A test data analysis system and method applied to an automobile LED module

Abstract

The application discloses a kind of test data analysis system and method applied to automobile LED module, it is related to data test technical field, the change information of the electric data, thermal data and light data in historical automobile LED module is collected, the influence relationship between three kinds of physical data is analyzed, and the detection model of three kinds of physical data is generated based on influence relationship;According to weight, the detection model of three kinds of physical data is coupled to obtain electric-optical-thermal multi-physical field coupling model;Real-time data is constructed state space using acquisition and calculation, and the weight of three detection models is used as action space;Reward function is set;Strategy network and value network are constructed, and dynamic weight mechanism is obtained;Three detection models are reconstructed using real-time weight, and electric-optical-thermal multi-physical field coupling model is optimized.

Description

Technical Field

[0001] This invention relates to the field of data testing technology, specifically to a test data analysis system and method for automotive LED modules. Background Technology

[0002] The automotive industry is continuously evolving towards intelligence, electrification, and connectivity, placing higher demands on the performance, reliability, and safety of automotive components. As a core component of automotive lighting systems, LED modules directly impact driving safety and the driving experience. For example, the realization of intelligent lighting functions such as adaptive headlights and adaptive high beams relies on the precise control and stable performance of LED modules. LEDs possess advantages such as high luminous efficiency, long lifespan, fast response speed, and small size, gradually replacing traditional light sources as the mainstream technology for automotive lighting. However, the performance of LED modules is affected by various factors, including current, temperature, and luminous intensity, and these factors are interconnected. For instance, changes in current cause LEDs to heat up, thus affecting their light output and color characteristics. Therefore, precise monitoring and analysis of these physical quantities are necessary to optimize the design and control of LED modules.

[0003] However, existing multi-physical quantity data testing models have limited interaction with specific automotive scenarios, and only provide simple combinations of multiple physical quantities. In automotive operations, the working conditions change extremely rapidly, and testing models with simple combinations in some extreme conditions and special scenarios have large testing errors. Therefore, how to set combination weights and adjust them in real time is crucial to the effectiveness of data testing. Summary of the Invention

[0004] The purpose of this invention is to provide a test data analysis system and method for automotive LED modules to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A test data analysis method for automotive LED modules, the method comprising the following steps:

[0007] S100: Collect historical information on changes in electrical, thermal, and optical data in automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship.

[0008] Furthermore, the specific steps for generating the detection model based on the influence relationship of the three types of physical data are as follows:

[0009] S101. Collect electrical, thermal, and optical data from the automotive LED module. The electrical data includes current and voltage data. Analyze the Joule heating generated when current passes through the LED module, finding a correlation between temperature change and the square of the current. Utilize the influence relationship between thermal and electrical data to construct the electro-thermal physical equation:

[0010]

[0011] In the formula, T represents the LED module temperature, t represents time, I represents the LED module current, k1 represents the electrothermal conversion coefficient, k2 represents the heat dissipation coefficient, and T amb Indicates ambient temperature;

[0012] When the temperature in the LED module changes, the change in light intensity is observed to obtain the influence relationship between thermal and optical data, and the thermal-optical physical equation is constructed as follows:

[0013]

[0014] In the formula, L represents the light intensity after the temperature change, L0 represents the light intensity before the temperature change, α represents the temperature attenuation coefficient, T0 represents the initial temperature, and T represents the LED module temperature.

[0015] When current flows through the LED module, the light intensity is observed to obtain the relationship between current and light intensity, and the electro-optical physical equation is constructed as follows:

[0016]

[0017] In the formula, β represents the photoelectric efficiency, and r represents the temperature modulation coefficient;

[0018] S102. Based on the physical equations of the three constructed influence relationships, parameter fitting is performed respectively to obtain the electrical detection model as follows: The thermal detection model is The light detection model is

[0019] By analyzing the relationships between these three types of physical data, we can gain a deeper understanding of the working principles and performance characteristics of automotive LED modules. During product design and manufacturing, these relationships can be used for optimization, improving the quality and reliability of LED modules. For example, by controlling electrical and thermal data, the stability of optical data can be ensured, reducing problems such as light decay and color shift.

[0020] The detection model can monitor and predict the performance of LED modules in real time. When abnormal data is detected, potential faults can be identified in a timely manner, and corresponding measures can be taken to repair them, avoid failures, and improve product reliability and safety.

[0021] S200. Set weights for the detection equations of the three physical data, and couple the detection models of the three physical data according to the weights to obtain the electro-optic-thermal multiphysics coupling model.

