A method and device for uniformly controlling the light-emitting color of an LED lamp bead and the LED lamp bead

Abstract

The present invention relates to the technical field of LED lamp bead light emission, and discloses a method, device and LED lamp bead for uniformly controlling the light emission color of an LED lamp bead. The method includes: arranging LED wafers in a triangular ring manner, and preparing an LED lamp bead through centrifugal treatment of a double-layer encapsulation glue; collecting an initial optical parameter matrix after applying rated working parameters, and establishing a standard data model; performing real-time monitoring on the LED lamp bead to obtain a monitoring data set; calculating a multi-dimensional deviation vector by using a spectral similarity algorithm, an Euclidean distance algorithm and a ratio algorithm; obtaining a comprehensive deviation evaluation value through weighted summation; generating a correction amount based on the evaluation value for closed-loop control until a control termination condition is reached. The method realizes precise control of the light emission color of the LED lamp bead, and improves the uniformity and stability of the light emission color of the LED lamp bead.

Description

Technical Field

[0001] The present invention relates to the technical field of LED lamp bead light emission, and particularly to a method and device for uniformly controlling the light emission color of LED lamp beads and an LED lamp bead. Background Art

[0002] The light emission uniformity and color stability of LED lamp beads have become key factors affecting product quality. Currently, during the production process of LED lamp beads, due to the influence of packaging processes, material properties, and environmental factors, problems such as uneven light emission color, obvious color spots, and spectral drift are widespread, which seriously affect the application quality of LED products in the fields of high-end display and lighting.

[0003] Traditional LED lamp bead packaging technologies mainly adopt a single-layer packaging structure, which is difficult to effectively control the distribution uniformity of phosphor and lacks a real-time monitoring and precise regulation mechanism for light emission color. Existing color control methods mostly use a single parameter as the control basis, and do not fully consider the synergistic relationship between spectral characteristics, chromaticity parameters, and luminous flux, resulting in insufficient control accuracy and stability and being unable to meet the technical requirements of high-quality LED products. Traditional solutions are often limited to static optimization and lack dynamic compensation capabilities. Especially during long-term use, due to the influence of factors such as temperature changes and device aging, the light emission characteristics of LED lamp beads will drift, and existing technologies are difficult to achieve real-time precise regulation, severely restricting the application range and service life of LED products. Summary of the Invention

[0004] The present invention provides a method and device for uniformly controlling the light emission color of LED lamp beads and an LED lamp bead, and the present invention improves the uniformity and stability of the light emission color of LED lamp beads.

[0005] In a first aspect, the present invention provides a method for uniformly controlling the light emission color of LED lamp beads, and the method for uniformly controlling the light emission color of LED lamp beads includes:

[0006] Placing LED wafers in a triangular ring pattern on an LED bracket, connecting wires to the LED wafers, and sequentially injecting a first encapsulant and a second encapsulant and performing centrifugation to obtain an LED lamp bead with a double-layer packaging structure;

[0007] Applying rated operating parameters to the LED lamp bead and collecting an initial optical parameter matrix;

[0008] Based on the initial optical parameter matrix, establishing a standard data model, and performing multi-parameter real-time monitoring on the LED lamp bead to obtain a monitoring data set;

[0009] Input the monitored data set into the data processing unit, calculate the spectral deviation coefficient through the spectral similarity algorithm, calculate the chromaticity deviation coefficient through the Euclidean distance algorithm, and calculate the luminous flux deviation coefficient through the ratio algorithm to obtain a multi-dimensional deviation vector;

[0010] Perform a weighted summation operation on the multi-dimensional deviation vector to obtain a comprehensive deviation evaluation value;

[0011] Based on the comprehensive deviation evaluation value, generate a current correction amount and a voltage correction amount, and perform closed-loop control on the LED lamp beads until the comprehensive deviation evaluation value is lower than the set threshold to obtain a control termination signal.

[0012] In a second aspect, the present invention provides a device for uniformly controlling the light-emitting color of an LED lamp bead. The device for uniformly controlling the light-emitting color of an LED lamp bead includes:

[0013] A packaging module for arranging LED wafers in a triangular ring manner on an LED bracket, connecting wires to the LED wafers, injecting a first packaging glue and a second packaging glue in sequence and performing centrifugal treatment to obtain an LED lamp bead with a double-layer packaging structure;

[0014] An acquisition module for applying rated operating parameters to the LED lamp bead and acquiring an initial optical parameter matrix;

[0015] A monitoring module for establishing a standard data model based on the initial optical parameter matrix, performing multi-parameter real-time monitoring on the LED lamp bead to obtain a monitored data set;

[0016] A calculation module for inputting the monitored data set into the data processing unit, calculating the spectral deviation coefficient through the spectral similarity algorithm, calculating the chromaticity deviation coefficient through the Euclidean distance algorithm, and calculating the luminous flux deviation coefficient through the ratio algorithm to obtain a multi-dimensional deviation vector;

[0017] A summation module for performing a weighted summation operation on the multi-dimensional deviation vector to obtain a comprehensive deviation evaluation value;

[0018] A control module for generating a current correction amount and a voltage correction amount based on the comprehensive deviation evaluation value, and performing closed-loop control on the LED lamp bead until the comprehensive deviation evaluation value is lower than the set threshold to obtain a control termination signal.

[0019] In a third aspect of the present invention, an LED lamp bead is provided. When the LED lamp bead is used, it is used to implement the above method for uniformly controlling the light-emitting color of the LED lamp bead.

[0020] In the technical solution provided by the present invention, a triangular annular wafer placement method and a double-layer packaging structure are adopted, significantly improving the uniformity and luminous efficiency of light. By precisely controlling the ratio and distribution of the phosphor and the diffusing powder, the problem of color spots in traditional LED lamp beads is effectively solved. A multi-dimensional parameter real-time monitoring system including spectrum, chromaticity, and luminous flux is established. Through high-precision sensors and synchronous acquisition technology, all-round dynamic monitoring of the optical characteristics of LED lamp beads is achieved, and the monitoring accuracy is improved to 0.1 nm. By integrating a bidirectional gated recurrent neural network and a PID control algorithm, adaptive adjustment of the luminous characteristics of LED lamp beads is realized, and the control accuracy reaches ±0.1%, greatly improving the uniformity and stability of the emitted light color. Through multi-dimensional deviation vector analysis and comprehensive evaluation mechanism, a complete closed-loop control system is established, which can compensate for the performance drift caused by environmental changes and device aging in real time, and extends the service life of LED lamp beads. By adopting a high-precision centrifugal packaging process and an automatic control technology, the production efficiency and product consistency are significantly improved, and the production cost is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 It is a schematic diagram of the steps of the method for uniformly controlling the emitted light color of the LED lamp bead in the embodiment of the present invention;

[0023] Figure 2 It is a schematic diagram of the structure of the device for uniformly controlling the emitted light color of the LED lamp bead in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] The embodiments of the present invention provide a method and a device for uniformly controlling the emitted light color of an LED lamp bead, and an LED lamp bead. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device including a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0025] For ease of understanding, the specific process of the embodiment of the present invention will be described below. Please refer to Figure 1 An embodiment of the method for uniformly controlling the light-emitting color of LED lamp beads in the embodiment of the present invention includes:

[0026] Step S1: Place the LED wafers on the LED bracket in a triangular ring pattern, connect the wires to the LED wafers, inject the first encapsulant and the second encapsulant in sequence and perform centrifugation to obtain an LED lamp bead with a double-layer encapsulation structure;

[0027] It can be understood that the execution subject of the present invention can be a device for uniformly controlling the light-emitting color of LED lamp beads, or a terminal or a server. Specifically, it is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example.

[0028] Specifically, the surface of the LED bracket is treated to ensure that the subsequent coating can adhere firmly. An alumina reflective layer is coated on the inner wall of the LED bracket, and a high-temperature sintering process is used to make the reflective layer closely combined with the bracket structure to improve the reflection performance of the bracket. Place the processed LED wafers on the bracket structure to ensure that the LED wafers form an angle of 60° with the horizontal direction. This design helps to achieve a good optical layout in the triangular ring array. Through an accurate positioning system, ensure that the wafers are evenly arranged in a triangular pattern to form a wafer array structure. Perform wire bonding on the wafer array structure. Connect the LED wafers to the conductive pads of the bracket structure through an automatic wire bonding system to ensure the reliability of the electrical connection to obtain a good conduction structure. Mix the phosphor with a particle size of 10 microns and epoxy resin in a mass ratio of 4:6, and use a high-speed stirring device for sufficient dispersion treatment to ensure that the phosphor is evenly distributed in the resin and improve the optical properties of the encapsulant. Inject the processed first encapsulant into the conduction structure, and strictly control the injection amount through a weight sensor to ensure that the injection amount error is controlled within the range of ±0.1 mg to obtain a uniform and appropriate thickness of the first encapsulation layer. After completing the first layer of encapsulation, place it in a centrifuge with a rotation speed of 3000 revolutions per minute for 300 seconds to enable the phosphor to be evenly distributed around the LED wafers to form an ideal coating structure. Mix the diffusion powder and epoxy resin in a mass ratio of 1.2:98.8, and also use a high-speed stirring device to ensure sufficient dispersion of the two to obtain a uniform second encapsulant. When injecting the second encapsulant into the coating structure, adjust the injection amount and injection angle through a numerical control system to ensure that the formed second encapsulant has a concave structure with a curvature radius of 1.5 times the diameter of the encapsulation surface, enhance the light scattering effect, and effectively improve the uniformity of the light emission of the LED lamp bead. After completing these steps, the finally formed LED lamp bead has a double-layer encapsulation structure.

