A Melting Monitoring System and Method for Solid Dyes

Abstract

An embodiment of the present application provides a melting monitoring system and method for solid dyes. Specifically, it includes setting a light source and a light sensor to generate an initial optical path, recording and processing light intensity values to form a time-light intensity curve, using a support vector machine model to identify the melting starting point, calculating the melting rate through a Kalman filter and generating a quality assessment report. When the melting rate reaches a preset threshold, a PID controller triggers a cooling program, and analyzes the uniformity and purity of the dye by combining deep learning techniques, and finally generates a melting monitoring report. The technical solution provided by the embodiment of the present application improves the product quality and production efficiency of solid dyes.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of automatic control, and particularly to a melting monitoring system and method for solid dyes. Background Art

[0002] In many industrial fields such as textiles, plastics, and coatings, the melting process of solid dyes is a key step to ensure the quality of the final product. Accurately monitoring the melting state of solid dyes is crucial for controlling the production process, optimizing process parameters, and ensuring product consistency.

[0003] Currently, there have been some technologies attempting to solve the problem of melting monitoring of solid dyes. For example, optical sensors are used to record the change in light intensity passing through the dye sample to form a time-light intensity curve; or temperature sensors are used to monitor the temperature change in the heating container.

[0004] The existing solutions have several significant drawbacks: First, most of them rely on simple threshold judgments or linear regression models to identify the melting starting point and rate, which are inadequate in the face of complex non-linear melting processes and are prone to misjudgment. Second, traditional denoising methods (such as mean filtering) can reduce noise interference, but may also blur key feature information, affecting the accuracy of subsequent analysis. Finally, existing systems rarely have intelligent adjustment functions and cannot automatically adjust production process parameters (such as cooling procedures) according to real-time monitoring data, limiting their application potential in automated production lines. Summary of the Invention

[0005] The embodiments of the present application provide a melting monitoring system and method for solid dyes to solve the problems of defective product quality and low production efficiency of solid dyes in the prior art.

[0006] In a first aspect, the embodiments of the present application provide a melting monitoring method for solid dyes, including:

[0007] Placing the solid dye to be monitored in a transparent heating container, arranging at least one light source on one side of the transparent heating container, arranging at least one light sensor on the opposite side, selecting a wavelength combination according to the absorption spectrum of the solid dye, and using a multi-wavelength light source to make the light path pass through the maximum cross-section of the solid dye to generate an initial optical path;

[0008] Recording the light intensity values transmitted through the solid dye received by the light sensor at preset time intervals to form an initial time-light intensity curve, and performing denoising processing on the data of the initial optical path by using a wavelet transform algorithm to obtain a time-light intensity curve;

[0009] Based on the changing trend of the time-light intensity curve, use a preset support vector machine model for observation to obtain the turning point where the slope of the light intensity curve changes. Based on the turning point, indicate the moment when the solid dye starts to melt to generate a melting start signal;

[0010] Based on the data of the time-light intensity curve of the melting start signal within a fixed time, use a Kalman filter to calculate the melting rate, and combine with a thermodynamic analysis algorithm to predict the stability of the solid dye under different environmental conditions to generate a quality assessment report;

[0011] When it is detected that the melting rate reaches a preset threshold, use a PID controller to automatically trigger a cooling program, and combine with image analysis technology of deep learning to predict and analyze the uniformity and purity of the solid dye to obtain key parameters, and integrate the key parameters and quality assessment results of the solid dye during the melting process to generate a melting monitoring report.

[0012] Optionally, based on the changing trend of the time-light intensity curve, use a preset support vector machine model for observation to obtain the turning point where the slope of the light intensity curve changes. Based on the turning point, indicate the moment when the solid dye starts to melt to generate a melting start signal, including:

[0013] Use the five-point cubic smoothing algorithm to smooth the time-light intensity curve to obtain a smoothed curve, and calculate the slope of the smoothed curve according to the central difference method combined with the least square fitting technology to obtain the slope values corresponding to each time point;

[0014] Use principal component analysis to perform dimensionality reduction on the slope values to obtain reduced-dimensional slope values, input the reduced-dimensional slope values into a pre-trained support vector machine model, classify and identify the reduced-dimensional slope values to obtain the initial turning point where the slope of the light intensity curve changes;

[0015] Based on the initial turning point, use the dynamic time warping algorithm to match the time-light intensity curve within a specified time period, and use an anomaly detection algorithm to monitor the model output to predict and locate the turning point where the slope of the light intensity curve changes;

[0016] Based on the turning point, use a Kalman filter to optimize the estimation of the melting start moment to generate a melting start signal.

[0017] Optionally, use principal component analysis to perform dimensionality reduction on the slope values to obtain reduced-dimensional slope values, including:

[0018] Optimize the slope value using a robust regression algorithm, organize and form a high-dimensional slope value matrix, and standardize the features of the high-dimensional slope value matrix according to the Z-Score normalization method to generate a standardized slope value matrix, where the features include zero mean and unit variance;

[0019] According to the standardized slope value matrix, calculate a predefined covariance matrix based on Mahalanobis distance, and use a randomized singular value decomposition algorithm to accelerate the decomposition process of the covariance matrix to obtain eigenvalues and eigenvectors;

[0020] Based on the eigenvalues and eigenvectors, use cross-validation techniques to evaluate the performance of the support vector machine model under different numbers of principal components, select N principal components that can retain the maximum information and verify the performance of the support vector machine model. When determining the value of N, use the least absolute shrinkage and selection operator regression to screen out the principal components helpful for the classification task to generate a principal component set;

[0021] Use the t-distributed stochastic neighbor embedding algorithm to perform preliminary dimensionality reduction on the standardized slope value matrix to obtain a preliminarily reduced slope value, and project the preliminarily reduced slope value onto a new coordinate system defined by the principal component set to generate a reduced slope value.

[0022] Optionally, when it is detected that the melting rate reaches a preset threshold, use a PID controller to automatically trigger a cooling program, and combine deep learning-based image analysis techniques to predict and analyze the uniformity and purity of the solid dye to obtain key parameters. Integrate the key parameters of the solid dye and the quality assessment results during the melting process to generate a melting monitoring report, including:

[0023] When the melting rate reaches the preset threshold, use fuzzy logic to optimize the parameter settings of the PID controller and automatically trigger a cooling program according to a preset temperature control strategy;

[0024] During the cooling process, use a high-resolution camera to obtain an initial image of the solid dye, and process the initial image based on multi-scale image fusion technology to obtain an optimized image;

[0025] Preprocess the optimized image according to the superpixel segmentation algorithm to obtain a processed image, and use an ensemble learning method to analyze features such as color distribution, particle size, and distribution extracted from the processed image to obtain key parameters including the uniformity index, purity grade, and impurity content of the solid dye;

[0026] According to the key parameters, collect external data during the melting process, and based on time series analysis, predict changes in the melting rate and environmental conditions within a future fixed time to obtain an analysis result, where the external data includes the melting rate, environmental temperature, and humidity;

[0027] Based on the analysis results, apply natural language processing technology to automatically generate a melting monitoring report.

[0028] Optionally, preprocess the optimized image according to the superpixel segmentation algorithm to obtain a processed image, including:

[0029] Use a multi-view dataset and a structural similarity index to perform weighted average fusion processing on the initial image of the solid stain obtained by a high-resolution camera to generate an optimized image;

[0030] According to the optimized image, initialize a superpixel grid, and based on an improved version of the simple linear iterative clustering algorithm combined with spectral clustering technology, perform iterative optimization processing on the boundaries of the superpixel grid to obtain a preliminary segmentation result;

[0031] Based on the preliminary segmentation result, apply an active contour model and a graph cut algorithm to refine and smooth the boundaries of the superpixel grid, and at the same time introduce shape regularity constraints such as ellipse fitting to generate an intermediate segmentation image;

[0032] According to the intermediate segmentation image, use a connected component labeling algorithm to remove isolated regions, and apply bilateral filtering and smoothing operations for morphological operations to generate a processed image, where the morphological operations include opening and closing operations.