[0022] Furthermore, the specific steps for coupling the detection models of the three physical data types according to their weights to obtain the electro-optical-thermal multiphysics coupling model are as follows:

[0023] S201. Set coupling weights for the detection models of the three types of physical data respectively. Let the weight of the electrical detection model be w. e The weights of the thermal detection model are w. t The weights of the optical detection model are w p ;

[0024] By using the set weights to weighted couple the detection models of the three physical data types, the electro-optical-thermal multiphysics coupling model is obtained as follows:

[0025]

[0026] In the formula, Cou(I, T, L) represents the output of the electro-optical-thermal multiphysics coupling model.

[0027] By using a coupled model, the complex interactions between these factors can be comprehensively considered, providing a more complete description of the LED module's operating state. For example, changes in current not only directly affect light output but also indirectly affect light characteristics by generating heat; the coupled model can simultaneously consider these direct and indirect effects.

[0028] During product operation, coupled models can serve as a tool for fault diagnosis. When anomalies are detected in certain physical data, combining them with coupled models can more accurately determine the cause and location of the fault. For example, when optical data is abnormal, by analyzing the changes in electrical and thermal data in the coupled model, it can be determined whether the abnormal current is caused by a circuit fault or whether the temperature rise is caused by a heat dissipation problem, thus affecting the optical output, thereby improving the accuracy and efficiency of fault diagnosis.

[0029] Under varying driving conditions, LED modules in automobiles face a variety of complex operating scenarios, such as different ambient temperatures, humidity levels, and power supply voltage fluctuations. Coupled models can take into account the combined effects of these factors on the electro-optical-thermal multiphysics fields, more accurately predicting the performance of LED modules under various operating conditions and providing assurance for the reliability design of automotive lighting systems.

[0030] S300: Uses sensors to collect real-time data on current, temperature and light intensity in automotive LED modules, calculates the real-time heating rate and light intensity deviation in automotive LED modules, and standardizes the collected data.

[0031] Furthermore, the specific steps for calculating the real-time heating rate and light intensity deviation in automotive LED modules are as follows:

[0032] S301. Utilize sensors to collect real-time data on current, temperature, and light intensity in the automotive LED module, and calculate the real-time heating rate of the automotive LED module using the following formula:

[0033]

[0034] In the formula, rT(t) represents the real-time heating rate of the automotive LED module, T(t) represents the temperature of the automotive LED module at time t, and Δt represents the heating time.

[0035] The formula for calculating the real-time light intensity deviation of automotive LED modules is: ΔL(t)=|L(t)-L ref |;In the formula, △L(t) represents the real-time light intensity deviation of the automotive LED module, L ref L(t) represents the nominal value of the luminous intensity of the automotive LED module; L(t) represents the collected luminous intensity of the automotive LED module.

[0036] S302. Use the Z-score standardization formula to standardize the calculated and collected data of different dimensions and transform them into data of a unified dimension.

[0037] S400: Construct a state space using real-time collected and computed data, and use the weights of the three detection models as the action space; set a reward function;

[0038] Furthermore, the specific steps for setting the reward function are as follows:

[0039] S401. Set up the state space, storing all standardized data in the state space. Let the state space be st = {I(t), T(t), L(t), rT(t), ΔL(t)}; where I(t) represents the collected current data of the automotive LED module. Set up the action space, which represents the adjustment actions on the weights of the electro-optical-thermal multiphysics coupling model. Let the action space be at = {Δw e , △w t , △w p}; Set restrictions: w e '+w t '+w p ' = 1, where w e '、w t 'and w p 'These represent the weights of the three detection models after adjustment;

[0040] S402. When testing automotive LED module data using an electro-optical-thermal multiphysics coupling model, different weight combinations are output in the action space. For each weight combination, the joint mean square error of the electro-optical-thermal multiphysics coupling model and the weight fluctuation variance are calculated. The reward function is designed using the model's joint mean square error and weight fluctuation variance:

[0041] R = -λ1×MSE - λ2×Var(w);

[0042] In the formula, R represents the reward result, λ1 and λ2 represent the hyperparameters set, MSE represents the joint mean square error of the model, and Var(w) represents the weight variance of the model.

[0043] S500: Using the state space as input and the action space as output, construct a policy network and use a neural network to construct structural layers; similarly, using the state space and action space as input and predicting accumulated rewards as output, construct a value network; use the policy network and value network to construct a dynamic weight mechanism.

[0044] Furthermore, the specific steps for constructing a dynamic weight mechanism using policy networks and value networks are as follows:

[0045] S501. Construct a policy network using the state space as input and the action space as output; construct a structural layer using a neural network; similarly, construct a value network using the state space and action space as input and the predicted cumulative reward as output.

[0046] S502. The policy network is trained using the data in the state space of the historical car LED module and the weight changes to obtain the mapping relationship between different state spaces and weight changes as at = H(st); When the car LED module is tested in real time, the data in the real-time state space is input into the policy network and the weight adjustment action is output.

