A method and system for dynamic monitoring of color steel plate coating thickness based on artificial intelligence

Abstract

The present application provides a method and system for dynamic monitoring of the thickness of color-coated steel plates based on artificial intelligence. The method obtains temperature distribution data on the surface of the color-coated steel plate coating and generates a thermal response map using a pulsed eddy current excitation method. The method then performs temporal correlation processing on the temperature distribution data and the thermal response map to extract heat conduction characteristics. The method then performs synchronous analysis of the heat conduction characteristics, distinguishes the type of color-coated steel plate coating defects and the level of thickness anomalies through a dynamic weight distribution mechanism, and generates a detection result containing the defect location, size, and thickness deviation value. Finally, a dynamic compensation instruction for the coating thickness is generated based on the detection result. The technical solution provided by the present application significantly improves the online control accuracy and production efficiency of the coating quality, forming a closed-loop feedback system from detection to control.

Description

Technical Field

[0001] The present application relates to the technical field of material surface quality detection, and in particular to a method and system for dynamic monitoring of the coating thickness of color-coated steel plates based on artificial intelligence. Background Art

[0002] With the widespread application of color-coated steel sheets in construction, automobiles, and home appliances, the uniformity and quality of their surface coatings directly impact the product's corrosion resistance, mechanical strength, and appearance. In high-speed, continuous production lines, coating thickness must be monitored in real time to address thickness anomalies caused by fluctuations in the spraying process, substrate deformation, or subtle subsurface defects. Furthermore, the detection technology must be resistant to high temperatures and electromagnetic interference, and capable of converting test results into process parameter adjustment instructions within milliseconds to ensure stable and consistent coating quality.

[0003] The industry currently primarily uses infrared thermal imaging, single-frequency eddy current testing, laser interferometry, or ultrasonic pulse echo technology. Infrared thermal imaging indirectly estimates thickness by analyzing the surface temperature distribution, but it struggles to distinguish between temperature anomalies caused by thickness deviations and subsurface defects. Single-frequency eddy current testing measures the thickness of conductive coatings based on the principle of electromagnetic induction, but has poor adaptability to non-conductive coatings or complex electromagnetic environments. Laser interferometry relies on surface topography detection and cannot penetrate the coating to identify internal defects. Ultrasonic testing is limited by detection efficiency and struggles to meet the real-time demands of high-speed production lines. Summary of the Invention

[0004] The present application provides a method and system for dynamic monitoring of the coating thickness of color steel plates based on artificial intelligence, which is used to solve the problems in the prior art of high misjudgment rate due to the single detection dimension, inability to determine the correlation between thickness anomalies and sub-surface defects, and lack of dynamic compensation mechanism.

[0005] In a first aspect, the present application provides a method for dynamic monitoring of coating thickness of color-coated steel plates based on artificial intelligence, comprising:

[0006] Continuously obtain temperature distribution data on the surface of the color-coated steel plate coating by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating;

[0007] Generating a thermal response spectrum of the color steel plate coating based on the principle of electromagnetic induction by pulsed eddy current excitation, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material;

[0008] Performing time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation;

[0009] The heat conduction characteristics are analyzed synchronously, and the color steel plate coating defect type and thickness abnormality level are distinguished through a dynamic weight distribution mechanism, and the detection results including the defect location, size and thickness deviation value are output;

[0010] A coating thickness dynamic compensation instruction is generated according to the detection result.

[0011] Optionally, generating a thermal response spectrum of the color steel plate coating based on the electromagnetic induction principle by pulsed eddy current excitation includes:

[0012] Dynamically setting the initial frequency and pulse width of the pulsed eddy current excitation according to the magnetic permeability and electrical conductivity of the color steel plate coating material, wherein the initial frequency is segmentedly modulated based on the critical value of the electromagnetic coupling between the color steel plate substrate and the coating;

[0013] The color steel plate coating is laterally scanned by a multi-band pulsed eddy current excitation signal, wherein the multi-band pulsed eddy current excitation signal includes attenuation waveforms of at least three different frequency bands, and the amplitude of the attenuation waveform decreases nonlinearly with increasing thickness of the color steel plate coating;

[0014] Collecting the eddy current attenuation component generated by the multi-band eddy current excitation signal when the color steel plate coating is horizontally scanned, performing phase delay compensation on the eddy current attenuation component using an orthogonal decomposition algorithm, and extracting eddy current phase delay parameters related to the thickness distribution of the color steel plate coating;

[0015] The eddy current phase delay parameter and the local temperature gradient change in the temperature distribution data are integrated in the frequency domain to generate a thermal response map that integrates the electromagnetic thermal characteristics. The multi-thermal response map includes the eddy current impedance mutation characteristics of the sub-surface defect area of ​​the color steel plate coating and the time-varying trajectory of the thermal resistance caused by thickness deviation.

[0016] Optionally, collecting the eddy current attenuation component generated internally when the multi-band eddy current excitation signal is transversely scanned on the color steel plate coating, performing phase delay compensation on the eddy current attenuation component using an orthogonal decomposition algorithm, and extracting eddy current phase delay parameters related to the thickness distribution of the color steel plate coating, including:

[0017] Decomposing the eddy current attenuation component of the multi-band pulsed eddy current excitation signal into an in-phase component and an orthogonal component, wherein the in-phase component is related to the electromagnetic properties of the color steel plate substrate, and the orthogonal component is related to the nonlinear change of the coating thickness;

[0018] Based on the attenuation waveforms of different frequency bands in the multi-band pulsed eddy current excitation signal, respectively calculating the phase delay factors of the orthogonal components in the coating thickness direction;

[0019] constructing a coating thickness distribution function according to the phase delay factor and the amplitude attenuation ratio of the in-phase component;

[0020] The thickness distribution function is normalized by a frequency band superposition algorithm to generate an eddy current phase delay parameter, wherein the eddy current phase delay parameter includes a phase mutation gradient caused by thickness deviation of the color steel plate coating and an impedance matching error in the sub-surface defect area.

[0021] Optionally, performing time-series correlation processing on the temperature distribution data and the thermal response map to extract heat conduction characteristics between the coating surface and the subsurface of the color steel plate includes:

[0022] Performing time stamp matching on the local temperature gradient change in the temperature distribution data and the time-varying trajectory of the thermal resistance in the thermal response map to generate a timing alignment parameter, wherein the timing alignment parameter is used to compensate for the signal acquisition time difference between the infrared thermal imager and the pulsed eddy current excitation device;

[0023] Based on the timing alignment parameters, the sampling frequency of the temperature distribution data is synchronously modulated with the pulse excitation frequency of the thermal response spectrum to generate a synchronous timing window, and the temperature gradient change within the synchronous timing window is fused with the eddy current impedance mutation characteristics using a frequency domain integration algorithm to obtain an electromagnetic thermal coupling characteristic vector;

[0024] According to the thermal diffusion coefficient of the material of the color steel plate coating, the heat conduction equation is corrected for the electromagnetic thermal coupling characteristic vector to generate a corrected non-uniform thermal diffusion model;

[0025] Extracting the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion mode, and converting the heat flux density distribution difference into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm;

[0026] The thermal resistance mutation gradient and the phase correlation in the electromagnetic thermal coupling eigenvector are multi-dimensionally fused to generate a heat conduction feature, wherein the heat conduction feature includes the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation.

[0027] Optionally, extracting the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion mode, and converting the heat flux density distribution difference into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm, including:

[0028] Decomposing the non-uniform heat diffusion pattern into a plurality of heat flux density gradient layers along the thickness direction of the color steel plate coating;

[0029] Based on the heat flux density difference between adjacent layers in the heat flux density gradient layer, a thermal resistance distribution matrix in the thickness direction of the color steel plate coating is constructed;

[0030] Calculating the thermal resistance mutation amplitude between the coating surface and the subsurface of the color steel plate according to the spatial gradient distribution of the thermal resistance variation in the thermal resistance distribution matrix, and normalizing the thermal resistance mutation amplitude by the heat flow path distortion caused by the thickness deviation to generate a normalized thermal resistance mutation coefficient;

[0031] The normalized thermal resistance mutation coefficient is dynamically weighted matched with the phase correlation in the electromagnetic thermal coupling eigenvector to generate a thermal resistance mutation gradient corresponding to the thickness anomaly level, wherein the thermal resistance mutation gradient includes the local thermal resistance jump value of the defective area of ​​the color steel plate coating and the cumulative error of the thickness deviation along the production line movement direction.

[0032] Optionally, the heat conduction characteristics are analyzed synchronously, and the color steel plate coating defect type and thickness anomaly level are distinguished through a dynamic weight distribution mechanism, and the detection results including the defect location, size and thickness deviation value are output, including:

[0033] Perform multi-dimensional feature decoupling on the thermal resistance variation caused by the non-uniform heat diffusion mode in the heat conduction feature and the thickness deviation to generate a defect feature vector and a thickness feature vector, wherein the defect feature vector includes the temperature gradient distortion coefficient and the heat flux density mutation amplitude of the defect area, and the thickness feature vector includes the thermal resistance mutation gradient and eddy current phase delay parameter at different locations of the coating;

[0034] According to the magnetic permeability and electrical conductivity of the color steel plate coating material, the defect thickness correlation matrix is ​​constructed;

[0035] Performing nonlinear superposition processing on the defect feature vector and the thickness feature vector to generate a comprehensive feature vector;

[0036] In the multidimensional feature space, based on the prior distribution data of the coating defect type and the thickness anomaly level, the confidence parameter of the defect area and the cumulative probability value of the thickness deviation are calculated;

[0037] According to the joint constraints of the confidence parameter and the cumulative probability value, a detection result including the defect position coordinates, the defect coverage area and the thickness deviation compensation value is generated, wherein the detection result is synchronized in real time with the motion trajectory of the coating equipment of the color steel plate production line.