[0033] Optionally, based on the data of the time-light intensity curve of the melting start signal within a fixed time, use a Kalman filter to calculate the melting rate, including:

[0034] Use the melting start signal and combine it with an adaptive threshold algorithm to calculate the fixed time when the solid stain starts to melt. Based on the fixed time, adopt a dynamic window selection method to automatically adjust the time period length according to the change trend of the melting rate, so as to extract the initial time-light intensity curve data starting from the fixed time;

[0035] Use a median filtering method to remove the noise and outliers in the initial time-light intensity curve data, and at the same time introduce a locally weighted regression scatter smoothing technique to retain the characteristic information in the initial time-light intensity curve data, and further apply a robust statistical method to automatically detect and remove outliers to obtain smooth time-light intensity curve data;

[0036] Based on the smooth time-light intensity curve data, use a five-point cubic smoothing algorithm to reduce noise interference, and use the least squares fitting technique combined with an adaptive weight adjustment strategy to calculate the slope value corresponding to each time point, expressed as the preliminary melting rate;

[0037] Use an unscented Kalman filter to optimize the preliminary melting rate to obtain the melting rate.

[0038] Optionally, a median filtering method is used to remove the noise and outliers from the initial time-light intensity curve data, while introducing a locally weighted regression scatterplot smoothing technique to retain the characteristic information in the initial time-light intensity curve data, and further applying a robust statistical method to automatically detect and remove outliers, obtaining smoothed time-light intensity curve data, including:

[0039] The median filtering method is used to perform preliminary denoising on the initial time-light intensity curve data, and the noise influence is reduced by replacing each data point with the median in the neighborhood, obtaining the preliminarily denoised data;

[0040] Based on the preliminarily denoised data, a locally weighted regression scatterplot smoothing technique is introduced, and the curve is smoothed by performing weighted least squares fitting on local data points, retaining the characteristic information and local change trend, generating the smoothed curve data;

[0041] According to the smoothed curve data, a robust statistical method is applied to automatically detect and remove outliers, obtaining the detection result;

[0042] Based on the detection result, the preliminarily denoised data, and the smoothed curve data, they are integrated into the smoothed time-light intensity curve data.

[0043] In a second aspect, an embodiment of the present application provides a melting monitoring system for a solid dye, including:

[0044] A generation module, configured to place the solid dye to be monitored in a transparent heating container, set at least one light source on one side of the transparent heating container, set at least one light sensor on the opposite side, select a wavelength combination according to the absorption spectrum of the solid dye, and use a multi-wavelength light source to make the optical path pass through the maximum cross-section of the solid dye, generating an initial optical path;

[0045] A denoising module, configured to record the light intensity values transmitted through the solid dye received by the light sensor at preset time intervals, form an initial time-light intensity curve, and perform denoising processing on the data of the initial optical path by using a wavelet transform algorithm, obtaining a time-light intensity curve;

[0046] An indication module, configured to perform observation based on the change trend of the time-light intensity curve by using a preset support vector machine model, obtain a turning point where the slope of the light intensity curve changes, and based on the turning point, indicate the moment when the solid dye starts to melt, so as to generate a melting start signal;

[0047] An evaluation module, configured to calculate a melting rate using a Kalman filter based on data of a time-light intensity curve of the melting start signal within a fixed time period, predict the stability of the solid dye under different environmental conditions by combining a thermodynamic analysis algorithm, and generate a quality evaluation report;

[0048] A monitoring module, configured to automatically trigger a cooling program using a PID controller when it detects that the melting rate reaches a preset threshold, predict and analyze the uniformity and purity of the solid dye by combining an image analysis technique of deep learning to obtain key parameters, and integrate the key parameters and the quality evaluation results of the solid dye during the melting process to generate a melting monitoring report.

[0049] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a melting monitoring method for a solid dye as described in the first aspect above.

[0050] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, where when the computer program is executed by a computer, it implements a melting monitoring method for a solid dye as described in the first aspect.

[0051] In the embodiment of the present application, the solid dye to be monitored is placed in a transparent heating container, at least one light source is arranged on one side of the transparent heating container, at least one light sensor is arranged on the opposite side, and a wavelength combination is selected according to the absorption spectrum of the solid dye. A multi-wavelength light source is used to make the optical path pass through the maximum cross-section of the solid dye to generate an initial optical path; the light intensity values transmitted through the solid dye received by the light sensor are recorded at preset time intervals to form an initial time-light intensity curve, and the wavelet transform algorithm is used to denoise the data of the initial optical path to obtain a time-light intensity curve; based on the change trend of the time-light intensity curve, a preset support vector machine model is used for observation to obtain a turning point where the slope of the light intensity curve changes. Based on the turning point, the moment when the solid dye starts to melt is indicated to generate a melting start signal; based on the data of the time-light intensity curve of the melting start signal within a fixed time period, a Kalman filter is used to calculate the melting rate, and the stability of the solid dye under different environmental conditions is predicted by combining a thermodynamic analysis algorithm to generate a quality evaluation report; when it is detected that the melting rate reaches a preset threshold, a cooling program is automatically triggered using a PID controller, and the uniformity and purity of the solid dye are predicted and analyzed by combining an image analysis technique of deep learning to obtain key parameters, and the key parameters and the quality evaluation results of the solid dye during the melting process are integrated to generate a melting monitoring report.

[0052] The technical solution of this application has the following beneficial effects:

[0053] This application generates an initial optical path through a multi-wavelength light source and an optical sensor, applies wavelet transform denoising processing to obtain an accurate time-light intensity curve; uses a support vector machine model to accurately locate the starting moment of melting, and adopts a Kalman filter to calculate the melting rate, combines a thermodynamic analysis algorithm to predict stability, ensuring precise monitoring of the melting process. When the melting rate reaches a preset threshold, the cooling program is automatically triggered, and the uniformity and purity of the dye are analyzed through deep learning technology, and finally a detailed melting monitoring report is generated. This method not only improves the monitoring accuracy, but also optimizes the production process and enhances the product quality control ability, with significant technical advantages and application value.

[0054] Furthermore, this application provides a method for accurately identifying the moment when the solid dye starts to melt based on the change trend of the time-light intensity curve. First, use the five-point cubic smoothing algorithm to smooth the time-light intensity curve, and calculate the slope value through the central difference method combined with the least square fitting technique; then, use principal component analysis to reduce the dimension of the slope value, and input the dimension-reduced slope value into a pre-trained support vector machine model for classification and identification of the initial turning point; then, based on the initial turning point, use the dynamic time warping algorithm to match the curve within a specified time period, and apply an anomaly detection algorithm to monitor the model output to predict and accurately locate the actual turning point; finally, use the Kalman filter to optimize the estimation of the melting starting moment and generate a melting starting signal.

[0055] Through the above method, integrating a variety of advanced algorithms, including smoothing processing, slope calculation, dimension reduction processing, support vector machine classification, dynamic time warping, and Kalman filter optimization, etc., realizes high-precision identification of the melting starting point of the solid dye. Compared with traditional methods, this technology not only improves the monitoring accuracy, but also enhances the adaptability to complex non-linear melting processes, can effectively reduce misjudgment and noise interference, ensures accurate estimation of the melting starting moment, thereby improving product quality and production efficiency. In addition, this method provides reliable data support for subsequent automatic control and further optimizes the production process flow.

[0056] These aspects or other aspects of this application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0058] Figure 1 The flowchart of a method for monitoring the melting of a solid dye provided by the present application is shown;

[0059] Figure 2 The structural schematic diagram of a system for monitoring the melting of a solid dye provided by the present application is shown;

[0060] Figure 3 The structural schematic diagram of a computing device provided by the present application is shown. Detailed implementation manners

[0061] To enable those skilled in the art to better understand the solutions of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application.

[0062] In some processes described in the specification, claims, and the above-mentioned drawings of the present application, multiple operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0064] Figure 1 A flowchart of a method for monitoring the melting of a solid dye provided for the embodiments of the present application is as follows Figure 1 As shown, the method includes:

[0065] 101. Place the solid dye to be monitored in a transparent heating container. Set at least one light source on one side of the transparent heating container and at least one light sensor on the opposite side. Select a wavelength combination according to the absorption spectrum of the solid dye, and use a multi-wavelength light source to make the light path pass through the maximum cross-section of the solid dye to generate an initial optical path;

[0066] Among them, a light sensor: a device that can detect and quantify the light intensity passing through an object; in this system, it is used to monitor the change in the light intensity passing through the solid dye, so as to provide data on the state of the dye;

[0067] Absorption spectrum: It represents the absorption degree of a substance to light of different wavelengths. By analyzing the absorption spectrum, the most suitable wavelength combination for monitoring a specific substance (such as a solid dye) can be selected;

[0068] Multi-wavelength light source: A light source that can emit light of multiple wavelengths. By selecting a suitable wavelength combination, the change in the state of the dye can be more accurately reflected, and the monitoring accuracy can be improved;

[0069] Initial optical path: The light path formed by the multi-wavelength light source passing through the maximum cross-section of the solid dye. This path contains key information about the state of the dye and is the basis for subsequent analysis.