[0047] The value network takes the real-time state space and weight adjustment actions (st, at) as inputs, calculates the accumulated reward according to the reward function, and outputs the result. It compares the outputs of all rewards and selects the input corresponding to the maximum reward as the optimal combination (st, at). The dynamic weight mechanism is constructed by combining the policy network and the value network.

[0048] During vehicle operation, environmental conditions and operating conditions are complex and constantly changing. By constructing a state space using real-time data and dynamically adjusting the weights of the detection model, the coupled model can continuously adapt to these changes, ensuring accurate description of the electro-optical-thermal characteristics of the LED module under various conditions.

[0049] By using the weights of the three detection models as the action space and leveraging the dynamic weighting mechanism of reinforcement learning, the weights are continuously adjusted based on real-time states and reward functions. This approach allows the model to automatically find the optimal weight combination under different operating conditions, enabling the coupled model to more accurately reflect the interrelationships between the three physical quantities. For example, in high-temperature environments, appropriately increasing the weights of the thermal model can more accurately predict the impact of temperature on light intensity and current.

[0050] S600 uses a dynamic weighting mechanism to obtain the real-time weights of three physical models, uses the real-time weights to reconstruct the three detection models, optimizes the electro-optical-thermal multi-physics coupling model, and uses the optimized electro-optical-thermal multi-physics coupling model to test automotive LED module data.

[0051] Furthermore, the specific steps for testing automotive LED module data using the optimized electro-optical-thermal multiphysics coupling model are as follows:

[0052] S601. When testing automotive LED modules, a dynamic weighting mechanism is used to calculate real-time weight adjustments. The weights of the electro-optical-thermal multiphysics coupling model are adjusted in real-time based on the time series. The weight adjustment formula for each physical detection model is as follows:

[0053] w i (t)=w i (t-1)+Δw i ;

[0054] In the formula, W i (t) represents the weight of the i-th physical detection model at time t, w i (t-1) represents the weight of the i-th physical detection model at time t-1, Δw i This represents the weight adjustment amount for the i-th physical detection model; i belongs to {e, t, p};

[0055] The automotive LED module was tested using a weighted electro-optical-thermal multiphysics coupling model.

[0056] By reconstructing the three detection models using real-time weights obtained through a dynamic weighting mechanism, the electro-optical-thermal multiphysics coupling model can be optimized. The optimized model can more accurately simulate the actual operation of LED modules, providing a more reliable basis for the design, control, and fault diagnosis of automotive LED modules.

[0057] A test data analysis system for automotive LED modules includes a test model construction module, a multiphysics coupling module, a data acquisition module, a spatial setting module, a dynamic weighting mechanism construction module, and a real-time weight adjustment module.

[0058] The detection model construction module is used to collect information on changes in electrical, thermal, and optical data in historical automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship.

[0059] The multiphysics coupling module is used to set weights for the detection equations of the three types of physical data, and to couple the detection models of the three types of physical data according to the weights to obtain an electro-optic-thermal multiphysics coupling model.

[0060] The data acquisition module is used to collect current, temperature and light intensity data in the automotive LED module in real time using sensors, calculate the real-time heating rate and light intensity deviation in the automotive LED module, and standardize the collected data.

[0061] The space setting module is used to construct a state space using the collected and calculated real-time data, and to use the weights of the three detection models as the action space; and to set the reward function.

[0062] The dynamic weighting mechanism construction module is used to generate weight adjustment actions using a policy network, judge weight adjustment actions using a value network, and construct a dynamic weighting mechanism using the policy network and the value network.

[0063] The real-time weight adjustment module is used to adjust the weights of the three physical detection models in the electro-optic-thermal multiphysics coupling model by using the weight adjustment action output by the dynamic weight mechanism.

[0064] The detection model construction module includes an electrical data detection model unit, a thermal data detection model unit, and an optical data detection model unit;

[0065] The electrical data detection model unit is used to analyze the Joule heat generated when current passes through the LED module, and obtain the correlation between temperature change and the square of current. The electro-thermal physical equation is constructed by utilizing the influence relationship between thermal data and electrical data.

[0066] The thermal data detection model unit is used to observe the change in light intensity when the temperature changes in the LED module, obtain the influence relationship between thermal data and light data, and construct the thermal-optical physical equation.

[0067] The optical data detection model unit is used to observe the light intensity when current passes through the LED module, obtain the influence relationship between current and light intensity, and construct the electro-optical physical equation.

[0068] The dynamic weighting mechanism construction module includes a policy network unit and a value network unit;

[0069] The policy network unit is used to construct a policy network by taking the state space as input and the action space as output.

[0070] The value network unit is used to construct a value network by taking the state space and action space as inputs and predicting the accumulated reward as output.

