A method for analyzing heat dissipation of LED backlight sources

Abstract

The present application provides a method for analyzing the heat dissipation of an LED backlight source, comprising: obtaining the local light decay degree estimate of the LED backlight source in the area with excessive heat concentration based on the real-time operating temperature data of the LED backlight source in the area with excessive heat concentration, combining with a pre-established temperature and light decay model, and using a support vector machine regression algorithm to obtain the local light decay degree estimate of the LED backlight source in the area at the current operating temperature; obtaining the overall average brightness value of the LED backlight source, and comparing it with the local light decay degree estimate, and if the local light decay degree exceeds a preset threshold value of the overall average brightness, it is determined that the LED backlight source has a local brightness unevenness phenomenon; and dynamically adjusting the heat dissipation efficiency of the radiator by controlling the rotation speed of the cooling fan and the opening and closing size of the vents according to the abnormal temperature range of the components, so that the temperature of the components returns to a normal range.

Description

Technical Field

[0001] The invention relates to the field of information technology, and in particular to a method for analyzing heat dissipation of an LED backlight source. Background Art

[0002] When LED backlight sources are used at different angles, the heat dissipation effect faces many challenges. When the backlight source is in a horizontal state, the heat generated by the LED light bar can form a stable heat plume through natural convection, and the heat can be well taken away to form a stable thermal boundary layer. The existence of the thermal boundary layer can effectively isolate the high-temperature area from the surrounding environment, so that the temperature of the LED light bar is maintained within a safe working range. However, when the backlight source is tilted or placed vertically, the change in the direction of gravity disrupts the formation of the heat plume. When tilted or placed vertically, the heat plume is affected by the direction of gravity and can no longer rise vertically and effectively take away the heat. The thermal boundary layer becomes unstable, and the heat exchange efficiency between the high-temperature area and the surrounding environment decreases. This causes the operating temperature of the LED light bar to increase and the heat dissipation efficiency to decrease. The decrease in heat dissipation efficiency brings a series of problems. The increase in operating temperature accelerates the light decay of the LED light bar and shortens its service life. At the same time, high temperature may also cause color deviation and uneven brightness of the LED light bar, affecting the display effect. In addition, PCB boards and other electronic components may also be affected by continuous high temperatures, resulting in reduced reliability of the backlight source. In summary, when the LED backlight source is tilted or placed vertically, the heat plume disturbance caused by the change in gravity direction leads to a decrease in heat dissipation efficiency, which in turn causes a series of problems such as light decay, color deviation, uneven brightness and reduced reliability, bringing challenges to the design and application of the backlight source. Summary of the invention

[0003] The present invention provides a method for analyzing heat dissipation of an LED backlight source, which mainly includes:

[0004] The tilt angle information of the LED backlight source is obtained, and according to the preset angle threshold, it is determined whether the LED backlight source is in a horizontal state that is conducive to natural convection heat dissipation or in an inclined state that causes the formation of heat plumes to be blocked. If it is determined to be in an inclined state, the temperature distribution data of the radiator surface is obtained; according to the obtained radiator surface temperature distribution data, the thermal balance equation is used to calculate the thermal resistance distribution of the radiator surface, and the area with the largest local thermal resistance is obtained. The area is determined to be an excessive heat concentration area, and the real-time operating temperature data of the LED backlight source in the area is obtained by infrared temperature measurement; according to the obtained real-time operating temperature data of the LED backlight source in the excessive heat concentration area, combined with the pre-established temperature and light decay model, the support vector machine regression algorithm is used to obtain the LED backlight source in the area at the current The estimated value of the local light decay degree at the working temperature; the average brightness value of the entire LED backlight source is obtained, and it is compared with the estimated value of the local light decay degree. If the local light decay degree exceeds the preset threshold value of the overall average brightness, it is determined that the LED backlight source has local brightness unevenness; if the LED backlight source has local brightness unevenness, the real-time temperature data of the PCB board and electronic components corresponding to the excessive heat concentration area are obtained, and the BP neural network algorithm is used to determine the abnormal temperature range of the components that cause the local brightness unevenness in combination with the pre-established component temperature and brightness uniformity model; according to the abnormal temperature range of the components, the heat dissipation efficiency of the radiator is dynamically adjusted by controlling the speed of the cooling fan and the opening and closing size of the vents to return the component temperature to the normal range.

[0005] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0006] The present invention discloses a method for heat dissipation analysis of an LED backlight source. The method obtains the inclination angle of the LED backlight source and the temperature distribution of the radiator, calculates the thermal resistance distribution and determines the area of ​​excessive heat concentration, thereby obtaining the real-time operating temperature of the LED in the area. The temperature and light decay model are combined to estimate the degree of local light decay and compare it with the overall average brightness to determine whether there is a local brightness unevenness phenomenon. If so, the temperature of the relevant PCB and components is further analyzed to determine the abnormal temperature range that causes the unevenness. Finally, the temperature of the components is returned to the normal range by dynamically adjusting the heat dissipation efficiency. The present invention realizes intelligent detection and automatic adjustment of the local brightness unevenness of the LED backlight source, effectively improving the service life and display quality of the LED backlight source. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 The present invention is a flow chart of a method for analyzing heat dissipation of an LED backlight source. DETAILED DESCRIPTION

[0008] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0009] like Figure 1 In this embodiment, a method for analyzing heat dissipation of an LED backlight source may specifically include:

[0010] S101. Obtain the tilt angle information of the LED backlight source, and determine, based on a preset angle threshold, whether the LED backlight source is in a horizontal state that is conducive to natural convection heat dissipation or in an inclined state that causes obstruction of heat plume formation; if it is determined to be in an inclined state, obtain temperature distribution data on the surface of the heat sink.

[0011] The tilt angle of the LED backlight source is obtained, and the tilt angle is measured in real time by the MEMS acceleration sensors installed at the four corners of the LED backlight source; the current state of the LED backlight source is judged according to the comparison between the tilt angle and the preset angle threshold; if the tilt angle is greater than or equal to the preset angle threshold, the LED backlight source is judged to be in a tilted state; for the LED backlight source judged to be in a tilted state, the heat sink surface temperature distribution data acquisition program is started; the heat sink surface temperature distribution data acquisition program collects real-time temperature data through 16 PT100 platinum resistance temperature sensors evenly arranged on the heat sink surface; the collected temperature data is processed by median filtering to generate a temperature data matrix reflecting the temperature distribution status of the heat sink surface; the temperature The method comprises the following steps: performing hot spot analysis and temperature gradient calculation based on the temperature data matrix, and generating a high-resolution temperature distribution thermogram using the bilinear interpolation method; identifying local hot spot areas with abnormally high temperature through the temperature distribution thermogram, and calculating the temperature gradient between the hot spot area and the surrounding area using the finite difference method; quantitatively evaluating the severity of local heat accumulation based on the difference between the temperature gradient and a preset threshold; adjusting the heat dissipation control strategy of the LED backlight source based on the evaluation result of the local heat accumulation; the heat dissipation control strategy comprises increasing the fan blade speed of the radiator corresponding to the high-temperature hot spot area, increasing the coolant flow of the liquid cooling system, and adjusting the airflow direction of the radiator; adjusting the airflow direction based on the hot spot position through a variable angle guide plate controlled by a servo motor, and guiding the cold air to flow to the high-temperature area.