[0029] Step S2: Apply the rated operating parameters to the LED lamp beads and collect the initial optical parameter matrix;

[0030] Specifically, control the environmental parameters of the LED lamp beads. Under standard environmental conditions, that is, at a temperature of 25°C and a relative humidity of 60%, perform a preheating treatment to ensure that the LED lamp beads reach a stable operating state. At this stage, the purpose of the preheating treatment is to eliminate the optical property instability caused by temperature fluctuations or humidity changes, ensuring the reliability and consistency of subsequent test results. Apply the rated operating parameters to the LED lamp beads in a stable operating state, set the voltage to 3.0V and the current to 350mA. These parameters are supplied and controlled by a high-precision DC power supply, aiming to ensure the normal operation of the LED lamp beads under the specified electrical conditions and maintain their stable power supply state. After the LED lamp beads reach a stable power supply state, use a spectral analyzer with a resolution of 0.1nm to align with the light-emitting surface of the LED lamp beads and perform a scan acquisition in the wavelength range of 380 - 780nm. The spectral analyzer can obtain the spectral intensity distribution data of the LED lamp beads at different wavelengths, reflecting the spectral characteristics of the lamp beads. At the same time, measure the chromaticity parameters of the light-emitting surface of the LED lamp beads. Measure at a distance of 100mm from the light-emitting surface through a colorimeter to collect the CIE chromaticity coordinate values and obtain the chromaticity distribution data. Place the LED lamp beads at the center of the integrating sphere and use a luminous flux sensor to collect the luminous flux value to obtain the luminous flux data. Divide the spectral intensity distribution data into wavelength intervals, and perform numerical fitting on the discrete wavelength points through an interpolation algorithm to obtain a continuous spectral curve. Fuse the chromaticity distribution data with the luminous flux data, and obtain the normalized parameter values through normalization processing to form the normalized data. Use matrix synthesis operation to combine the continuous spectral curve with the normalized data, and generate a three-dimensional data structure through a data reconstruction algorithm, and finally obtain the initial optical parameter matrix.

[0031] Step S3: Establish a standard data model based on the initial optical parameter matrix, perform multi-parameter real-time monitoring on the LED lamp beads, and obtain a monitoring data set;

[0032] Specifically, parameter decomposition is performed on the initial optical parameter matrix. Using matrix decomposition algorithms, spectral data, chromaticity data, and luminous flux data are converted into a feature matrix to extract the standard optical characteristics of the LED lamp beads and reduce the complexity of the data. For example, the feature matrix contains key indicators such as color temperature and color rendering index, accurately reflecting the optical characteristics of the LED lamp beads under different operating conditions. Based on the standard optical characteristics, a three-dimensional parameter space is constructed to visualize the variation relationships of spectra, chromaticity, and luminous flux. By fitting spectral curve models, chromaticity distribution models, and luminous flux variation models using the least squares method, standard mathematical models are formed to describe the performance of the LED lamp beads under ideal conditions and provide a prediction basis for various situations encountered in the actual application of the lamp beads. The constructed standard mathematical model is input into a Kalman filter, and a dynamic prediction model is established through a state prediction algorithm. This model can be updated in real time, continuously adjusting the predicted values according to the real-time collected data to improve the monitoring and control of the lamp bead performance. For example, if it is found that the spectral data deviates from the expected value, the system can quickly give feedback and adjust the current or voltage to restore its normal state. Parameter thresholds are set based on the standard data model. Through interval control algorithms, the spectral fluctuation range, chromaticity deviation range, and luminous flux variation range are determined to form clear parameter control intervals. The setting of the control intervals directly affects the luminous effect and stability of the LED lamp beads. Reasonable threshold setting can effectively prevent performance degradation caused by factors such as temperature changes and aging, ensuring that the lamp beads always maintain an efficient luminous effect under different operating conditions. To achieve efficient real-time monitoring, a multi-channel data acquisition system with a sampling frequency of 1000 Hz is used to synchronously capture spectral, chromaticity, and luminous flux data. High-frequency data acquisition can timely capture the instantaneous changes occurring in the lamp beads during use, ensuring that the system can quickly respond when the performance changes. The real-time sampled data is input into the standard data model, and the deviation between the current state and the predicted state is calculated through a state estimation algorithm. The quantification of the deviation enables the system to identify whether the performance of the LED lamp beads is abnormal. If the deviation exceeds the set range, an alarm is issued or automatic adjustment is made to ensure the normal operation of the lamp beads. In the anomaly detection link, the Mahalanobis distance algorithm is used to determine whether the collected data falls within the set parameter control intervals. Through this method, potential abnormal data is quickly identified, and the data is marked for validity. The data marked as valid is reorganized in time series. Through a data encapsulation algorithm, spectral data, chromaticity data, and luminous flux data are integrated into a unified format to form a complete monitoring data set.

[0033] Step S4: Input the monitoring data set into the data processing unit, calculate the spectral deviation coefficient through the spectral similarity algorithm, calculate the chromaticity deviation coefficient through the Euclidean distance algorithm, calculate the luminous flux deviation coefficient through the ratio algorithm, and obtain a multi-dimensional deviation vector;

[0034] Specifically, perform Fourier transform on the real-time spectral data in the monitoring dataset to convert the spectral signal from the time domain to the frequency domain, extract the spectral frequency components, obtain the spectral frequency-domain features, and reveal the frequency components of the light-emitting characteristics of the lamp beads, which helps analyze their optical performance. Calculate the cosine similarity between the spectral frequency-domain features and the spectral data in the standard data model. Through this algorithm, judge the similarity degree between the current spectrum and the standard spectrum, and quantify the spectral deviation. Through normalization processing, adjust the calculation result to the range of 0 to 1 to eliminate the influence brought by different dimensions, and obtain the spectral deviation coefficient.

[0035] At the same time, convert the real-time chromaticity data into three-dimensional chromaticity space coordinates. Through the coordinate mapping algorithm, generate chromaticity space distribution points to form a chromaticity space vector, which accurately reflects the chromaticity performance of the lamp beads. Calculate the Euclidean distance between the chromaticity space vector and the chromaticity data in the standard data model, and map this distance value to the interval of 0 to 1 through standardization processing to obtain the chromaticity deviation coefficient. Reconstruct the real-time luminous flux data according to the time series. Through the sliding window algorithm, generate the luminous flux time series curve to obtain the luminous flux sequence, which helps identify the law of the change of the luminous flux over time. Perform a ratio operation on the luminous flux sequence and the luminous flux values in the standard data model, and convert the ratio result into a linear relationship through logarithmic transformation to more clearly show the degree of the luminous flux deviation, and obtain the luminous flux deviation coefficient. Construct a three-dimensional feature space based on the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient. Based on these three deviation coefficients, perform orthogonalization processing to generate the standard eigenbasis. By constructing the eigenbasis matrix, provide a unified reference framework for subsequent data analysis. Perform projection transformation on the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient on the eigenbasis matrix. Through this matrix operation, generate a multi-dimensional deviation vector, which comprehensively reflects the performance deviations of the LED lamp beads in the three dimensions of spectrum, chromaticity, and luminous flux.