[0038] Optionally, generating a coating thickness dynamic compensation instruction according to the detection result includes:

[0039] Calculating a spray flow rate compensation amount and a roller pressure compensation amount of the coating material based on the thickness deviation value and the defect coverage area in the detection result;

[0040] According to the real-time motion trajectory of the color steel plate coating equipment, the spray flow compensation amount and the roller pressure compensation amount are temporally and spatially matched to generate dynamic compensation parameters synchronized with the position coordinates of the coating equipment, wherein the dynamic compensation parameters include the spray flow correction coefficient and the roller pressure gradient adjustment amount;

[0041] Performing delay compensation correction on the dynamic compensation parameter by using the timestamp difference between the defect position coordinates and the preset production line motion speed parameter to generate a delay compensation correction coefficient;

[0042] The delay compensation correction coefficient is superimposed and integrated with the dynamic compensation parameter to generate a coating thickness dynamic compensation instruction.

[0043] In a second aspect, the present application provides an artificial intelligence-based dynamic monitoring system for color steel plate coating thickness, comprising:

[0044] An acquisition module is used to continuously acquire temperature distribution data of the coating surface of the color steel plate by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating;

[0045] A first generating module is configured to generate a thermal response spectrum of the color steel plate coating based on the principle of electromagnetic induction by means of a pulsed eddy current excitation method, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material;

[0046] An extraction module is used to perform time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation;

[0047] An analysis module is used to synchronously analyze the heat conduction characteristics, distinguish the color steel plate coating defect type and thickness anomaly level through a dynamic weight distribution mechanism, and output a detection result including the defect location, size and thickness deviation value;

[0048] The second generating module is used to generate a coating thickness dynamic compensation instruction according to the detection result.

[0049] In a third aspect, an embodiment of the present application provides a computing device comprising 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 an artificial intelligence-based dynamic monitoring method for the coating thickness of color steel plates 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. When the computer program is executed by a computer, it implements a dynamic monitoring method for the thickness of the color steel plate coating based on artificial intelligence as described in the first aspect.

[0051] The embodiment method of the present application uses multimodal data fusion of infrared thermal imaging and pulsed eddy current excitation to synchronously collect temperature distribution data of the coating surface of the color steel plate and the electromagnetic response characteristics of the sub-surface, and uses time series correlation processing to extract heat conduction characteristics, decoupling the coating thickness deviation from the thermal-electromagnetic coupling effect of the sub-surface defect, thereby solving the misjudgment problem caused by the data dimension fragmentation of a single physical field detection technology (such as infrared or eddy current). At the same time, a dynamic weight allocation mechanism is used to achieve accurate classification of defect types and thickness anomalies, providing high-confidence detection results for real-time dynamic compensation.

[0052] Furthermore, through the orthogonal decomposition algorithm and frequency domain integration processing, the correlation between the eddy current attenuation component inside the coating and the surface temperature gradient is quantified, and the signal contrast in the sub-surface defect area is enhanced, which solves the problem of insufficient sensitivity of traditional single-frequency eddy current detection to thin coatings and tiny defects. At the same time, through the extraction of phase mutation gradient and impedance matching error, a quantifiable electromagnetic-thermal coupling characteristic input is provided for the dynamic weight allocation mechanism.

[0053] In summary, by combining the macroscopic temperature field analysis of infrared thermal imaging with the microscopic electromagnetic response detection of pulsed eddy current excitation, under the collaborative mechanism of time synchronization, frequency domain fusion and dynamic parameter correction, the synchronous detection and correlation judgment of color steel plate coating thickness anomalies and sub-surface defects are achieved, breaking through the detection bottleneck of traditional methods in complex industrial environments (such as high temperature and strong electromagnetic interference). At the same time, the detection results are converted into spray flow and roller pressure compensation instructions of the coating material in real time, forming a closed-loop feedback system from detection to control, which significantly improves the online control accuracy and production efficiency of coating quality.

[0054] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0056] Figure 1 The flowchart of the method for dynamic monitoring of the coating thickness of color-coated steel plates based on artificial intelligence provided by the present application is shown;

[0057] Figure 2 The present invention provides a schematic structural diagram of a color steel plate coating thickness dynamic monitoring system based on artificial intelligence;

[0058] Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0060] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but 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 between different operations, and the serial numbers themselves do not represent any order of execution. 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 of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0061] The researchers found that the existing color steel plate coating thickness detection technology has a high misjudgment rate due to its reliance on single physical field data (such as infrared or eddy current only), and is unable to synchronously correlate the coupling effects of thickness anomalies and sub-surface defects. At the same time, it lacks a real-time dynamic compensation mechanism to adjust the production line process parameters. Based on this, an artificial intelligence-based dynamic monitoring method for color steel plate coating thickness is provided. This method establishes a thermal-electromagnetic coupling feature analysis model by fusing multi-modal data of infrared thermal imaging and pulsed eddy current excitation, realizes accurate decoupling judgment of thickness deviation and defects, and adjusts the coating equipment parameters in real time through dynamic compensation instructions.

[0062] The technical solution of the present application can be applied to online monitoring of coating thickness and process closed-loop control scenarios in high-speed continuous color steel plate production lines, especially to complex working conditions where there are sub-surface micro defects or sudden changes in coating material properties.

[0063] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0064] Figure 1The present invention provides a flowchart of a method for dynamic monitoring of the coating thickness of a color steel plate based on artificial intelligence, as shown in FIG. Figure 1 As shown, the method includes:

[0065] Step 101: continuously acquiring temperature distribution data of the coating surface of the color steel plate by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating;

[0066] In this step, infrared thermal imaging refers to the technology of using an infrared thermal imager to convert the invisible infrared radiation on the surface of an object into a visual temperature distribution image. Its principle is based on the differences in thermal conductivity characteristics of different materials or structures. The temperature distribution data is a dynamic record of the temperature values ​​at various positions on the coating surface, which can reflect the thermal field anomalies caused by sub-surface defects. The local temperature gradient change refers to the area of ​​sudden change in surface temperature distribution caused by sub-surface defects (such as bubbles and debonding) that hinder or accelerate heat conduction. The gradient value is calculated through the temperature difference of adjacent areas and is used to locate the defect boundary.

[0067] In this embodiment, first, an infrared thermal imager is deployed in the cooling section of the continuous production line of color-coated steel plates, and its scanning range is adjusted to synchronize with the moving speed of the production line to ensure complete coverage of the coating surface and real-time temperature data collection. Secondly, the thermal imager is started to continuously scan the moving coating, record the changes in surface temperature over time, and generate a dynamic thermal field data set containing horizontal, vertical and time dimensions. Then, the raw data is preprocessed: the environmental interference noise is eliminated through a filtering algorithm, and the temperature change signal caused by the heat conduction of the coating itself is retained. Finally, the spatial differential algorithm is used to calculate the temperature gradient between adjacent areas. When the gradient value exceeds the preset threshold, it is marked as a potential defect area, and the coordinates are associated with the gradient features and stored for subsequent detection steps.

[0068] For example, in a continuous coating production line for color-coated steel sheets, heat conduction is hindered in a certain section of the coating due to the presence of tiny bubbles within it. During the cooling phase, the infrared thermal imager detected a small, abnormally high temperature on the surface of this area, with the temperature distribution showing a significant gradient change at its boundaries. Through gradient analysis, the system identified that the gradient value in this area was significantly higher than the normal range, and the gradient direction exhibited a circular distribution characteristic. Combined with electromagnetic detection data from subsequent steps (abnormal eddy current response in this area), the system determined this to be a sub-surface bubble defect and triggered automatic adjustment of the spraying process parameters to restore coating uniformity. This case demonstrates the real-time defect capture capability of infrared thermal imaging and its closed-loop linkage with production control.

[0069] Step 102: generating a thermal response spectrum of the color steel plate coating by pulsed eddy current excitation based on the principle of electromagnetic induction, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material;

[0070] In this step, the pulsed eddy current excitation method refers to the technology of exciting an alternating magnetic field on the coating surface through periodic pulse current, generating eddy currents inside the material based on the principle of electromagnetic induction, and analyzing the electromagnetic parameters of the coating by detecting the eddy current attenuation characteristics; the thermal response spectrum is a composite data set that integrates the temperature distribution and electromagnetic impedance changes caused by the eddy current thermal effect, and is used to characterize the electromagnetic-thermal coupling characteristics of defects inside the coating; the magnetic permeability characterizes the magnetization ability of the material in the magnetic field and affects the distribution of the eddy current magnetic field; the electrical conductivity reflects the conductive properties of the material and determines the penetration depth and attenuation rate of the eddy current in the coating.

[0071] In this embodiment, an adaptive algorithm dynamically matches the initial frequency and pulse width of pulsed eddy current excitation based on the material properties (magnetic permeability and electrical conductivity) of the color-coated steel plate coating, ensuring the excitation signal's sensitivity to surface and subsurface defects. Secondly, an eddy current probe array is deployed downstream of the coating station on the production line to generate a multi-band pulse excitation signal, which scans the coating laterally: high-frequency signals focus on minor surface defects, while low-frequency signals penetrate the substrate to detect deeper anomalies. Next, the eddy current decay signal within the coating is collected. An orthogonal decomposition algorithm is used to separate the in-phase component (reflecting the substrate's magnetic permeability) from the quadrature component (reflecting the coating's electrical conductivity and thickness), and phase delay is dynamically compensated. Finally, the compensated eddy current phase parameters are combined with temperature gradient data for multi-physics fusion to generate a thermal response map. Regions of sudden electromagnetic impedance changes overlapping with abnormal temperature gradients are marked as high-confidence defects.

[0072] For example, in the color steel plate production line of the above-mentioned automobile manufacturing plant, step 101 detects the presence of an annular high-temperature gradient anomaly in a certain area. Step 102 then starts pulsed eddy current detection: based on the magnetic permeability and electrical conductivity of the coating material, the system automatically adjusts the excitation frequency to the mid-frequency band to match the coating thickness. The scan found that the amplitude of the orthogonal component of the eddy current in this area dropped significantly, and the phase delay exceeded the normal range. Combined with the temperature gradient data of step 101, it was determined to be a composite defect of debonding of sub-surface bubbles and substrates. Based on the detection results, the system simultaneously optimizes the spray flow rate and roller pressure parameters to repair coating defects and improve substrate bonding strength, ultimately ensuring that the coating thickness and adhesion meet process standards.