[0071] Compared with the prior art, the beneficial effects of the present invention are:

[0072] 1. This invention utilizes real-time weights obtained through a dynamic weighting mechanism to reconstruct three detection models, thereby optimizing the electro-optical-thermal multiphysics coupling model. The optimized model can more accurately simulate the actual operation of LED modules, providing a more reliable basis for the design, control, and fault diagnosis of automotive LED modules.

[0073] 2. The policy network of this invention takes the state space as input and outputs an action space, enabling it to make optimal decisions based on real-time states. The value network predicts accumulated rewards, providing a reference for the policy network's decisions and helping it obtain the maximum reward in the long run. This reinforcement learning-based decision-making mechanism can quickly find the optimal weight adjustment strategy in complex environments, improving decision-making efficiency. Attached Figure Description

[0074] Figure 1 This is a module distribution diagram of a test data analysis system for automotive LED modules according to the present invention;

[0075] Figure 2 This is a schematic diagram illustrating the steps of a test data analysis method for automotive LED modules according to the present invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Example: Figures 1-2 As shown, the present invention provides a technical solution.

[0078] A test data analysis method for automotive LED modules, the method comprising the following steps:

[0079] S100: Collect historical information on changes in electrical, thermal, and optical data in automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship.

[0080] The specific steps for generating a detection model based on influence relationships to produce three types of physical data are as follows:

[0081] S101. Collect electrical, thermal, and optical data from the automotive LED module. The electrical data includes current and voltage data. Analyze the Joule heating generated when current passes through the LED module, finding a correlation between temperature change and the square of the current. Utilize the influence relationship between thermal and electrical data to construct the electro-thermal physical equation:

[0082]

[0083] In the formula, T represents the LED module temperature, t represents time, I represents the LED module current, k1 represents the electrothermal conversion coefficient, k2 represents the heat dissipation coefficient, and T amb Indicates ambient temperature;

[0084] When the temperature in the LED module changes, the change in light intensity is observed to obtain the influence relationship between thermal and optical data, and the thermal-optical physical equation is constructed as follows:

[0085]

[0086] In the formula, L represents the light intensity after the temperature change, L0 represents the light intensity before the temperature change, α represents the temperature attenuation coefficient, T0 represents the initial temperature, and T represents the LED module temperature.

[0087] When current flows through the LED module, the light intensity is observed to obtain the relationship between current and light intensity, and the electro-optical physical equation is constructed as follows:

[0088]

[0089] In the formula, β represents the photoelectric efficiency, and r represents the temperature modulation coefficient;

[0090] S102. Based on the physical equations of the three constructed influence relationships, parameter fitting is performed respectively to obtain the electrical detection model as follows: The thermal detection model is The light detection model is

[0091] By analyzing the relationships between these three types of physical data, we can gain a deeper understanding of the working principles and performance characteristics of automotive LED modules. During product design and manufacturing, these relationships can be used for optimization, improving the quality and reliability of LED modules. For example, by controlling electrical and thermal data, the stability of optical data can be ensured, reducing problems such as light decay and color shift.

[0092] The detection model can monitor and predict the performance of LED modules in real time. When abnormal data is detected, potential faults can be identified in a timely manner, and corresponding measures can be taken to repair them, avoid failures, and improve product reliability and safety.

[0093] S200. Set weights for the detection equations of the three physical data, and couple the detection models of the three physical data according to the weights to obtain the electro-optic-thermal multiphysics coupling model.

[0094] The specific steps for coupling the detection models of the three physical data according to the weights to obtain the electro-optic-thermal multiphysics coupling model are as follows:

[0095] S201. Set coupling weights for the detection models of the three types of physical data respectively. Let the weight of the electrical detection model be w. e The weights of the thermal detection model are w. t The weights of the optical detection model are w p ;

[0096] By using the set weights to weighted couple the detection models of the three physical data types, the electro-optical-thermal multiphysics coupling model is obtained as follows:

[0097]

[0098] In the formula, Cou(I, T, L) represents the output of the electro-optical-thermal multiphysics coupling model.

[0099] By using a coupled model, the complex interactions between these factors can be comprehensively considered, providing a more complete description of the LED module's operating state. For example, changes in current not only directly affect light output but also indirectly affect light characteristics by generating heat; the coupled model can simultaneously consider these direct and indirect effects.

[0100] During product operation, coupled models can serve as a tool for fault diagnosis. When anomalies are detected in certain physical data, combining them with coupled models can more accurately determine the cause and location of the fault. For example, when optical data is abnormal, by analyzing the changes in electrical and thermal data in the coupled model, it can be determined whether the abnormal current is caused by a circuit fault or whether the temperature rise is caused by a heat dissipation problem, thus affecting the optical output, thereby improving the accuracy and efficiency of fault diagnosis.