[0036] Step S5: Perform a weighted summation operation on the multi-dimensional deviation vector to obtain a comprehensive deviation evaluation value;

[0037] Specifically, perform principal component analysis on the multi-dimensional deviation vector, extract the main feature components and secondary feature components through the singular value decomposition algorithm, and obtain the feature component sequence. Identify the most representative features in multiple dimensions such as spectrum, chromaticity, and luminous flux. Through singular value decomposition, effectively reduce the high-dimensional data to a lower dimension while retaining as much original information as possible. Perform an orthogonal transformation on the feature component sequence, use the Schmidt orthogonalization algorithm to construct the standard orthogonal basis vectors, and form the orthogonal basis matrix. The construction of the orthogonal basis matrix helps to eliminate the correlation between features and improve the stability and accuracy of subsequent calculations. Each column of the orthogonal basis matrix represents an independent feature direction, enabling more accurate capture of the changes in the light-emitting characteristics of the LED beads when performing data processing. Perform dimensionality reduction on the orthogonal basis matrix according to the main and secondary feature contribution rates. Adopt the method of truncating singular values, retain the main feature dimensions, and generate the dimensionality-reduced feature matrix. Through the dimensionality reduction process, the computational complexity can be effectively reduced, and at the same time, the risk of overfitting can be avoided, ensuring that only the features with the greatest impact on performance are retained in feature selection. Based on the dimensionality-reduced feature matrix, construct a weight assignment model. Calculate the initial weight values of the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient through the coefficient of variation method. Assign an initial relative importance index to each feature to reflect its role in the overall performance evaluation. The coefficient of variation can effectively measure the fluctuation degree of each feature and help determine which feature occupies a more important position in the overall performance. Perform hierarchical analysis on the initial weight values. Verify the rationality of the weight assignment through the consistency test algorithm, and obtain the corrected weight coefficients to ensure that the weight relationship between different features is reasonable and avoid errors caused by subjectively setting the weight ratio. After passing the consistency test, if it is found that the weights of some features deviate significantly from the expected values, adjust the corrected weight coefficients to make them more in line with the actual situation. Linearly combine the corrected weight coefficients with the preset weight ratio (such as 0.4:0.4:0.2) to generate the target weight coefficients. Based on the generated target weight coefficients, perform weighted fusion on the multi-dimensional deviation vector. Generate the weighted deviation value through matrix multiplication operations, which can comprehensively consider the influence of features in different dimensions and form a comprehensive evaluation index. Perform a non-linear mapping on the normalized deviation value, use the sigmoid function to compress it into the interval from 0 to 1 to improve the stability of the data, making the final comprehensive deviation evaluation value easier to understand and use. The comprehensive deviation evaluation value calculated through the above steps can comprehensively reflect the light-emitting performance of the LED beads. Through this evaluation value, the control system monitors and adjusts the working state of the LED beads in real time to ensure that its light-emitting color remains within a uniform range.

[0038] Step S6: Based on the comprehensive deviation evaluation value, generate a current correction amount and a voltage correction amount, and perform closed-loop control on the LED beads until the comprehensive deviation evaluation value is lower than the set threshold, and obtain a control termination signal.

[0039] Specifically, the comprehensive deviation evaluation value is input into the parameter prediction model, which adopts a bidirectional gated recurrent neural network. The network consists of an input layer, a hidden layer, and a fully connected layer, where the hidden layer contains a forward gated recurrent unit and a backward gated recurrent unit. After the input layer receives the comprehensive deviation evaluation value, the forward gated recurrent unit extracts the forward time series features, while the backward gated recurrent unit extracts the backward time series features. After being processed by these two units, the feature information is fused into the fully connected layer to obtain the initial compensation parameters. The initial compensation parameters are processed by the self-attention mechanism. By calculating the dot product operation of the query matrix, key matrix, and value matrix, and through the softmax normalization process, the attention weight matrix is obtained. The weight matrix helps to emphasize the more important features in the initial compensation parameters, thus improving the effectiveness and accuracy of subsequent calculations. Based on the attention weight matrix, the initial compensation parameters are weighted, and feature enhancement is performed through the multi-head attention mechanism. In this process, each attention head independently calculates its attention score and extracts features to obtain the enhanced feature vector. The enhanced feature vector is input into the PID controller, and the parameter adjustment is carried out using a proportional coefficient of 0.8, an integral coefficient of 0.3, and a derivative coefficient of 0.1 to obtain the PID compensation parameters. The adjustment mechanism of the PID controller can perform adaptive adjustment based on the current deviation and historical data, enabling the control process to respond more quickly and accurately to the deviation change and realizing the effective control of the LED lamp beads. Through adjustment, appropriate current correction amounts and voltage correction amounts can be generated to ensure that the lamp beads emit light evenly under different working conditions. The current and voltage correction amounts are input into a high-precision digital-to-analog converter, usually a 16-bit DAC, to achieve signal conversion. This process also includes PWM modulation to output the drive signal and obtain the corrected working parameters. The corrected working parameters are monitored and adjusted in real time through a closed-loop control system. By comparing the corrected working parameters with the set threshold, the comprehensive deviation evaluation value is monitored in real time to determine whether the control reaches the expected effect. If the comprehensive deviation evaluation value is lower than the set threshold, a control termination signal is issued. At the same time, the comparison result is fed back to the input layer of the bidirectional gated recurrent neural network to ensure that the network can self-adjust according to the latest data and further optimize the control process. The closed-loop control mechanism ensures that the LED lamp beads emit light evenly in different working states, improving its overall performance and user experience.

[0040] In the embodiments of the present invention, a triangular annular wafer placement method and a double-layer encapsulation structure are adopted, significantly improving the light uniformity and luminous efficiency. By precisely controlling the ratio and distribution of the phosphor and the diffusing powder, the color spot problem of traditional LED lamp beads is effectively solved. A multi-dimensional parameter real-time monitoring system including spectrum, chromaticity, and luminous flux is established. Through high-precision sensors and synchronous acquisition technology, all-round dynamic monitoring of the optical characteristics of LED lamp beads is realized, and the monitoring accuracy is improved to 0.1 nm. By integrating a bidirectional gated recurrent neural network and a PID control algorithm, adaptive adjustment of the luminous characteristics of LED lamp beads is achieved, and the control accuracy reaches ±0.1%, greatly improving the uniformity and stability of the emitted light color. Through multi-dimensional deviation vector analysis and a comprehensive evaluation mechanism, a complete closed-loop control system is established, which can compensate for the performance drift caused by environmental changes and device aging in real time, and extend the service life of LED lamp beads. By adopting a high-precision centrifugal encapsulation process and an automatic control technology, the production efficiency and product consistency are significantly improved, and the production cost is reduced.

[0041] In a specific embodiment, the process of performing step S1 may specifically include the following steps:

[0042] Perform surface treatment on the LED bracket, coat an alumina reflective layer on the inner wall of the LED bracket, and make the reflective layer combine with the bracket through a high-temperature sintering process to obtain a bracket structure;

[0043] Place the LED wafer on the bracket structure, make the LED wafer form a 60° angle with the horizontal direction, and perform triangular annular array arrangement through a positioning system to obtain a wafer array structure;

[0044] Perform gold wire bonding on the wafer array structure, and connect the LED wafer and the conductive pads of the bracket structure through an automatic wire bonding system to obtain a conduction structure;

[0045] Mix phosphor with a particle size of 10 microns and epoxy resin in a mass ratio of 4:6, and perform dispersion treatment through a high-speed stirring device to obtain a first encapsulation adhesive. Then inject the first encapsulation adhesive into the conduction structure, and control the injection amount through a weight sensor to make the injection amount error controlled within the range of ±0.1 mg to obtain a first encapsulation layer;

[0046] Place the first encapsulation layer in a centrifuge with a rotation speed of 3000 revolutions per minute, and perform centrifugal treatment for 300 seconds to make the phosphor evenly distributed around the LED wafer to obtain a coated structure. Then mix diffusing powder and epoxy resin in a mass ratio of 1.2:98.8, and perform dispersion treatment through a high-speed stirring device to obtain a second encapsulation adhesive;

[0047] Inject the second encapsulation adhesive into the coated structure, and adjust the injection amount and injection angle through a numerical control system to make the second encapsulation adhesive form a concave structure with a curvature radius 1.5 times the diameter of the encapsulation surface, thereby obtaining an LED lamp bead with a double-layer encapsulation structure.