[0012] Specifically, the tilt angle information of the LED backlight source is obtained, and the current state of the LED backlight source is determined according to the preset angle threshold. MEMS acceleration sensors are installed at the four corners of the LED backlight source to measure the tilt angle data in real time, and the measured angle value is compared with the preset angle threshold stored in the system. If the measured angle value is less than the preset threshold, it is determined that the LED backlight source is in a horizontal state that is conducive to natural convection heat dissipation; if the measured angle value is greater than or equal to the preset threshold, it is determined that the LED backlight source is in a tilted state that may cause the formation of heat plumes to be blocked. For the LED backlight source determined to be in a tilted state, the radiator surface temperature distribution data acquisition program is started. 16 PT100 platinum resistance temperature sensors are evenly arranged on the surface of the radiator to form a 4x4 matrix, and real-time temperature data of each area of ​​the radiator is collected at a preset sampling frequency. The collected temperature data is subjected to noise elimination and outlier removal by the median filtering algorithm to generate a temperature data matrix reflecting the temperature distribution status of the radiator surface. Hot spot analysis and temperature gradient calculation are performed on the obtained radiator surface temperature distribution data matrix. The temperature data of the 16 sampling points are interpolated using the bilinear interpolation method to generate a high-resolution temperature distribution thermodynamic map. The local hotspot areas with abnormally high temperature are identified through thermal map analysis, and the temperature gradient between the hotspot area and the surrounding area is calculated using the finite difference method. The severity of local heat accumulation is quantitatively evaluated based on the difference between the temperature gradient and the preset threshold. Based on the evaluation results of local heat accumulation, the heat dissipation control strategy of the LED backlight is adjusted. For the high-temperature hotspot area obtained by the evaluation, the fan speed of the radiator corresponding to the area is increased or the coolant flow of the liquid cooling system is increased. At the same time, the airflow direction of the radiator is adjusted according to the temperature distribution data. The variable angle guide plate controlled by the servo motor is used to adjust the airflow direction according to the hotspot position to guide the cold air to the high-temperature area. The heat dissipation parameters are continuously adjusted through closed-loop feedback control to suppress the local heat accumulation phenomenon and make the temperature distribution of the LED backlight uniform. The LED backlight tilt angle is obtained using a high-precision MEMS acceleration sensor with a measurement accuracy of ±0.1°. The sensor sampling frequency is set to 100Hz, and the raw data is smoothed by the sliding window averaging algorithm to reduce the impact of instantaneous fluctuations. The preset angle threshold is 5°, and the tilt state judgment is triggered when the measured angle exceeds the threshold for 5 consecutive seconds. The radiator surface temperature distribution data is collected using 16 PT100 platinum resistance temperature sensors with a measurement range of -50°C to 200°C and an accuracy of ±0.1°C. The sensors are arranged in a 4x4 matrix, covering the core area of ​​the radiator of 100mmx100mm. The sampling frequency is set to 1Hz, and the collected data is preprocessed using the median filter algorithm to filter out abnormal values. The temperature data matrix is ​​used to generate a high-resolution thermal map of 400x400 pixels through bilinear interpolation, and the spatial resolution after interpolation reaches 0.25mm / pixel.Hot spot analysis uses a local maximum detection algorithm to identify areas with temperatures higher than the surrounding area by more than 3°C as hot spots. The temperature gradient is calculated using the finite difference method, and the x- and y-direction gradients of the center point are calculated with a 3x3 pixel window, and the gradient threshold is set to 2°C / mm. The heat dissipation control strategy is dynamically adjusted according to the hot spot location and temperature gradient. The radiator is equipped with 4 adjustable speed fans with a speed range of 800-3000RPM, and the speed is adjusted linearly according to the hot spot temperature. The flow rate of the liquid cooling system can be adjusted in the range of 0.5-2L / min, and precise control is achieved through a proportional integral control algorithm. The variable angle guide plate is controlled by 4 servo motors with an angle adjustment range of ±30° and a response time of less than 100ms. The closed-loop feedback control uses a fuzzy PID algorithm, and parameter adjustment is performed every 5 seconds. The control target is to make the difference between the maximum temperature and the average temperature of the radiator surface less than 5°C.

[0013] S102. Based on the obtained radiator surface temperature distribution data, the heat balance equation is used to calculate the thermal resistance distribution on the radiator surface, and the area with the largest local thermal resistance is obtained. The area is determined to be an area of ​​excessive heat concentration, and the real-time operating temperature data of the LED backlight source in the area is obtained by infrared temperature measurement.

[0014] A heat balance equation calculation model is established based on the acquired radiator surface temperature distribution data, and the calculation model divides the radiator surface into calculation grids. The heat conduction equation is discretized using the central difference format to construct a heat balance equation group including heat conduction, convection and radiation heat transfer modes. The heat balance equation group is solved by the Gauss-Seidel iteration method to obtain the heat flux density distribution of each grid point. According to the heat flux density distribution and temperature distribution data, the thermal resistance distribution on the radiator surface is calculated to generate a thermal resistance distribution matrix. The thermal resistance distribution matrix is ​​converted into a high-resolution thermal resistance distribution map using a bicubic spline interpolation algorithm. The thermal resistance distribution map is analyzed, and the sliding window method is used to determine the area with the largest local thermal resistance. If the average thermal resistance value of the window exceeds a preset threshold and is higher than a preset ratio of the surrounding area, the area is marked as an excessive heat concentration area. For the excessive heat concentration area, an infrared temperature measurement device is deployed for real-time temperature measurement. A boundary extraction algorithm based on threshold segmentation is used to automatically lock the boundary of the excessive heat concentration area and extract the temperature data in the area. The infrared thermal image is registered with the thermal resistance distribution map to ensure that the temperature measurement area corresponds accurately to the excessive heat concentration area. The temperature distribution data of the excessive heat concentration area is transmitted in real time through a data acquisition interface.

[0015] Specifically, a heat balance equation calculation model is established based on the obtained radiator surface temperature distribution data. The radiator surface is divided into 100x100 calculation grids, each grid corresponding to a temperature data point. The heat conduction equation is discretized using the central difference format to construct a heat balance equation group. Three heat transfer modes of heat conduction, convection and radiation are considered in the equation, and the thermal conductivity, ambient temperature and convection heat transfer coefficient of the radiator material are introduced as known parameters. The heat balance equation group is solved by the Gauss-Seidel iteration method to obtain the heat flux density distribution of each grid point. The thermal resistance distribution of the radiator surface is calculated using the calculated heat flux density distribution and temperature distribution data. For each grid point, the local thermal resistance value is calculated according to the thermal resistance definition formula R=ΔT / q, where ΔT is the difference between the point and the ambient temperature, and q is the heat flux density of the point. A thermal resistance distribution matrix is ​​generated, and the thermal resistance data is converted into a high-resolution thermal resistance distribution map by the bicubic spline interpolation algorithm. The thermal resistance distribution map is analyzed to determine the area with the largest local thermal resistance. The sliding window method is used to scan the thermal resistance distribution map using a 5x5 window, and the window movement step is set to 1 grid unit. The average thermal resistance value in each window is calculated. If the average thermal resistance value of the window exceeds the preset threshold and is more than 20% higher than the surrounding area, the area is marked as an excessive heat concentration area. The position coordinates and area of ​​the excessive heat concentration area are recorded. For the determined excessive heat concentration area, an infrared temperature measurement device is deployed. An infrared thermal imager with a resolution of 640x480 pixels and a temperature measurement accuracy of ±0.5℃ is selected to measure the temperature of the excessive heat concentration area in real time. The emissivity parameters of the infrared thermal imager are set to match the actual emissivity of the surface material of the LED backlight source. The boundary extraction algorithm based on threshold segmentation is used to automatically lock the boundary of the excessive heat concentration area and extract the temperature data in the area. The infrared thermal image is image-aligned with the thermal resistance distribution map to ensure that the temperature measurement area corresponds accurately to the excessive heat concentration area. The temperature distribution data of the excessive heat concentration area is transmitted in real time through the data acquisition interface as the working temperature data of the LED backlight source. Based on the surface temperature distribution data of the radiator, a heat balance equation model with a grid size of 100x100 was constructed. The heat conduction equation was discretized using the central difference format, considering an aluminum radiator with a thermal conductivity of 237W / (m·K), an ambient temperature of 25℃, and a convection heat transfer coefficient of 15W / (m²·K). The Gauss-Seidel iteration method was used to solve the equations, with the convergence threshold set to 10^-6 and the maximum number of iterations set to 1000. The heat flux distribution obtained ranged from 1000 to 5000W / m². The thermal resistance distribution was then calculated, and the local thermal resistance value was between 0.01-0.1℃ / W. A high-resolution thermal resistance distribution map with a resolution of 1000x1000 pixels was generated using the bicubic spline interpolation algorithm, with an interpolation node spacing of 0.1mm. A 5x5 pixel sliding window with a step size of 1 pixel was used to scan the thermal resistance distribution map.The thermal resistance threshold is set to 0.08℃ / W. If the average thermal resistance of the window exceeds this value and is 20% higher than the surrounding area, it is marked as an excessive heat concentration area. The center coordinates and area of ​​such areas are recorded. The typical area of ​​excessive heat concentration areas is about 25-100 square millimeters. Deploy FLIRA655sc infrared thermal imager with a resolution of 640x480 pixels, a temperature measurement accuracy of ±0.5℃, and a sampling rate of 50Hz. Set the emissivity parameter to 0.95 to match the surface coating of the LED backlight. Use the Otsu threshold segmentation algorithm to extract the boundary of the excessive heat concentration area, and the threshold selection range is between 30%-70% of the grayscale value of the thermal image. The infrared thermal image and the thermal resistance distribution map are aligned by affine transformation, and the alignment error is controlled within ±2 pixels. The temperature data of the excessive heat concentration area is transmitted in real time with a data transmission rate of 10MB / s, and is transmitted to the data processing unit through the Ethernet interface. The whole process forms a closed-loop control, and the thermal resistance distribution and the position of the excessive heat concentration area are updated every 5 seconds to ensure the real-time monitoring accuracy of the working temperature of the LED backlight.