[0073] Step 103: performing time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation;

[0074] In this step, time series correlation processing refers to the time synchronization alignment and spatial coordinate matching of the temperature distribution data of step 101 and the thermal response map of step 102, and the analysis of the coupling relationship between the two as they evolve over time; the heat conduction characteristics are analyzed by fusing electromagnetic and thermal data to determine the internal heat transfer law of the coating, including the distortion of the heat diffusion path (non-uniform pattern) caused by structural anomalies (such as bubbles and cracks) in defective areas and the quantitative changes in thermal resistance caused by fluctuations in coating thickness.

[0075] In this embodiment, the temperature distribution data collected by the infrared thermal imager in step 101 is first matched with the thermal response map generated by the eddy current probe in step 102 by timestamp and spatial coordinates, ensuring that the two types of data for the same coating area are fully synchronized in terms of time sequence. Secondly, a multi-physics field coupling algorithm is used to analyze the interaction between the temperature field and the electromagnetic field: the rate of temperature drop over time (reflecting the thermal diffusion capacity of the coating) and the eddy current phase delay (reflecting the coating thickness and the bonding state with the substrate) are extracted to establish a correlation model between thermal conduction and electromagnetic impedance. Next, areas of abnormal thermal conduction are identified: if the temperature drop rate in a certain area is significantly lower than that of the surrounding area (non-uniform diffusion) and the eddy current phase delay exceeds the normal range (abnormal thermal resistance), it is determined to be a defective area, and the change in thermal resistance is calculated to quantify the severity of the defect. Finally, the thermal conduction characteristic data is mapped to the production line process parameter library to generate compensation instructions such as spray flow rate and roller pressure.

[0076] For example, in the color-coated steel plate production line of an automobile manufacturing plant, step 101 detects an abnormal temperature gradient in a certain area, and step 102 verifies the presence of an eddy current phase delay in this area. Step 103, through time-series correlation analysis, finds that the temperature drop rate in this area is slower than that in normal areas, and the thermal resistance value increases due to insufficient coating thickness. The system determines that this is a combined defect of coating thickness deviation and sub-surface bubbles, and automatically triggers a compensation mechanism: the spray station increases the paint flow in this area, and the roller pressing station increases the local pressure to restore the coating thickness to uniformity and eliminate bubbles. This case verifies the ability of multi-physics field data fusion to accurately diagnose complex defects and the effectiveness of process closed-loop control.

[0077] Step 104: synchronously analyzing the heat conduction characteristics, distinguishing the color steel plate coating defect type and thickness anomaly level through a dynamic weight allocation mechanism, and outputting a detection result including the defect location, size, and thickness deviation value;

[0078] In this step, synchronous analysis refers to the multi-dimensional cross-validation of the heat conduction characteristics extracted in step 103 (such as the non-uniform heat diffusion pattern and the thermal resistance change), and a comprehensive judgment is made based on the temporal and spatial evolution of the defects. The dynamic weight allocation mechanism dynamically adjusts the contribution weight of each characteristic parameter based on the sensitivity of different defect types (such as bubbles, cracks, and debonding) and thickness anomalies to the heat conduction characteristics to optimize classification accuracy. The thickness anomaly level is based on the deviation range of the quantified thermal resistance change from the standard thickness, and the coating defects are divided into mild, moderate, and severe levels, providing a priority basis for process compensation.

[0079] In this embodiment, a defect classification model is first constructed based on a thermal conductivity characteristic database (including historical defect data and process parameters). Differentiated weight coefficients are assigned to typical defects such as bubbles (localized sudden increases in thermal resistance), cracks (linear distribution of thermal resistance), and insufficient thickness (global increase in thermal resistance). Secondly, the degree of match between the thermal conductivity characteristics of the current coating area and the standard model is calculated in real time. If a region exhibits both non-uniform thermal diffusion (high weight ratio) and a significant increase in thermal resistance (medium weight ratio), it is determined to be a bubble defect; if the thermal resistance increases uniformly but there are no localized diffusion anomalies (high weight ratio), it is determined to be insufficient thickness. Next, the thickness anomaly is classified according to the thermal resistance deviation: a deviation below a threshold is classified as a minor level (only a warning), while a deviation exceeding the threshold is classified as a severe level (triggering a shutdown for repairs). Finally, the defect location, type, size, and thickness deviation value are integrated into a structured inspection report, which is simultaneously pushed to the production line control system and quality inspection platform.

[0080] For example, in a color-coated steel plate production line at an automobile manufacturer, step 103 detected that the thermal resistance value in a certain area increased compared to the standard value and exhibited a non-uniform diffusion pattern. Step 104 analyzed the dynamic weight distribution: the non-uniform diffusion weight accounted for a high proportion, and the thermal resistance increment weight accounted for a medium proportion, indicating a bubble defect (moderate level). Meanwhile, another area on the same production line showed a uniform increase in thermal resistance but no diffusion anomalies, indicating insufficient thickness (severe level). Based on this, the system outputs the detection results: the bubble defect location (coordinates XY), diameter range, and thickness deviation percentage. It also sends flow compensation instructions to the spray station to focus on repairing areas with insufficient thickness. This case demonstrates the dynamic weighting mechanism's ability to accurately classify and grade complex defects.

[0081] Step 105, generating a coating thickness dynamic compensation instruction according to the detection result;

[0082] In this step, dynamic compensation instructions refer to real-time control instructions generated by the process parameter mapping model based on the coating defect type (such as bubbles, insufficient thickness) and abnormality level. They are used to adjust the operating parameters of process equipment such as spraying and rolling (such as paint flow and rolling pressure) to eliminate coating thickness deviations and repair defects to ensure that the coating uniformity meets the process standards.

[0083] In this embodiment, the central control system first receives the defect location, type, and thickness deviation value output from step 104 and calls the compensation rule library preset in the process parameter library: for bubble defects, the spray flow rate increment ratio and roller pressure correction coefficient are matched; for insufficient thickness defects, the spray speed adjustment amplitude and roller pressure strength compensation value are matched. Secondly, the execution position of the spray robot arm and the hydraulic roller press is located according to the defect coordinates, and the compensation parameters are distributed to the corresponding equipment via the industrial bus protocol. Next, the spray robot arm increases the paint spray volume in the defect area according to the compensation value, and the hydraulic roller press simultaneously increases the local pressure to compact the coating. Finally, the re-inspection process (steps 101-104) is triggered to verify the repair effect. If the test results meet the standards, production continues; otherwise, the compensation parameters are iteratively optimized until the defect is eliminated.

[0084] For example, in a color-coated steel production line at an automobile manufacturer, step 104 detected a severe thickness deficiency in a certain area (the highest deviation level). Based on the compensation rule base, the system generated instructions: the spray robot increased the paint spray volume in that area to 120% of the standard value, and the hydraulic roller press simultaneously increased the local pressure to 130% of the standard value. After compensation, re-inspection showed that the coating thickness in that area returned to within process tolerances, and the thermal conductivity characteristics and electromagnetic response data returned to normal. The system recorded the compensation parameters and updated the process knowledge base to optimize the adaptive control capabilities of subsequent production lines.

[0085] Since the traditional single-frequency eddy current testing method has poor adaptability to coating thickness changes and material property fluctuations due to fixed electromagnetic parameters, it is difficult to extract high-contrast defect signals in thin layers or non-uniform coatings. To solve the above problem, in some embodiments, according to step 102, a pulsed eddy current excitation method is used to generate a thermal response spectrum of the color-coated steel plate coating based on the electromagnetic induction principle, including:

[0086] Step 201, dynamically setting the initial frequency and pulse width of pulsed eddy current excitation according to the magnetic permeability and electrical conductivity of the color steel plate coating material, wherein the initial frequency is segmentedly modulated based on the electromagnetic coupling threshold between the color steel plate substrate and the coating;

[0087] In this step, magnetic permeability refers to the ability of the material to be magnetized in a magnetic field, which determines the penetration depth of the eddy current magnetic field in the coating; electrical conductivity refers to the physical quantity of the material's conductive properties, which affects the distribution and attenuation rate of eddy currents; the electromagnetic coupling threshold refers to the threshold of electromagnetic field interaction at the interface between the substrate and the coating, which determines whether the eddy current signal can penetrate the coating to reach the substrate.

[0088] In this embodiment, the magnetic permeability and electrical conductivity parameters of the color-coated steel plate coating are first obtained from a material database. Combined with the electromagnetic properties of the substrate, an adaptive temperature algorithm is used to calculate the electromagnetic coupling threshold and determine the initial frequency segmentation range (e.g., low frequency penetrates the substrate, medium frequency matches the coating thickness, and high frequency focuses on the surface). Secondly, the frequency band is dynamically divided based on the critical value: when the coating thickness is thin, the high frequency band is selected to enhance surface defect sensitivity; when the thickness increases, the low frequency band is switched to penetrate the substrate. Next, based on the nonlinear relationship between electrical conductivity and magnetic permeability, the pulse width is optimized using a gradient descent algorithm to ensure that the energy distribution of the excitation signal matches the thermal response characteristics of the coating. Finally, the initial frequency and pulse width parameters are transmitted to the eddy current probe array, completing the dynamic parameter adaptation.

[0089] Step 202: performing a transverse scan on the color steel plate coating using a multi-band pulsed eddy current excitation signal, wherein the multi-band pulsed eddy current excitation signal includes attenuation waveforms in at least three different frequency bands, and the amplitude of the attenuation waveform decreases nonlinearly with increasing thickness of the color steel plate coating;

[0090] In this step, the multi-band pulsed eddy current excitation signal refers to a pulse current signal containing multiple frequency components, which covers the detection depth from the coating surface to the substrate through different frequency bands; the attenuation waveform refers to the signal amplitude attenuation characteristics caused by material impedance when the eddy current propagates in the coating, and its attenuation rate is related to the coating thickness and electromagnetic parameters.

