Dynamic testing system for heat dissipation performance of heat sink

Abstract

The application relates to the technical field of radiator monitoring, and particularly discloses a dynamic testing system based on the heat dissipation performance of a radiator, which is characterized in that a heat source simulation body is arranged on the heat transfer interface of the radiator, and a high-fidelity dynamic electric power load is applied to the heat source simulation body to simulate actual working conditions; a microwave detection signal is radiated to the space where the heat source simulation body is located, and a plurality of scattering response signals are received; a three-dimensional dielectric constant dynamic distribution of the heat source simulation body is reconstructed by solving a fusion space-time constraint inverse scattering problem; a three-dimensional temperature field dynamic distribution is generated according to a material dielectric constant-temperature relationship mapping; finally, the migration track and temperature change curve of the highest temperature point in space are intelligently identified and extracted from continuous time-series temperature field data, which are used as direct evaluation bases for the dynamic heat dissipation performance of the radiator; the application realizes nondestructive, full-field and real-time visualization and quantitative analysis of the internal dynamic hot spot migration process of a hidden heat source, and improves the authenticity of dynamic heat testing.

Description

Technical Field

[0001] This invention relates to the field of radiator monitoring technology, and more specifically to a dynamic testing system for radiator heat dissipation performance. Background Technology

[0002] As electronic devices rapidly evolve towards high power density and high-frequency dynamic loads, the operating states of their key internal chips (such as CPUs, GPUs, and power modules) exhibit high transientity and spatial non-uniformity. Chip power consumption can fluctuate dramatically within microseconds to milliseconds. Simultaneously, due to differences in task scheduling and kernel activation, heat dissipation areas (hot spots) are no longer fixed on the chip surface or internally, but migrate rapidly with time and load. This presents unprecedented challenges to the design of thermal management systems: heat sinks not only need sufficient thermal resistance under steady-state conditions, but also must be able to promptly and effectively suppress localized temperature spikes caused by these moving heat sources under dynamic operating conditions, preventing chip failure or frequency throttling due to transient overheating. Therefore, dynamic thermal performance testing of heat sinks to evaluate their performance under real, varying operating conditions has become an indispensable and crucial step in thermal design verification.

[0003] The existing technology lacks a way to achieve real-time, lag-free, high spatial resolution tracking and quantitative characterization of transient hot spots that migrate rapidly in three-dimensional space caused by dynamic loads inside the heat source simulation body below the radiator, under non-contact and without physical probe intrusion, so as to accurately evaluate the ultimate heat dissipation performance of the radiator for dynamically moving heat sources. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic testing system for the heat dissipation performance of radiators to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A dynamic testing system for radiator heat dissipation performance includes:

[0007] The dynamic thermal load simulation module is used to set a heat source simulation body on the heat transfer interface of the heat sink and apply electrical power to the heat source simulation body according to a preset time sequence to simulate the dynamic thermal load under actual working conditions.

[0008] The probe wave signal processing module is used to radiate probe wave signals into the space where the heat source simulator is located while applying dynamic thermal load, and simultaneously receive the multi-channel response wave signals generated after being acted upon by the heat source simulator.

[0009] The dielectric constant reconstruction module reconstructs the three-dimensional dynamic distribution of the dielectric constant of the heat source simulation body under dynamic thermal load by solving the inverse scattering problem based on the difference between the multi-channel response wave signal and a preset reference wave signal.

[0010] The temperature field generation module maps and generates the three-dimensional temperature field dynamic distribution of the heat source simulation body based on the dynamic distribution of the three-dimensional dielectric constant and the known material dielectric constant-temperature correspondence.

[0011] The performance evaluation and analysis module identifies and extracts the location migration trajectory and temperature change curve of the highest temperature point in space from the continuous time-series data of the dynamic distribution of the three-dimensional temperature field, which serves as the basis for evaluating the heat dissipation performance of the radiator on dynamically moving heat sources.