[0016] S103, based on the acquired real-time operating temperature data of the LED backlight source in the excessive heat concentration area, combined with a pre-established temperature and light decay model, a support vector machine regression algorithm is used to obtain an estimated value of the local light decay degree of the LED backlight source in the area at the current operating temperature.

[0017] Acquire the real-time working temperature data of the LED backlight source; pre-process the real-time working temperature data, including removing abnormal values; according to the pre-processed temperature data, use the sliding average filtering method to smooth the temperature data to obtain the filtered temperature data; perform normalization processing on the filtered temperature data to obtain normalized temperature data; according to the normalized temperature data and the pre-established temperature and light decay model, extract the features related to temperature and light decay, including the average temperature, the maximum temperature, the temperature fluctuation range and the temperature change rate; use the support vector machine regression algorithm to establish the LED backlight source light decay degree estimation The invention discloses a model, wherein the support vector machine regression algorithm uses a radial basis function as a kernel function; the hyperparameter C value and gamma value of the support vector machine regression algorithm are determined by a grid search method; the normalized temperature data and the characteristics related to temperature and light decay are input into the LED backlight source light decay degree estimation model to obtain the light decay degree estimation value of the LED backlight source; if the light decay degree estimation value exceeds a preset light decay degree threshold, an alarm signal is triggered; the normalized temperature data and the light decay degree estimation value are added to a historical data set; and the LED backlight source light decay degree estimation model is retrained at preset intervals.

[0018] Specifically, the real-time operating temperature data of the LED backlight source in the heat over-concentration area is obtained from the data acquisition system, and the temperature data is preprocessed, including removing outliers and smoothing. The numpy library of Python is used to implement sliding average filtering, and the arithmetic mean is taken for every 5 consecutive data points to eliminate the impact of short-term temperature fluctuations on subsequent analysis. The processed temperature data is normalized and the temperature value is mapped to the range of 0-1 for subsequent model processing. Combined with the pre-established temperature and light decay model, the features related to temperature and light decay are extracted. Statistical features are extracted from the temperature data, including average temperature, maximum temperature, temperature fluctuation range and temperature change rate. According to the temperature increase characteristics caused by thermal plume disturbance, the proportion of time that the temperature exceeds the normal operating temperature and the temperature over-limit amplitude are calculated. Considering the temperature fluctuation caused by thermal plume disturbance, the standard deviation of the temperature is calculated as an additional feature to reflect the impact of thermal plume on temperature stability. The activation energy Ea, pre-exponential factor A and luminous flux decay rate β in the temperature light decay model are extracted as features. These features are combined to construct the input feature vector for the training of the support vector machine regression model. The support vector machine regression algorithm is used to establish the LED backlight light decay estimation model. The radial basis function RBF is used as the kernel function, and the optimal hyperparameters C and gamma are determined by the grid search method. The C value is in the range of 10^-2 to 10^2, and the gamma value is in the range of 10^-4 to 1. Both are searched on a logarithmic scale with a step size of multiples of 10. The preprocessed temperature data and the extracted features are used as input, and the historical light decay data is used as the training label to train the support vector machine regression model. During the training process, the K-fold cross-validation method is used to evaluate the model performance and select the model parameters with the smallest root mean square error. The real-time temperature data is input into the trained support vector machine regression model to obtain the local light decay estimation value of the LED backlight in the area at the current operating temperature. According to the specifications provided by the LED manufacturer, the light flux drops to 70% of the initial value as the light decay threshold. When the estimated value exceeds the threshold, an alarm signal is triggered. At the same time, the new temperature data and the corresponding light decay estimation value are added to the historical data set, and the model is updated regularly to improve the prediction accuracy. The support vector machine regression model is retrained every 100 hours of operation to ensure that the model can adapt to the characteristics of LED backlight performance changing over time. In the area where the heat of the LED backlight is excessively concentrated, the operating temperature data is collected in real time, and the sampling frequency is set to 10Hz. The sliding average filter is implemented using Python's numpy library, and the arithmetic mean is taken for every 5 consecutive data points to filter out high-frequency noise. The temperature data range is usually between 60℃ and 90℃, and it is mapped to the 0-1 interval through linear normalization. Features are extracted from the processed temperature data, including the average temperature, maximum temperature, temperature fluctuation range, and temperature change rate within 1 minute. The proportion of time when the temperature exceeds 80℃ and the average value of the excess amplitude are calculated. The temperature standard deviation is used as an indicator of thermal plume disturbance.Combined with the temperature light decay model, the activation energy Ea, pre-exponential factor A and luminous flux decay rate β are extracted. A 17-dimensional feature vector is constructed as input to the support vector machine regression model. The grid search method is used to optimize the hyperparameters, with the C value in the range of [0.01, 0.1, 1, 10, 100] and the gamma value in the range of [0.0001, 0.001, 0.01, 0.1, 1]. A 5-fold cross validation is used to select the parameter combination with the smallest root mean square error. The training data set contains 1000 hours of historical data, with one sample per hour. The model outputs an estimate of the degree of light decay, ranging from 0 to 100%. The light decay threshold is set to 30%, corresponding to the light flux falling to 70% of the initial value. The light decay estimate is updated every 10 minutes, and an alarm is triggered when the estimated value exceeds the threshold for three consecutive times. New data is added to the historical data set in a sliding window manner, keeping the data set size constant at 1000 samples. The model is automatically retrained every 100 hours of running time to ensure that the model adapts to changes in LED performance.

[0019] Use an infrared thermal imager to scan the area where the heat of the LED backlight source is excessively concentrated, obtain a temperature distribution diagram, take multiple temperature sampling points, record the temperature data at different time points, analyze the temperature change trend, input the temperature data into a pre-trained light decay model, and obtain the estimated light decay degree of the LED in the area.

[0020] A high-precision infrared thermal imager is used to scan the excessive heat concentration area of ​​the LED backlight source to obtain a temperature distribution map. According to the temperature distribution map, the threshold-based regional growing algorithm is used to divide the LED chip and phosphor layer areas, where the area with the largest temperature gradient is determined as the seed point. The temperature data obtained from the LED chip and phosphor layer area is analyzed in time series, and the temperature change trend is calculated by the sliding window method. If the temperature mean of three consecutive windows shows an upward trend and the rising rate exceeds the preset threshold, the temperature abnormality alarm is triggered. Based on the spatial distribution characteristics of the temperature data, the temperature spatial correlation of the excessive heat concentration area is calculated. The spatial autocorrelation index Moran's I is used to evaluate the degree of temperature aggregation in space. If the Moran's I index is greater than the preset threshold, it is determined that there is a significant temperature spatial aggregation phenomenon. The processed temperature data is input into a pre-trained light decay model, which adopts a long short-term memory network structure and includes two layers of LSTM units. If the estimated light decay degree exceeds the preset threshold, the light decay warning is triggered, and the current temperature state and the estimated light decay degree are recorded.