[0091] In this embodiment, first, a three-segment pulse excitation signal including low frequency, medium frequency and high frequency is generated by an eddy current probe array, wherein the low frequency signal is used to detect the bonding state of the substrate, the medium frequency signal matches the coating thickness, and the high frequency signal captures tiny surface defects. Secondly, during the lateral scanning process, the eddy current signal amplitude attenuation rate is monitored in real time: when the coating thickness increases, the high frequency signal amplitude decays rapidly, and the low frequency signal maintains high penetration. Then, a mapping model of amplitude attenuation and coating thickness is established through a nonlinear fitting algorithm, and the probe spacing and scanning speed are dynamically adjusted to ensure the signal-to-noise ratio of different thickness areas. Finally, the scanning data is classified and stored by frequency band for subsequent phase analysis.

[0092] Step 203: collecting the eddy current attenuation component generated by the multi-band eddy current excitation signal when the color steel plate coating is horizontally scanned, performing phase delay compensation on the eddy current attenuation component using an orthogonal decomposition algorithm, and extracting the eddy current phase delay parameter related to the thickness distribution of the color steel plate coating;

[0093] In this step, the eddy current attenuation component refers to the signal amplitude and phase change components caused by material impedance when the eddy current propagates in the coating; the orthogonal decomposition algorithm refers to the mathematical method of decomposing the eddy current signal into an in-phase component (reflecting magnetic permeability) and an orthogonal component (reflecting electrical conductivity and thickness).

[0094] In this embodiment, a high-speed data acquisition card is first used to capture the multi-band eddy current decay signal, extracting its amplitude, phase, and time-domain waveform. Next, an orthogonal decomposition algorithm is used to decouple the signal: the in-phase component aligns with the excitation signal and is used to analyze the substrate's magnetic permeability; the quadrature component lags the excitation signal and is used to quantify the coating's conductivity and thickness. Next, a phase compensation algorithm is used to eliminate the inherent delay error of the probe coil's inductance and capacitance, while retaining the phase lag caused by the coating itself. Finally, a calibration curve comparing phase delay and coating thickness is used to generate a thickness distribution map, marking abnormal areas.

[0095] Step 204: performing frequency domain integration processing on the eddy current phase delay parameter and the local temperature gradient change in the temperature distribution data to generate a thermal response map integrating electromagnetic thermal characteristics, wherein the thermal response map includes the eddy current impedance mutation characteristics of the subsurface defect area of ​​the color steel plate coating and the time-varying trajectory of the thermal resistance caused by the thickness deviation;

[0096] In this step, frequency domain integration processing refers to the fusion analysis of electromagnetic data and thermal field data in the frequency domain to extract the coupling characteristics of the two; the time-varying trajectory of thermal resistance refers to the quantitative curve of the change of thermal resistance over time caused by coating thickness deviation.

[0097] In this embodiment, the eddy current phase parameters from steps 201-203 are first spatiotemporally aligned with the temperature gradient data from step 101 to ensure consistency between data coordinates and timestamps. Next, the two types of data are converted to the frequency domain using a fast Fourier transform (FFT). A frequency-domain integration algorithm is then used to extract electromagnetic-thermal coupling features: the frequency band of eddy current impedance mutations (reflecting defects) and the frequency band of thermal resistance variation (reflecting thickness deviation). Next, a convolutional neural network (CNN) is used to identify abnormal patterns in the coupling features. For example, the overlapping region of high-frequency eddy current impedance mutations and medium-frequency thermal resistance jumps is identified as a subsurface bubble defect. Finally, a three-dimensional thermal response map is generated, integrating the electromagnetic impedance and thermal resistance parameters, and annotating the defect location, type, and thickness deviation level.

[0098] The following is a specific embodiment:

[0099] In the continuous production line for color-coated steel sheets at an automobile manufacturer, the coating material is a galvanized aluminum alloy with a designed medium thickness. The system first dynamically sets the initial frequency of the pulsed eddy current excitation to a low frequency range (for penetrating the substrate to detect bonding), a medium frequency range (for matching coating thickness), and a high frequency range (for capturing small surface defects) based on the coating material's magnetic permeability (medium magnetization) and electrical conductivity (high conductivity). The pulse width is then optimized based on the electromagnetic coupling threshold to match the coating's thermal response characteristics. Subsequently, a multi-band eddy current probe array scans the coating laterally. The high-frequency signal detects a region of nonlinear amplitude attenuation. The medium-frequency signal mapping indicates that the coating thickness in this area is below the design standard, and the low-frequency signal verifies that there is no debonding of the substrate. After acquiring the eddy current attenuation signal, an orthogonal decomposition algorithm isolates the phase delay component. Combined with the local temperature gradient anomaly (significant increase in thermal resistance) detected in the infrared thermal imaging data, a thermal response map is generated through frequency-domain integration, integrating the electromagnetic and thermal characteristics. This region is marked as a combined defect of insufficient thickness and subsurface microcracks. The system then triggered a dynamic compensation command to adjust the spray flow and roller pressure. Re-inspection after repair showed that the coating thickness and thermal conductivity characteristics were restored to the process standard range.

[0100] This embodiment realizes full-dimensional defect detection from the coating surface to the substrate of the color steel plate by dynamically setting the multi-band pulse eddy current excitation parameters and combining electromagnetic-thermal data fusion analysis. The segmented modulation and lateral scanning of the electromagnetic excitation signal cover the sensitive frequency bands of defects of different depths. The orthogonal decomposition and phase compensation improve the thickness detection accuracy; the frequency domain integral processing associates the eddy current impedance mutation with the temperature gradient change, effectively distinguishing complex defects such as bubbles, cracks, and thickness deviations. The thermal response map finally generated and the dynamic compensation instructions form a closed-loop control, which realizes real-time defect positioning, accurate classification and adaptive adjustment of process parameters in high-speed continuous production lines, significantly improving the uniformity of coating quality and production yield, while reducing material waste and equipment downtime losses.

[0101] Since the attenuation component generated by the multi-band eddy current excitation signal inside the coating causes the thickness distribution characteristics to be blurred due to phase delay, the traditional decomposition method cannot accurately quantify the coupling effect of thickness deviation and defects. Therefore, in order to improve the phase delay compensation accuracy, as another embodiment, according to step 203, the eddy current attenuation component generated inside the multi-band eddy current excitation signal when the color steel plate coating is horizontally scanned is collected, and the orthogonal decomposition algorithm is used to perform phase delay compensation on the eddy current attenuation component, and the eddy current phase delay parameters related to the thickness distribution of the color steel plate coating are extracted, including:

[0102] Step 301: decompose the eddy current attenuation component of the multi-band pulsed eddy current excitation signal into an in-phase component and a quadrature component, wherein the in-phase component is related to the electromagnetic properties of the color steel plate substrate, and the quadrature component is related to the nonlinear change of the coating thickness;

[0103] In this step, the in-phase component refers to the component in the eddy current signal that is consistent with the phase of the excitation current, which is used to reflect the magnetic permeability and electromagnetic coupling state of the substrate; the orthogonal component refers to the component in the eddy current signal that lags behind the excitation current in phase, which is used to characterize the eddy current impedance difference caused by the change in coating thickness.

[0104] In this embodiment, a high-speed data acquisition card is used to capture multi-band eddy current decay signals. A quadrature demodulator then synchronously demodulates the signals to separate the in-phase component (reflecting the substrate's magnetic permeability) and the quadrature component (reflecting the coating's conductivity and thickness). A digital filter is then used to eliminate high-frequency noise interference, preserving the signal characteristics of the effective frequency band. Finally, the decomposed in-phase and quadrature components are stored by frequency band for subsequent phase delay analysis.

[0105] Step 302 , calculating the phase delay factors of the orthogonal components in the coating thickness direction based on the attenuation waveforms of different frequency bands in the multi-band pulsed eddy current excitation signal;

[0106] In this step, the phase delay factor refers to the phase lag of the orthogonal component relative to the excitation signal. Its value changes nonlinearly with the increase of coating thickness and is used to quantify the electromagnetic response difference caused by thickness deviation.

[0107] In this example, time-frequency analysis is first performed on the attenuation waveform for each frequency band to extract the phase lag of the orthogonal components. Next, a nonlinear mapping relationship between the phase lag factor and coating thickness is established using a calibration curve (based on phase lag data from samples of known thickness). Next, an interpolation algorithm is used to compensate for phase errors caused by probe spacing and scanning speed, ensuring comparability of data at different locations. Finally, a phase lag factor matrix corresponding to each frequency band is generated and used as an input parameter for thickness distribution calculation.

[0108] Step 303: constructing a coating thickness distribution function according to the phase delay factor and the amplitude attenuation ratio of the in-phase component;

[0109] In this step, the coating thickness distribution function is a mathematical model established by fusing the phase delay factor (reflecting the thickness change) and the in-phase component amplitude attenuation ratio (reflecting the electromagnetic properties of the substrate) to describe the spatial distribution of the coating thickness and abnormal areas.

[0110] In this example, the phase retardation factor and the amplitude attenuation ratio of the in-phase component are first aligned according to spatial coordinates to form a multidimensional dataset. Next, a multivariate regression algorithm is used to construct a thickness distribution function, using the phase retardation factor as the primary variable and the amplitude attenuation ratio as a correction term to eliminate the influence of fluctuations in the substrate's electromagnetic properties on the thickness calculation. Next, a nonlinear optimization algorithm is used to iteratively solve the function parameters to ensure the model's adaptability to varying thickness deviations. Finally, a 3D visualization of the thickness distribution function is output, highlighting abnormal areas.

[0111] Step 304: normalize the thickness distribution function using a frequency band superposition algorithm to generate an eddy current phase delay parameter, wherein the eddy current phase delay parameter includes a phase jump gradient caused by thickness deviation of the color steel plate coating and an impedance matching error in the sub-surface defect area;

[0112] In this step, the frequency band superposition algorithm refers to the weighted fusion of thickness distribution functions of different frequency bands to eliminate the random errors of single frequency band detection; the phase mutation gradient refers to the spatial change rate of the phase delay factor in the thickness deviation area, which is used to locate the defect boundary; the impedance matching error refers to the deviation of the eddy current impedance caused by sub-surface defects from the theoretical value, which is used to distinguish the defect type.