[0012] As a further aspect of the present invention: the simulation of dynamic thermal load under actual working conditions specifically includes:

[0013] Real-time acquisition of electrical parameters of actual electronic devices during operation, and calculation of instantaneous power data stream;

[0014] Dynamic heat transfer characteristics are corrected for instantaneous power data streams to eliminate power response delay caused by the thermal capacitance of electronic device packages and generate corrected power timing.

[0015] Based on the corrected power timing and the resistance-temperature characteristics of the heat source simulator, the voltage control sequence for directly driving the heat source simulator is generated.

[0016] By executing a voltage control sequence through a high-bandwidth power output unit, the heat source simulator generates a spatial dynamic thermal load consistent with the thermal behavior of actual electronic devices.

[0017] As a further aspect of the present invention: the receiving process of the multi-channel response wave signal is as follows:

[0018] Based on the real-time power value of the dynamic heat load and the historical spatial temperature distribution, the estimated location of the high-temperature region on the heat source simulator at the current moment is calculated.

[0019] Based on the estimated location, multiple beamforming units arranged around the target are controlled to adaptively adjust the energy focus of the probe wave signal and direct it towards the spatial region where the estimated location is located for radiation.

[0020] At the same time, multiple beamforming units synchronously receive response wave signals from different scattering paths of the heat source simulation body, and separate and buffer the response wave signals according to the corresponding spatial paths to form a set of multi-path response wave signals associated with spatial orientation.

[0021] As a further aspect of the present invention: the calculation of the estimated location of the high-temperature region generated on the heat source simulator at the current moment specifically includes:

[0022] Based on the real-time power values ​​and the known electrical connection layout applied to the heat source simulator, the spatial pattern of the current power density distribution is identified.

[0023] Extract the migration paths and evolution rates of historical high-temperature regions that have the strongest spatiotemporal correlation with the currently identified spatial patterns from historical spatial temperature distribution;

[0024] The migration path is extrapolated based on the evolution rate, and combined with spatial patterns to generate a dynamic spatial probability distribution map representing the probability of high temperature occurrence.

[0025] From the dynamic spatial probability distribution map, continuous spatial regions with probability values ​​exceeding a preset threshold are extracted, and the geometric center coordinates of the continuous spatial regions are determined as the estimated locations.

[0026] As a further aspect of the present invention: the reconstructing of the three-dimensional dynamic distribution of dielectric constant of the heat source simulator under dynamic thermal load specifically includes:

[0027] The complex domain difference between the multi-channel response wave signal and the preset reference wave signal is performed to obtain a complex signal difference set that contains only the dielectric disturbance information caused by dynamic thermal load.

[0028] Based on the known spatial path and incident angle of each signal in the complex signal difference set, a set of nonlinear equations is constructed with the dielectric constant perturbation of the discrete voxels of the entire heat source simulation volume as the unknown.

[0029] The physical structure and geometric contour of the heat source simulation body are introduced as spatial constraints, and the three-dimensional dielectric constant distribution obtained from the previous time step is used as the initial value of the temporal constraints. The solution of the nonlinear equation system is transformed into an iterative optimization process with temporal and spatial constraints.

[0030] By minimizing the residual between the calculated and measured values ​​of the complex signal difference set and satisfying the spatiotemporal constraints, the dielectric constant perturbation of each spatial voxel in the heat source simulation body at the current moment is iteratively solved, and then the three-dimensional dynamic distribution of dielectric constant is synthesized.

[0031] As a further aspect of the present invention: the process of constructing the nonlinear equation system is as follows:

[0032] For each signal in the complex signal difference set, based on the spatial path and incident angle of each signal, the additional phase shift and amplitude attenuation accumulated when the wavefront of the probe signal passes through each spatial voxel are calculated, and the initial contribution coefficients of all voxels under the corresponding signal are generated.

[0033] Based on the initial contribution coefficients and incorporating the estimated influence of multiple internal scattering paths caused by the geometry of the heat source simulation body, the contribution coefficients of each spatial voxel are iteratively corrected to obtain the final spatial weight coefficients.