[0048] Specifically, the surface of the LED bracket is treated, and an alumina reflective layer is coated on the inner wall of the LED bracket. As an excellent reflective material, alumina can effectively enhance the light efficiency of the LED. The treatment method of its inner wall usually involves a high-temperature sintering process. In this process, the alumina reflective layer combines with the inner wall of the bracket in a high-temperature environment to form a strong and well-reflective bracket structure. Assuming the thickness of the bracket is , and the thickness of the alumina layer is , then during the sintering process, the bonding strength between the reflective layer and the bracket is described by the following formula:

[0049] ;

[0050] Among them, is the bonding strength, is the reflectivity of alumina, is the reflectivity of the bracket material. This bonding strength directly affects the overall performance of the LED lamp bead. Place the LED wafer on the bracket structure, ensuring that the LED wafer forms an angle of 60° with the horizontal direction. This design helps to improve the light divergence efficiency and the light uniformity. During this process, a positioning system is used to ensure that the wafers are arranged in a triangular ring array to form a wafer array structure. This layout can effectively disperse the LED light sources in space and improve the overall brightness. Perform wire bonding on the wafer array structure. Through an automatic wire bonding system, connect the LED wafer to the conductive pads of the bracket structure to form a conductive structure. During this process, the accuracy and consistency of the wire bonding are crucial for ensuring current flow. Through precise control, the reliability of the LED can be greatly improved. Mix phosphor with a particle size of 10 microns and epoxy resin in a mass ratio of 4:6, and then disperse them through a high-speed stirring device to ensure that the phosphor is evenly distributed in the colloid to form a high-performance first encapsulation glue. Inject the first encapsulation glue into the conductive structure, and control the injection volume through a weight sensor to ensure that the error is controlled within the range of ±0.1 mg to obtain the first encapsulation layer. Place the first encapsulation layer in a centrifuge with a rotation speed of 3000 revolutions per minute and perform centrifugation for 300 seconds to make the phosphor evenly distributed around the LED wafer to form a coating structure. The uniform distribution can ensure that the LED can provide a stable light output during operation. Mix the diffusion powder and epoxy resin in a mass ratio of 1.2:98.8, and also disperse them through a high-speed stirring device to obtain the second encapsulation glue to improve the light diffusion and enhance the light uniformity. When injecting the second encapsulation glue into the coating structure, use a numerical control system to adjust the injection volume and injection angle to ensure that the second encapsulation glue forms a concave structure with a curvature radius 1.5 times the diameter of the encapsulation surface. The relationship between the set curvature radius and the diameter of the encapsulation surface is expressed by the following formula:

[0051] ;

[0052] The concave design can enhance the focusing effect of light, thereby improving the overall luminous efficiency and lighting effect.

[0053] In a specific embodiment, the process of performing step S2 may specifically include the following steps:

[0054] Control the environmental parameters of the LED lamp beads, and perform preheating treatment in a standard environment with a temperature of 25 °C and a relative humidity of 60% to obtain LED lamp beads in a stable working state;

[0055] Apply the rated working parameters of a voltage of 3.0V and a current of 350mA to the LED lamp beads in a stable working state, and perform power supply control through a high-precision DC power supply to obtain a stable power supply state;

[0056] Align the spectral analyzer with a resolution of 0.1nm to the light-emitting surface of the LED lamp beads, and perform scanning and acquisition in the wavelength range of 380 - 780nm to obtain spectral intensity distribution data;

[0057] Measure the chromaticity parameters of the light-emitting surface of the LED lamp beads, and collect CIE chromaticity coordinate values at a distance of 100mm from the light-emitting surface through a chromaticity meter to obtain chromaticity distribution data;

[0058] Place the LED lamp beads at the center position of the integrating sphere, collect the luminous flux value through a luminous flux sensor to obtain luminous flux data, divide the wavelength interval of the spectral intensity distribution data, and perform numerical fitting on the discrete wavelength points through an interpolation algorithm to obtain a continuous spectral curve;

[0059] Fuse the chromaticity distribution data and the luminous flux data, obtain the normalized parameter value through normalization processing to obtain the normalized data, and perform matrix synthesis operation on the continuous spectral curve and the normalized data, and generate a three-dimensional data structure through a data reconstruction algorithm to obtain the initial optical parameter matrix.

[0060] Specifically, control the environmental parameters of the LED lamp beads. Perform preheating treatment in a standard environment with a temperature of 25 °C and a relative humidity of 60% to eliminate the photoelectric performance fluctuations caused by environmental factors and ensure that the LED lamp beads can exert their best performance in a stable environment. The preheating stage is a thermal equilibrium process, and its temperature and humidity are described by the following formula:

[0061] ;

[0062] Where is the environmental effect, and are the influence coefficients of temperature and humidity respectively, and are the reference temperature and humidity of the environment. By adjusting these parameters, the LED lamp beads can reach a stable working state. After the LED lamp beads reach the stable working state, the rated working parameters are applied. A voltage of 3.0V and a current of 350mA are provided and controlled by a high-precision DC power supply. This power supply can ensure the accuracy of the current and voltage, thus avoiding performance fluctuations caused by unstable power supply. The stability of the power supply state is crucial for the light output and color stability of the LED lamp beads. A spectral analyzer with a resolution of 0.1nm is used to align with the light-emitting surface of the LED lamp beads and scan and collect in the wavelength range of 380 - 780nm to obtain spectral intensity distribution data. The spectral intensity distribution is expressed as:

[0063] ;

[0064] where is the spectral output of the light source, is the response function of the spectral analyzer. While obtaining the spectral data, the chromaticity parameters of the light-emitting surface of the LED lamp beads are measured. The CIE chromaticity coordinate values are collected at a position 100mm away from the light-emitting surface through a chromaticity meter to obtain chromaticity distribution data. In this way, the color performance of the LED lamp beads under specific conditions can be obtained, strengthening the understanding of its performance. The LED lamp beads are placed at the center of the integrating sphere, and the luminous flux value is collected by a luminous flux sensor to obtain luminous flux data, ensuring that the light output ability of the LED can be accurately measured. At the same time, in order to analyze the spectral intensity distribution data, the wavelength interval is divided, and the discrete wavelength points are numerically fitted through an interpolation algorithm to obtain a continuous spectral curve. The mathematical expression of this process is represented by the following formula:

[0065] ;

[0066] where is the continuous curve of the spectral intensity, is the spectral intensity at the discrete wavelength and is the weight of each wavelength. By fusing the chromaticity distribution data and the luminous flux data, more comprehensive optical characteristics are obtained. The data with different dimensions are unified through standardization processing to obtain normalized parameter values. On this basis, through matrix synthesis operation, the continuous spectral curve is combined with the standardized data, and a three-dimensional data structure is generated using a data reconstruction algorithm to obtain the initial optical parameter matrix.

[0067] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0068] Perform parameter decomposition on the initial optical parameter matrix, and convert spectral data, chromaticity data, and luminous flux data into a feature matrix through a matrix decomposition algorithm to obtain the standard optical characteristics of the LED lamp beads;

[0069] Construct a three-dimensional parameter space based on the standard optical characteristics, and fit the spectral curve model, chromaticity distribution model, and luminous flux change model through the least squares method to obtain the standard mathematical model of the LED lamp beads;

[0070] Input the standard mathematical model into the Kalman filter, and establish a dynamic prediction model for spectrum, chromaticity, and luminous flux through the state prediction algorithm to obtain the standard data model;

[0071] Set parameter thresholds based on the standard data model, and determine the spectral fluctuation range, chromaticity offset range, and luminous flux change range through the interval control algorithm to obtain the parameter control interval;

[0072] Perform real-time monitoring on the LED lamp beads, and synchronously collect spectral data, chromaticity data, and luminous flux data through a multi-channel data acquisition system with a sampling frequency of 1000 Hz to obtain real-time sampling data;

[0073] Input the real-time sampling data into the standard data model, and calculate the deviation between the current state and the predicted state through the state estimation algorithm to obtain the state deviation data;

[0074] Perform anomaly detection based on the state deviation data, and determine whether the data falls within the parameter control interval through the Mahalanobis distance algorithm to obtain the data validity flag;

[0075] Perform time series recombination on the data with a valid data validity flag, and integrate the spectral data, chromaticity data, and luminous flux data into a unified format through the data encapsulation algorithm to obtain the monitoring data set.

[0076] Specifically, perform parameter decomposition on the initial optical parameter matrix, and convert spectral data, chromaticity data, and luminous flux data into a feature matrix through a matrix decomposition algorithm to extract the standard optical characteristics of the LED lamp beads. Using the singular value decomposition technique, assume the initial optical parameter matrix is , then its decomposition is expressed as:

[0077] ;

[0078] Among them, is the left singular vector matrix, is the diagonal matrix, which contains singular values, is the right singular vector matrix. Through this decomposition, the feature matrix , for subsequent model construction and analysis. A three-dimensional parameter space is constructed based on standard optical features, and spectral curve models, chromaticity distribution models, and luminous flux change models are fitted using the least squares method. By minimizing the sum of squared errors , expressed as:

[0079] ;

[0080] where is the observed data, is the fitting function, is the independent variable. Through this method, a standard mathematical model of the LED lamp bead can be obtained, reflecting the changes in its optical characteristics under different conditions. The standard mathematical model is input into the Kalman filter, and dynamic prediction models of spectrum, chromaticity, and luminous flux are established through the state prediction algorithm to obtain the standard data model. During this process, the state transition equation of the Kalman filter is expressed as:

[0081] ;

[0082] where is the current state, is the state transition matrix, is the control input matrix, is the control input, is the process noise. By continuously updating the state estimate, the changes in the optical characteristics of the LED lamp bead during operation can be accurately predicted. Based on the standard data model, parameter thresholds are set, and the spectral fluctuation range, chromaticity shift range, and luminous flux change range are determined through the interval control algorithm. The standard deviations and the mean of each parameter are obtained through statistical analysis to set the control interval, for example:

[0083] ;

[0084] where is a constant used to determine the strictness of the control. The establishment of the control interval is to ensure that the LED lamp bead is always in a controllable state during operation and avoid excessive fluctuations. The LED lamp bead is monitored in real time, and spectral, chromaticity, and luminous flux data are synchronously collected using a multi-channel data acquisition system with a sampling frequency of 1000 Hz to obtain real-time sampling data, ensuring that during the operation of the LED lamp bead, various parameter changes can be promptly responded to to maintain performance stability. The real-time sampling data is input into the standard data model, and the deviation between the current state and the predicted state is calculated through the state estimation algorithm to obtain the state deviation data. The state deviation is expressed as:

[0085] ;

[0086] By analyzing this deviation, it is identified whether there is an abnormal situation. Anomaly detection is carried out based on the state deviation data, and the Mahalanobis distance algorithm is used to determine whether the data falls within the set parameter control interval. Mahalanobis distance The calculation formula is:

[0087] ;

[0088] where is the current state vector, is the mean vector, is the covariance matrix. By judging whether the distance exceeds the set threshold, a data validity flag is obtained. The data marked as valid is reorganized in time series, and the spectral data, chromaticity data, and luminous flux data are integrated into a unified format through a data encapsulation algorithm to obtain a monitoring data set.