[0021] Specifically, a high-precision infrared thermal imager is used to scan the excessive heat concentration area of ​​the LED backlight source to obtain the temperature distribution map. The scanning resolution of the infrared thermal imager is set to 640x480 pixels, the temperature accuracy is ±0.1℃, and the scanning frequency is 10Hz. A threshold-based regional growing algorithm is used to divide the LED chip and phosphor layer areas with the area with the largest temperature gradient as the seed point. The average temperature, maximum temperature and temperature gradient of the two areas are recorded respectively. Five temperature sampling points are selected in each area, and the temperature data of these sampling points at different times are recorded. The acquired temperature data is analyzed in time series to calculate the temperature change trend of the LED chip and phosphor layer. Using the sliding window method, the window size is set to 60 seconds and the step size is 10 seconds. The temperature mean, variance and linear regression slope in each window are calculated. By comparing these statistics of adjacent windows, it is determined whether the temperature is on an upward trend. According to the maximum allowable temperature change rate in the LED specification, the temperature abnormality alarm threshold is set, and the typical value is 0.5-1℃ / minute. If the temperature mean of three consecutive windows shows an upward trend and the rising rate exceeds the set threshold, the temperature abnormality alarm is triggered. Based on the spatial distribution characteristics of temperature data, the temperature spatial correlation of the excessive heat concentration area is calculated. The spatial autocorrelation index Moran's I is used to evaluate the degree of temperature aggregation in space. If the Moran's I index is greater than 0.5, it indicates that there is a significant temperature spatial aggregation phenomenon, and the hot spot area is further divided. The Getis-OrdGi* statistic is used to calculate the hot spot score of each pixel point. A score greater than 2 indicates a significant hot spot area. The identified hot spot area information is input as a feature into the subsequent light decay model. The processed temperature data is input into the pre-trained light decay model to obtain the estimated light decay degree of the LED in the area. The light decay model adopts a long short-term memory network LSTM structure, which contains two layers of LSTM units, each layer has 64 neurons, and is followed by a fully connected layer to output the light decay degree prediction value. The input features include LED chip temperature, phosphor layer temperature, temperature change rate, temperature spatial aggregation index and hot spot area information. The model outputs a light decay degree value between 0 and 100, where 0 indicates no light decay and 100 indicates complete light decay. If the estimated light decay exceeds the preset threshold, the light decay warning is triggered, and the current temperature status and estimated light decay are recorded for subsequent model updates and maintenance decisions. In the area where the heat of the LED backlight source is excessively concentrated, the FLIRA8580 infrared thermal imager is deployed to scan the temperature distribution, and the resolution of 640x480 pixels, the temperature accuracy of ±0.1°C, and the scanning frequency of 10Hz are set. The threshold-based regional growing algorithm is used to divide the LED chip and phosphor layer with the area with a temperature gradient greater than 5°C / mm as the seed point. The average temperature of the two areas is recorded as 85°C for the chip and 70°C for the phosphor, the highest temperature is 95°C for the chip and 80°C for the phosphor, and the temperature gradient is 10°C / mm for the chip and 5°C / mm for the phosphor. Five temperature sampling points are selected in each area and recorded once per second.The temperature trend is analyzed using a sliding window method with a 60-second window and a 10-second step size, and the mean, variance, and linear regression slope are calculated. If the mean temperature of three consecutive windows rises and the rate exceeds 0.8°C / minute, an abnormal alarm is triggered. The Moran's I index is calculated to evaluate the spatial correlation of temperature. When the index exceeds 0.5, the Getis-OrdGi* statistic is used to identify hot spots. A score greater than 2 is defined as a significant hot spot. The temperature data, trend characteristics, and hot spot information are input into the LSTM light decay model, which contains two layers of 64-neuron LSTM units and a fully connected output layer. The input features include LED chip temperature, phosphor temperature, temperature change rate, Moran's I index, and hot spot area ratio. The model outputs a light decay value of 0-100, and an early warning is triggered when it exceeds 20.

[0022] S104, obtaining the overall average brightness value of the LED backlight source, and comparing it with the estimated value of the local light decay degree. If the local light decay degree exceeds a preset threshold value of the overall average brightness, it is determined that the LED backlight source has a local brightness unevenness phenomenon.

[0023] Obtain the overall brightness distribution diagram of the LED backlight source, wherein the overall brightness distribution diagram is obtained by scanning the LED backlight source comprehensively with a high-precision brightness meter. Establish a brightness-light decay correspondence diagram based on the overall brightness distribution diagram, wherein the brightness-light decay correspondence diagram is generated based on the previously obtained local light decay degree estimation value. Divide the surface of the LED backlight source into several areas, each area corresponding to a local light decay degree estimation value. Convert the discrete light decay degree estimation value into a continuous light decay distribution diagram through an interpolation algorithm. Use an exponential decay model

[0024] ,

[0025] Where L0 is the initial brightness, β is the attenuation coefficient, and t is the usage time. The degree of light decay is converted into the corresponding brightness attenuation ratio. ANSYS Fluent software is used to simulate the thermal flow field to obtain a thermal plume distribution map. The thermal plume distribution map is aligned with the brightness-light decay correspondence map to identify abnormal areas. If the brightness attenuation degree of the abnormal area exceeds the preset threshold of the overall average brightness, it is determined that there is significant brightness unevenness in the abnormal area. The correlation coefficient between the thermal plume distribution map and the abnormal area is calculated. If the correlation coefficient is greater than the preset value, it is determined that the thermal plume disturbance is the main factor causing the local brightness unevenness. For the abnormal area, an adaptive histogram equalization algorithm is used to enhance the local contrast and generate a brightness unevenness heat map.

[0026] Specifically, the LED backlight source is fully scanned by a high-precision brightness meter to obtain the overall brightness distribution map. The grid division method is used to divide the surface of the LED backlight source into 20×20 measurement units, each unit is 5mm×5mm in size, to balance the measurement accuracy and efficiency. The stepper motor is used to control the brightness meter to move in the XY plane to accurately locate and measure the brightness of each measurement unit. The brightness value of each measurement unit is recorded to form a two-dimensional brightness data matrix. The arithmetic average of all brightness values ​​in the matrix is ​​performed to obtain the average brightness value of the entire LED backlight source. Combined with the previously obtained local light decay degree estimation value, a brightness-light decay correspondence diagram is established. By consulting the LED specification sheet or conducting new product testing, the theoretical initial brightness value L0 of the LED is obtained, and the light decay degree is converted into the corresponding brightness decay ratio using an exponential decay model. The ratio of the actual brightness value of each area to the theoretical initial brightness value is calculated to obtain the local brightness decay degree. The thermal flow field simulation is performed using ANSYS Fluent software to obtain the thermal plume distribution map, and the mutual information method is used to align it with the light decay distribution map, and the influence of the thermal plume on the light decay is comprehensively considered. Compare the local brightness attenuation with the overall average brightness to identify abnormal areas. Set brightness uniformity evaluation indicators, such as relative standard deviation. Calculate the deviation of the brightness value of each area relative to the overall average brightness and calculate the RSD value. If the RSD value exceeds the preset threshold, it is determined that there is brightness unevenness. At the same time, the adaptive histogram equalization algorithm is used to enhance the local contrast and generate a brightness unevenness heat map. Analyze the impact of thermal plume disturbance on local brightness. Spatial registration of the previously obtained thermal plume simulation results and the brightness unevenness heat map is performed, and the correlation coefficient between the two is calculated. If the correlation coefficient is greater than 0.8, it is determined that the thermal plume disturbance is the main factor causing the local brightness unevenness. Based on the registration results, identify the areas most severely affected by the thermal plume and mark these areas as key monitoring objects. Directly compare the local light decay degree with the overall average brightness. If the local light decay degree exceeds the preset threshold of the overall average brightness, it is further confirmed that there is significant brightness unevenness in the area. In the process of LED backlight brightness uniformity detection, Konica Minolta CA-310 display measuring instrument is used for high-precision brightness scanning, setting a 20x20 grid, each grid unit is 5mmx5mm. Use a 2-axis stepper motor to control the measuring instrument, XY plane resolution 0.1mm, scanning speed 10mm / s. The brightness range is 300-500cd / m², forming a 400x400 pixel brightness data matrix, and the average brightness is calculated to be 420cd / m². Combined with the estimated value of local light decay, the bicubic spline interpolation algorithm is used to generate a continuous light decay distribution diagram. The exponential decay model L(t)=500*exp(-0.0001t) is used, and the initial brightness L0 is obtained from the LED specification as 500cd / m², and t is the use time.ANSYS Fluent 19.2 software was used to simulate the heat plume, with 1 million grids, k-ε as the turbulence model, and the boundary condition as the heat flux density of the LED chip of 1W / cm². The heat plume distribution was superimposed on the light decay distribution by the mutual information registration algorithm. The calculated relative standard deviation was 7.2%, which exceeded the 5% threshold, and it was determined that there was brightness unevenness. The CLAHE algorithm was used with a block size of 8x8 and a contrast limit threshold of 0.02 to generate an enhanced brightness unevenness heat map. The Pearson correlation coefficient between the heat plume distribution and the brightness unevenness heat map was 0.87, which exceeded the 0.8 threshold, confirming that the heat plume was the main influencing factor. Three areas severely affected by the heat plume were identified, with areas of 25mm², 30mm² and 20mm², respectively. The local light decay levels were 28%, 25% and 22% lower than the overall average brightness, respectively, all exceeding the 20% threshold and were marked as key monitoring objects.