[0034] The difference of each complex signal is represented as the superposition of the perturbation of the dielectric constant of all spatial voxels and the corresponding final spatial weight coefficient, and each signal is established as a nonlinear equation.

[0035] The nonlinear equations established by collecting all signals in the set of complex signal differences constitute a system of nonlinear equations.

[0036] As a further aspect of the present invention: the dynamic distribution of the three-dimensional temperature field of the mapped heat source simulation body specifically includes:

[0037] For each spatial voxel in the three-dimensional dynamic distribution of dielectric constant, based on the material dielectric constant-temperature correspondence, the preliminary temperature value corresponding to the dielectric constant of the spatial voxel is calculated independently to form a preliminary three-dimensional temperature distribution;

[0038] Based on the known physical structure of the heat source simulation body, the heat conduction path of the material between adjacent spatial voxels is identified, and the spatial continuity of the preliminary three-dimensional temperature distribution is corrected by the constraint of the heat conduction path to obtain the intermediate three-dimensional temperature distribution.

[0039] The intermediate three-dimensional temperature distribution is compared with the temperature change trend predicted based on historical time series data, and the temperature values ​​that exceed the prediction range are optimized for time series smoothing based on the physical heat transfer law.

[0040] By integrating the results of spatial continuity correction and temporal smoothing optimization, a final three-dimensional temperature field dynamic distribution that is continuous in space and time is synthesized.

[0041] As a further aspect of the present invention: the identification and extraction of the location migration trajectory and temperature change curve of the highest temperature point in space specifically includes:

[0042] For the dynamic distribution of the three-dimensional temperature field at each moment, identify all spatial regions that satisfy the local spatial temperature gradient exceeding the preset threshold, and take the geometric center of the spatial region as the candidate point of highest spatial temperature.

[0043] Based on the historically determined location migration trajectory, the spatial coherence between the candidate spatial highest temperature point and the trajectory in time is calculated, and candidate points that can maintain a smooth trajectory movement direction are selected first.

[0044] By combining the real-time changes of dynamic heat load, the physical rationality of temperature changes at the selected candidate points is verified, and outliers caused by instantaneous measurement disturbances are eliminated.

[0045] The highest spatial temperature point and its corresponding temperature value at each confirmed moment are connected in chronological order to generate the location migration trajectory and temperature change curve, respectively.

[0046] As a further aspect of the present invention: the calculation of the spatial coherence of the candidate spatial highest temperature point and the trajectory in terms of time sequence specifically includes:

[0047] Based on the spatial coordinates of the most recent points on the historical migration trajectory, the estimated direction and speed of movement of the highest temperature point at the current moment are calculated.

[0048] Starting from the end point of the historical trajectory, a spatial sector prediction area is calculated along the predicted direction of movement and based on the movement rate.

[0049] Calculate the vertical distance from the highest temperature point in each candidate space to the center line of the spatial sector prediction area, and use the vertical distance as a quantitative indicator of spatial coherence.

[0050] The candidate point with the lowest quantification index value is selected first as the final choice to maintain a smooth trajectory at the current moment.

[0051] The beneficial effects of this invention are:

[0052] (1) This invention, by introducing non-contact microwave detection and tomographic imaging principles, achieves non-destructive, full-field, real-time inversion of the dynamic distribution of three-dimensional dielectric constant inside a heat source simulation body, and further maps it into a three-dimensional temperature field with high spatiotemporal resolution. This technical approach breaks free from the spatial anchoring and thermal inertia constraints of fixed sensors, and can track and record the entire process of the generation, evolution, and spatial drift of dynamic hotspots without lag. The resulting "spatiotemporal four-dimensional thermal information field" provides a real and complete data foundation for accurately evaluating the heat dissipation capacity of the radiator under the worst operating conditions (i.e., following the moving hotspot), avoiding misjudgment of thermal failure risk due to measurement blind spots.