[0113] In this embodiment, dynamic weights are first assigned to the low-frequency, mid-frequency, and high-frequency thickness distribution functions (higher weights for low frequencies enhance substrate penetration, and lower weights for high frequencies reduce noise interference). Next, a normalized thickness distribution map is generated using a frequency band superposition algorithm, and the phase jump gradient is extracted (a defect boundary is marked when the gradient exceeds a threshold). Next, the impedance matching error in the defect region is calculated: if the error and the phase jump gradient increase synchronously, it is determined to be a thickness deviation; if only the error increases significantly, it is determined to be a subsurface defect. Finally, the eddy current phase delay parameter, which includes the phase jump gradient and the impedance error, is generated and fed to the data fusion module.

[0114] The following is a specific embodiment:

[0115] In the continuous production line of color-coated steel plates, the coating is designed to have a medium thickness. The system first performs an orthogonal decomposition on the multi-band eddy current signal to separate the in-phase component (normal magnetic permeability of the substrate) and the orthogonal component (abnormal phase delay in a certain area). The phase delay factor of the area is calculated, and combined with the amplitude attenuation ratio of the in-phase component, a thickness distribution function is constructed to show that the thickness of the area is lower than the standard value. After normalization processing using the frequency band superposition algorithm, the eddy current phase delay parameters are generated: the phase mutation gradient is significant and the impedance matching error increases synchronously, which is determined to be a defect of insufficient coating thickness. Based on this, the system triggers the spray flow compensation instruction. After repair, re-inspection shows that the thickness has returned to the standard range, and the phase parameters and impedance errors have returned to normal.

[0116] This embodiment achieves high-precision detection and defect classification of color-coated steel plate coating thickness through orthogonal decomposition and frequency band superposition algorithms. The decomposition of multi-band eddy current signals and the calculation of phase delay factors effectively distinguish the influence of substrate electromagnetic properties and coating thickness variations, while normalization eliminates random errors. The generated eddy current phase delay parameter incorporates the phase gradient and impedance matching error, accurately locating thickness deviations and sub-surface defects, providing a quantitative basis for dynamic compensation and significantly improving the reliability of coating quality control and the response speed of production line process adjustments.

[0117] In order to solve the problem of loss of correlation between surface temperature gradient and subsurface thermal resistance time-varying trajectory caused by the time-series asynchrony between infrared thermal imaging and eddy current testing, and the difficulty of extracting heat conduction characteristics by traditional data fusion methods, in some embodiments, according to step 103, the temperature distribution data and the thermal response map are subjected to time-series correlation processing to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, including:

[0118] Step 401: performing time stamp matching on the local temperature gradient change in the temperature distribution data and the time-varying trajectory of the thermal resistance in the thermal response map to generate a timing alignment parameter, wherein the timing alignment parameter is used to compensate for the signal acquisition time difference between the infrared thermal imager and the pulsed eddy current excitation device;

[0119] In this step, the timing alignment parameters are compensation coefficients generated by matching the data acquisition timestamps of the infrared thermal imager and the eddy current probe, which are used to eliminate data timing misalignment caused by device response delay or sampling frequency difference.

[0120] In this embodiment, a high-precision clock synchronization module first records the raw data acquisition timestamps of the infrared thermal imager and eddy current probe, extracting the time offset between them. Next, an interpolation algorithm is used to resample the temperature gradient and thermal resistance time-varying trajectories on the time axis, ensuring a one-to-one correspondence between the sampling points of the two data types. Next, a least-squares method is used to fit the functional relationship between the time offset and spatial position to generate a dynamic time alignment parameter table. Finally, this parameter table is used to perform real-time compensation on subsequent data streams, ensuring the spatiotemporal consistency of electromagnetic and thermal data.

[0121] Step 402: Based on the timing alignment parameters, the sampling frequency of the temperature distribution data is synchronously modulated with the pulse excitation frequency of the thermal response spectrum to generate a synchronous timing window, and the temperature gradient change within the synchronous timing window is fused with the eddy current impedance mutation feature using a frequency domain integration algorithm to obtain an electromagnetic thermal coupling feature vector.

[0122] In this step, the synchronized timing window is a fixed-length data segment divided by the aligned time axis, which is used to constrain the fusion range of electromagnetic and thermal data; the frequency domain integration algorithm is a method of converting the time domain signal to the frequency domain and extracting the electromagnetic-thermal coupling characteristics through integration operations.

[0123] In this embodiment, the sampling frequency of the infrared thermal imager is first adjusted according to the timing alignment parameters to ensure strict synchronization with the repetition frequency of the eddy current excitation pulse. Secondly, the continuous data stream is segmented into synchronized timing windows of fixed duration (e.g., 100ms), each containing complete temperature gradient and eddy current impedance data. Next, the time-domain signal is converted to the frequency domain using a fast Fourier transform (FFT). The temperature gradient spectrum and eddy current impedance spectrum are integrated in the frequency domain to extract the energy coupling characteristics of the two in the same frequency band. Finally, the integration results are superimposed according to the frequency band weights to generate an electromagnetic thermal coupling feature vector containing temperature-eddy current correlation information.

[0124] Step 403, performing a heat conduction equation correction on the electromagnetic thermal coupling eigenvector according to the thermal diffusion coefficient of the color steel plate coating material to generate a corrected non-uniform thermal diffusion model;

[0125] In this step, the correction of the heat conduction equation is a mathematical optimization process that physically constrains the electromagnetic thermal coupling characteristics by introducing the material thermal diffusion coefficient (a physical quantity that characterizes the thermal conductivity of the material); the non-uniform heat diffusion pattern is a quantitative model that reflects the abnormal heat conduction path in the defective area after correction.

[0126] In this example, the thermal diffusivity of the color-coated steel plate coating is first retrieved from the material database to construct a three-dimensional partial differential equation for heat conduction. Secondly, the electromagnetic thermal coupling eigenvector is substituted into the equation as a boundary condition, and the theoretical heat diffusion path is calculated using the finite element analysis method (FEM). Next, the deviation between the theoretical and measured values ​​is compared. If the heat flux density in a certain area deviates significantly from the theoretical model (for example, due to an increase in thermal resistance caused by bubbles), it is marked as a non-uniform diffusion area. Finally, an iterative optimization algorithm is used to correct the abnormal components in the eigenvector, generating a non-uniform heat diffusion pattern map that includes defective heat conduction characteristics.

[0127] Step 404: extracting the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion mode, and converting the heat flux density distribution difference into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm;

[0128] In this step, the difference in heat flux density distribution is the quantitative deviation of the surface and subsurface heat conduction paths; the dynamic thermal resistance mapping algorithm is a mathematical method that calculates the change in thermal resistance through the difference in heat flux density and maps it to the thickness anomaly level.

[0129] In this embodiment, the heat flux gradient in the non-uniform heat diffusion pattern is first extracted through spatial differential operations, and the heat flux difference between the surface and subsurface is calculated. Secondly, based on the thermal resistance definition formula (thermal resistance = temperature difference / heat flux) and combined with the coating thickness calibration curve, a mapping relationship between the heat flux difference and thickness deviation is established. Next, a dynamic weight allocation model (such as a neural network) is used to classify the thermal resistance mutation gradient into mild, moderate, and severe levels. Finally, a thermal resistance mutation gradient map is generated, including the thickness anomaly level and defect location, and pushed to the process compensation system.

[0130] Step 405: Perform multi-dimensional feature fusion on the thermal resistance sudden change gradient and the phase correlation in the electromagnetic thermal coupling feature vector to generate a heat conduction feature, wherein the heat conduction feature includes the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation;

[0131] In this step, multidimensional feature fusion is the process of merging the thermal resistance mutation gradient and the electromagnetic phase correlation into a unified feature vector through principal component analysis (PCA) or weighted fusion algorithm, which is used to comprehensively characterize the thermal-electromagnetic coupling characteristics of the defect.

[0132] In this example, the thermal resistance gradient data and the phase parameters in the electromagnetic thermal coupling feature vector are first normalized to eliminate dimensional differences. Second, principal component analysis (PCA) is used to extract the core characteristic components of the two data types and construct a low-dimensional fusion feature space. Next, a weighted fusion algorithm is used to combine the thermal resistance gradient and phase correlation according to their contribution, generating a thermal conduction feature that includes the non-uniform heat diffusion pattern and the thermal resistance variation. Finally, the fused feature is input into a classification model to output a comprehensive diagnostic result for defect type and thickness deviation.

[0133] The following is a specific embodiment:

[0134] In the continuous production line of color-coated steel plates, a certain section of the coating had an abnormal local temperature gradient due to subsurface bubbles, and the eddy current detection showed a sudden change in impedance. The system first compensated for the time difference between the infrared and eddy current data through the timing alignment parameters, generated a synchronous timing window, and fused the electromagnetic-thermal feature vector. After correction based on the thermal diffusion coefficient, the non-uniform thermal diffusion pattern showed that the heat flux density in this area was significantly lower than the theoretical value, and the dynamic thermal resistance mapping algorithm calculated that its thermal resistance mutation gradient reached a severe level. Subsequently, the thermal resistance gradient was combined with the electromagnetic phase correlation through multi-dimensional feature fusion, and it was determined to be a composite defect of bubbles and insufficient thickness. The system triggered the spray flow and roller pressure compensation instructions. After repair, re-inspection showed that the thermal resistance gradient returned to the normal range and the coating thickness met the process standards.

[0135] This embodiment achieves precise alignment and deep coupled analysis of electromagnetic and thermal data through time synchronization, frequency domain integration, and multidimensional feature fusion, transforming non-uniform heat diffusion characteristics and thermal resistance gradients into quantifiable evidence for defect diagnosis. Time alignment eliminates data misalignment between devices, while frequency domain integration enhances the signal-to-noise ratio of weak signals. Correction to the heat conduction equation introduces material physical property constraints to reduce misjudgments. Dynamic thermal resistance mapping and multidimensional fusion enhance defect classification accuracy, ultimately enabling real-time positioning, precise classification, and closed-loop process control of coating defects in high-speed production lines, significantly improving production quality and resource utilization.