[0089] Among them, the real-time sampling data is input into the standard data model, and the deviation between the current state and the predicted state is calculated through the state estimation algorithm to obtain the state deviation data; anomaly detection is carried out based on the state deviation data, and the Mahalanobis distance algorithm is used to determine whether the data falls within the parameter control interval to obtain the data validity flag, including the following steps: separating the real-time sampling data into three groups of characteristic data, namely spectral data, chromaticity data, and luminous flux data, to obtain the current state vector; performing data smoothing processing on the current state vector, eliminating random fluctuations through the moving average method, to obtain the smoothed state data; inputting the smoothed state data into the standard data model, calculating the predicted value of the next moment according to the historical data, to obtain the predicted state vector; calculating the difference between the current state vector and the predicted state vector to obtain the state deviation matrix; performing normalization processing on the state deviation matrix, converting it into a standard score through normalization operation, to obtain the normalized deviation value; calculating the Mahalanobis distance based on the normalized deviation value to obtain the distance evaluation value; comparing the distance evaluation value with the preset parameter control interval to obtain the out-of-bounds judgment result; setting a marking rule based on the out-of-bounds judgment result, marking the data with the distance evaluation value exceeding the control interval as abnormal to obtain the initial marking sequence; performing time series correlation analysis on the initial marking sequence to judge the abnormal duration to obtain the time series correlation mark; converting the time series correlation mark into a binary sequence to generate 0-1 marked data to obtain the data validity flag.

[0090] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0091] Performing Fourier transform on the real-time spectral data in the monitoring data set, and extracting the spectral frequency components through the spectral analysis algorithm to obtain the spectral frequency domain characteristics;

[0092] Calculate the cosine similarity between the spectral frequency domain features and the spectral data in the standard data model. Through normalization, obtain a spectral similarity value ranging from 0 to 1, and get the spectral deviation coefficient;

[0093] Convert the real-time chromaticity data in the monitoring dataset into three-dimensional chromaticity space coordinates. Generate chromaticity space distribution points through the coordinate mapping algorithm, and obtain the chromaticity space vector;

[0094] Calculate the Euclidean distance between the chromaticity space vector and the chromaticity data in the standard data model. Map the distance value to the interval of 0 - 1 through standardization, and obtain the chromaticity deviation coefficient;

[0095] Reconstruct the real-time luminous flux data in the monitoring dataset according to the time series. Generate the luminous flux time series curve through the sliding window algorithm, obtain the luminous flux sequence, and perform a ratio operation on the luminous flux sequence and the luminous flux values in the standard data model. Convert the ratio result into a linear relationship through logarithmic transformation, and obtain the luminous flux deviation coefficient;

[0096] Construct a three-dimensional feature space based on the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient. Generate the standard feature basis through orthogonalization, and obtain the feature basis matrix;

[0097] Perform a projection transformation on the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient on the feature basis matrix. Generate a multi-dimensional deviation vector through matrix operations.

[0098] Specifically, perform a Fourier transform on the real-time spectral data in the monitoring dataset. Extract the spectral frequency components through the spectral analysis algorithm to obtain the spectral frequency domain features. The basic formula for the Fourier transform is:

[0099] ;

[0100] where, represents the frequency domain signal, is the time domain signal, is the frequency, is the imaginary unit. By performing a Fourier transform on the real-time spectral data, identify the intensity and phase information of each frequency component in the spectrum, and obtain the spectral frequency domain features. These features can provide important information about the light-emitting characteristics of the LED lamp beads. Calculate the cosine similarity between the extracted spectral frequency domain features and the spectral data in the standard data model to obtain the spectral deviation coefficient. The formula for the cosine similarity is:

[0101] ;

[0102] where, and respectively represent two spectral frequency domain feature vectors, is the norm of a vector. Through normalization, the cosine similarity value is restricted between 0 and 1 to obtain the spectral deviation coefficient, which effectively quantifies the difference between the current spectrum and the standard spectrum, helping the system determine whether the luminous performance of the LED lamp bead meets the expectations. When processing chromaticity data, the real-time chromaticity data is converted into three-dimensional chromaticity space coordinates, which usually relies on the CIE chromaticity coordinate system. Chromaticity space distribution points are generated through a coordinate mapping algorithm to obtain a chromaticity space vector. The Euclidean distance between the chromaticity space vector and the chromaticity data in the standard data model is calculated to obtain the chromaticity deviation coefficient. The formula for the Euclidean distance is:

[0103] ;

[0104] where represents the distance, and are the corresponding coordinate values in the chromaticity space, is the number of dimensions. The obtained distance value is mapped to the range of 0 to 1 through standardization to generate the chromaticity deviation coefficient. The real-time luminous flux data in the monitoring dataset is reconstructed according to the time series, and the sliding window algorithm is used to generate the luminous flux time series curve to obtain the luminous flux sequence. The ratio operation is performed on the luminous flux sequence and the luminous flux value in the standard data model, and the ratio result is converted into a linear relationship through logarithmic transformation to obtain the luminous flux deviation coefficient. The ratio operation is expressed by the following formula:

[0105] ;

[0106] where is the ratio, is the measured luminous flux value, is the standard luminous flux value. The logarithmic transformation is expressed as:

[0107] ;

[0108] Through the above steps, the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient are calculated. These coefficients together construct a three-dimensional feature space. To better utilize these deviation coefficients, orthogonalization processing is performed to generate the standard eigenbasis to obtain the eigenbasis matrix. The eigenbasis matrix is a linear space containing the relationships between various deviation coefficients, expressed as:

[0109] ;

[0110] In this matrix, each row represents a deviation coefficient. The spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient are subjected to projection transformation on the eigenbasis matrix, and a multi-dimensional deviation vector is generated through matrix operations.

[0111] In a specific embodiment, the process of executing step S5 may specifically include the following steps:

[0112] Perform principal component analysis on the multi-dimensional deviation vector, extract the main feature components and secondary feature components through the singular value decomposition algorithm, and obtain a sequence of feature components;

[0113] Perform an orthogonal transformation on the sequence of feature components, construct a standard orthogonal basis vector through the Schmidt orthogonalization algorithm, obtain an orthogonal basis matrix, and perform dimensionality reduction processing on the orthogonal basis matrix according to the main and secondary feature contribution rates. Retain the main feature dimensions through the truncated singular value method to obtain a dimensionality-reduced feature matrix;

[0114] Construct a weight assignment model based on the dimensionality-reduced feature matrix, and calculate the initial weight values of the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient through the coefficient of variation method;

[0115] Perform hierarchical analysis on the initial weight values, verify the rationality of the weight assignment through the consistency test algorithm, obtain the corrected weight coefficients, and linearly combine the corrected weight coefficients with the preset weight ratio of 0.4:0.4:0.2. Generate the target weight coefficient through the weighted average algorithm;

[0116] Perform weighted fusion on the multi-dimensional deviation vector based on the target weight coefficient, generate a weighted deviation value through matrix multiplication, obtain a normalized deviation value, and perform a non-linear mapping on the normalized deviation value. Compress the deviation value into the 0-1 interval through the sigmoid function to obtain a comprehensive deviation evaluation value.