[0027] S105. If the LED backlight source has local uneven brightness, the real-time temperature data of the PCB board and electronic components corresponding to the excessive heat concentration area are obtained, and the abnormal temperature range of the components that causes the local uneven brightness is determined by using the BP neural network algorithm in combination with the pre-established component temperature and brightness uniformity model.

[0028] A temperature distribution diagram of an LED backlight PCB board is obtained, wherein the temperature distribution diagram is obtained by performing a real-time temperature scan of the PCB board by a high-precision infrared thermal imager; a component area is divided according to the temperature distribution diagram, wherein the component area is obtained by dividing the temperature distribution diagram into different areas by a regional growing algorithm, wherein each area corresponds to an electronic component on the PCB board; a relationship model between component temperature and local brightness uniformity is established, wherein the relationship model is established by a polynomial regression method, wherein the local brightness uniformity is used as a dependent variable and the component temperature is used as an independent variable; a BP neural network model is trained, wherein the BP neural network model is used to determine an abnormal temperature range of a component that causes local brightness unevenness, wherein the input data of the BP neural network model includes features extracted from the relationship model between component temperature and local brightness uniformity; and a degree of abnormal temperature of a component is determined, wherein the degree of abnormal temperature is obtained by inputting component temperature data obtained in real time into the BP neural network model, and if the degree of abnormal temperature exceeds a preset threshold, the corresponding component is marked as an abnormal component.

[0029] Specifically, a high-precision infrared thermal imager is used to perform real-time temperature scanning on the PCB board of the LED backlight source to obtain the temperature distribution map of the excessive heat concentration area. The scanning resolution of the infrared thermal imager is set to 640x480 pixels, the temperature accuracy is ±0.1℃, and the scanning frequency is 10Hz. The adaptive threshold segmentation algorithm is used to define the area where the temperature is 2 standard deviations higher than the average temperature as the excessive heat concentration area. The temperature distribution map is divided into different regions by the regional growing algorithm, and each region corresponds to an electronic component on the PCB board. The average temperature, maximum temperature and temperature gradient of each component area are recorded to form a component temperature data set. Combined with the pre-established component temperature and brightness uniformity model, the input feature vector is constructed. Key features are extracted from the temperature data set, including the average temperature, maximum temperature, temperature gradient of each component, and the temperature difference between adjacent components. The polynomial regression model is used to establish the relationship between component temperature and local brightness uniformity. The model form is U=a0+a1T+a2T^2, where U is brightness uniformity, T is component temperature, and a0, a1, and a2 are regression coefficients. According to the model, the theoretical brightness uniformity value corresponding to the temperature of each component is calculated. These features are combined with the actual measured local brightness unevenness to form the input data of the BP neural network. The BP neural network model is designed and trained to determine the abnormal temperature range of the components that cause local brightness unevenness. A three-layer BP neural network is constructed. The number of neurons in the input layer is the same as the number of features. The hidden layer uses 50 neurons and adopts the ReLU activation function. The output layer has 1 neuron and uses the Sigmoid activation function to represent the abnormal temperature degree of the component. The model is trained using historical data, the mean square error is used as the loss function, and the network parameters are optimized using the Adam optimizer. The initial learning rate is set to 0.001, and the learning rate decays by 10% every 100 epochs. The number of training rounds is 1000 times until the model converges. The component temperature data obtained in real time is input into the trained BP neural network to obtain the abnormal temperature degree score of each component. The abnormal temperature degree threshold is set. When the score exceeds the threshold, the component is marked as abnormal. According to the maximum operating temperature Tmax in the component specification, the temperature range is divided into normal <0.8Tmax, mild abnormality 0.8Tmax-0.9Tmax, moderate abnormality 0.9Tmax-Tmax, and severe abnormality >Tmax. According to the actual temperature value of the abnormal component, determine the temperature range to which each abnormal component belongs. Generate an abnormal component report, including the component location, current temperature, temperature range and corresponding local brightness unevenness. In the process of detecting local brightness unevenness of LED backlight sources, the FLIRT1020 infrared thermal imager is used to scan the PCB board, setting 640x480 pixel resolution, ±0.1℃ temperature accuracy, and 10Hz scanning frequency.Using the Otsu adaptive threshold algorithm, the area with a temperature higher than the average value of 98.5℃ plus 2 standard deviations, that is, the area above 103.7℃, is defined as an excessive heat concentration area. The regional growing algorithm is applied, and the temperature gradient is greater than 5℃ / mm as the seed point to divide 15 areas corresponding to components. The average temperature, maximum temperature and temperature gradient of each component area are recorded. The relationship between temperature and brightness uniformity is established using the quadratic polynomial regression model U=-0.05T^2+3T-120, where T is the component temperature and U is the brightness uniformity. A BP neural network is constructed with 20 neurons in the input layer, 50 neurons in the hidden layer using the ReLU activation function, and 1 neuron in the output layer using the Sigmoid function. The Adam optimizer is used, with an initial learning rate of 0.001, a 10% decay every 100 epochs, and 1000 rounds of training until the loss function converges to below 0.001. Input real-time temperature data to obtain the abnormal temperature degree score, and set the threshold to 0.7. According to the maximum operating temperature Tmax of the components, the temperature range is divided into normal <96℃, mild abnormality 96-108℃, moderate abnormality 108-120℃, and severe abnormality >120℃. The generated report shows that the three abnormal components are LED driver chip 1, location X23Y45, temperature 112.5℃, moderate abnormality, local brightness unevenness 18.2%; capacitor C5, location X56Y78, temperature 98.3℃, mild abnormality, unevenness 12.5%; resistor R12, location X89Y34, temperature 122.1℃, severe abnormality, unevenness 25.7%.

[0030] Using a thermocouple array, multiple temperature measurement points are arranged on the surface of the PCB board to form a temperature distribution grid. An infrared camera is used to capture the thermal map of the entire PCB board. The collected temperature data is compared with the LED brightness distribution map to find the associated areas of temperature anomaly and uneven brightness. Based on historical data, the normal operating temperature range of various components is set, and the measured temperature is compared with the normal range to identify the components with abnormal temperature and their specific temperature range.

[0031] The invention relates to a method for obtaining surface layout data of a PCB board including position information of a thermocouple array, wherein the thermocouple array is used to form a temperature distribution grid; setting sampling parameters of a high-precision data acquisition module according to the surface layout data of the PCB board, wherein the high-precision data acquisition module is used to record real-time temperature data of each temperature measurement point; receiving real-time temperature data sent by the high-precision data acquisition module, and generating a temperature distribution matrix of the PCB board; performing thermal imaging of the PCB board using a high-resolution infrared camera to obtain thermal imaging data; performing image processing on the thermal imaging data, and identifying an area whose temperature is higher than a preset threshold of the surrounding area as a hot spot; obtaining brightness distribution data of an LED backlight source, and performing pixel-level alignment of the brightness distribution data with the temperature distribution matrix; calculating a correlation coefficient between the brightness distribution data and the temperature distribution matrix, and if the correlation coefficient of a local area exceeds a preset threshold, marking the area as an associated area of ​​abnormal temperature and uneven brightness; receiving database information including a normal working temperature range of components; comparing the real-time temperature data with the normal temperature range in the database information, and if the temperature of a component exceeds its normal range, marking it as an abnormal component; and determining the abnormal level of the abnormal component according to the temperature deviation degree of the abnormal component.