[0070] In actual operation, place the solid dye to be monitored in a transparent heating container, and install a light source and a light sensor on both sides respectively. Select a suitable wavelength combination according to the absorption spectrum of the dye, and use a multi-wavelength light source to make the light path pass through the maximum cross-section of the dye to generate an initial optical path. This step provides basic data for the subsequent time-light intensity curve.

[0071] For example, in an automated production line, the system automatically identifies and loads a specified type of solid dye into the transparent heating container. Subsequently, the system configures the multi-wavelength light source according to the preset absorption spectrum and starts the light source to irradiate the dye. At the same time, the light sensor starts to collect the light intensity values passing through in real time and initially constructs the initial optical path. This process ensures the accuracy of data collection and lays the foundation for subsequent steps.

[0072] 102. Record the light intensity values passing through the solid dye received by the light sensor at preset time intervals to form an initial time-light intensity curve, and perform denoising processing on the data of the initial optical path using the wavelet transform algorithm to obtain a time-light intensity curve;

[0073] Among them, the time-light intensity curve: shows the light intensity values changing with time, reflecting the dynamic change trend of the solid dye gradually melting during the heating process;

[0074] Wavelet transform algorithm: A signal processing technique that can effectively remove noise interference while retaining the important features of the original signal, and is suitable for processing complex time series data.

[0075] In actual operation, the light intensity values received by the optical sensor are recorded at a preset time interval to form an initial time-light intensity curve. Then, the wavelet transform algorithm is used to denoise the curve to obtain a smoother and more accurate time-light intensity curve for subsequent analysis.

[0076] Based on the above embodiment, the system records the light intensity values at a frequency of once per second to generate a time-light intensity curve. Then, the wavelet transform algorithm is applied to denoise the curve to eliminate the influence of random noise. The processed data is stored and prepared for the next step of analysis, ensuring the quality and reliability of the data.

[0077] 103. Based on the change trend of the time-light intensity curve, use a preset support vector machine model for observation to obtain the turning point where the slope of the light intensity curve changes. Based on the turning point, indicate the moment when the solid dye starts to melt to generate a melting start signal;

[0078] Among them, the support vector machine model: A machine learning model that is particularly suitable for classification tasks; in this application scenario, it is used to identify the turning point where the slope changes on the time-light intensity curve, that is, the moment when the solid dye starts to melt;

[0079] Melting start signal: Marks the exact moment when the solid dye starts to melt, providing a key reference point for subsequent rate calculation and quality assessment.

[0080] In actual operation, based on the change trend of the time-light intensity curve, use a pre-trained support vector machine model to observe the slope change, determine the initial turning point, indicate the moment when the solid dye starts to melt, and generate a melting start signal.

[0081] Continuing the above embodiment, the system uses the support vector machine model to analyze the denoised time-light intensity curve and identify the turning point with a significant slope change. Once this turning point is determined, the system immediately generates a melting start signal and transmits it to the control system to prepare for subsequent rate calculation and cooling procedures.

[0082] 104. Based on the data of the time-light intensity curve within a fixed time of the melting start signal, use a Kalman filter to calculate the melting rate, and combine the thermodynamic analysis algorithm to predict the stability of the solid dye under different environmental conditions to generate a quality assessment report;

[0083] Among them, the Kalman filter: a recursive filter used to estimate the state variables of a system from a series of incomplete and noisy measurement data; in this application, it is used to optimize the estimation of the melting rate;

[0084] Thermodynamic analysis algorithm: Based on thermodynamic principles, it predicts the stability of solid dyes under different environmental conditions to help evaluate product quality;

[0085] Quality assessment report: A report generated by comprehensively analyzing factors such as the melting rate and environmental conditions, used to evaluate the quality and stability of solid dyes.

[0086] In actual operation, based on the melting start signal, the Kalman filter is used to calculate the melting rate on the data of the time-light intensity curve within a fixed time. Combining the thermodynamic analysis algorithm to predict the stability of solid dyes under different environmental conditions, a quality assessment report is generated.

[0087] According to the above embodiments, after receiving the melting start signal, the system uses the Kalman filter to calculate the melting rate and combines the thermodynamic analysis algorithm to predict the stability of the dye under different environmental conditions. The system automatically generates a detailed quality assessment report containing key parameters such as the melting rate and temperature change for production management personnel to refer to in order to optimize the production process.

[0088] 105. When it is detected that the melting rate reaches the preset threshold, the PID controller is used to automatically trigger the cooling program, and the uniformity and purity of the solid dye are predicted and analyzed by combining the image analysis technology of deep learning to obtain key parameters. Integrate the key parameters of the solid dye and the quality assessment results during the melting process to generate a melting monitoring report.

[0089] Among them, the PID controller: a proportional-integral-derivative controller used to automatically adjust system parameters such as temperature or speed to reach the set target value; in this application, it is used to control the cooling program;

[0090] Deep learning image analysis technology: Using a deep neural network to analyze images to identify and quantify the uniformity and purity of dyes;

[0091] Melting monitoring report: The final report integrating all key parameters and quality assessment results during the melting process, used to comprehensively understand the entire melting process.

[0092] In actual operation, when it is detected that the melting rate reaches the preset threshold, the PID controller is used to automatically trigger the cooling program. Combining the image analysis technology of deep learning to predict and analyze the uniformity and purity of the solid dye to obtain key parameters. Integrate all key parameters of the dye and the quality assessment results during the melting process to generate a melting monitoring report.

[0093] Continuing with the above embodiment, when the system detects that the melting rate reaches the preset threshold, the PID controller automatically starts the cooling program to prevent product quality problems caused by overheating. Meanwhile, the system uses deep learning technology to analyze the uniformity and purity of the dye, ensuring product quality. Finally, the system integrates all the data to generate a detailed melting monitoring report for further analysis and improvement by production personnel.

[0094] Through the implementation of steps 101 to 105, by using an integrated optical sensor, a multi-wavelength light source, a support vector machine model, a Kalman filter, and deep learning image analysis technology, this method achieves high-precision monitoring and intelligent control of the melting process of solid dye. This method not only improves the accuracy of monitoring, but also enhances the adaptability to complex non-linear melting processes, reduces misjudgment and noise interference, and ensures accurate estimation of the melting start time, rate, and end point. In addition, this method provides reliable data support for subsequent automatic control, further optimizes the production process flow, and improves product consistency and production efficiency.

[0095] To solve the problems of low accuracy and high misjudgment rate in identifying the melting start point of solid dye by traditional methods and further improve the accuracy and reliability of melting monitoring, in some embodiments, in step 103, based on the change trend of the time-light intensity curve, a preset support vector machine model is used for observation to obtain the turning point where the slope of the light intensity curve changes. Based on the turning point, the moment when the solid dye starts to melt is indicated to generate a melting start signal, including:

[0096] The five-point cubic smoothing algorithm is used to smooth the time-light intensity curve to obtain a smooth curve, and the slope of the smooth curve is calculated according to the central difference method combined with the least square fitting technique to obtain the slope values corresponding to each time point; the principal component analysis is used to perform dimensionality reduction processing on the slope values to obtain the dimensionality-reduced slope values, and the dimensionality-reduced slope values are input into a pre-trained support vector machine model to classify and identify the dimensionality-reduced slope values to obtain the initial turning point where the slope of the light intensity curve changes; based on the initial turning point, the dynamic time warping algorithm is used to match the time-light intensity curve within a specified time period, and the anomaly detection algorithm is used to monitor the model output to predict and locate the turning point where the slope of the light intensity curve changes; based on the turning point, the Kalman filter is used to optimize the estimation of the melting start time to generate a melting start signal.

[0097] In this embodiment, the five-point cubic smoothing algorithm: a data smoothing technique that reduces the influence of noise by using the weighted average of adjacent five data points and is applicable to the smoothing processing of time series data; this algorithm can effectively remove high-frequency noise while retaining the main trend of the data;

[0098] Central difference method: A numerical differentiation method used to calculate the derivative (i.e., slope) of discrete data points; in this application scenario, it is used to estimate the slope of each point on the time-light intensity curve to help identify key change points during the melting process;

[0099] Least squares fitting technique: A statistical method that determines the best-fit line or curve by minimizing the sum of the squared errors between the observed data and the model predictions; this method is often used to extract useful feature information from noisy data;

[0100] Dynamic Time Warping algorithm (DTW): An algorithm used to measure the similarity between two time series, even if these series may have different speeds or lengths; in this scenario, DTW is used to match the time-light intensity curves within a specified time period to ensure the accuracy of turning point localization;

[0101] Anomaly detection algorithm: A technique used to identify data points in a dataset that do not conform to the expected pattern; in this application scenario, it is used to monitor the quality of the output of the support vector machine model to improve the reliability of turning point localization.