[0117] Specifically, perform principal component analysis on the multi-dimensional deviation vector, extract the main feature components and secondary feature components through the singular value decomposition algorithm, and obtain a sequence of feature components. The basic formula for singular value decomposition is:

[0118] ;

[0119] Among them, represents the data matrix, is the left singular matrix, is a diagonal matrix, where the diagonal elements are singular values, is the transpose of the right singular matrix. Through singular value decomposition, the original data is decomposed into different feature components, the main variation directions in the data are found, the main feature components are extracted, and the secondary feature components are screened out to form a sequence of feature components. Perform an orthogonal transformation on the sequence of feature components to ensure the independence between each feature component, and use the Schmidt orthogonalization algorithm. Through Schmidt orthogonalization, construct a standard orthogonal basis vector to obtain an orthogonal basis matrix. Represent the sequence of feature components as:

[0120] ;

[0121] Among them, represents the th orthogonal basis vector. The orthogonal basis matrix is dimensionally reduced, screened according to the contribution rates of the primary and secondary features, and the main feature dimensions are retained by the method of truncated singular values to obtain a dimensionally reduced feature matrix. The dimensional reduction process is represented by the following relationship:

[0122] ;

[0123] Among them, represents the data matrix after dimensional reduction, is the selected orthogonal basis vector matrix, is the corresponding singular value matrix. Based on the dimensionally reduced feature matrix, a weight assignment model is constructed. The initial weight values of the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient are calculated by the coefficient of variation method. The formula for the coefficient of variation is:

[0124] ;

[0125] Among them, represents the coefficient of variation, is the standard deviation, is the mean. By calculating the coefficient of variation of each deviation coefficient, the corresponding initial weight values are obtained, reflecting the relative importance of each deviation. Hierarchical analysis is performed on the initial weight values, and a consistency test algorithm is used to verify the rationality of the weight assignment. When performing the consistency test, the consistency ratio (CR) is used to judge whether the weight assignment is appropriate. The formula is:

[0126] ;

[0127] Among them, is the consistency index, is the random consistency index. If , it means that the weight assignment is reasonable; otherwise, the weight assignment needs to be re-evaluated. After obtaining the corrected weight coefficients, they are linearly combined with the preset weight ratio (such as 0.4:0.4:0.2), and the target weight coefficients are generated through the weighted average algorithm. It is expressed as:

[0128] ;

[0129] Among them, is the target weight coefficient, , , are the weight values of the corrected deviation coefficients. Based on the target weight coefficients, the multi-dimensional deviation vectors are weighted and fused. This step involves matrix multiplication operations to generate weighted deviation values, which is expressed as:

[0130] ;

[0131] Among them, is the weighted deviation value, is the target weight coefficient matrix, is the multi-dimensional deviation vector. Normalize the weighted deviation value and perform a non-linear mapping. Compress the deviation value into the range of 0 - 1 through the sigmoid function. The formula is:

[0132] ;

[0133] Among them, is the output of the sigmoid function, is the normalized deviation value of the input. Through this process, the comprehensive deviation evaluation value is obtained, which can effectively reflect the performance and stability of the LED lamp beads in all aspects.

[0134] Among them, a weight allocation model is constructed based on the dimensionality-reduced feature matrix, and the initial weight values of the spectral deviation coefficient, chromaticity deviation coefficient, and luminous flux deviation coefficient are calculated by the coefficient of variation method; the initial weight values are analyzed hierarchically, and the rationality of the weight allocation is verified through the consistency test algorithm to obtain the corrected weight coefficients, and the corrected weight coefficients and the preset weight ratio of 0.4:0.4:0.2 are linearly combined, and the target weight coefficients are generated through the weighted average algorithm, including the following steps: statistically analyze the dimensionality-reduced feature matrix, and obtain the coefficient of variation matrix by calculating the mean and standard deviation of the spectral data, chromaticity data, and luminous flux data; input the coefficient of variation matrix into the feature normalization module, and normalize the coefficient of variation by the maximum-minimum normalization method to obtain the normalized coefficient of variation; construct a judgment matrix based on the normalized coefficient of variation, and calculate the relative importance of the spectral, chromaticity, and luminous flux parameters by the pairwise comparison method to obtain the feature importance matrix; perform eigenvalue decomposition on the feature importance matrix, and solve the maximum eigenvalue and its corresponding eigenvector by the power iteration method to obtain the initial weight values; input the initial weight values into the hierarchical analysis model, and obtain the consistency ratio by calculating the ratio of the consistency index CR and the random consistency index RI; perform weight correction based on the consistency ratio, and adjust the weight allocation by the iterative optimization algorithm until the consistency ratio is less than 0.1 to obtain the corrected weight coefficients; perform linear interpolation on the corrected weight coefficients and the preset weight ratio of 0.4:0.4:0.2, and set the interpolation coefficient α = 0.6 by the convex combination algorithm to obtain the combined weight vector; input the combined weight vector into the weight optimization module, and solve the optimal weight allocation by the Lagrange multiplier method to obtain the balanced weight coefficients; perform normalization on the balanced weight coefficients, and map the weight values to the interval (0, 1) and ensure that the sum is 1 by the softmax function to obtain the normalized weight coefficients; construct a weighted function based on the normalized weight coefficients, and fuse the multi-dimensional features by the linear weighted combination algorithm to obtain the target weight coefficients.

[0135] In a specific embodiment, the process of executing step S6 may specifically include the following steps:

[0136] Input the comprehensive deviation evaluation value into the parameter prediction model, and the parameter prediction model is a bidirectional gated recurrent neural network, and the bidirectional gated recurrent neural network includes an input layer, a hidden layer, and a fully connected layer, and the hidden layer includes a forward gated recurrent unit and a backward gated recurrent unit;

[0137] Input the comprehensive deviation evaluation value into the input layer of the parameter prediction model, extract the forward time series features through the forward gated recurrent unit, extract the backward time series features through the backward gated recurrent unit, and fuse them through the fully connected layer to obtain the initial compensation parameters;

[0138] Perform self-attention mechanism processing on the initial compensation parameters. By calculating the dot product operations of the query matrix, key matrix, and value matrix, and through softmax normalization processing, an attention weight matrix is obtained;

[0139] Weight the initial compensation parameters based on the attention weight matrix, and perform feature enhancement through the multi-head attention mechanism. Each attention head independently calculates the attention score and performs feature extraction to obtain an enhanced feature vector;

[0140] Input the enhanced feature vector into the PID controller, and adjust the parameters through a proportional coefficient of 0.8, an integral coefficient of 0.3, and a differential coefficient of 0.1 to obtain the PID compensation parameters;

[0141] Perform adaptive adjustment according to the PID compensation parameters to achieve the collaborative optimization of deep learning and PID control, obtain the current correction amount and voltage correction amount, and input the current correction amount and voltage correction amount into a high-precision digital-to-analog converter. Perform signal conversion through a 16-bit DAC, and output a drive signal through PWM modulation to obtain the corrected working parameters;

[0142] Perform closed-loop control on the corrected working parameters. By real-time monitoring the comparison result between the comprehensive deviation evaluation value and the set threshold, and feedback the comparison result to the input layer of the bidirectional gated recurrent neural network to obtain a control termination signal.

[0143] Specifically, input the comprehensive deviation evaluation value into the parameter prediction model. The parameter prediction model is a bidirectional gated recurrent neural network, and its network structure includes an input layer, a hidden layer, and a fully connected layer. Among them, the hidden layer is divided into a forward gated recurrent unit and a backward gated recurrent unit. The former extracts forward temporal features, and the latter extracts backward temporal features. This structure can effectively capture the context information in the sequence data and enhance the expression ability of the model. After inputting the comprehensive deviation evaluation value into the input layer of the parameter prediction model, the forward gated recurrent unit processes the input time series data, and the generated output is expressed as:

[0144] ;

[0145] where, represents the forward hidden state, is the current input, is the weight matrix of the forward unit, is the weight matrix related to the hidden state of the previous time step, is the bias term, is the activation function, using the sigmoid or tanh function. Similarly, the output of the backward gated recurrent unit is expressed as:

[0146] ;

[0147] Among them, represents the reverse hidden state, , and are the weights and bias terms of the reverse unit respectively. The forward and reverse hidden states are fused in the fully connected layer to form the initial compensation parameter, expressed as:

[0148] ;

[0149] Among them, concat means concatenating the forward and reverse hidden states, is the weight matrix of the fully connected layer, is the bias term, and the finally obtained is the initial compensation parameter. The initial compensation parameter is processed by the self-attention mechanism, and the dot product operations of the query matrix, key matrix and value matrix are calculated to extract features and enhance important information. The query, key and value matrices are defined as follows:

[0150] ;

[0151] Among them, , and are the weight matrices of the query, key and value respectively. The formula for calculating the attention score is:

[0152] ;

[0153] Among them, represents the attention output, is the dimension of the key, used for scaling to avoid the gradient vanishing caused by too large dot product. After softmax normalization, the attention weight matrix is obtained, and the initial compensation parameter is weighted to obtain the enhanced feature vector. The enhanced feature vector is input into the PID controller for adjustment. The output of the PID controller is calculated according to the following formula:

[0154] ;