[0101] Under varying driving conditions, LED modules in automobiles face a variety of complex operating scenarios, such as different ambient temperatures, humidity levels, and power supply voltage fluctuations. Coupled models can take into account the combined effects of these factors on the electro-optical-thermal multiphysics fields, more accurately predicting the performance of LED modules under various operating conditions and providing assurance for the reliability design of automotive lighting systems.

[0102] S300: Uses sensors to collect real-time data on current, temperature and light intensity in automotive LED modules, calculates the real-time heating rate and light intensity deviation in automotive LED modules, and standardizes the collected data.

[0103] The specific steps for calculating the real-time heating rate and light intensity deviation in automotive LED modules are as follows:

[0104] S301. Utilize sensors to collect real-time data on current, temperature, and light intensity in the automotive LED module, and calculate the real-time heating rate of the automotive LED module using the following formula:

[0105]

[0106] In the formula, rT(t) represents the real-time heating rate of the automotive LED module, T(t) represents the temperature of the automotive LED module at time t, and Δt represents the heating time.

[0107] The formula for calculating the real-time light intensity deviation of automotive LED modules is: ΔL(t)=|L(t)-L ref |;In the formula, △L(t) represents the real-time light intensity deviation of the automotive LED module, L ref L(t) represents the nominal value of the luminous intensity of the automotive LED module; L(t) represents the collected luminous intensity of the automotive LED module.

[0108] S302. Use the Z-score standardization formula to standardize the calculated and collected data of different dimensions and transform them into data of a unified dimension.

[0109] S400: Construct a state space using real-time collected and computed data, and use the weights of the three detection models as the action space; set a reward function;

[0110] The specific steps for setting the reward function are as follows:

[0111] S401. Set up the state space, storing all standardized data in the state space. Let the state space be st = {I(t), T(t), L(t), rT(t), ΔL(t)}; where I(t) represents the collected current data of the automotive LED module. Set up the action space, which represents the adjustment actions on the weights of the electro-optical-thermal multiphysics coupling model. Let the action space be at = {Δw e , △w t , △w p}; Set restrictions: w e '+w t '+w p ' = 1, where w e '、w t 'and w p 'These represent the weights of the three detection models after adjustment;

[0112] S402. When testing automotive LED module data using an electro-optical-thermal multiphysics coupling model, different weight combinations are output in the action space. For each weight combination, the joint mean square error of the electro-optical-thermal multiphysics coupling model and the weight fluctuation variance are calculated. The reward function is designed using the model's joint mean square error and weight fluctuation variance:

[0113] R = -λ1×MSE - λ2×Var(w);

[0114] In the formula, R represents the reward result, λ1 and λ2 represent the hyperparameters set, MSE represents the joint mean square error of the model, and Var(w) represents the weight variance of the model.

[0115] S500: Construct a policy network using the state space as input and the action space as output, and use a neural network to construct the structural layer; similarly, construct a value network using the state space and action space as input and the predicted accumulated reward as output; and use the policy network and value network to construct a dynamic weight mechanism.

[0116] The specific steps for constructing a dynamic weight mechanism using policy networks and value networks are as follows:

[0117] S501. Construct a policy network using the state space as input and the action space as output; construct a structural layer using a neural network; similarly, construct a value network using the state space and action space as input and the predicted cumulative reward as output.

[0118] S502. The policy network is trained using the data in the state space of the historical car LED module and the weight changes to obtain the mapping relationship between different state spaces and weight changes as at = H(st); When the car LED module is tested in real time, the data in the real-time state space is input into the policy network and the weight adjustment action is output.

[0119] The value network takes the real-time state space and weight adjustment actions (st, at) as inputs, calculates the accumulated reward according to the reward function, and outputs the result. It compares the outputs of all rewards and selects the input corresponding to the maximum reward as the optimal combination (st, at). The dynamic weight mechanism is constructed by combining the policy network and the value network.

[0120] During vehicle operation, environmental conditions and operating conditions are complex and constantly changing. By constructing a state space using real-time data and dynamically adjusting the weights of the detection model, the coupled model can continuously adapt to these changes, ensuring accurate description of the electro-optical-thermal characteristics of the LED module under various conditions.

[0121] By using the weights of the three detection models as the action space and leveraging the dynamic weighting mechanism of reinforcement learning, the weights are continuously adjusted based on real-time states and reward functions. This approach allows the model to automatically find the optimal weight combination under different operating conditions, enabling the coupled model to more accurately reflect the interrelationships between the three physical quantities. For example, in high-temperature environments, appropriately increasing the weights of the thermal model can more accurately predict the impact of temperature on light intensity and current.

[0122] S600 uses a dynamic weighting mechanism to obtain the real-time weights of three physical models, uses the real-time weights to reconstruct the three detection models, optimizes the electro-optical-thermal multi-physics coupling model, and uses the optimized electro-optical-thermal multi-physics coupling model to test automotive LED module data.