[0102] In the embodiments of this application, first, the five-point cubic smoothing algorithm is used to smooth the time-light intensity curve to eliminate noise interference, and then the central difference method is combined with the least squares fitting technique to calculate the slope value corresponding to each time point.

[0103] Next, principal component analysis (PCA) is used to reduce the dimensionality of the slope values to obtain the reduced-dimensional slope values, which are then input into a pre-trained support vector machine model for classification and identification to find the preliminary turning points. Based on the preliminary turning points, the Dynamic Time Warping algorithm is used to match the curves within a specified time period, and the anomaly detection algorithm is used to monitor the model output to accurately predict and locate the actual turning points.

[0104] Finally, based on the located turning points, the Kalman filter is used to optimize the estimation of the melting start time to generate a melting start signal.

[0105] The following is a specific example:

[0106] For example, in an automated dye production line, the system first collects the time-light intensity curve data of the solid dye. To improve the data quality, the system applies the five-point cubic smoothing algorithm to smooth the original curve, significantly reducing the influence of random noise. Subsequently, the system uses the central difference method combined with the least squares fitting technique to calculate the slope value of each point and identify the key regions of slope change.

[0107] Next, the system inputs these slope values into a pre-trained support vector machine model. After dimensionality reduction through principal component analysis, the model can more efficiently identify the initial turning points of the slope changes. Based on this initial turning point, the system further applies the dynamic time warping algorithm to match the curve within a specified time period to ensure the accuracy of the turning point positioning. Meanwhile, an anomaly detection algorithm is used to monitor the model output to ensure the reliability of the recognition results.

[0108] Once the system determines the final turning point, i.e., the moment when the solid dye starts to melt, the Kalman filter is used to optimize the estimation of this moment, generating an accurate melting start signal. This signal not only provides the basis for subsequent rate calculations but also triggers a series of automated control programs, such as automatically adjusting the heating power or starting the cooling program, thus achieving highly automated and intelligent management of the entire production process.

[0109] By comprehensively applying this method of multiple advanced algorithms and technologies, the system not only improves the accuracy and reliability of melting monitoring but also enhances its adaptability to complex non-linear melting processes, significantly improving product quality and production efficiency.

[0110] To address the complexity issues in processing and analyzing the high-dimensional slope value matrix and further improve the accuracy and efficiency of the support vector machine model classification, in some embodiments, principal component analysis is used to perform dimensionality reduction on the slope values to obtain reduced-dimensional slope values, including:

[0111] Optimizing the slope values using a robust regression algorithm, organizing them to form a high-dimensional slope value matrix, and standardizing the features of the high-dimensional slope value matrix according to the Z-Score normalization method to generate a standardized slope value matrix, where the features include zero mean and unit variance; calculating a pre-defined covariance matrix based on Mahalanobis distance according to the standardized slope value matrix, and using a randomized singular value decomposition algorithm to accelerate the decomposition process of the covariance matrix to obtain eigenvalues and eigenvectors; based on the eigenvalues and eigenvectors, using cross-validation techniques to evaluate the performance of the support vector machine model under different numbers of principal components, selecting N principal components that can retain the maximum information and verify the performance of the support vector machine model. When determining the value of N, the least absolute shrinkage and selection operator regression is used to screen out the principal components helpful for the classification task to generate a principal component set; using the t-distributed stochastic neighbor embedding algorithm to perform preliminary dimensionality reduction on the standardized slope value matrix to obtain preliminary reduced-dimensional slope values, and projecting the preliminary reduced-dimensional slope values onto a new coordinate system defined by the principal component set to generate reduced-dimensional slope values.

[0112] In this embodiment, the robust regression algorithm: a regression analysis method for dealing with outliers in a dataset; it reduces the impact of outliers on the estimation of model parameters to ensure the stability and accuracy of the regression results; in this application scenario, the robust regression algorithm is used to optimize the slope value to improve the reliability of subsequent analysis;

[0113] High-dimensional slope value matrix: a data matrix composed of slope values corresponding to multiple time points, usually having a high dimension; this matrix contains key information at different time points during the melting process of the solid dye;

[0114] Z-Score normalization method: a data normalization technique that transforms data into a form with zero mean and unit variance; this helps to eliminate the dimensional differences between different features and facilitates subsequent statistical analysis and machine learning model training;

[0115] Mahalanobis distance: a measure of the similarity between data points that takes into account the covariance structure of the dataset; in this application scenario, calculating the covariance matrix based on the Mahalanobis distance can more accurately capture the correlation between data points;

[0116] Randomized Singular Value Decomposition (RSVD): an algorithm for accelerating the decomposition of large matrices; through random sampling techniques, RSVD can significantly reduce the computational complexity while maintaining high accuracy, and is suitable for large-scale data processing;

[0117] Cross-validation technique: a method for evaluating the performance of machine learning models, which divides the dataset into multiple subsets for multiple training and testing to select the best model parameters or structure; in this application, it is used to determine the optimal value of the number of principal components;

[0118] Least Absolute Shrinkage and Selection Operator regression (LASSO): a regularization regression method that compresses unimportant regression coefficients to zero by introducing a penalty term, thereby achieving feature selection; in this scenario, LASSO is used to screen out the principal components helpful for the classification task;

[0119] t-Distributed Stochastic Neighbor Embedding (t-SNE): a dimensionality reduction technique, especially suitable for the visualization and clustering analysis of high-dimensional data; it maps the probability distribution in high-dimensional space to low-dimensional space by simulating the probability distribution in high-dimensional space, retaining the local structure.

[0120] In the embodiment of this application, first, the robust regression algorithm is used to optimize the slope value and organize it into a high-dimensional slope value matrix. Then, the Z-Score normalization method is used to normalize the high-dimensional slope value matrix to generate a normalized slope value matrix. Next, the covariance matrix based on the Mahalanobis distance is calculated according to the normalized slope value matrix, and the Randomized Singular Value Decomposition algorithm is used to accelerate its decomposition process to obtain eigenvalues and eigenvectors.

[0121] Next, based on cross-validation technology, evaluate the performance of the support vector machine model under different numbers of principal components, select N principal components that can maximize the retained information and verify the model performance, and at the same time use LASSO to screen out the principal components helpful for the classification task to generate a set of principal components.

[0122] Finally, use the t-SNE algorithm to perform preliminary dimensionality reduction on the standardized slope value matrix, project the preliminary dimensionality-reduced slope values onto the new coordinate system defined by the set of principal components, and generate the final dimensionality-reduced slope values.

[0123] The following is a specific example:

[0124] According to the above embodiments, the system first collects the time-light intensity curve data during the melting process and calculates the slope values at each time point to form a high-dimensional slope value matrix. To improve the quality and stability of the data, the system uses a robust regression algorithm to optimize the slope values and reduce the influence of outliers.

[0125] Subsequently, the system normalizes the optimized slope values into the form of zero mean and unit variance to generate a standardized slope value matrix. Based on this standardized matrix, the system calculates the covariance matrix based on Mahalanobis distance and uses the random singular value decomposition algorithm to quickly decompose the matrix to extract eigenvalues and eigenvectors.

[0126] To determine the optimal number of principal components, the system uses cross-validation technology to evaluate the performance of the support vector machine model under different numbers of principal components, and at the same time uses LASSO regression to screen out the principal components most helpful for the classification task to generate a set of principal components. This process not only improves the accuracy of the model but also ensures the generalization ability of the model.

[0127] Next, the system uses the t-SNE algorithm to perform preliminary dimensionality reduction on the standardized slope value matrix, projects the preliminary dimensionality-reduced slope values onto the new coordinate system defined by the set of principal components, and generates dimensionality-reduced slope values. These dimensionality-reduced data not only reduce the computational complexity but also retain the key features of the original data, providing high-quality input data for the subsequent classification of the support vector machine model.

[0128] By this method of comprehensively applying a variety of advanced algorithms and technologies, the system not only improves the accuracy and efficiency of data processing but also enhances the adaptability to complex non-linear melting processes, significantly improving product quality and production efficiency. This method provides strong support for the intelligent management of automated control systems.