[0155] Among them, is the control output, is the current error, , and They are the proportional coefficient, integral coefficient, and derivative coefficient respectively. The set proportional coefficient is 0.8, the integral coefficient is 0.3, and the derivative coefficient is 0.1. The control output is dynamically adjusted through the feedback error. According to the compensation parameters output by the PID controller, the current correction amount and voltage correction amount are adaptively adjusted. Combining the collaborative optimization of deep learning and PID control can ensure the fast response and stability of the system. On this basis, the current correction amount and voltage correction amount are input into a high-precision digital-to-analog converter (DAC), and signal conversion is performed through a 16-bit DAC, which is expressed as:

[0156] ;

[0157] Among them, is the output voltage, is the digital input value, is the resolution of the DAC (16 bits here), is the reference voltage. After signal conversion, through PWM modulation, a drive signal is output to obtain the corrected working parameters. Closed-loop control is performed on the corrected working parameters. The real-time monitored comprehensive deviation evaluation value is compared with the set threshold. If the deviation is within a reasonable range, the current state is maintained; otherwise, the result is fed back to the input layer of the bidirectional threshold recurrent neural network for readjustment to obtain a control termination signal. The feedback mechanism is expressed by the formula:

[0158] ;

[0159] Among them, represents the comprehensive deviation evaluation value, is the set threshold. When the deviation is within the set range, the feedback is 1, indicating that the system is stable and the control process ends; otherwise, the system enters an adjustment state. The closed-loop control strategy ensures the continuous optimization of the performance of the LED lamp beads and can effectively cope with different working conditions.

[0160] The method for uniformly controlling the light-emitting color of the LED lamp beads in the embodiments of the present invention has been described above. Next, the device for uniformly controlling the light-emitting color of the LED lamp beads in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the device for uniformly controlling the light-emitting color of the LED lamp beads in the embodiments of the present invention includes:

[0161] A packaging module for arranging LED wafers on an LED bracket in a triangular ring shape, connecting wires to the LED wafers, and sequentially injecting a first encapsulant and a second encapsulant and performing centrifugal treatment to obtain an LED lamp bead with a double-layer packaging structure;

[0162] An acquisition module for applying rated working parameters to the LED lamp beads and acquiring an initial optical parameter matrix;

[0163] A monitoring module, which is used to establish a standard data model based on the initial optical parameter matrix, perform real-time multi-parameter monitoring on the LED lamp beads, and obtain a monitoring data set;

[0164] A calculation module, which is used to input the monitoring data set into a data processing unit, calculate the spectral deviation coefficient through a spectral similarity algorithm, calculate the chromaticity deviation coefficient through an Euclidean distance algorithm, calculate the luminous flux deviation coefficient through a ratio algorithm, and obtain a multi-dimensional deviation vector;

[0165] A summation module, which is used to perform a weighted summation operation on the multi-dimensional deviation vector to obtain a comprehensive deviation evaluation value;

[0166] A control module, which is used to generate a current correction amount and a voltage correction amount based on the comprehensive deviation evaluation value, perform closed-loop control on the LED lamp beads until the comprehensive deviation evaluation value is lower than a set threshold, and obtain a control termination signal.

[0167] Through the collaborative cooperation of the above-mentioned various components, adopting a triangular ring-shaped wafer placement method and a double-layer packaging structure, the uniformity and luminous efficiency of light are significantly improved. By precisely controlling the ratio and distribution of the phosphor and the diffusing powder, the color spot problem of traditional LED lamp beads is effectively solved. A multi-dimensional parameter real-time monitoring system including spectrum, chromaticity, and luminous flux is established. Through high-precision sensors and synchronous acquisition technology, the all-round dynamic monitoring of the optical characteristics of LED lamp beads is realized, and the monitoring accuracy is improved to 0.1 nm. By integrating a bidirectional gated recurrent neural network and a PID control algorithm, the adaptive adjustment of the luminous characteristics of LED lamp beads is realized, and the control accuracy reaches ±0.1%, greatly improving the uniformity and stability of the luminous color. Through multi-dimensional deviation vector analysis and a comprehensive evaluation mechanism, a complete closed-loop control system is established, which can compensate for the performance drift caused by environmental changes and device aging in real time, and extend the service life of LED lamp beads. By adopting a high-precision centrifugal packaging process and an automatic control technology, the production efficiency and product consistency are significantly improved, and the production cost is reduced.

[0168] An embodiment of the present invention also provides an LED lamp bead, which is used to realize the method for uniformly controlling the luminous color of the above-mentioned LED lamp bead in the claims when in use.

[0169] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0170] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0171] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs, etc., which can store program codes.

[0172] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for uniformly controlling the light color of an LED lamp bead, characterized in that: The method comprises: The LED chips are placed on the LED bracket in a triangular ring shape, the LED chips are connected by wires, and the first packaging glue and the second packaging glue are injected and centrifuged in sequence to obtain LED lamp beads with a double-layer packaging structure; Applying rated operating parameters to the LED lamp beads and collecting an initial optical parameter matrix; A standard data model is established based on the initial optical parameter matrix, and multi-parameter real-time monitoring of the LED lamp beads is performed to obtain a monitoring data set; The monitoring data set is input into a data processing unit, and the spectral deviation coefficient is calculated by a spectral similarity algorithm, the chromaticity deviation coefficient is calculated by a Euclidean distance algorithm, and the luminous flux deviation coefficient is calculated by a ratio algorithm to obtain a multidimensional deviation vector; specifically comprising: performing Fourier transform on the real-time spectral data in the monitoring data set, extracting spectral frequency components by a spectrum analysis algorithm, and obtaining spectral frequency domain features; performing cosine similarity calculation on the spectral frequency domain features and the spectral data in the standard data model, obtaining a spectral similarity value with a numerical range of 0-1 by normalization processing, and obtaining a spectral deviation coefficient; converting the real-time chromaticity data in the monitoring data set into three-dimensional chromaticity space coordinates, generating chromaticity space distribution points by a coordinate mapping algorithm, and obtaining a chromaticity space vector; comparing the chromaticity space vector with the spectral data in the standard data model, and obtaining a spectral similarity value with a numerical range of 0-1 by normalization processing, and obtaining a spectral deviation coefficient. The Euclidean distance is calculated based on the chromaticity data, and the distance value is mapped to the interval of 0-1 through standardization processing to obtain the chromaticity deviation coefficient; the real-time luminous flux data in the monitoring data set is reconstructed according to the time series, and the luminous flux time series curve is generated through the sliding window algorithm to obtain the luminous flux sequence, and the luminous flux sequence is ratioed with the luminous flux value in the standard data model, and the ratio result is converted into a linear relationship through logarithmic transformation to obtain the luminous flux deviation coefficient; a three-dimensional feature space is constructed based on the spectral deviation coefficient, the chromaticity deviation coefficient and the luminous flux deviation coefficient, and a standard feature basis is generated through orthogonalization processing to obtain a feature basis matrix; the spectral deviation coefficient, the chromaticity deviation coefficient and the luminous flux deviation coefficient are projected on the feature basis matrix, and a multi-dimensional deviation vector is generated through matrix operation; The multidimensional deviation vector is weighted and summed to obtain a comprehensive deviation evaluation value; specifically, the method comprises: performing principal component analysis on the multidimensional deviation vector, extracting the main characteristic component and the secondary characteristic component by a singular value decomposition algorithm, and obtaining a characteristic component sequence; performing an orthogonal transformation on the characteristic component sequence, constructing a standard orthogonal basis vector by a Schmidt orthogonalization algorithm, and obtaining an orthogonal basis matrix; and performing dimensionality reduction processing on the orthogonal basis matrix according to the primary and secondary characteristic contribution rates, retaining the main characteristic dimensions by a truncated singular value method, and obtaining a reduced-dimensional characteristic matrix; constructing a weight distribution model based on the reduced-dimensional characteristic matrix, and calculating the spectral deviation coefficient and the chromaticity deviation by a coefficient of variation method. The initial weight value of the coefficient and the luminous flux deviation coefficient is obtained by performing a hierarchical analysis on the initial weight value, verifying the rationality of the weight distribution through a consistency check algorithm, obtaining a modified weight coefficient, and linearly combining the modified weight coefficient with a preset weight ratio of 0.4:0.4:0.2, and generating a target weight coefficient through a weighted average algorithm; based on the target weight coefficient, weighted fusion is performed on the multidimensional deviation vector, a weighted deviation value is generated through a matrix multiplication operation, a normalized deviation value is obtained, and a nonlinear mapping is performed on the normalized deviation value, and the deviation value is compressed to a range of 0-1 through a sigmoid function to obtain a comprehensive deviation evaluation value; Based on the comprehensive deviation evaluation value, a current correction amount and a voltage correction amount are generated, and the LED lamp bead is closed-loop controlled until the comprehensive deviation evaluation value is lower than a set threshold value, and a control termination signal is obtained.