[0053] (2) This invention applies a high-fidelity dynamic power load and an adaptively oriented probe wave signal simultaneously, and deeply integrates multiple physical constraints such as heat source structure, material properties, and historical evolution in signal processing. This enables the entire testing process to highly reproduce the thermal behavior characteristics of complex electronic devices (such as GPUs and high-performance processors) under real workloads. The generated "spatial maximum temperature point migration trajectory" and "temperature change curve" intuitively and quantitatively reveal the comprehensive performance shortcomings of the heat sink when dealing with spatially non-uniform and temporally unsteady thermal loads, such as insufficient response to hot spot movement in a specific direction or limited ability to suppress sudden power spikes. This allows the design optimization and reliability verification of the heat sink to leap from the traditional "point" or "line" evaluation based on steady-state or simple cyclic loads to the evaluation based on the "field" and "behavior" of real complex scenarios, enhancing the correlation between test results and actual product application conditions, and providing refined analysis tools and decision-making basis for the precise thermal design of high power density electronic devices. Attached Figure Description

[0054] The invention will now be further described with reference to the accompanying drawings.

[0055] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0056] 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.

[0057] Please see Figure 1 As shown, the present invention is a dynamic testing system for the heat dissipation performance of a radiator, comprising:

[0058] The dynamic thermal load simulation module is used to set a heat source simulation body on the heat transfer interface of the heat sink and apply electrical power to the heat source simulation body according to a preset time sequence to simulate the dynamic thermal load under actual working conditions.

[0059] The probe wave signal processing module is used to radiate probe wave signals into the space where the heat source simulator is located while applying dynamic thermal load, and simultaneously receive the multi-channel response wave signals generated after being acted upon by the heat source simulator.

[0060] The dielectric constant reconstruction module reconstructs the three-dimensional dynamic distribution of the dielectric constant of the heat source simulation body under dynamic thermal load by solving the inverse scattering problem based on the difference between the multi-channel response wave signal and a preset reference wave signal.

[0061] The temperature field generation module maps and generates the three-dimensional temperature field dynamic distribution of the heat source simulation body based on the dynamic distribution of the three-dimensional dielectric constant and the known material dielectric constant-temperature correspondence.

[0062] The performance evaluation and analysis module identifies and extracts the location migration trajectory and temperature change curve of the highest temperature point in space from the continuous time-series data of the dynamic distribution of the three-dimensional temperature field, which serves as the basis for evaluating the heat dissipation performance of the radiator on dynamically moving heat sources.

[0063] In the dynamic thermal load simulation module, voltage and current sensors connected to the power supply circuit of the actual electronic device are used to synchronously acquire the instantaneous voltage and current values ​​of the device during operation. The voltage sensor is connected in parallel across the load, and the current sensor is connected in series in the power supply line. The analog signal outputs of both sensors are connected to the same multi-channel synchronous data acquisition card to ensure strict alignment of the voltage and current signal acquisition times. The sampling rate of the data acquisition card is set to no less than 10 times the fundamental frequency of the operating current of the electronic device under test. The acquisition card converts the synchronously acquired analog voltage and current signals into digital signals. For each set of synchronously acquired instantaneous digital voltage and current values, they are multiplied, and the result is the instantaneous power value at that sampling moment. The instantaneous power values ​​calculated from all sampling moments are arranged in chronological order into an ordered data sequence, which is the instantaneous power data stream.

[0064] Dynamic heat transfer characteristics are corrected on the instantaneous power data stream to eliminate the power response delay caused by the thermal capacity of the electronic device package, resulting in a corrected power timing sequence. The core of this correction lies in eliminating the hysteresis effect of the internal thermal inertia of the device's structure on the surface heat flow response. First, the surface temperature rise response curve of the actual electronic device package structure under a known step power input is obtained through experimental calibration. Based on this curve, a dynamic transfer relationship characterizing the process from heat generation in the chip junction region to the formation of equivalent heat flow on the package surface is established. Then, using this dynamic transfer relationship as a benchmark, a deconvolution operation is performed on the instantaneous power data stream. The purpose of this operation is to inversely calculate the original, unfiltered power input sequence that the chip junction region should have if the heat flow on the package surface is to strictly follow the changes in the instantaneous power data stream. The sequence obtained from this calculation is the corrected power timing sequence, which more directly corresponds to the heat flow change ultimately transferred to the heat sink.