[0032] Specifically, a 20x20 array of thermocouples is arranged on the surface of the PCB board to form a temperature distribution grid, with a spacing of 2.5mm between each thermocouple to ensure that all key areas are covered. A high-precision data acquisition module is used, and the sampling frequency is set to 1Hz, while recording the real-time temperature data of each temperature measurement point. For LED driver ICs, capacitors, and resistors, micro temperature probes are installed on their surfaces, and the sampling frequency is increased to 10Hz to obtain more accurate temperature changes. All temperature data are aggregated to the central processing unit to generate a real-time temperature distribution matrix of the PCB board. A high-resolution infrared camera is used to perform thermal imaging of the entire PCB board, and the shooting frequency is set to 1Hz. FLIRResearchIR software is used for image acquisition and preprocessing, and the thermal map is analyzed, including median filtering denoising and Otsu adaptive threshold segmentation. Through the hotspot detection algorithm, areas with temperatures higher than 2°C above the surrounding area are identified as hotspots. The identified hotspot areas are spatially aligned with the PCB component layout diagram to determine the specific components corresponding to each hotspot. A hotspot area report is generated, including the hotspot location coordinates, area, average temperature, and maximum temperature. Obtain the brightness distribution map of the LED backlight source, and use a high-precision brightness meter to perform scanning measurements under the same grid division as the thermocouple array. Align the brightness distribution map with the temperature distribution matrix at the pixel level and calculate the correlation coefficient between the two. If the correlation coefficient between the temperature and brightness of a local area exceeds 0.8, the area is marked as a correlation area between temperature anomaly and uneven brightness. Use the watershed algorithm to extract the contour of the associated area and mark it on the original PCB board image. According to historical operating data and component specifications, establish a database of normal operating temperature ranges for various components. For LED driver ICs, the normal temperature range is set to 60-85℃; the normal temperature range of capacitors is 40-70℃; and the normal temperature range of resistors is 50-80℃. Use the average temperature value of a 5-minute sliding window for abnormal judgment to eliminate the impact of instantaneous fluctuations. Compare the measured temperature data with the preset normal temperature range. If the temperature of a component exceeds its normal range, it is marked as abnormal. According to the degree of temperature deviation, the abnormal temperature is divided into three intervals: mild abnormality exceeding 0-5℃, moderate abnormality exceeding 5-10℃, and severe abnormality exceeding 10℃. Generate an abnormal component report, including component type, location, measured temperature, normal temperature range and degree of abnormality. Deploy a 20×20=400-point thermocouple array on the LED backlight PCB board, and use an Agilent34970A data logger to collect temperature data at a frequency of 1Hz. At the same time, install OmegaSA1XL-K-SRTC micro-thermocouples on the surface of LED driver IC, capacitors and resistors to collect key component temperatures at a frequency of 10Hz. Use a FLIRA655sc infrared camera to capture thermal images at a frequency of 1Hz with a resolution of 640x480 pixels.The heat map was denoised by 5x5 median filtering using FLIR Research IR software, and then the Otsu algorithm was applied for adaptive threshold segmentation to identify hotspots with a temperature higher than 2°C around the surrounding area. The hotspot area and the PCB layout were spatially registered by affine transformation with an accuracy of ±0.5mm. The Konica Minolta CA-310 display brightness meter was used to scan the LED backlight brightness under the same 20x20 grid with a resolution of 0.01cd / m². The brightness distribution map and the temperature distribution matrix were aligned by bilinear interpolation, and the Pearson correlation coefficient was calculated. For areas with a correlation coefficient greater than 0.8, the watershed algorithm was applied to extract the contour. According to the component specifications, the normal temperature range of the LED driver ICTPS61169 was set to 60-85°C, the capacitor C1210 was set to 40-70°C, and the resistor R0805 was set to 50-80°C. The average temperature was calculated using a 5-minute sliding window and compared with the preset range. The abnormal components identified are LED driver IC, location X23Y45, temperature 88.5℃, mild abnormality; capacitor C5, location X56Y78, temperature 72.3℃, mild abnormality; resistor R12, location X89Y34, temperature 83.1℃, moderate abnormality.

[0033] S106. According to the abnormal temperature range of the components, the heat dissipation efficiency of the radiator is dynamically adjusted by controlling the speed of the cooling fan and the opening and closing size of the vents, so that the temperature of the components returns to the normal range.

[0034] A heat dissipation control model is established, and a fuzzy control algorithm is used. The temperature deviation of components, the temperature change rate and the ambient temperature are used as input variables, and the speed of the cooling fan and the opening degree of the vents are used as output variables. The fuzzy set of temperature deviation is set to {high, medium, low}, the fuzzy set of temperature change rate is set to {fast, medium, slow}, and the fuzzy set of ambient temperature is set to {high, medium, low}. According to the control quantity obtained by the fuzzy control algorithm, ANSYS Fluent is used for CFD simulation, and the k-ε turbulence model is used to analyze the heat dissipation requirements of the local area. The velocity field and temperature field distribution of the heat plume are obtained through the CFD simulation, and the heat accumulation area and airflow dead corner are identified. According to the CFD analysis results, an adjustable radiator structure is designed, and the radiator structure includes a multi-section independently controlled cooling fan array, and the speed of each fan can be adjusted separately. For the heat accumulation area, the speed of the fan at the corresponding position is increased, and the angle of the corresponding guide plate is adjusted. Monitor the temperature changes of components in real time, use the PID control algorithm, take the deviation between the target temperature and the actual temperature as input, and calculate the adjustment of the fan speed and the guide plate angle. If the temperature of the component continues to exceed the normal range, trigger the adaptive control mechanism, and use the particle swarm optimization algorithm to adjust the PID parameters in real time. Minimize the temperature control error as the objective function, search for the optimal PID parameters, and set the number of particles of the particle swarm optimization algorithm to the preset value, and the number of iterations to the preset value.

[0035] Specifically, a heat dissipation control model is established based on the abnormal temperature range data of components. A fuzzy control algorithm is used, with component temperature deviation, temperature change rate and ambient temperature as input variables, and cooling fan speed and vent opening degree as output variables. The fuzzy set of temperature deviation is set to {high, medium, low}, the fuzzy set of temperature change rate is set to {fast, medium, slow}, the fuzzy set of ambient temperature is set to {high, medium, low}, the fuzzy set of fan speed is set to {high, medium, low}, and the fuzzy set of vent opening degree is set to {large, medium, small}. A fuzzy rule base is formulated, such as "If the temperature deviation is large, the change rate is fast and the ambient temperature is high, then the fan speed is high and the vent opening degree is large". The centroid method is used for defuzzification to obtain specific control quantities. ANSYS Fluent is used for CFD simulation, and the k-ε turbulence model is used to analyze the heat dissipation requirements of the local area. A three-dimensional model of the PCB board and its surrounding environment is established. Boundary conditions are set, including component heat source, ambient temperature and initial airflow velocity. Run CFD simulation to obtain the velocity field and temperature field distribution of the heat plume. According to the simulation results, identify the heat accumulation area and airflow dead corner, and determine the local area where heat dissipation needs to be strengthened. Use the CFD analysis results directly to guide the next step of heat sink structure design. Design an adjustable heat sink structure to achieve dynamic enhancement of local heat dissipation capacity. Use a multi-segment independently controlled cooling fan array, and the speed of each fan can be adjusted individually. Install a variable angle guide plate on the surface of the heat sink, and control the inclination of the guide plate by a servo motor, ranging from -30 to 60 degrees to cope with more complex heat dissipation requirements. According to the CFD analysis results, increase the speed of the fan at the corresponding position of the heat accumulation area, and adjust the angle of the corresponding guide plate to guide the cold air to the hot spot area, effectively dealing with the impact of the heat plume on local heat dissipation. Monitor the temperature changes of components in real time and dynamically optimize the heat dissipation strategy. Use the PID control algorithm to calculate the adjustment amount of the fan speed and the guide plate angle with the deviation between the target temperature and the actual temperature as input. The initial PID parameters are set to proportional coefficient Kp=0.5, integral coefficient Ki=0.1, and differential coefficient Kd=0.05. The temperature data is collected every 10 seconds to update the PID control output. If the temperature continues to exceed the normal range, the adaptive control mechanism is triggered, and the particle swarm optimization algorithm is used to adjust the PID parameters in real time. The objective function is to minimize the temperature control error and search for the optimal PID parameters. Parameter optimization is performed every 100 control cycles, the number of particles is set to 20, and the number of iterations is 50. The heat dissipation effect is regularly evaluated, and the temperature reduction rate and energy efficiency are calculated as the basis for optimizing the heat dissipation strategy. In the LED backlight heat dissipation control system, the TSic506F temperature sensor is first used to monitor the component temperature in real time with an accuracy of ±0.1℃. The fuzzy control algorithm takes temperature deviation, change rate and ambient temperature as inputs, and sets 7 membership functions.For example, when the temperature deviation is 5°C, the rate of change is 2°C / minute, and the ambient temperature is 35°C, the fan speed is output as 3000RPM and the vent opening is 80% according to the rule base. ANSYS Fluent 2021R1 was used for CFD simulation, with a grid cell number of 1 million and an RNGk-ε turbulence model. The simulation results show that there is a hot spot of 92°C around the LED driver IC, and the air flow velocity is only 0.2m / s. Based on this, the radiator structure is designed, and an array of 5 DeltaAFB0612EH-F00 fans is used with a maximum speed of 6000RPM. Five 50mm×30mm aluminum guide plates are installed, and the angle is controlled by the MG996R servo, ranging from -30° to 60°. According to the CFD results, the fan speed corresponding to the hot spot area is increased to 5500RPM, and the guide plate angle is adjusted to 45° so that the cold airflow directly covers the hot spot. The initial parameters of PID control are set to Kp=0.5, Ki=0.1, and Kd=0.05. Temperature data is collected every 10 seconds. If the temperature exceeds 90°C for 5 consecutive times, the adaptive mechanism is triggered. The particle swarm optimization algorithm uses 20 particles and iterates 50 times. The search range is Kp=0.1-1.0, Ki=0.05-0.5, and Kd=0.01-0.1. The optimized PID parameters are Kp=0.7, Ki=0.15, and Kd=0.03, which stabilizes the temperature at 85±2°C. The system calculates the heat dissipation effect every hour, and the temperature reduction rate reaches 15°C / hour, and the energy efficiency is 0.8°C / W·h.