[0136] In order to improve the mapping accuracy, as another embodiment, according to step 404, the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion pattern is extracted, and the heat flux density distribution difference is converted into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm, including:

[0137] Step 501, decomposing the non-uniform heat diffusion pattern into a plurality of heat flux density gradient layers along the thickness direction of the color steel plate coating;

[0138] In this step, the heat flux gradient layer is a segmented model that divides the coating into several virtual layers along the thickness direction, and the heat flux density in each layer changes uniformly and continuously, which is used to quantify the abnormal distribution of the heat conduction path inside the coating.

[0139] In this example, based on the three-dimensional data of the non-uniform heat diffusion pattern, the coating is first sliced ​​along the thickness direction (Z-axis) at a preset resolution (e.g., one layer per micron). Next, the average heat flux density of each layer is calculated using finite element analysis (FEM), generating a heat flux gradient distribution curve along the thickness direction. Next, a clustering algorithm is used to identify critical layers with sudden changes in heat flux density (e.g., a sudden drop in heat flux density in a layer). The coating is then decomposed into multiple gradient layers, and the locations of the abnormal layers and their heat flux differences are marked.

[0140] Step 502: constructing a thermal resistance distribution matrix in the thickness direction of the color steel plate coating based on the heat flux density difference between adjacent layers in the heat flux density gradient layer;

[0141] In this step, the thermal resistance distribution matrix is ​​a two-dimensional matrix that calculates the thermal resistance value of each layer by the difference in heat flux density between adjacent layers, and is used to describe the spatial distribution characteristics of thermal resistance in the thickness direction of the coating.

[0142] In this example, the thermal resistance between each pair of adjacent layers is first calculated according to the thermal resistance definition formula (thermal resistance = temperature difference / heat flux). Next, the thermal resistance values ​​of each layer are arranged along the thickness direction according to spatial coordinates to construct a two-dimensional thermal resistance distribution matrix. Next, an interpolation algorithm is used to fill in matrix holes caused by missing data or noise, ensuring the continuity and integrity of the matrix. Finally, the matrix is ​​smoothed (e.g., using Gaussian filtering) to eliminate local abnormal fluctuations.

[0143] Step 503: Calculate the thermal resistance mutation amplitude between the coating surface and the subsurface of the color steel plate based on the spatial gradient distribution of the thermal resistance variation in the thermal resistance distribution matrix, and normalize the thermal resistance mutation amplitude by the heat flow path distortion caused by the thickness deviation to generate a normalized thermal resistance mutation coefficient.

[0144] In this step, the thermal resistance mutation amplitude is the maximum difference between the surface and subsurface thermal resistance values; the normalized thermal resistance mutation coefficient is to standardize the thermal resistance mutation amplitude into a dimensionless parameter by eliminating the influence of thickness deviation on the heat flow path.

[0145] In this embodiment, the thermal resistance difference between the surface layer and the deepest subsurface layer in the thermal resistance distribution matrix is ​​first calculated as the initial thermal resistance mutation amplitude. Next, based on the thickness deviation calibration curve (which maps thickness deviation to heat flow path distortion), the contribution of heat flow path distortion to the thermal resistance mutation amplitude is calculated. Next, a normalization algorithm (such as the maximum-minimum method) is used to convert the thermal resistance mutation amplitude into a coefficient within the range of 0–1, eliminating dimensional differences between regions of different thicknesses. Finally, a table of normalized thermal resistance mutation coefficients is generated and associated with the coating's spatial coordinates.

[0146] Step 504: Dynamically weight matching is performed on the normalized thermal resistance mutation coefficient and the phase correlation in the electromagnetic thermal coupling eigenvector to generate a thermal resistance mutation gradient corresponding to the thickness anomaly level, wherein the thermal resistance mutation gradient includes the local thermal resistance jump value of the defective area of ​​the color steel plate coating and the cumulative error of the thickness deviation along the production line movement direction;

[0147] In this step, dynamic weight matching is to assign differentiated weight ratios based on the physical correlation between the normalized thermal resistance mutation coefficient and the electromagnetic phase parameters to generate a thermal resistance mutation gradient that integrates thermal-electromagnetic characteristics.

[0148] In this embodiment, first, a correlation analysis is performed on the normalized thermal resistance mutation coefficient and the electromagnetic phase parameter to determine the weight ratio of the two in the defect area (for example, the thermal resistance coefficient accounts for 70% and the phase parameter accounts for 30%). Secondly, a weighted fusion algorithm is used to generate a thermal resistance mutation gradient map, marking the local thermal resistance jump value (for example, the gradient value exceeds the threshold) and the cumulative thickness deviation along the production line movement direction. Next, a sliding window statistical method is used to calculate the cumulative error of the thickness deviation in each area and classify it by severity level (such as mild, moderate, and severe). Finally, the thermal resistance mutation gradient is bound to the cumulative error and pushed to the process compensation system.

[0149] The following is a specific embodiment:

[0150] In the continuous production line of color-coated steel plates, a section of the coating had an abnormal non-uniform thermal diffusion pattern due to subsurface bubbles. The system first decomposed the thermal diffusion pattern into multiple gradient layers along the thickness and identified a sudden drop in the heat flux density in the third layer. After constructing the thermal resistance distribution matrix, the amplitude of the surface and subsurface thermal resistance mutations was calculated as a high value, and the mutation coefficient was generated after thickness deviation normalization. Dynamic weight matching was performed in combination with electromagnetic phase parameters to generate a thermal resistance mutation gradient, which showed that there was a serious local thermal resistance jump in this area, and the cumulative error along the direction of movement of the production line exceeded the standard. The system determined that it was a combined defect of bubbles and insufficient thickness, triggering the spray flow compensation and roller pressure increase instructions. After repair, re-inspection showed that the thermal resistance gradient returned to the normal range.

[0151] This embodiment achieves high-precision quantification of thermal resistance gradient changes in color-coated steel plate coatings and dynamic tracking of cumulative defect errors through layered modeling and dynamic weight matching. Heat flux gradient layer decomposition reveals thermal conduction anomalies along the thickness direction, while thermal resistance distribution matrix construction provides visual analysis of spatial thermal resistance distribution. Normalization eliminates thickness deviation interference, and dynamic weight matching integrates thermal and electromagnetic characteristics. The resulting thermal resistance gradient provides a precise quantitative basis for production line process adjustments, significantly improving coating defect repair efficiency and quality stability.

[0152] In order to solve the problem that conventional detection methods have high classification errors due to the aliasing of defect features and thickness features, and lack a dynamic correction mechanism for the defect-thickness correlation based on material properties, in some embodiments, according to step 104, the heat conduction features are simultaneously analyzed, and the color steel plate coating defect type and thickness anomaly level are distinguished through a dynamic weight allocation mechanism, and a detection result including the defect location, size, and thickness deviation value is output, including:

[0153] Step 601: Perform multi-dimensional feature decoupling on the thermal resistance variation caused by the non-uniform heat diffusion pattern in the heat conduction feature and the thickness deviation to generate a defect feature vector and a thickness feature vector, wherein the defect feature vector includes the temperature gradient distortion coefficient and the heat flux density mutation amplitude of the defect area, and the thickness feature vector includes the thermal resistance mutation gradient and eddy current phase delay parameter at different locations of the coating.

[0154] In this step, multidimensional feature decoupling is the process of separating the non-uniform heat diffusion pattern and the thermal resistance variation into independent feature vectors through principal component analysis or independent component analysis. The defect feature vector characterizes the local heat conduction anomaly caused by the defect, and the thickness feature vector quantifies the global impact of coating thickness deviation on thermal resistance.

[0155] In this example, the temperature gradient distortion coefficient and heat flux mutation amplitude in the non-uniform thermal diffusion model are first normalized. Principal component analysis is then used to extract defect-sensitive components and generate defect eigenvectors. Next, independent component analysis is performed on the thermal resistance mutation gradient and eddy current phase delay parameters to eliminate redundant information and generate thickness eigenvectors. Finally, the two eigenvectors are bound according to spatial coordinates to form a decoupled, independent dataset.

[0156] Step 602: construct a defect thickness correlation matrix based on the magnetic permeability and electrical conductivity of the color steel plate coating material;

[0157] In this step, the defect thickness correlation matrix is ​​a mathematical mapping model established through the physical relationship between magnetic permeability (material magnetization ability) and electrical conductivity (electrical conductivity). It is used to describe the coupling mechanism between defect types (such as bubbles and cracks) and thickness deviations.

[0158] In this example, the magnetic permeability and electrical conductivity parameters in the material database are first retrieved, and the correlation weights between defects and thickness are derived based on the electromagnetic-thermal coupling equation. Next, a nonlinear regression algorithm is used to establish a mapping relationship between magnetic permeability, electrical conductivity, and thermal resistance, generating a defect-thickness correlation matrix. Finally, the matrix is ​​orthogonalized to eliminate multicollinearity interference and improve its physical interpretability.

[0159] Step 603, performing nonlinear superposition processing on the defect feature vector and the thickness feature vector to generate a comprehensive feature vector;

[0160] In this step, the nonlinear superposition processing is to merge the defect and thickness feature vectors into a unified comprehensive feature vector through a neural network or a weighted fusion algorithm, which is used to comprehensively reflect the coupling effect of the defect and thickness.

[0161] In this example, the defect and thickness feature vectors are first fed into a pretrained deep neural network, where a nonlinear activation function is used to extract high-order correlation features between them. Next, an attention mechanism is employed to dynamically assign weights to the defect and thickness features, generating a fused composite feature vector. Finally, the composite feature vector undergoes dimensionality reduction, retaining its core features for subsequent analysis.

[0162] Step 604, in the multidimensional feature space, based on the prior distribution data of the coating defect type and the thickness anomaly level, calculate the confidence parameter of the defect area and the cumulative probability value of the thickness deviation;

[0163] In this step, the confidence parameter is a probability score of defect existence calculated by a Bayesian probability model; the cumulative probability value is a quantified probability of the severity of thickness deviation based on historical data statistics.

