total loss of the model in the current batch data; s227, calculating the gradient of the Gaussian mixture variation self-encoder model parameters based on the total loss, and updating the parameters of the Gaussian mixture variation self-encoder model through a back propagation algorithm.
  4. 4. A method of artificial intelligence based lining cloth dyeing control according to claim 3, characterized in that the calculating of the KL-divergence of the latent spatial parameters and gaussian mixture variation self-encoder model comprises the steps of: S2251, obtaining the mean and variance of the potential spatial representation of each data point in the multi-source data from the encoder output; S2252, setting the mean value, variance and mixing coefficient of each Gaussian distribution component of the Gaussian mixture model; S2253, for potential representation of data points, calculating KL divergence between posterior distribution of data points and prior distribution of Gaussian mixture variation self-encoder model; the calculation formula of the KL divergence is as follows: ; in the formula, Representing the mean; Representing the variance; Representing the mean value of the jth gaussian distribution component; representing the variance of the jth gaussian distribution component; a weight representing the jth gaussian distribution component; Representing posterior distribution With a priori distribution KL divergence between; Indicating the amount of the components in the gaussian mixture model.
  5. 5. The method for dyeing control of a lining cloth based on artificial intelligence according to claim 1, wherein the introducing a model predictive control layer predicts a state of a dyeing process of the lining cloth based on an analysis result, and adjusts a dyeing parameter in real time according to the prediction result comprises the steps of: s51, constructing an introduction model prediction control layer; S52, inputting dyeing parameters and parameters of the current dyeing process into a model prediction control layer, and predicting the state change of the lining cloth dyeing process in one hour in the future; s53, analyzing the influence of different dyeing parameters according to the prediction result, and selecting the optimal dyeing parameters and parameters of the dyeing process.
  6. 6. An artificial intelligence-based lining cloth dyeing control system for realizing the artificial intelligence-based lining cloth dyeing control method according to any one of claims 1 to 5, which is characterized by comprising a data collection and monitoring module, a feature extraction module, a video stream monitoring module, a feature analysis module, a control layer prediction module and an intelligent feedback module; The data collection and monitoring module is connected with the video stream monitoring module through the feature extraction module, the video stream monitoring module is connected with the control layer prediction module through the feature analysis module, and the control layer prediction module is connected with the intelligent feedback module; The data collection and monitoring module is used for monitoring multi-source data in the lining cloth dyeing vat in real time and capturing video streams in the lining cloth dyeing vat at the same time; The characteristic extraction module is used for carrying out characteristic extraction on the multi-source data by utilizing the Gaussian mixture variation self-encoder model to obtain multi-source characteristic data; The video stream monitoring module is used for acquiring the change moment of the fabric state in the video stream and extracting the characteristic data of the fabric state change according to the observed fabric color change and the heterogeneity of dye distribution in the video stream; The characteristic analysis module is used for fusing the multi-source characteristic data and the fabric state change characteristic data, creating a comprehensive data set, analyzing the comprehensive data set by utilizing a depth deterministic strategy gradient algorithm in combination with the self-adaptive control theory, and obtaining an analysis result; the control layer prediction module is used for introducing a model prediction control layer, predicting the state of the lining cloth dyeing process based on the analysis result, and adjusting the dyeing parameters in real time according to the prediction result; The intelligent feedback module is used for establishing an intelligent feedback mechanism, collecting deviation between the lining cloth dyeing effect and an expected target, feeding back in real time, and adjusting parameters of the lining cloth dyeing process according to the feedback result.

Description

Lining cloth dyeing control method and system based on artificial intelligence Technical Field The invention relates to the field of control, in particular to an artificial intelligence-based lining cloth dyeing control method and system. Background Lining dyeing is a textile processing process for dyeing a lining (typically a cloth used to make shirts or other garments), and lining dyeing control is a complex process involving precise adjustment and monitoring of multiple variables to ensure high quality dyeing results. This includes temperature control, pH adjustment, accurate dosing of dyes and chemicals, time management, and the like. The temperature during dyeing has a significant effect on the dye absorption rate. Different types of dyes and fibres show the best dyeing effect at different temperatures. Therefore, accurate control of the heating and cooling processes is critical to achieving the desired color and color fastness, the pH value of the dyeing solution directly influences the affinity of the dye to the fibers, and the adsorption rate and uniformity of the dye on the fibers can be controlled by accurately adjusting the pH value, so that the final dyeing quality is influenced, and the type, concentration and addition mode of the dye influence the uniformity and effect of dyeing. The use of the auxiliary agent can improve dyeing effect, such as improving color fastness, improving cloth hand feeling and the like, and the correct selection and addition of the dye and the auxiliary agent are important for achieving ideal dyeing effect, and the dyeing time is important for achieving color depth and uniformity. The different dyeing stages (e.g. pretreatment, dyeing and post-treatment) require precise control times to ensure optimal results for each stage. The traditional dyeing process monitoring may not fully cover all important physical and chemical parameters, so that the overall view of the dyeing process cannot be accurately mastered, the dyeing effect and the stability of the product quality are affected, an effective mechanism is lacked to synchronize information of different data sources, the data quality is ensured, the basis of data analysis is unstable, the accuracy and the reliability of decision making are further affected, when the traditional control method faces to a complex and changeable dyeing process, accurate prediction and effective adjustment are difficult to be made due to insufficient learning and generalization capability of a model, an effective real-time monitoring means and a quick response mechanism are lacked, when the heterogeneous fabric color change and dye distribution are faced, the dyeing continuity and consistency are affected, and in the traditional dyeing process control, the comprehensive utilization of multi-source data and fabric state change moments is insufficient, so that the comprehensive understanding and accurate control capability of the dyeing process is affected. For the problems in the related art, no effective solution has been proposed at present. Disclosure of Invention In order to overcome the problems, the invention aims to provide an artificial intelligence-based lining cloth dyeing control method and system, and aims to solve the problems that the dyeing continuity and consistency are affected due to the fact that an effective real-time monitoring means and a quick response mechanism are lacking, the dyeing continuity and the dyeing consistency cannot be timely adjusted when the fabric is subjected to color change and dye distribution heterogeneity are faced, and the comprehensive utilization of multi-source data and fabric state change time is insufficient in the traditional dyeing process control, so that the comprehensive understanding and accurate control capability of a dyeing process is affected. For this purpose, the invention adopts the following specific technical scheme: According to an aspect of the present invention, there is provided an artificial intelligence-based lining cloth dyeing control method including the steps of: s1, monitoring multi-source data in an lining cloth dyeing vat in real time, and capturing video streams in the lining cloth dyeing vat at the same time; S2, performing feature extraction on the multi-source data by utilizing a Gaussian mixture variation self-encoder model to obtain multi-source feature data; s3, acquiring change time of fabric state in the video stream, and extracting fabric state change characteristic data according to fabric color change and dye distribution heterogeneity observed in the video stream; S4, merging the multi-source characteristic data and the fabric state change characteristic data, creating a comprehensive data set, and analyzing the comprehensive data set by utilizing a depth deterministic strategy gradient algorithm and combining with an adaptive control theory to obtain an analysis result; s5, introducing a model prediction control layer, predicting the dyeing process