[0065] Based on the corrected power timing sequence and the resistance-temperature characteristics of the heat source simulator, a voltage control sequence for directly driving the heat source simulator is generated. The resistance-temperature characteristics of the heat source simulator refer to the relationship between its resistance value and its body temperature, which has been pre-calibrated. For each target power value in the corrected power timing sequence, the real-time resistance value at that temperature is calculated using the resistance-temperature characteristics, based on the real-time or estimated temperature of the heat source simulator obtained at the current moment through independent temperature measurement (such as infrared thermometry). Subsequently, according to the fundamental relationship that electrical power equals the square of voltage divided by resistance, the precise voltage value to be applied across the heat source simulator is derived from the target power value and the real-time resistance value. This calculation is performed on all target power values ​​in chronological order to obtain the voltage control sequence.

[0066] The voltage control sequence is executed by a high-bandwidth power output unit, causing the heat source simulator to generate a spatial dynamic thermal load consistent with the thermal behavior of the actual electronic device. The high-bandwidth power output unit is a programmable power supply capable of rapidly and accurately following voltage command changes. The output of this unit is connected to the heat source simulator, and the corresponding voltage is output strictly according to the voltage values ​​and time points specified in the voltage control sequence. Since the voltage control sequence comprehensively considers the target heat flux and the dynamic electrothermal characteristics of the heat source simulator itself, the instantaneous heating power generated by the heat source simulator and the resulting change in surface heat flux density after applying this voltage sequence can highly reproduce the dynamic thermal load characteristics transferred from the surface of the actual electronic device package to the heat sink in both time and space.

[0067] In the probe wave signal processing module, the estimated location of the high-temperature region on the heat source simulator at the current moment is calculated. This calculation is based on the real-time power value of the dynamic heat load and the historical spatial temperature distribution. The first step is to identify the spatial pattern of the current power density distribution. Specifically, based on the real-time power value and the pre-calibrated electrical connection layout diagram on the heat source simulator, the instantaneous current distribution ratio of each current path is determined. Combining this with the known geometric region corresponding to each path on or inside the heat source simulator, the total power is distributed according to this ratio and the region area, thus obtaining a spatial pattern diagram describing how the power is distributed in space. The second step is to extract relevant historical high-temperature region migration information. From the stored historical spatial temperature distribution data, the temperature distribution maps recorded at all historical moments with the highest similarity to the currently identified spatial pattern are selected. "Highest similarity" means that the power density ratios at corresponding locations in the spatial pattern maps of the two moments are most closely related. From these selected historical temperature distribution maps, the highest recorded spatial temperature points are extracted, and these points are connected chronologically to form one or more historical high-temperature region migration paths. Simultaneously, the average spatial displacement rate of the temperature points along each path is calculated as its evolution rate. The third step is to generate a dynamic spatial probability distribution map. The historical migration path with the highest similarity to the current spatial pattern is selected, and its last recorded temperature point is used as the starting point. Following the direction and speed indicated by the calculated evolution rate, the path is extended into the future for a predetermined time length to form a predicted path. Then, using this predicted path as the center line, a decay rule is set where the probability decreases with increasing distance from the center line. For example, the probability decreases linearly or exponentially with increasing distance, thereby generating a dynamic spatial probability distribution map covering a portion of the heat source simulation body and characterizing the probability of high-temperature occurrence. The fourth step is to determine the estimated location. In the dynamic spatial probability distribution map, all pixels with a probability value greater than a pre-set fixed value (e.g., 70% of the maximum probability value) are marked. The spatial three-dimensional coordinates of the geometric center point of the circumscribed polygon of the region formed by these consecutive pixels are calculated, and these coordinates are determined as the estimated position.