[0036] It should be noted that the above examples are only some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.

Claims

1. A method for analyzing heat dissipation of an LED backlight source, characterized in that: The method comprises: Obtaining the tilt angle information of the LED backlight source, and judging whether the LED backlight source is in a horizontal state that is conducive to natural convection heat dissipation or in an inclined state that causes the formation of heat plumes to be blocked according to a preset angle threshold; if it is judged to be in an inclined state, obtaining the temperature distribution data of the heat sink surface; Based on the obtained radiator surface temperature distribution data, the heat balance equation is used to calculate the thermal resistance distribution on the radiator surface, and the area with the largest local thermal resistance is obtained. This area is determined to be an area of ​​excessive heat concentration, and the real-time operating temperature data of the LED backlight source in this area is obtained by infrared temperature measurement; Based on the real-time operating temperature data of the LED backlight source in the excessive heat concentration area, combined with the pre-established temperature and light decay model, the support vector machine regression algorithm is used to obtain the estimated value of the local light decay degree of the LED backlight source in the area at the current operating temperature; Obtain the overall average brightness value of the LED backlight source, and compare it with the estimated value of the local light decay degree. If the local light decay degree exceeds the preset threshold value of the overall average brightness, it is determined that the LED backlight source has a local brightness unevenness phenomenon; If the LED backlight source has local uneven brightness, the real-time temperature data of the PCB board and electronic components corresponding to the excessive heat concentration area is obtained, and the BP neural network algorithm is used to determine the abnormal temperature range of the components that cause the local uneven brightness, combined with the pre-established component temperature and brightness uniformity model; According to the abnormal temperature range of components, the heat dissipation efficiency of the radiator is dynamically adjusted by controlling the speed of the cooling fan and the opening and closing size of the vents, so that the temperature of the components returns to the normal range; The method of obtaining the real-time operating temperature data of the LED backlight source in the excessive heat concentration area, combining the pre-established temperature and light decay model, and using the support vector machine regression algorithm to obtain the estimated value of the local light decay degree of the LED backlight source in the area at the current operating temperature includes: Get the real-time working temperature data of LED backlight source; Preprocessing the real-time operating temperature data, including removing abnormal values; According to the preprocessed temperature data, a sliding average filtering method is used to smooth the temperature data to obtain filtered temperature data; Performing normalization processing on the filtered temperature data to obtain normalized temperature data; Extracting temperature- and light-attenuation-related features according to the normalized temperature data and a pre-established temperature and light-attenuation model, including an average temperature, a maximum temperature, a temperature fluctuation range, and a temperature change rate; A support vector machine regression algorithm is used to establish a light decay degree estimation model for LED backlight sources, wherein the support vector machine regression algorithm uses a radial basis function as a kernel function; Determine the hyperparameter C value and gamma value of the support vector machine regression algorithm by grid search method; Inputting the normalized temperature data and the characteristics related to temperature and light decay into the LED backlight source light decay degree estimation model to obtain a light decay degree estimation value of the LED backlight source; If the estimated value of the light decay degree exceeds a preset light decay degree threshold, an alarm signal is triggered; Adding the normalized temperature data and the light attenuation level estimation value to a historical data set; Retraining the LED backlight source light decay estimation model at preset intervals; also comprising: using an infrared thermal imager to scan the LED backlight source heat over-concentration area, obtaining a temperature distribution diagram, taking multiple temperature sampling points, recording temperature data at different time points, analyzing the temperature change trend, inputting the temperature data into a pre-trained light decay model, and obtaining an estimated light decay degree of the LED in the area; The method includes using an infrared thermal imager to scan the excessive heat concentration area of ​​the LED backlight source, obtaining a temperature distribution diagram, taking multiple temperature sampling points, recording temperature data at different time points, analyzing the temperature change trend, and inputting the temperature data into a pre-trained light decay model to obtain an estimated light decay degree of the LED in the area, including: Use a high-precision infrared thermal imager to scan the excessive heat concentration area of ​​the LED backlight source to obtain the temperature distribution map; According to the temperature distribution map, the LED chip and phosphor layer regions are divided using a threshold-based region growing algorithm, wherein the region with the largest temperature gradient is determined as a seed point; Performing time series analysis on the temperature data obtained from the LED chip and phosphor layer area, and calculating the temperature change trend by a sliding window method; If the temperature average of three consecutive windows shows an upward trend and the rising rate exceeds the preset threshold, the temperature abnormality alarm is triggered; Calculating the temperature spatial correlation of the heat over-concentration area based on the spatial distribution characteristics of the temperature data; The spatial autocorrelation index Moran's I is used to evaluate the degree of temperature aggregation in space. If the Moran's I index is greater than a preset threshold, it is determined that there is a significant temperature spatial aggregation phenomenon; The processed temperature data is input into a pre-trained light attenuation model, which adopts a long short-term memory network structure and includes two layers of LSTM units; If the estimated light decay exceeds the preset threshold, a light decay warning is triggered, and the current temperature status and estimated light decay are recorded; If the LED backlight source has a local uneven brightness phenomenon, the real-time temperature data of the PCB board and the electronic components corresponding to the excessive heat concentration area are obtained, and the abnormal temperature range of the components that cause the local uneven brightness is determined by using the BP neural network algorithm in combination with the pre-established component temperature and brightness uniformity model, including: Obtaining a temperature distribution diagram of an LED backlight PCB board, wherein the temperature distribution diagram is obtained by performing a real-time temperature scan of the PCB board by a high-precision infrared thermal imager; Dividing component regions according to the temperature distribution map, wherein the component regions are obtained by dividing the temperature distribution map into different regions using a region growing algorithm, wherein each region corresponds to an electronic component on the PCB; Establishing a relationship model between component temperature and local brightness uniformity, wherein the relationship model is established using a polynomial regression method, wherein the local brightness uniformity is used as a dependent variable and the component temperature is used as an independent variable; Training a BP neural network model, the BP neural network model is used to determine the abnormal temperature range of the component that causes the local uneven brightness, wherein the input data of the BP neural network model includes features extracted from the relationship model between the component temperature and the local brightness uniformity; Determine the abnormal temperature degree of the component, the abnormal temperature degree is obtained by inputting the component temperature data acquired in real time into the BP neural network model, and if the abnormal temperature degree exceeds a preset threshold, the corresponding component is marked as an abnormal component; It also includes: using a thermocouple array to arrange multiple temperature measurement points on the surface of the PCB board to form a temperature distribution grid, using an infrared camera to capture a thermal map of the entire PCB board, comparing the collected temperature data with the LED brightness distribution map, finding the associated areas of temperature anomalies and uneven brightness, setting the normal operating temperature range of various components based on historical data, comparing the measured temperature with the normal range, and identifying components with abnormal temperatures and their specific temperature ranges.