[0129] To solve the problem that it is difficult to precisely control the uniformity and purity of solid dyes during the melting process and further improve product quality and production efficiency, in some embodiments, when it is detected that the melting rate reaches a preset threshold in step 105, a cooling program is automatically triggered using a PID controller, and the uniformity and purity of the solid dye are predicted and analyzed by combining the image analysis technology of deep learning to obtain key parameters. The key parameters of the solid dye and the quality evaluation results during the melting process are integrated to generate a melting monitoring report, including:

[0130] When the melting rate reaches the preset threshold, the parameter settings of the PID controller are optimized using fuzzy logic, and a cooling program is automatically triggered according to the preset temperature control strategy; during the cooling process, a high-resolution camera is used to obtain the initial image of the solid dye, and the initial image is processed based on the multi-scale image fusion technology to obtain an optimized image; the optimized image is preprocessed according to the superpixel segmentation algorithm to obtain a processed image, and an ensemble learning method is used to analyze features such as color distribution, particle size, and distribution extracted from the processed image to obtain key parameters including the uniformity index, purity level, and impurity content of the solid dye; according to the key parameters, external data during the melting process are collected, and the melting rate and environmental condition changes within a future fixed time are predicted based on time series analysis to obtain an analysis result, where the external data includes the melting rate, environmental temperature, and humidity; based on the analysis result, a melting monitoring report is automatically generated using natural language processing technology.

[0131] In this embodiment, the fuzzy logic optimizes the PID controller: a control strategy based on fuzzy logic for optimizing the parameter settings of the PID (Proportional-Integral-Derivative) controller; it imitates the human decision-making process to process uncertain and imprecise information, thereby achieving more flexible and robust control;

[0132] Multi-scale image fusion technology: a technology that combines images of multiple scales or resolutions into a single high-quality image; this method can improve the detail clarity and overall quality of the image, especially suitable for complex scene images obtained by high-resolution cameras;

[0133] Superpixel segmentation algorithm: an image segmentation technology that simplifies the image representation by grouping pixels into superpixels; each superpixel consists of a group of adjacent pixels with similar color or texture characteristics, which helps subsequent feature extraction and analysis;

[0134] Ensemble learning method: a machine learning strategy that improves the overall performance by combining the prediction results of multiple models; common ensemble learning methods include random forests, gradient boosting trees, etc., and are used in this application to extract and analyze key features from the processed image;

[0135] Time series analysis: A statistical method used to analyze data sequences that change over time to identify trends, periodicity, and other patterns; in this application scenario, it is used to predict future melting rates and changes in environmental conditions;

[0136] Natural Language Processing Technology (NLP): A field of computer science that involves enabling computers to understand, interpret, and generate human language; in this application scenario, NLP is used to automatically generate detailed melting monitoring reports.

[0137] In the embodiments of this application, when it is detected that the melting rate of the solid dye reaches a preset threshold, the system uses fuzzy logic to optimize the parameter settings of the PID controller and automatically triggers a cooling program according to a preset temperature control strategy. During the cooling process, a high-resolution camera is used to capture the initial images of the solid dye, and multi-scale image fusion technology is used to process these images to obtain optimized images. Then, the superpixel segmentation algorithm is applied to preprocess the optimized images to generate processed images. Next, an ensemble learning method is used to extract features such as color distribution, particle size, and distribution from the processed images, and key parameters such as the uniformity index, purity grade, and impurity content of the solid dye are calculated. Based on these key parameters, the system collects external data during the melting process (such as melting rate, environmental temperature, and humidity), and predicts the melting rate and changes in environmental conditions within a future fixed time through time series analysis. Finally, natural language processing technology is applied to automatically generate a melting monitoring report containing all key parameters and quality assessment results.

[0138] The following is a specific example:

[0139] Continuing the above embodiment, when the system detects that the melting rate reaches the preset threshold, the fuzzy logic optimized PID controller automatically adjusts its parameter settings and starts the cooling program. To ensure the cooling effect and product quality, the system uses a high-resolution camera to capture the status images of the solid dye in real time.

[0140] These initial images are first processed by multi-scale image fusion technology to enhance the details and clarity of the images and generate optimized images. Next, the system applies the superpixel segmentation algorithm to preprocess the optimized images, decomposing the images into multiple superpixel regions for subsequent feature extraction.

[0141] Based on the processed images, the system uses an ensemble learning method to analyze the features such as color distribution, particle size, and distribution extracted from the images, and calculates key parameters such as the uniformity index, purity grade, and impurity content of the solid dye. For example, by analyzing the color differences in different regions, the uniformity of the dye can be quantified; by measuring the particle size and distribution, its purity and impurity content can be evaluated.

[0142] Meanwhile, the system collects external data during the melting process (such as melting rate, ambient temperature, and humidity), and uses time series analysis methods to predict the melting rate and changes in environmental conditions over a period of time in the future. This step not only helps production personnel make preparations in advance but also provides important feedback information for optimizing the production process.

[0143] Finally, the system applies natural language processing technology to automatically generate a detailed melting monitoring report. The report includes all key parameters, quality assessment results, and predictions of future trends, providing comprehensive support for production management and quality control. This solution that comprehensively applies a variety of advanced technologies and methods significantly improves the automation level and product quality of solid dye production.

[0144] To solve the problems of low accuracy and edge information loss of traditional image segmentation methods when dealing with solid dye images and further improve the accuracy and detail retention of image segmentation, in some embodiments, the optimized image is preprocessed according to the superpixel segmentation algorithm to obtain a processed image, including:

[0145] Using a multi-view dataset and the structural similarity index, perform weighted average fusion processing on the initial image of the solid dye obtained by a high-resolution camera to generate an optimized image; according to the optimized image, initialize a superpixel grid, and based on an improved simple linear iterative clustering algorithm combined with spectral clustering technology, perform iterative optimization processing on the boundaries of the superpixel grid to obtain a preliminary segmentation result; based on the preliminary segmentation result, apply an active contour model and a graph cut algorithm to refine and smooth the boundaries of the superpixel grid, and at the same time introduce shape regularity constraints such as ellipse fitting to generate an intermediate segmentation image; according to the intermediate segmentation image, use a connected component labeling algorithm to remove isolated regions, and apply bilateral filtering and smoothing operations for morphological operations to generate a processed image, where the morphological operations include opening and closing operations.

[0146] In this embodiment, a multi-view dataset: a data set composed of images taken from multiple different perspectives, which can provide more comprehensive target information; in this application scenario, it is used to enhance the quality and details of solid dye images;

[0147] Structural similarity index (SSIM): a metric for measuring the similarity between two images, considering differences in brightness, contrast, and structure; in this scenario, SSIM is used to guide the weighted average fusion processing to generate a high-quality optimized image;

[0148] Superpixel grid: a grid structure that divides an image into multiple superpixels, and each superpixel consists of a group of adjacent pixels with similar color or texture characteristics; the superpixel grid helps to simplify the image representation and accelerate subsequent processing;

[0149] Simple Linear Iterative Clustering (SLIC): A commonly used superpixel segmentation algorithm that simplifies the image representation by dividing the image into a fixed number of superpixels; The improved SLIC combined with spectral clustering technology can further improve the segmentation accuracy;

[0150] Spectral clustering technology: A graph-theory-based clustering method that performs clustering analysis by constructing a similarity matrix between image pixels; This technology can effectively capture complex structural features in the image;

[0151] Active contour model (Snake Model): An image segmentation method that gradually approximates the target boundary by minimizing an energy function; In this application, it is used to refine the boundaries of the superpixel grid;

[0152] Graph Cut algorithm: A graph-theory-based image segmentation method that divides the image by finding the minimum cut; This method is commonly used to refine and optimize the segmentation results;

[0153] Shape regularity constraints (such as ellipse fitting): A method for constraining the shape of the segmentation result to ensure that the segmented region meets specific geometric shape requirements; In this scenario, ellipse fitting is used to smooth and adjust the segmentation boundary;

[0154] Connected component labeling algorithm: An algorithm for identifying and labeling connected regions in an image; It can remove isolated small regions and improve the accuracy of the segmentation result;

[0155] Bilateral filtering: A non-linear filtering method that can smooth image noise while preserving edges; In this application scenario, it is used for preprocessing before morphological operations;

[0156] Morphological operations (opening and closing): A series of set-theory-based image processing operations used to remove small objects or fill holes; Opening is usually used to remove small noise points, while closing is used to fill holes.

[0157] In the embodiments of this application, first, the initial image obtained by the high-resolution camera is processed by weighted average fusion using the multi-view data set and the structural similarity index to generate an optimized image. Then, the superpixel grid is initialized based on the optimized image, and the boundaries of the superpixel grid are iteratively optimized using the improved simple linear iterative clustering algorithm combined with spectral clustering technology to obtain a preliminary segmentation result.