[0164] In this embodiment, prior distribution data (such as the probability of bubble and crack occurrence and thickness deviation distribution curves) is first loaded from the historical defect database. Next, a Bayesian inference algorithm is used to calculate the probability of the current defect feature vector matching known defect types, generating a confidence parameter. Next, a Monte Carlo simulation method is used to calculate the cumulative probability of the thickness feature vector and classify the thickness deviation abnormality level. Finally, the confidence parameter is associated with the cumulative probability value and stored, forming a quantitative basis for defect diagnosis.

[0165] Step 605: Generate a detection result including defect location coordinates, defect coverage area, and thickness deviation compensation value based on the combined constraints of the confidence parameter and the cumulative probability value, wherein the detection result is synchronized in real time with the motion trajectory of the coating equipment of the color steel plate production line;

[0166] In this step, the joint constraint condition is used to generate executable detection results through a comprehensive threshold judgment rule of the confidence parameter (the credibility of the defect existence) and the cumulative probability value (the severity of the thickness deviation).

[0167] In this embodiment, a confidence threshold (e.g., ≥90%) and a cumulative probability threshold (e.g., ≥95%) are set. If both are met, a high-confidence defect is identified. Next, a spatial clustering algorithm is used to mark the coordinates and coverage area of ​​the defective area. Compensation values ​​are generated based on a thickness deviation compensation rule base. Finally, the inspection results are transmitted to the coating equipment in real time via an industrial communication protocol. The trajectory of the spray robot and the paint flow rate are adjusted to ensure that compensation actions are synchronized with production line movements.

[0168] The following is a specific embodiment:

[0169] In the continuous production line for color-coated steel sheets, an area experienced an abnormal temperature gradient distortion coefficient due to subsurface bubbles, while the thickness feature vector indicated an excessively high thermal resistance gradient. The system separated defects and thickness features through multi-dimensional feature decoupling, constructed a defect-thickness correlation matrix, and then generated a comprehensive feature vector through nonlinear superposition. The Bayesian model calculated its confidence parameter to be 92% and its cumulative probability value to be 97%, determining it to be a high-confidence bubble defect. The system generated the test results: defect coordinates (XY), coverage area, and spray flow compensation value, and synchronized them to the coating equipment in real time. The robotic arm precisely increased the flow rate when the production line moved to that coordinate, and re-inspection after repair showed that both the coating thickness and thermal conductivity characteristics met the standards.

[0170] This embodiment achieves independent analysis and integrated diagnosis of defects and thickness characteristics through multi-dimensional feature decoupling and nonlinear superposition. Combining Bayesian probability and Monte Carlo simulation to quantify defect confidence and thickness deviation probability, it ultimately generates highly accurate inspection results and integrates them with production line equipment in real time. The coordinated optimization of defect location and compensation value calculation significantly improves coating repair efficiency, reduces material waste, and ensures quality stability and process controllability for high-speed, continuous production lines.

[0171] Researchers have found that traditional compensation instructions result in adjustment lag due to the spatiotemporal mismatch between the motion trajectory of the coating equipment and the detection results. This leads to significant cumulative errors, especially in high-speed production lines. To improve the real-time performance of compensation, in some embodiments, according to step 105, a dynamic compensation instruction for coating thickness is generated based on the detection results, including:

[0172] Step 701, calculating a spray flow compensation amount and a roller pressure compensation amount of the coating material based on the thickness deviation value and the defect coverage area in the detection result;

[0173] In this step, the spray flow compensation is the paint injection amount correction value calculated based on the thickness deviation value and the defect area, which is used to fill the area with insufficient coating thickness; the roller pressure compensation is the roller equipment pressure correction value adjusted according to the mechanical strength requirements of the defect area, which is used to compact the coating and eliminate internal defects.

[0174] In this embodiment, first, the compensation model in the process rule library is called to calculate the spray flow compensation ratio (such as a 10% increase in flow) and the roller pressure increment (such as a 15% increase in pressure) based on the thickness deviation value (such as the deviation percentage) and the defect coverage area (such as the defect diameter).

[0175] Step 702: performing spatiotemporal matching processing on the spray flow compensation amount and the roller pressure compensation amount based on the real-time motion trajectory of the color steel plate coating equipment to generate dynamic compensation parameters synchronized with the position coordinates of the coating equipment, wherein the dynamic compensation parameters include a spray flow correction coefficient and a roller pressure gradient adjustment amount;

[0176] In this step, the spatiotemporal matching process is the process of dynamically associating the compensation parameters with the real-time position of the coating equipment through coordinate mapping and time axis synchronization algorithm; the dynamic compensation parameters are the action instructions sent to the equipment in real time to ensure that the compensation action is accurately matched with the production line rhythm.

[0177] In this embodiment, the real-time coordinates and velocity vectors of the coating equipment are first acquired through the production line motion controller to establish a time-space mapping model of the equipment's motion trajectory. Next, the defect location coordinates in the compensation parameter table are converted to the target position in the equipment coordinate system. An interpolation algorithm is then used to generate the spray flow correction coefficient (e.g., the flow correction value corresponding to the XY position) and the roller pressure gradient adjustment (e.g., the pressure value that increases linearly along the motion direction). Finally, the dynamic compensation parameters are timestamped and bound to the equipment control command queue to ensure that compensation actions are synchronized with equipment motion.

[0178] Step 703 , performing delay compensation correction on the dynamic compensation parameter by using the timestamp difference between the defect position coordinates and the preset production line motion speed parameter to generate a delay compensation correction coefficient;

[0179] In this step, the delay compensation correction is an optimization process of calibrating the dynamic compensation parameters on the time axis by calculating the signal transmission delay and equipment response time difference between the detection system and the coating equipment; the delay compensation correction coefficient is a dynamic adjustment factor used to correct the instruction issuance time.

[0180] In this embodiment, a high-precision clock synchronization module first records the difference between the timestamp generated by the inspection result and the device's current timestamp. Next, the displacement offset of the defect location within the delay time is calculated based on the production line speed (e.g., 2 m / s) (e.g., a 0.1 second delay corresponds to a 0.2 m offset). Next, a Kalman filter is used to predict the device's future position and generate a delay compensation correction factor (e.g., a time advance of 0.1 seconds). Finally, the correction factor is used to adjust the timestamp of the dynamic compensation parameter to eliminate positioning errors caused by delays.

[0181] Step 704: superimpose and fuse the delay compensation correction coefficient and the dynamic compensation parameter to generate a coating thickness dynamic compensation instruction;

[0182] In this step, superposition fusion is the process of merging the delay compensation correction coefficient and dynamic compensation parameters into executable instructions through time axis calibration and spatial coordinate alignment; the coating thickness dynamic compensation instruction is the control signal finally sent to the coating equipment, which includes time, position, flow and pressure parameters.

[0183] In this embodiment, a delay compensation correction coefficient is first applied to the timestamp of the dynamic compensation parameters to adjust the timing of command issuance. Secondly, a spatial interpolation algorithm is used to re-align the corrected compensation parameters with the equipment's real-time coordinates, ensuring that the spray flow rate and roller pressure are precisely triggered at the target location. Finally, a dynamic compensation command containing time, space, and motion parameters is generated and issued in real time to the coating equipment's execution unit via the industrial bus.

[0184] The following is a specific embodiment:

[0185] On a continuous production line for color-coated steel sheets, thickness deviations and bubble defects were detected in a certain area. The system calculated that the spray flow rate needed to increase by 12% and the roller pressure by 18%. Dynamic compensation parameters were generated based on the equipment's real-time motion trajectory, but due to signal transmission delays, the command was delayed by 0.1 seconds. The delay compensation correction factor was used to predict the equipment's future position, and the command issuance timestamp was adjusted. The dynamic compensation command generated after superposition and fusion accurately triggered the spray robot arm to increase the flow rate at the target position, and the roller press simultaneously increased the pressure. Re-inspection after repair showed that the coating thickness was uniform, the bubble defects had been eliminated, and the compensation action was fully synchronized with the production line rhythm.

[0186] This embodiment achieves precise synchronization of coating defect repair instructions with the high-speed production line rhythm through spatiotemporal matching and delay compensation correction. The dynamic compensation parameters for spray flow and roller pressure, after spatiotemporal calibration, eliminate signal transmission delays and equipment response errors, ensuring that compensation actions are precisely triggered at the target location. The dynamic compensation instructions generated by superposition and fusion significantly improve coating repair efficiency and quality consistency, reduce material waste and equipment downtime losses, and are suitable for the industrial needs of high-speed continuous production.

[0187] Figure 2 The present invention provides a schematic diagram of a dynamic monitoring system for color steel plate coating thickness based on artificial intelligence, as shown in FIG. Figure 2 As shown, the system includes:

[0188] An acquisition module is used to continuously acquire temperature distribution data of the coating surface of the color steel plate by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating;

[0189] The first generating module 21 is used to generate a thermal response spectrum of the color steel plate coating based on the principle of electromagnetic induction by means of pulsed eddy current excitation, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material;

[0190] An extraction module 22 is configured to perform time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation;

[0191] An analysis module 23 is configured to simultaneously analyze the heat conduction characteristics, distinguish the color steel plate coating defect type and thickness anomaly level through a dynamic weight allocation mechanism, and output a detection result including the defect location, size, and thickness deviation value;

[0192] The second generating module 24 is configured to generate a coating thickness dynamic compensation instruction according to the detection result.

[0193] Figure 2 The artificial intelligence-based dynamic monitoring system for color steel plate coating thickness can be executed Figure 1 The implementation principle and technical effects of the AI-based dynamic monitoring method for color-coated steel plate coating thickness described in the illustrated embodiment will not be elaborated on here. The specific manner in which each module and unit performs operations in the AI-based dynamic monitoring system for color-coated steel plate coating thickness in the above embodiment has been described in detail in the embodiments of the method and will not be elaborated on here.