2. The method according to claim 1, characterized in that: The step of obtaining the tilt angle information of the LED backlight source and judging whether the LED backlight source is in a horizontal state that is conducive to natural convection heat dissipation or in a tilted state that causes the formation of heat plumes to be blocked according to a preset angle threshold, and if it is judged to be in a tilted state, obtaining the temperature distribution data of the heat sink surface includes: Obtaining the tilt angle of the LED backlight source, wherein the tilt angle is measured in real time by MEMS acceleration sensors installed at the four corners of the LED backlight source; According to the comparison between the tilt angle and a preset angle threshold, the current state of the LED backlight source is determined; If the tilt angle is greater than or equal to a preset angle threshold, it is determined that the LED backlight source is in a tilted state; For the LED backlight source determined to be in a tilted state, a heat sink surface temperature distribution data collection program is started; The radiator surface temperature distribution data acquisition program collects real-time temperature data through 16 PT100 platinum resistance temperature sensors evenly arranged on the radiator surface; Perform median filtering on the collected temperature data to generate a temperature data matrix reflecting the temperature distribution of the radiator surface; Performing hot spot analysis and temperature gradient calculation on the temperature data matrix, and generating a high-resolution temperature distribution thermodynamic map using a bilinear interpolation method; Identify local hot spots where the temperature is abnormally elevated through the temperature distribution thermodynamic map, and calculate the temperature gradient between the hot spots and surrounding areas using a finite difference method; Quantitatively assess the severity of local heat accumulation based on the difference between the temperature gradient and a preset threshold; Based on the evaluation result of the local heat accumulation, adjusting the heat dissipation control strategy of the LED backlight source; The heat dissipation control strategy includes increasing the speed of the radiator fan blades corresponding to the high-temperature hot spot area, increasing the coolant flow of the liquid cooling system, and adjusting the radiator airflow direction; The variable angle guide plate controlled by the servo motor adjusts the airflow direction according to the hot spot location, guiding the cold air to the high temperature area.

3. The method according to claim 1, characterized in that The method comprises: calculating the thermal resistance distribution of the heat sink surface by using a heat balance equation based on the obtained surface temperature distribution data of the heat sink, obtaining the area with the largest local thermal resistance, determining the area as an area of ​​excessive heat concentration, and obtaining the real-time operating temperature data of the LED backlight source in the area by infrared temperature measurement, including: A heat balance equation calculation model is established according to the acquired radiator surface temperature distribution data, wherein the calculation model divides the radiator surface into calculation grids; The heat conduction equation is discretized using the central difference format to construct a set of heat balance equations including heat conduction, convection and radiation heat transfer modes; Solving the heat balance equations by Gauss-Seidel iteration method to obtain the heat flux density distribution of each grid point; Calculating the thermal resistance distribution on the surface of the heat sink according to the heat flux density distribution and temperature distribution data, and generating a thermal resistance distribution matrix; The thermal resistance distribution matrix is ​​converted into a high-resolution thermal resistance distribution map using a bicubic spline interpolation algorithm; Analyze the thermal resistance distribution diagram and determine the area with the largest local thermal resistance using a sliding window method; If the average thermal resistance of the window exceeds a preset threshold and is higher than a preset ratio of the surrounding area, the area is marked as an excessive heat concentration area; For the excessively concentrated heat area, an infrared temperature measuring device is deployed to perform real-time temperature measurement; A boundary extraction algorithm based on threshold segmentation is used to automatically lock the boundary of the excessive heat concentration area and extract temperature data in the area; Performing image registration on the infrared thermal image and the thermal resistance distribution map to ensure that the temperature measurement area corresponds accurately to the excessive heat concentration area; The temperature distribution data of the excessive heat concentration area is transmitted in real time through the data acquisition interface.

4. The method according to claim 1, characterized in that The method of obtaining the average brightness value of the entire LED backlight source and comparing it with the estimated value of the local light decay degree, and if the local light decay degree exceeds a preset threshold value of the overall average brightness, then determining that the LED backlight source has a local brightness unevenness phenomenon, includes: Obtaining an overall brightness distribution diagram of the LED backlight source, wherein the overall brightness distribution diagram is obtained by performing a comprehensive scan of the LED backlight source with a high-precision brightness meter; Establishing a brightness-light decay correspondence diagram according to the overall brightness distribution diagram, wherein the brightness-light decay correspondence diagram is generated based on a previously obtained local light decay degree estimation value; Dividing the surface of the LED backlight source into a plurality of regions, each region corresponding to a local light decay degree estimation value; Through the interpolation algorithm, the discrete light attenuation estimation value is converted into a continuous light attenuation distribution diagram; Using exponential decay model ， Where L0 is the initial brightness, β is the attenuation coefficient, and t is the usage time. The light attenuation degree is converted into the corresponding brightness attenuation ratio; ANSYS Fluent software was used to simulate the thermal flow field and obtain the thermal plume distribution map; Aligning the thermal plume distribution map with the brightness-light attenuation correspondence map to identify abnormal areas; If the brightness attenuation of the abnormal area exceeds the preset threshold of the overall average brightness, it is determined that there is a significant brightness non-uniformity phenomenon in the abnormal area; Calculating a correlation coefficient between the thermal plume distribution map and the abnormal area, and if the correlation coefficient is greater than a preset value, determining that the thermal plume disturbance is the main factor causing the local brightness unevenness; For the abnormal area, an adaptive histogram equalization algorithm is used to enhance the local contrast and generate a brightness non-uniformity heat map.

5. The method according to claim 1, characterized in that The method uses a thermocouple array to arrange multiple temperature measurement points on the surface of the PCB board to form a temperature distribution grid, uses an infrared camera to capture a thermal map of the entire PCB board, compares the collected temperature data with the LED brightness distribution map, finds the associated area of ​​abnormal temperature and uneven brightness, sets the normal operating temperature range of various components based on historical data, compares the measured temperature with the normal range, and identifies the components with abnormal temperature and their specific temperature range, including: Acquiring PCB board surface layout data including position information of a thermocouple array, wherein the thermocouple array is used to form a temperature distribution grid; The sampling parameters of the high-precision data acquisition module are set according to the surface layout data of the PCB board, and the high-precision data acquisition module is used to record the real-time temperature data of each temperature measurement point; the real-time temperature data sent by the high-precision data acquisition module is received to generate a temperature distribution matrix of the PCB board; Using a high-resolution infrared camera to perform thermal imaging on the PCB board to obtain thermal imaging data; Performing image processing on the thermal imaging data to identify areas with a temperature higher than a preset threshold of surrounding areas as hot spots; Acquire brightness distribution data of the LED backlight source, and perform pixel-level alignment on the brightness distribution data and the temperature distribution matrix; Calculating the correlation coefficient between the brightness distribution data and the temperature distribution matrix, and if the correlation coefficient of a local area exceeds a preset threshold, marking the area as an associated area of ​​temperature anomaly and brightness unevenness; Receiving database information containing normal operating temperature ranges of components; The real-time temperature data is compared with the normal temperature range in the database information. If the temperature of a component exceeds its normal range, it is marked as an abnormal component; according to the temperature deviation degree of the abnormal component, the abnormal level of the abnormal component is determined.

6. The method according to claim 1, characterized in that The method of dynamically adjusting the heat dissipation efficiency of the radiator by controlling the speed of the cooling fan and the opening and closing size of the vents according to the abnormal temperature range of the components to return the temperature of the components to the normal range includes: A heat dissipation control model was established, using a fuzzy control algorithm, taking the temperature deviation and temperature change rate of components and the ambient temperature as input variables, and the speed of the cooling fan and the opening degree of the vents as output variables; According to the control quantity obtained by the fuzzy control algorithm, CFD simulation is performed using ANSYS Fluent, and the k-ε turbulence model is used to analyze the heat dissipation demand of the local area; The velocity field and temperature field distribution of the heat plume are obtained through the CFD simulation, and the heat accumulation area and airflow dead corner are identified; According to the CFD analysis results, an adjustable radiator structure is designed, wherein the radiator structure includes a multi-section independently controlled cooling fan array, and the speed of each fan can be adjusted individually; For the heat accumulation area, increase the speed of the fan at the corresponding position and adjust the angle of the corresponding guide plate; Monitor component temperature changes in real time, use PID control algorithm, take the deviation between target temperature and actual temperature as input, and calculate the adjustment of fan speed and deflector angle; If the temperature of the component continues to exceed the normal range, the adaptive control mechanism is triggered and the PID parameters are adjusted in real time using the particle swarm optimization algorithm; Taking minimization of temperature control error as the objective function, the optimal PID parameters are searched, the number of particles of the particle swarm optimization algorithm is set to a preset value, and the number of iterations is set to a preset value.

Images