[0158] Next, based on the preliminary segmentation results, an active contour model and a graph cut algorithm are applied to refine and smooth the boundaries of the superpixel grid. Meanwhile, shape regularity constraints (such as ellipse fitting) are introduced to generate an intermediate segmentation image. Finally, according to the intermediate segmentation image, the connected component labeling algorithm is used to remove isolated regions, and bilateral filtering and smoothing operations are applied for morphological operations (including opening and closing operations) to generate the final processed image. The following is a specific example:

[0159] Continuing with the above embodiment, the system uses a high-resolution camera to capture images of the solid dye from multiple angles to form a multi-view dataset. To generate a high-quality optimized image, the system performs weighted average fusion processing on these images using the structural similarity index. This step not only improves the clarity of the image but also enhances the detailed information in the image.

[0160] Next, the system initializes the superpixel grid based on the optimized image and uses an improved version of the simple linear iterative clustering algorithm combined with spectral clustering technology to iteratively optimize the boundaries of the superpixel grid. This method can accurately capture the subtle structural changes inside the dye to generate a preliminary segmentation result.

[0161] To further refine and optimize the segmentation result, the system applies an active contour model and a graph cut algorithm to adjust and smooth the boundaries of the superpixel grid and introduces shape regularity constraints (such as ellipse fitting). These steps ensure that the boundaries of the segmented regions are smoother and more regular, generating an intermediate segmentation image.

[0162] Subsequently, the system uses the connected component labeling algorithm to remove isolated regions in the intermediate segmentation image to avoid interference from these small regions in subsequent analysis. To further improve the image quality, the system applies bilateral filtering and smoothing operations for morphological operations (including opening and closing operations) to generate the final processed image.

[0163] Through this solution that comprehensively uses a variety of advanced image processing techniques and methods, the system not only improves the accuracy and reliability of image segmentation but also enhances the adaptability to complex non-linear melting processes, significantly improving product quality and production efficiency. This method provides strong support for the intelligent management and optimization of the automated control system.

[0164] To solve the problems of noise interference and the influence of outliers in the melting rate calculation process and further improve the accuracy and reliability of the melting rate calculation, in some embodiments, using the data of the time-light intensity curve of the melting start signal within a fixed time in step 104, calculating the melting rate using a Kalman filter includes:

[0165] Using the melting start signal and combining with an adaptive threshold algorithm, calculate the fixed moment when the solid dye starts to melt. Based on the fixed moment, adopt a dynamic window selection method to automatically adjust the time period length according to the changing trend of the melting rate, so as to extract the initial time-light intensity curve data starting from the fixed moment; use a median filtering method to remove the noise and outliers in the initial time-light intensity curve data, and at the same time introduce a locally weighted regression scatterplot smoothing technique to retain the characteristic information in the initial time-light intensity curve data, and further apply a robust statistical method to automatically detect and remove outliers to obtain smoothed time-light intensity curve data; based on the smoothed time-light intensity curve data, use a five-point cubic smoothing algorithm to reduce noise interference, and use a least squares fitting technique combined with an adaptive weight adjustment strategy to calculate the slope value corresponding to each time point, expressed as the preliminary melting rate; use an unscented Kalman filter to optimize the preliminary melting rate to obtain the melting rate.

[0166] In this embodiment, the adaptive threshold algorithm: a method for dynamically adjusting the threshold, which automatically adjusts the threshold level according to the real-time change of the data; in this application scenario, it is used to determine the exact moment when the solid dye starts to melt;

[0167] The dynamic window selection method: a flexible data processing technique that automatically adjusts the length of the analysis time period according to the changing trend of the data to better capture the data characteristics; this helps to more accurately extract and analyze the time-light intensity curve data;

[0168] The five-point cubic smoothing algorithm: a data smoothing technique based on weighted average, which reduces noise interference by performing a cubic polynomial fitting on each data point and its adjacent four points; this method can effectively remove high-frequency noise while retaining the main trend of the data;

[0169] The least squares fitting technique combined with an adaptive weight adjustment strategy: a statistical method that determines the best fitting straight line or curve by minimizing the sum of the squares of the errors between the observed data and the model predictions, and combines an adaptive weight adjustment strategy to optimize the fitting effect; this strategy allows adjusting the influence of different data points on the final fitting result according to their importance;

[0170] The unscented Kalman filter (UKF): an improved version of the extended Kalman filter, suitable for nonlinear systems; the UKF approximates the probability distribution of the system state by using a set of carefully selected sample points (called Sigma points), thereby improving the accuracy of state estimation.

[0171] In the embodiments of the present application, first, a fixed moment when the solid dye starts to melt is calculated by using the melting start signal and combining with an adaptive threshold algorithm. Based on this fixed moment, a dynamic window selection method is adopted to automatically adjust the length of the time period according to the change trend of the melting rate, and the initial time-light intensity curve data starting from this moment is extracted.

[0172] Next, the median filtering method is used to remove the noise and outliers in these initial data. At the same time, the locally weighted regression scatterplot smoothing technique is introduced to retain the feature information, and the robust statistical method is further applied to automatically detect and remove the outliers, obtaining the smoothed time-light intensity curve data.

[0173] Then, based on the smoothed data, the five-point cubic smoothing algorithm is used to further reduce the noise interference, and the least squares fitting technique is combined with the adaptive weight adjustment strategy to calculate the slope value corresponding to each time point, which is expressed as the preliminary melting rate.

[0174] Finally, the unscented Kalman filter is used to optimize the preliminary melting rate to obtain a more accurate melting rate.

[0175] The following is a specific example:

[0176] Continuing the above embodiment, when the system detects the signal that the solid dye starts to melt, first, the adaptive threshold algorithm is used to determine the exact start moment of melting. Based on this moment, the system adopts the dynamic window selection method to automatically adjust the analysis time period according to the change trend of the melting rate, and extracts the initial time-light intensity curve data starting from the melting start moment.

[0177] To ensure the quality of the data, the system first uses the median filtering method to remove the noise and outliers in the initial time-light intensity curve data. Next, the locally weighted regression scatterplot smoothing technique is applied to retain the key feature information in the data, and the robust statistical method is further used to automatically detect and remove any remaining outliers, generating the smoothed time-light intensity curve data.

[0178] After obtaining the high-quality smoothed data, the system applies the five-point cubic smoothing algorithm to further reduce the noise interference, and uses the least squares fitting technique combined with the adaptive weight adjustment strategy to calculate the slope value of each time point as the preliminary melting rate. This process not only improves the accuracy of the data but also enhances the adaptability to complex non-linear melting processes.

[0179] Finally, the system uses an unscented Kalman filter to optimize the preliminary melting rate and obtain a more accurate melting rate. The optimized melting rate not only provides more reliable monitoring results but also provides an important basis for subsequent cooling procedures and other automated control operations. Through this solution that comprehensively applies a variety of advanced technologies and methods, the system significantly improves the accuracy and reliability of melting monitoring, further optimizes the production process flow, and improves product quality and production efficiency.

[0180] To address the problem that noise and outliers in the initial time-light intensity curve data affect the analysis accuracy and further improve the effect of data smoothing and the retention of characteristic information, in some embodiments, the median filtering method is used to remove the noise and outliers in the initial time-light intensity curve data, and at the same time, the locally weighted regression scatterplot smoothing technique is introduced to retain the characteristic information in the initial time-light intensity curve data, and the robust statistical method is further applied to automatically detect and remove outliers, obtaining the smoothed time-light intensity curve data, including:

[0181] The median filtering method is used to perform preliminary denoising on the initial time-light intensity curve data. By replacing each data point with the median in the neighborhood, the noise effect is reduced, and the preliminarily denoised data is obtained. Based on the preliminarily denoised data, the locally weighted regression scatterplot smoothing technique is introduced. By performing weighted least squares fitting on local data points, the curve is smoothed, and the characteristic information and local change trend are retained to generate the smoothed curve data. According to the smoothed curve data, the robust statistical method is applied to automatically detect and remove outliers to obtain the detection result. Based on the detection result, the preliminarily denoised data, and the smoothed curve data, they are integrated into the smoothed time-light intensity curve data.

[0182] In this embodiment, the median filtering method: a non-linear digital filtering technique that reduces noise by replacing each data point with the median of the data points in the neighborhood; this method is particularly suitable for removing impulse noise (also known as salt-and-pepper noise) while trying to retain the main features of the original signal;

[0183] Locally weighted regression scatterplot smoothing technique (LOESS): a non-parametric regression method that smooths the curve by performing weighted least squares fitting on local data points; it can effectively capture the local change trend in the data while retaining important characteristic information;

[0184] Robust statistical method: a statistical technique that can still provide reliable results in the presence of outliers or noise; common robust statistical methods include RANSAC (Random Sample Consensus algorithm), M-estimation, etc.; these methods can automatically detect and remove outliers in the data, thereby improving the accuracy of the analysis results.