2. The method for uniformly controlling the light emission color of an LED lamp bead according to claim 1, characterized in that: The LED chips are placed on the LED bracket in a triangular ring manner, the LED chips are connected by wires, and the first packaging glue and the second packaging glue are injected and centrifuged in sequence to obtain the LED lamp beads with a double-layer packaging structure, including: Performing surface treatment on the LED bracket, coating an aluminum oxide reflective layer on the inner wall of the LED bracket, and combining the reflective layer with the bracket through a high-temperature sintering process to obtain a bracket structure; The LED chips are placed on the support structure so that the LED chips form an angle of 60° with the horizontal direction, and arranged in a triangular ring array by a positioning system to obtain a chip array structure; Performing gold wire bonding on the chip array structure, connecting the LED chip and the conductive pad of the support structure through an automatic wire bonding system to obtain a conductive structure; A phosphor with a particle size of 10 microns is mixed with an epoxy resin in a mass ratio of 4:6, and dispersed by a high-speed stirring device to obtain a first packaging glue, and the first packaging glue is injected into the conductive structure, and the injection amount is controlled by a weight sensor so that the injection amount error is controlled within the range of ±0.1 mg to obtain a first packaging layer; The first packaging layer is placed in a centrifuge at a speed of 3000 rpm, and the phosphor is evenly distributed around the LED chip by centrifugation for 300 seconds to obtain a coating structure, and the diffusion powder and the epoxy resin are mixed at a mass ratio of 1.2:98.8, and dispersed by a high-speed stirring device to obtain a second packaging glue; The second packaging glue is injected into the coating structure, and the injection amount and injection angle are adjusted by the numerical control system so that the second packaging glue forms a concave structure with a curvature radius of 1.5 times the diameter of the packaging surface, thereby obtaining an LED lamp bead with a double-layer packaging structure.

3. The method for uniformly controlling the light emission color of an LED lamp bead according to claim 2, characterized in that: The step of applying rated operating parameters to the LED lamp beads and collecting an initial optical parameter matrix includes: The LED lamp beads are subjected to environmental parameter control and preheated in a standard environment with a temperature of 25° C. and a relative humidity of 60% to obtain LED lamp beads in a stable working state; Apply the rated working parameters of 3.0V voltage and 350mA current to the LED lamp beads in stable working state, and control the power supply through a high-precision DC power supply to obtain a stable power supply state; A spectrum analyzer with a resolution of 0.1 nm is aimed at the light-emitting surface of the LED lamp bead, and scanning and collecting data within the wavelength range of 380-780 nm to obtain spectral intensity distribution data; The chromaticity parameters of the light-emitting surface of the LED lamp bead are measured, and the CIE chromaticity coordinate values ​​are collected by a colorimeter at a distance of 100 mm from the light-emitting surface to obtain chromaticity distribution data; The LED lamp bead is placed at the center of an integrating sphere, and the luminous flux value is collected by a luminous flux sensor to obtain luminous flux data, and the spectral intensity distribution data is divided into wavelength intervals, and discrete wavelength points are numerically fitted by an interpolation algorithm to obtain a continuous spectrum curve; The chromaticity distribution data and the luminous flux data are fused, a normalized parameter value is obtained through standardization processing, and standardized data is obtained. A matrix synthesis operation is performed on the continuous spectral curve and the standardized data, and a three-dimensional data structure is generated through a data reconstruction algorithm to obtain an initial optical parameter matrix.

4. The method for uniformly controlling the light emission color of an LED lamp bead according to claim 3, characterized in that: The standard data model is established based on the initial optical parameter matrix, and multi-parameter real-time monitoring of the LED lamp beads is performed to obtain a monitoring data set, including: Performing parameter decomposition on the initial optical parameter matrix, converting the spectrum data, chromaticity data and luminous flux data into a feature matrix through a matrix decomposition algorithm, and obtaining the standard optical characteristics of the LED lamp beads; A three-dimensional parameter space is constructed based on the standard optical characteristics, and a spectral curve model, a chromaticity distribution model and a luminous flux variation model are fitted by the least squares method to obtain a standard mathematical model of the LED lamp bead; The standard mathematical model is input into a Kalman filter, and a dynamic prediction model of spectrum, chromaticity and luminous flux is established through a state prediction algorithm to obtain a standard data model; The parameter threshold is set based on the standard data model, and the spectral fluctuation range, chromaticity shift range and luminous flux variation range are determined by the interval control algorithm to obtain the parameter control interval; The LED lamp beads are monitored in real time, and spectrum, chromaticity and luminous flux data are synchronously collected through a multi-channel data acquisition system with a sampling frequency of 1000 Hz to obtain real-time sampling data; Inputting the real-time sampling data into the standard data model, calculating the deviation between the current state and the predicted state through a state estimation algorithm, and obtaining state deviation data; Perform anomaly detection based on the state deviation data, determine whether the data falls within the parameter control interval through a Mahalanobis distance algorithm, and obtain a data validity mark; The data marked as valid are reorganized in time sequence, and the spectral data, chromaticity data and luminous flux data are integrated into a unified format through a data encapsulation algorithm to obtain a monitoring data set.

5. The method for uniformly controlling the light emission color of an LED lamp bead according to claim 1, characterized in that: The method generates a current correction amount and a voltage correction amount based on the comprehensive deviation evaluation value, performs closed-loop control on the LED lamp bead, and obtains a control termination signal until the comprehensive deviation evaluation value is lower than a set threshold value, including: Inputting the comprehensive deviation evaluation value into a parameter prediction model, wherein the parameter prediction model is a bidirectional threshold recurrent neural network, wherein the bidirectional threshold recurrent neural network comprises an input layer, a hidden layer and a fully connected layer, wherein the hidden layer comprises a forward threshold recurrent unit and a reverse threshold recurrent unit; Input the comprehensive deviation evaluation value into the input layer of the parameter prediction model, extract the forward time series feature through the forward threshold recurrent unit, extract the reverse time series feature through the reverse threshold recurrent unit, and obtain the initial compensation parameter through the full connection layer fusion; The initial compensation parameters are processed by a self-attention mechanism, and an attention weight matrix is ​​obtained by calculating the dot product operation of the query matrix, the key matrix and the value matrix, and performing a softmax normalization process; The initial compensation parameters are weighted based on the attention weight matrix, and feature enhancement is performed through a multi-head attention mechanism, where each attention head independently calculates an attention score and performs feature extraction to obtain an enhanced feature vector; The enhanced characteristic vector is input into the PID controller, and the parameters are adjusted by using a proportional coefficient of 0.8, an integral coefficient of 0.3, and a differential coefficient of 0.1 to obtain a PID compensation parameter; Adaptive adjustment is performed according to the PID compensation parameters to achieve collaborative optimization of deep learning and PID control, and current correction and voltage correction are obtained. The current correction and voltage correction are input into a high-precision digital-to-analog converter, and signal conversion is performed through a 16-bit DAC. The drive signal is output through PWM modulation to obtain the corrected working parameters; The corrected working parameters are closed-loop controlled, and the comparison result between the comprehensive deviation evaluation value and the set threshold is monitored in real time, and the comparison result is fed back to the input layer of the bidirectional threshold recurrent neural network to obtain a control termination signal.

6. A device for uniformly controlling the light color of an LED lamp bead, characterized in that: Used to perform the method for uniformly controlling the light color of an LED lamp bead as described in any one of claims 1 to 5, the device for uniformly controlling the light color of an LED lamp bead comprises: The packaging module is used to place the LED chip in a triangular ring on the LED bracket, connect the LED chip with wires, inject the first packaging glue and the second packaging glue in sequence and perform centrifugal treatment to obtain an LED lamp bead with a double-layer packaging structure; A collection module, used for applying rated working parameters to the LED lamp beads and collecting an initial optical parameter matrix; A monitoring module, used to establish a standard data model based on the initial optical parameter matrix, perform multi-parameter real-time monitoring on the LED lamp beads, and obtain a monitoring data set; A calculation module, used for inputting the monitoring data set into a data processing unit, calculating the spectral deviation coefficient by a spectral similarity algorithm, calculating the chromaticity deviation coefficient by a Euclidean distance algorithm, calculating the luminous flux deviation coefficient by a ratio algorithm, and obtaining a multidimensional deviation vector; A summing module, used for performing a weighted summing operation on the multi-dimensional deviation vector to obtain a comprehensive deviation evaluation value; The control module is used to generate a current correction amount and a voltage correction amount based on the comprehensive deviation evaluation value, and perform closed-loop control on the LED lamp bead until the comprehensive deviation evaluation value is lower than a set threshold value, thereby obtaining a control termination signal.

7. An LED lamp bead, characterized in that: When used, the LED lamp bead is used to implement the method for uniformly controlling the luminous color of the LED lamp bead according to any one of claims 1 to 5.

Images