[0123] The specific steps for testing automotive LED module data using the optimized electro-optical-thermal multiphysics coupling model are as follows:

[0124] S601. When testing automotive LED modules, a dynamic weighting mechanism is used to calculate real-time weight adjustments. The weights of the electro-optical-thermal multiphysics coupling model are adjusted in real-time based on the time series. The weight adjustment formula for each physical detection model is as follows:

[0125] w i (t)=w i (t-1)+Δw i ;

[0126] In the formula, W i (t) represents the weight of the i-th physical detection model at time t, w i (t-1) represents the weight of the i-th physical detection model at time t-1, Δw i This represents the weight adjustment amount for the i-th physical detection model; i belongs to {e, t, p};

[0127] The automotive LED module was tested using a weighted electro-optical-thermal multiphysics coupling model.

[0128] By reconstructing the three detection models using real-time weights obtained through a dynamic weighting mechanism, the electro-optical-thermal multiphysics coupling model can be optimized. The optimized model can more accurately simulate the actual operation of LED modules, providing a more reliable basis for the design, control, and fault diagnosis of automotive LED modules.

[0129] A test data analysis system for automotive LED modules includes a test model construction module, a multiphysics coupling module, a data acquisition module, a spatial setting module, a dynamic weighting mechanism construction module, and a real-time weight adjustment module.

[0130] The detection model construction module is used to collect information on changes in electrical, thermal, and optical data in historical automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship.

[0131] The multiphysics coupling module is used to set weights for the detection equations of the three types of physical data, and to couple the detection models of the three types of physical data according to the weights to obtain an electro-optic-thermal multiphysics coupling model.

[0132] The data acquisition module is used to collect current, temperature and light intensity data in the automotive LED module in real time using sensors, calculate the real-time heating rate and light intensity deviation in the automotive LED module, and standardize the collected data.

[0133] The space setting module is used to construct a state space using the collected and calculated real-time data, and to use the weights of the three detection models as the action space; and to set the reward function.

[0134] The dynamic weighting mechanism construction module is used to generate weight adjustment actions using a policy network, judge weight adjustment actions using a value network, and construct a dynamic weighting mechanism using the policy network and the value network.

[0135] The real-time weight adjustment module is used to adjust the weights of the three physical detection models in the electro-optic-thermal multiphysics coupling model by using the weight adjustment action output by the dynamic weight mechanism.

[0136] The detection model construction module includes an electrical data detection model unit, a thermal data detection model unit, and an optical data detection model unit;

[0137] The electrical data detection model unit is used to analyze the Joule heat generated when current passes through the LED module, and obtain the correlation between temperature change and the square of current. The electro-thermal physical equation is constructed by utilizing the influence relationship between thermal data and electrical data.

[0138] The thermal data detection model unit is used to observe the change in light intensity when the temperature changes in the LED module, obtain the influence relationship between thermal data and light data, and construct the thermal-optical physical equation.

[0139] The optical data detection model unit is used to observe the light intensity when current passes through the LED module, obtain the influence relationship between current and light intensity, and construct the electro-optical physical equation.

[0140] The dynamic weighting mechanism construction module includes a policy network unit and a value network unit;

[0141] The policy network unit is used to construct a policy network by taking the state space as input and the action space as output.

[0142] The value network unit is used to construct a value network by taking the state space and action space as inputs and predicting the accumulated reward as output.

[0143] Example: When a certain type of automotive LED headlight module operates at high brightness for an extended period (such as during high-speed driving at night), the temperature rises rapidly due to increased current, and the light intensity gradually decreases due to thermal quenching. By dynamically adjusting the weights of the electrical, thermal, and optical models, the prediction accuracy of the multiphysics coupling model is optimized.

[0144] The initial weights for the electro-optical-thermal multiphysics coupling model were assigned as 0.4, 0.3, and 0.3, respectively.

[0145] Real-time data from automotive LED modules is collected. According to the time series, when time is 0, the current is 2.0, the temperature is 25, and the light intensity is 1000; when time is 1, the current is 2.1, the temperature is 26, and the light intensity is 990; when time is 2, the current is 2.2, the temperature is 28, and the light intensity is 975.

[0146] Calculate the heating rate = 1.0 and the light intensity deviation = 101; and standardize the data.

[0147] The standardized data is input into the policy network, which outputs the weight adjustment action. The value network is then used to calculate the reward and select the optimal weight adjustment action. For example, the optimal weight adjustment action is:

[0148] Input = [1.0, 1.0, -1.0, 1.0, 1.0], Output = Δw e =+0.05,△w t =-0.02,△w p = -0.03; The current prediction error (MSE) is 0.5, the weight variance is 0.01, and the hyperparameters λ1 = 1, λ2 = 0.1: Calculate the reward = R = -1 × 0.5 - 0.1 × 0.01 = -0.501;

[0149] The coupling model is updated with new weights, and the automotive LED module is tested using the optimized coupling model.