[0185] In the embodiments of the present application, first, a median filtering method is used to perform preliminary denoising on the initial time-light intensity curve data. The noise influence is reduced by replacing each data point with the median value within its neighborhood, generating the preliminarily denoised data.

[0186] Next, based on the preliminarily denoised data, the locally weighted scatterplot smoothing technique (LOESS) is introduced. The curve is smoothed by performing weighted least squares fitting on local data points, retaining the feature information and local change trends, and generating the smoothed curve data.

[0187] Then, a robust statistical method is applied to automatically detect and remove outliers in the smoothed curve data, obtaining the detection result. Finally, based on the detection result, the preliminarily denoised data, and the smoothed curve data, the final smoothed time-light intensity curve data is integrated and generated.

[0188] The following is a specific example:

[0189] Continuing with the above embodiment, the system collected the time-light intensity curve data of the solid dye during the melting process. To ensure the quality of this data, first, a median filtering method was used to perform preliminary denoising on the initial data. Specifically, for each data point, the system calculated the median value within its neighborhood and replaced the original data point with this median value. This can effectively remove random noise, especially impulse noise, while retaining the main trend information.

[0190] Next, based on the preliminarily denoised data, the system introduced the locally weighted scatterplot smoothing technique (LOESS). By assigning weights to the local regions near each data point and performing weighted least squares fitting, the system was able to smooth the curve while retaining important local change trends and feature information. For example, during certain key time periods, the light intensity of the dye may exhibit significant changes, and these change trends need to be accurately captured and retained.

[0191] Subsequently, the system applied a robust statistical method (such as RANSAC) to automatically detect and remove outliers in the smoothed curve data. These outliers may be caused by sensor failures or other external interferences, and their presence will affect the accuracy of subsequent analysis. Through the robust statistical method, the system can identify and exclude these outliers to ensure the authenticity and reliability of the remaining data.

[0192] Finally, the system integrates the initially denoised data, the smoothed curve data, and the detection results of the robust statistical method to generate the final smoothed time-light intensity curve data. These high-quality data not only provide a solid foundation for subsequent melting rate calculations but also enhance the automated control ability and intelligent management level of the entire production process. Through this solution that comprehensively applies a variety of advanced technologies and methods, the system significantly improves the accuracy and reliability of data processing, further optimizes the production process flow, and improves product quality and production efficiency.

[0193] Figure 2 FIG. [0000394] is a schematic structural diagram of a melting monitoring device (or system) for a solid dye provided by an embodiment of the present application. As Figure 2 shown, the device includes:

[0194] A generation module 21, configured to place the solid dye to be monitored in a transparent heating container, set at least one light source on one side of the transparent heating container, set at least one light sensor on the opposite side, select a wavelength combination according to the absorption spectrum of the solid dye, and use a multi-wavelength light source to make the optical path pass through the maximum cross-section of the solid dye to generate an initial optical path;

[0195] A denoising module 22, configured to record the light intensity values transmitted through the solid dye received by the light sensor at a preset time interval to form an initial time-light intensity curve, and perform denoising processing on the data of the initial optical path by using a wavelet transform algorithm to obtain a time-light intensity curve;

[0196] An indication module 23, configured to perform observation based on the change trend of the time-light intensity curve by using a preset support vector machine model to obtain a turning point where the slope of the light intensity curve changes, and based on the turning point, indicate the moment when the solid dye starts to melt to generate a melting start signal;

[0197] An evaluation module 24, configured to calculate the melting rate by using a Kalman filter based on the data of the time-light intensity curve of the melting start signal within a fixed time, and combine a thermodynamic analysis algorithm to predict the stability of the solid dye under different environmental conditions to generate a quality evaluation report;

[0198] A monitoring module 25, configured to automatically trigger a cooling program by using a PID controller when it is detected that the melting rate reaches a preset threshold, combine image analysis technology of deep learning to predict and analyze the uniformity and purity of the solid dye to obtain key parameters, and integrate the key parameters and quality evaluation results of the solid dye during the melting process to generate a melting monitoring report.

[0199] Figure 2 The described melting monitoring device for a solid dye can executeFigure 1 A method for monitoring the melting of a solid dye as described in the illustrated embodiment, the implementation principle and technical effects of which will not be elaborated further. For the device for monitoring the melting of a solid dye in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiments related to this method and will not be elaborated here.

[0200] In a possible design, Figure 2 The device for monitoring the melting of a solid dye in the illustrated embodiment can be implemented as a computing device, such as Figure 3 as shown, the computing device may include a storage component 31 and a processing component 32;

[0201] The storage component 31 stores one or more computer instructions, where the one or more computer instructions are called and executed by the processing component 32.

[0202] The processing component 32 is used for the Figure 1 method for monitoring the melting of a solid dye in the above

[0203] embodiment. Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0204] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0205] Of course, the computing device may also necessarily include other components, such as an input / output interface, a display component, a communication component, etc.

[0206] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above peripheral interface module may be an output device, an input device, etc.

[0207] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0208] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform. At this time, the computing device can refer to a cloud server. The above processing components, storage components, etc. can be basic server resources leased or purchased from a cloud computing platform.

[0209] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above Figure 1 XX method shown in the embodiment.

[0210] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0211] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0212] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0213] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A method for monitoring the melting of a solid colorant, characterized in that: include: The solid dye to be monitored is placed in a transparent heating container, at least one light source is arranged on one side of the transparent heating container, at least one light sensor is arranged on the other side opposite thereto, a wavelength combination is selected according to the absorption spectrum of the solid dye, and a multi-wavelength light source is used to make the light path pass through the maximum cross section of the solid dye to generate an initial optical path; Recording the light intensity value transmitted through the solid dye received by the light sensor according to a preset time interval to form an initial time-light intensity curve, and performing denoising processing on the data of the initial optical path using a wavelet transform algorithm to obtain a time-light intensity curve; Based on the change trend of the time-light intensity curve, a preset support vector machine model is used for observation to obtain a turning point where the slope of the light intensity curve changes, and based on the turning point, the moment when the solid dye begins to melt is indicated to generate a melting start signal; Based on the data of the time-light intensity curve of the melting start signal at a fixed time, the melting rate is calculated using a Kalman filter, and the stability of the solid dye under different environmental conditions is predicted in combination with a thermodynamic analysis algorithm to generate a quality assessment report; When it is detected that the melting rate reaches a preset threshold, a cooling program is automatically triggered by a PID controller, and the uniformity and purity of the solid dye are predicted and analyzed by combining deep learning image analysis technology to obtain key parameters, and the key parameters and quality assessment results of the solid dye during the melting process are integrated to generate a melting monitoring report; When it is detected that the melting rate reaches a preset threshold, the cooling program is automatically triggered by the PID controller, and the uniformity and purity of the solid dye are predicted and analyzed by the deep learning image analysis technology to obtain key parameters. The key parameters and quality assessment results of the solid dye during the melting process are integrated to generate a melting monitoring report, including: When the melting rate reaches a preset threshold, the parameter setting of the PID controller is optimized using fuzzy logic, and a cooling program is automatically triggered according to a preset temperature control strategy; During the cooling process, an initial image of the solid dye is obtained using a high-resolution camera, and the initial image is processed based on a multi-scale image fusion technology to obtain an optimized image; Preprocessing the optimized image according to a superpixel segmentation algorithm to obtain a processed image, and using an ensemble learning method to analyze the color distribution, particle size and distribution and other features extracted from the processed image to obtain key parameters including uniformity index, purity level and impurity content of the solid colorant; According to the key parameters, external data in the melting process are collected, and the melting rate and environmental condition changes in a fixed time in the future are predicted based on time series analysis to obtain analysis results, wherein the external data includes melting rate, environmental temperature and humidity; Based on the analysis results, natural language processing technology is used to automatically generate melting monitoring reports; Preprocessing the optimized image according to a superpixel segmentation algorithm to obtain a processed image includes: Using a multi-view data set and a structural similarity index, a weighted average fusion process is performed on the initial image of the solid dye acquired by a high-resolution camera to generate an optimized image; Initializing a superpixel grid according to the optimized image, and iteratively optimizing the boundary of the superpixel grid based on an improved simple linear iterative clustering algorithm combined with a spectral clustering technique to obtain a preliminary segmentation result; Based on the preliminary segmentation result, an active contour model and a graph cut algorithm are applied to refine and smooth the boundaries of the superpixel grid, and shape regularity constraints such as ellipse fitting are introduced to generate an intermediate segmented image; According to the intermediate segmented image, a connected component labeling algorithm is used to remove isolated areas, and bilateral filtering and smoothing operations are applied to perform morphological operations to generate a processed image, wherein the morphological operations include opening operations and closing operations.