[0194] In one possible design, Figure 2 The embodiment shown is a color steel plate coating thickness dynamic monitoring system based on artificial intelligence that 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;

[0195] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0196] The processing component 32 is used for the above Figure 1 The embodiment provides an artificial intelligence-based dynamic monitoring method for the thickness of color-coated steel plate coatings.

[0197] 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 as 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 to perform the above method.

[0198] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory 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.

[0199] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0200] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0201] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0202] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0203] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is an artificial intelligence-based dynamic monitoring method for the thickness of color-coated steel plate coating.

[0204] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0205] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0206] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0207] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for dynamic monitoring of color steel plate coating thickness based on artificial intelligence, characterized in that: include: Continuously obtain temperature distribution data on the surface of the color-coated steel plate coating by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating; Generating a thermal response spectrum of the color steel plate coating based on the principle of electromagnetic induction by pulsed eddy current excitation, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material; Performing time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation; The heat conduction characteristics are analyzed synchronously, and the color steel plate coating defect type and thickness abnormality level are distinguished through a dynamic weight distribution mechanism, and the detection results including the defect location, size and thickness deviation value are output; A coating thickness dynamic compensation instruction is generated according to the detection result.

2. The method according to claim 1, characterized in that The thermal response spectrum of the color steel plate coating is generated based on the electromagnetic induction principle by pulsed eddy current excitation, including: Dynamically set the initial frequency and pulse width of the multi-band pulsed eddy current excitation signal based on the magnetic permeability and electrical conductivity of the color steel plate coating material, wherein the initial frequency is segmented and modulated based on the critical value of the electromagnetic coupling between the color steel plate substrate and the coating; The color steel plate coating is laterally scanned by a multi-band pulsed eddy current excitation signal, wherein the multi-band pulsed eddy current excitation signal includes attenuation waveforms of at least three different frequency bands, and the amplitude of the attenuation waveform decreases nonlinearly with increasing thickness of the color steel plate coating; The eddy current attenuation component generated by the multi-band pulsed eddy current excitation signal when the color steel plate coating is horizontally scanned is collected, and the phase delay compensation of the eddy current attenuation component is performed using an orthogonal decomposition algorithm to extract the eddy current phase delay parameter related to the thickness distribution of the color steel plate coating; The eddy current phase delay parameter and the local temperature gradient change in the temperature distribution data are integrated in the frequency domain to generate a thermal response map that integrates the electromagnetic thermal characteristics. The thermal response map includes the eddy current impedance mutation characteristics of the sub-surface defect area of ​​the color steel plate coating and the time-varying trajectory of the thermal resistance caused by thickness deviation.

3. The method according to claim 2, characterized in that The eddy current attenuation component generated by the multi-band pulsed eddy current excitation signal when the color steel plate coating is horizontally scanned is collected, and the phase delay compensation of the eddy current attenuation component is performed using an orthogonal decomposition algorithm to extract the eddy current phase delay parameter related to the thickness distribution of the color steel plate coating, including: Decomposing the eddy current attenuation component of the multi-band pulsed eddy current excitation signal into an in-phase component and an orthogonal component, wherein the in-phase component is related to the electromagnetic properties of the color steel plate substrate, and the orthogonal component is related to the nonlinear change of the coating thickness; Based on the attenuation waveforms of different frequency bands in the multi-band pulsed eddy current excitation signal, respectively calculating the phase delay factors of the orthogonal components in the coating thickness direction; constructing a coating thickness distribution function according to the phase delay factor and the amplitude attenuation ratio of the in-phase component; The thickness distribution function is normalized by a frequency band superposition algorithm to generate an eddy current phase delay parameter, wherein the eddy current phase delay parameter includes a phase mutation gradient caused by thickness deviation of the color steel plate coating and an impedance matching error in the sub-surface defect area.

4. The method according to claim 1, wherein The temperature distribution data is subjected to time-series correlation processing with the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, including: Performing time stamp matching on the local temperature gradient change in the temperature distribution data and the time-varying trajectory of the thermal resistance in the thermal response map to generate a timing alignment parameter, wherein the timing alignment parameter is used to compensate for the signal acquisition time difference between the infrared thermal imager and the pulsed eddy current excitation device; Based on the timing alignment parameters, the sampling frequency of the temperature distribution data is synchronously modulated with the pulse excitation frequency of the thermal response spectrum to generate a synchronous timing window, and the temperature gradient change within the synchronous timing window is fused with the eddy current impedance mutation characteristics using a frequency domain integration algorithm to obtain an electromagnetic thermal coupling characteristic vector; According to the thermal diffusion coefficient of the material of the color steel plate coating, the heat conduction equation is corrected for the electromagnetic thermal coupling characteristic vector to generate a corrected non-uniform thermal diffusion model; Extracting the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion mode, and converting the heat flux density distribution difference into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm; The thermal resistance mutation gradient and the phase correlation in the electromagnetic thermal coupling eigenvector are multi-dimensionally fused to generate a heat conduction feature, wherein the heat conduction feature includes the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation.

5. The method according to claim 4, characterized in that Extracting the heat flux density distribution difference between the coating surface and the subsurface of the color steel plate in the non-uniform heat diffusion mode, and converting the heat flux density distribution difference into a thermal resistance mutation gradient corresponding to the thickness anomaly level through a dynamic thermal resistance mapping algorithm, including: Decomposing the non-uniform heat diffusion pattern into a plurality of heat flux density gradient layers along the thickness direction of the color steel plate coating; Based on the heat flux density difference between adjacent layers in the heat flux density gradient layer, a thermal resistance distribution matrix in the thickness direction of the color steel plate coating is constructed; Calculating the thermal resistance mutation amplitude between the coating surface and the subsurface of the color steel plate according to the spatial gradient distribution of the thermal resistance variation in the thermal resistance distribution matrix, and normalizing the thermal resistance mutation amplitude by the heat flow path distortion caused by the thickness deviation to generate a normalized thermal resistance mutation coefficient; The normalized thermal resistance mutation coefficient is dynamically weighted matched with the phase correlation in the electromagnetic thermal coupling eigenvector to generate a thermal resistance mutation gradient corresponding to the thickness anomaly level, wherein the thermal resistance mutation gradient includes the local thermal resistance jump value of the defective area of ​​the color steel plate coating and the cumulative error of the thickness deviation along the production line movement direction.

6. The method according to claim 1, characterized in that The heat conduction characteristics are analyzed synchronously, and the color steel plate coating defect type and thickness abnormality level are distinguished through a dynamic weight allocation mechanism. The detection results containing the defect location, size and thickness deviation value are output, including: Perform multi-dimensional feature decoupling on the thermal resistance variation caused by the non-uniform heat diffusion mode in the heat conduction feature and the thickness deviation to generate a defect feature vector and a thickness feature vector, wherein the defect feature vector includes the temperature gradient distortion coefficient and the heat flux density mutation amplitude of the defect area, and the thickness feature vector includes the thermal resistance mutation gradient and eddy current phase delay parameter at different locations of the coating; According to the magnetic permeability and electrical conductivity of the color steel plate coating material, the defect thickness correlation matrix is ​​constructed; Performing nonlinear superposition processing on the defect feature vector and the thickness feature vector to generate a comprehensive feature vector, wherein the comprehensive feature vector is mapped to a multidimensional feature space through a classification boundary function in a dynamic weight allocation mechanism; In the multidimensional feature space, based on the prior distribution data of the coating defect type and the thickness anomaly level, the confidence parameter of the defect area and the cumulative probability value of the thickness deviation are calculated; According to the joint constraints of the confidence parameter and the cumulative probability value, a detection result including the defect position coordinates, the defect coverage area and the thickness deviation compensation value is generated, wherein the detection result is synchronized in real time with the motion trajectory of the coating equipment of the color steel plate production line.

7. The method according to claim 1, characterized in that Generating a coating thickness dynamic compensation instruction according to the detection result, including: Calculating a spray flow rate compensation amount and a roller pressure compensation amount of the coating material based on the thickness deviation value and the defect coverage area in the detection result; According to the real-time motion trajectory of the color steel plate coating equipment, the spray flow compensation amount and the roller pressure compensation amount are temporally and spatially matched to generate dynamic compensation parameters synchronized with the position coordinates of the coating equipment, wherein the dynamic compensation parameters include the spray flow correction coefficient and the roller pressure gradient adjustment amount; Performing delay compensation correction on the dynamic compensation parameter by using the timestamp difference between the defect position coordinates and the preset production line motion speed parameter to generate a delay compensation correction coefficient; The delay compensation correction coefficient is superimposed and integrated with the dynamic compensation parameter to generate a coating thickness dynamic compensation instruction.

8. An artificial intelligence-based dynamic monitoring system for color steel plate coating thickness, characterized in that: include: An acquisition module is used to continuously acquire temperature distribution data of the coating surface of the color steel plate by infrared thermal imaging, wherein the temperature distribution data includes local temperature gradient changes corresponding to sub-surface defects of the coating; A first generating module is configured to generate a thermal response spectrum of the color steel plate coating based on the principle of electromagnetic induction by means of a pulsed eddy current excitation method, wherein the electromagnetic parameters of the pulsed eddy current excitation are dynamically adjusted according to the magnetic permeability and electrical conductivity of the color steel plate coating material; An extraction module is used to perform time-series correlation processing on the temperature distribution data and the thermal response map to extract the heat conduction characteristics between the coating surface and the subsurface of the color steel plate, wherein the heat conduction characteristics include the non-uniform heat diffusion pattern of the defect area and the thermal resistance change caused by the thickness deviation; An analysis module is used to synchronously analyze the heat conduction characteristics, distinguish the color steel plate coating defect type and thickness anomaly level through a dynamic weight distribution mechanism, and output a detection result including the defect location, size and thickness deviation value; The second generating module is used to generate a coating thickness dynamic compensation instruction according to the detection result.

9. A computing device, characterized in that It includes 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 an artificial intelligence-based dynamic monitoring method for color steel plate coating thickness as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, an artificial intelligence-based dynamic monitoring method for the thickness of a color steel plate coating is implemented as described in any one of claims 1 to 7.

Images