[0150] To assess the effectiveness of the optimized model, the actual temperature value was 28°C, the model's predicted value before optimization was 27.5, and the model's predicted value after optimization was 27.8. The results showed that the optimized model significantly improved its performance.

[0151] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A test data analysis method applied to automotive LED modules, characterized in that: The method includes the following steps: S100: Collect historical information on changes in electrical, thermal, and optical data in automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship. The specific steps for generating a detection model based on influence relationships to produce three types of physical data are as follows: S101. Collect electrical, thermal, and optical data from the automotive LED module. The electrical data includes current and voltage data. Analyze the Joule heating generated when current passes through the LED module, finding a correlation between temperature change and the square of the current. Utilize the influence relationship between thermal and electrical data to construct the electro-thermal physical equation: ； In the formula, T represents the LED module temperature, t represents time, I represents the LED module current, k1 represents the electrothermal conversion coefficient, k2 represents the heat dissipation coefficient, and T amb Indicates ambient temperature; When the temperature in the LED module changes, the change in light intensity is observed to obtain the influence relationship between thermal and optical data, and the thermal-optical physical equation is constructed as follows: ； In the formula, L represents the light intensity after the temperature change, L0 represents the light intensity before the temperature change, α represents the temperature attenuation coefficient, T0 represents the initial temperature, and T represents the LED module temperature. When current flows through the LED module, the light intensity is observed to obtain the relationship between current and light intensity, and the electro-optical physical equation is constructed as follows: ； In the formula, β represents the photoelectric efficiency, and r represents the temperature modulation coefficient; S102. Based on the physical equations of the three constructed influence relationships, parameter fitting is performed respectively to obtain the electrical detection model as follows: The thermal detection model is The light detection model is ; S200. Set weights for the detection equations of the three physical data, and couple the detection models of the three physical data according to the weights to obtain the electro-optic-thermal multiphysics coupling model. S300: Uses sensors to collect real-time data on current, temperature and light intensity in automotive LED modules, calculates the real-time heating rate and light intensity deviation in automotive LED modules, and standardizes the collected data. S400: Construct a state space using real-time collected and computed data, and use the weights of the three detection models as the action space; set a reward function; The specific steps for setting the reward function are as follows: S401. Set up the state space, storing all standardized data in the state space, denoted as st = {I(t), T(t), L(t), rT(t), ΔL(t)}; where I(t) represents the collected current data of the automotive LED module; set up the action space, which represents the adjustment actions for the weights of the electro-optical-thermal multiphysics coupling model, denoted as at = {Δw e , △w t , △w p }; Set restrictions: w e '+w t '+w p =1, where w e '、w t 'and w p 'These represent the weights of the three detection models after adjustment; S402. When testing automotive LED module data using an electro-optical-thermal multiphysics coupling model, different weight combinations are output in the action space. For each weight combination, the joint mean square error of the electro-optical-thermal multiphysics coupling model and the weight fluctuation variance are calculated. The reward function is designed using the model's joint mean square error and weight fluctuation variance: ； In the formula, R represents the reward result, λ1 and λ2 represent the set hyperparameters, MSE represents the joint mean square error of the model, and Var(w) represents the weight variance of the model. S500: Using the state space as input and the action space as output, construct a policy network and use a neural network to construct structural layers; similarly, using the state space and action space as input and predicting accumulated rewards as output, construct a value network; use the policy network and value network to construct a dynamic weight mechanism. The specific steps for constructing a dynamic weight mechanism using policy networks and value networks are as follows: S501. Construct a policy network using the state space as input and the action space as output; construct a structural layer using a neural network; similarly, construct a value network using the state space and action space as input and the predicted cumulative reward as output. S502. The policy network is trained using the data in the state space and weight changes in the historical automotive LED module to obtain the mapping relationship between different state spaces and weight changes as at=H(st); When the automotive LED module is tested in real time, the data in the real-time state space is input into the policy network, and the weight adjustment action is output. The value network takes the real-time state space and weight adjustment actions (st, at) as inputs, calculates the accumulated reward according to the reward function, and outputs the result. It compares the outputs of all rewards and selects the input corresponding to the maximum reward as the optimal combination (st, at). The dynamic weight mechanism is constructed by combining the policy network and the value network. S600 uses a dynamic weighting mechanism to obtain the real-time weights of three physical models, uses the real-time weights to reconstruct the three detection models, optimizes the electro-optical-thermal multi-physics coupling model, and uses the optimized electro-optical-thermal multi-physics coupling model to test automotive LED module data.