2. The method according to claim 1, characterized in that Based on the change trend of the time-light intensity curve, a preset support vector machine model is used for observation to obtain a turning point where the slope of the light intensity curve changes, and based on the turning point, the moment when the solid dye begins to melt is indicated to generate a melting start signal, including: The time-light intensity curve is smoothed by using a five-point cubic smoothing algorithm to obtain a smooth curve, and the slope of the smooth curve is calculated according to a central difference method combined with a least squares fitting technique to obtain the slope value corresponding to each time point; Performing dimensionality reduction processing on the slope value by using principal component analysis to obtain a dimensionality reduction slope value, inputting the dimensionality reduction slope value into a pre-trained support vector machine model, classifying and identifying the dimensionality reduction slope value, and obtaining an initial turning point where the slope of the light intensity curve changes; Based on the initial turning point, a dynamic time warping algorithm is used to match the time-light intensity curve within a specified time period, and an anomaly detection algorithm is used to monitor the model output to predict and locate the turning point where the slope of the light intensity curve changes; Based on the turning point, a Kalman filter is used to optimize the estimation of the melting start time to generate a melting start signal.

3. The method according to claim 2, characterized in that The slope value is subjected to dimensionality reduction processing by principal component analysis to obtain a dimensionality reduction slope value, including: Optimizing the slope value using a robust regression algorithm to form a high-dimensional slope value matrix, and standardizing the features of the high-dimensional slope value matrix according to a Z-Score standardization method to generate a standardized slope value matrix, wherein the features include a zero mean and a unit variance; According to the standardized slope value matrix, a predefined Mahalanobis distance-based covariance matrix is ​​calculated, and a random singular value decomposition algorithm is used to accelerate the decomposition process of the covariance matrix to obtain eigenvalues ​​and eigenvectors; Based on the eigenvalues ​​and the eigenvectors, a cross-validation technique is used to evaluate the performance of the support vector machine model under different numbers of principal components, and N principal components that can retain maximum information and verify the performance of the support vector machine model are selected. When determining the value of N, the least absolute shrinkage and selection operator regression is used to screen out the principal components that are helpful for the classification task, and a principal component set is generated; The standardized slope value matrix is ​​preliminarily reduced in dimension using a t-distributed random neighbor embedding algorithm to obtain preliminary reduced-dimensionality slope values, and the preliminary reduced-dimensionality slope values ​​are projected onto a new coordinate system defined by the principal component set to generate reduced-dimensionality slope values.

4. The method according to claim 1, characterized in that Based on the data of the time-light intensity curve of the melting start signal within a fixed time, the melting rate is calculated using a Kalman filter, including: The melting start signal is used in combination with an adaptive threshold algorithm to calculate a fixed time when the solid dye starts to melt, and based on the fixed time, a dynamic window selection method is used to automatically adjust the length of the time period according to the changing trend of the melting rate, thereby extracting initial time-light intensity curve data starting from the fixed time; The median filtering method is used to remove noise and outliers from the initial time-light intensity curve data, and the local weighted regression scatter point smoothing technology is introduced to retain the characteristic information in the initial time-light intensity curve data, and the robust statistical method is further applied to automatically detect and remove outliers to obtain smoothed time-light intensity curve data; Based on the smoothed time-light intensity curve data, a five-point cubic smoothing algorithm is used to reduce noise interference, and a least squares fitting technique combined with an adaptive weight adjustment strategy is used to calculate the slope value corresponding to each time point, which is expressed as a preliminary melting rate; The preliminary melting rate is optimized using an unscented Kalman filter to obtain a melting rate.

5. The method according to claim 4, characterized in that The median filtering method is used to remove the noise and outliers of the initial time-light intensity curve data, and the local weighted regression scatter point smoothing technology is introduced to retain the characteristic information in the initial time-light intensity curve data, and the robust statistical method is further applied to automatically detect and remove outliers to obtain smoothed time-light intensity curve data, including: Performing preliminary denoising on the initial time-light intensity curve data by using a median filtering method, reducing the influence of noise by replacing each data point with the median value in the neighborhood, and obtaining preliminary denoised data; Based on the data after preliminary denoising, a local weighted regression scatter point smoothing technique is introduced to smooth the curve by performing weighted least square fitting on local data points, retaining feature information and local change trends, and generating smoothed curve data; According to the smoothed curve data, a robust statistical method is applied to automatically detect and remove outliers to obtain a detection result; Based on the detection result, the data after preliminary denoising and the smoothed curve data, smoothed time-light intensity curve data are integrated.

6. A melting monitoring system for solid coloring agent, characterized in that: include: A generating module, for placing a solid dye to be monitored in a transparent heating container, arranging at least one light source on one side of the transparent heating container, arranging at least one light sensor on the other side opposite thereto, selecting a wavelength combination according to an absorption spectrum of the solid dye, and using a multi-wavelength light source to make a light path pass through a maximum cross section of the solid dye, so as to generate an initial optical path; A denoising module, for recording the light intensity value transmitted through the solid dye and received by the light sensor at preset time intervals to form an initial time-light intensity curve, and performing denoising processing on the data of the initial optical path by a wavelet transform algorithm to obtain a time-light intensity curve; An indication module, for observing, based on the changing trend of the time-light intensity curve, using a preset support vector machine model, obtaining a turning point at which the slope of the light intensity curve changes, and indicating, based on the turning point, the moment when the solid dye begins to melt, so as to generate a melting start signal; An evaluation module, for calculating the melting rate using a Kalman filter based on the data of the time-light intensity curve of the melting start signal at a fixed time, predicting the stability of the solid dye under different environmental conditions in combination with a thermodynamic analysis algorithm, and generating a quality evaluation report; A monitoring module, which is used to automatically trigger a cooling program using a PID controller when it is detected that the melting rate reaches a preset threshold, predict and analyze the uniformity and purity of the solid colorant in combination with deep learning image analysis technology to obtain key parameters, integrate the key parameters and quality assessment results of the solid colorant during the melting process, and generate a melting monitoring report; When it is detected that the melting rate reaches a preset threshold, the cooling program is automatically triggered by the PID controller, and the uniformity and purity of the solid dye are predicted and analyzed by the deep learning image analysis technology to obtain key parameters. The key parameters and quality assessment results of the solid dye during the melting process are integrated to generate a melting monitoring report, including: When the melting rate reaches a preset threshold, the parameter setting of the PID controller is optimized using fuzzy logic, and a cooling program is automatically triggered according to a preset temperature control strategy; During the cooling process, an initial image of the solid dye is obtained using a high-resolution camera, and the initial image is processed based on a multi-scale image fusion technology to obtain an optimized image; Preprocessing the optimized image according to a superpixel segmentation algorithm to obtain a processed image, and using an ensemble learning method to analyze the color distribution, particle size and distribution and other features extracted from the processed image to obtain key parameters including uniformity index, purity level and impurity content of the solid colorant; According to the key parameters, external data in the melting process are collected, and the melting rate and environmental condition changes in a fixed time in the future are predicted based on time series analysis to obtain analysis results, wherein the external data includes melting rate, environmental temperature and humidity; Based on the analysis results, natural language processing technology is used to automatically generate melting monitoring reports; Preprocessing the optimized image according to a superpixel segmentation algorithm to obtain a processed image includes: Using a multi-view data set and a structural similarity index, a weighted average fusion process is performed on the initial image of the solid dye acquired by a high-resolution camera to generate an optimized image; Initializing a superpixel grid according to the optimized image, and iteratively optimizing the boundary of the superpixel grid based on an improved simple linear iterative clustering algorithm combined with a spectral clustering technique to obtain a preliminary segmentation result; Based on the preliminary segmentation result, an active contour model and a graph cut algorithm are applied to refine and smooth the boundaries of the superpixel grid, and shape regularity constraints such as ellipse fitting are introduced to generate an intermediate segmented image; According to the intermediate segmented image, a connected component labeling algorithm is used to remove isolated areas, and bilateral filtering and smoothing operations are applied to perform morphological operations to generate a processed image, wherein the morphological operations include opening operations and closing operations.

7. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a melting monitoring method for solid colorants as described in any one of claims 1 to 5.

8. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a melting monitoring method for a solid colorant as described in any one of claims 1 to 5 is implemented.

Images