2. The test data analysis method for automotive LED modules according to claim 1, characterized in that: The specific steps in S200 to couple the detection models of the three physical data according to the weights to obtain the electro-optic-thermal multiphysics coupling model are as follows: S201. Set coupling weights for the detection models of the three types of physical data respectively. Let the weight of the electrical detection model be w. e The weights of the thermal detection model are w. t The weights of the optical detection model are w p ; By using the set weights to weighted couple the detection models of the three physical data types, the electro-optical-thermal multiphysics coupling model is obtained as follows: ； In the formula, Cou(I,T,L) represents the output of the electro-optical-thermal multiphysics coupling model.

3. The test data analysis method for automotive LED modules according to claim 1, characterized in that: The specific steps for calculating the real-time heating rate and light intensity deviation in the automotive LED module in S300 are as follows: S301. Utilize sensors to collect real-time data on current, temperature, and light intensity in the automotive LED module, and calculate the real-time heating rate of the automotive LED module using the following formula: ； In the formula, rT(t) represents the real-time heating rate of the automotive LED module, T(t) represents the temperature of the automotive LED module at time t, and Δt represents the heating time. The formula for calculating the real-time light intensity deviation of automotive LED modules is: In the formula, ΔL(t) represents the real-time light intensity deviation of the automotive LED module, and L... ref The nominal value of the luminous intensity of the automotive LED module is represented; L(t) represents the collected luminous intensity of the automotive LED module. S302. Use the Z-score standardization formula to standardize the calculated and collected data of different dimensions and transform them into data of a unified dimension.

4. The test data analysis method for automotive LED modules according to claim 1, characterized in that: The specific steps for testing automotive LED module data using the optimized electro-optic-thermal multiphysics coupling model in S600 are as follows: S601. When testing automotive LED modules, a dynamic weighting mechanism is used to calculate real-time weight adjustments. The weights of the electro-optical-thermal multiphysics coupling model are adjusted in real-time based on the time series. The weight adjustment formula for each physical detection model is as follows: ； In the formula, w i (t) represents the weight of the i-th physics detection model at time t, w i (t-1) represents the weight of the i-th physical detection model at time t-1, Δw i This represents the weight adjustment amount for the i-th physical detection model; i belongs to {e, t, p}; The automotive LED module was tested using a weighted electro-optical-thermal multiphysics coupling model.

5. A test data analysis system for automotive LED modules, employing the test data analysis method for automotive LED modules as described in any one of claims 1-4, characterized in that: The test data analysis system includes a detection model construction module, a multiphysics coupling module, a data acquisition module, a spatial setting module, a dynamic weighting mechanism construction module, and a real-time weight adjustment module. The detection model construction module is used to collect information on changes in electrical, thermal, and optical data in historical automotive LED modules, analyze the influence relationship between the three types of physical data, and generate a detection model for the three types of physical data based on the influence relationship. The multiphysics coupling module is used to set weights for the detection equations of the three types of physical data, and to couple the detection models of the three types of physical data according to the weights to obtain an electro-optic-thermal multiphysics coupling model. The data acquisition module is used to collect current, temperature and light intensity data in the automotive LED module in real time using sensors, calculate the real-time heating rate and light intensity deviation in the automotive LED module, and standardize the collected data. The space setting module is used to construct a state space using the collected and calculated real-time data, and to use the weights of the three detection models as the action space; and to set the reward function. The dynamic weighting mechanism construction module is used to generate weight adjustment actions using a policy network, judge weight adjustment actions using a value network, and construct a dynamic weighting mechanism using the policy network and the value network. The real-time weight adjustment module is used to adjust the weights of the three physical detection models in the electro-optic-thermal multiphysics coupling model by using the weight adjustment action output by the dynamic weight mechanism.

6. The test data analysis system for automotive LED modules according to claim 5, characterized in that: The detection model construction module includes an electrical data detection model unit, a thermal data detection model unit, and an optical data detection model unit. The electrical data detection model unit is used to analyze the Joule heat generated when current passes through the LED module, and obtain the correlation between temperature change and the square of current. The electro-thermal physical equation is constructed by utilizing the influence relationship between thermal data and electrical data. The thermal data detection model unit is used to observe the change in light intensity when the temperature changes in the LED module, obtain the influence relationship between thermal data and light data, and construct the thermal-optical physical equation. The optical data detection model unit is used to observe the light intensity when current passes through the LED module, obtain the influence relationship between current and light intensity, and construct the electro-optical physical equation.

7. The test data analysis system for automotive LED modules according to claim 5, characterized in that: The dynamic weighting mechanism construction module includes a policy network unit and a value network unit; The policy network unit is used to construct a policy network by taking the state space as input and the action space as output. The value network unit is used to construct a value network by taking the state space and action space as inputs and predicting the accumulated reward as output.

Images