[0068] Based on the estimated location, the radiation direction of the probe wave signal is adaptively adjusted. This is achieved by controlling multiple beamforming units arranged around the heat source simulator. Each beamforming unit includes an antenna and a corresponding phase and amplitude controller. By adjusting the relative phase and amplitude of the signals emitted by each unit, the probe wave signals emitted by all units interfere in space, and the focal region with the strongest energy after interference is controlled at the spatial coordinates of the estimated location. This process is accomplished by solving a set of equations with the preset focal position as the objective and the phase and amplitude of each unit as variables, ensuring that the energy focal point is accurately guided to the estimated region, thereby enhancing the probe signal strength in that region.

[0069] Simultaneously, response wave signals are received and processed. While radiating the probe wave signal, the multiple beamforming units also function as receiving units. They synchronously receive the response wave signals scattered back from the heat source simulator. Due to the different spatial positions of each unit, the signals they receive correspond to different scattering paths of the heat source simulator. Subsequently, based on the known spatial coordinates of each beamforming unit, the emission parameters of the probe wave signal, and the preset electromagnetic wave propagation principle, the response component belonging to its own spatial path is separated from the mixed signal received by each unit. This separation process can be achieved using spatial filtering algorithms in digital signal processing. All separated response wave signals corresponding to specific spatial orientations are buffered separately, ultimately forming a set of multiple response wave signals clearly associated with different spatial orientations.

[0070] In the dielectric constant reconstruction module, firstly, the signal difference containing only dielectric disturbance information caused by dynamic thermal load is acquired. Each of the currently received multi-channel response wave signals is then subjected to a complex-domain difference operation with its corresponding signal from the preset reference wave signal. Specifically, for each signal, the real part of the measured signal in its complex form is subtracted from the real part of the reference signal, and the imaginary part is subtracted from the imaginary part of the reference signal, resulting in a new complex number. This new complex number characterizes the amplitude and phase changes of that signal caused by the change in the temperature (i.e., dielectric constant) distribution inside the heat source simulator. After performing this operation on all signals, the set of complex signal differences is obtained.

[0071] Secondly, a set of nonlinear equations is constructed with the dielectric constant perturbation of discrete voxels in the entire heat source simulation space as unknowns. First, the heat source simulation and its adjacent space are divided into numerous tiny cubic units in three dimensions, each called a voxel. For each signal in the complex signal difference set, the wavefront propagation path of the probe wave signal is calculated based on the known spatial positions (i.e., known spatial path length and incident angle) of the corresponding transmitting and receiving units. For each voxel, the geometric length of the path passing through that voxel is calculated. Then, assuming a unit-intensity dielectric constant perturbation exists in the voxel, the additional phase shift and amplitude attenuation caused by this perturbation are calculated according to the physical laws of electromagnetic wave propagation in a medium. This calculated value is the initial contribution coefficient of the voxel to this signal. Next, considering that electromagnetic waves undergo multiple reflections and scatterings inside a non-uniform object, an estimate of the influence of multiple internal scattering paths is introduced. The method is as follows: Using the initial contribution coefficient distribution calculated in the previous step as the first estimate, the total signal change accumulated after the wavefront passes through all voxels under this distribution is simulated and compared with a simplified first-order scattering approximation. The difference is used to correct the initial contribution coefficient of each voxel. This correction process can be repeated several times (e.g., 3 to 5 times), each time using the previously corrected coefficient distribution to simulate the total signal change and update the correction amount, ultimately obtaining a more accurate spatial weighting coefficient for each voxel in this signal path. Then, the measured value of the current complex signal difference is expressed as the sum of the products of the unknown dielectric constant perturbation of all voxels and their corresponding final spatial weighting coefficients, thus establishing an equation for the perturbation of all voxels. Finally, these equations established for all signals in the complex signal difference set are compiled together to form a set of nonlinear equations equal in number to the number of signal paths.

[0072] Furthermore, the solution of the nonlinear equations is transformed into an iterative optimization process with spatiotemporal constraints. The spatial constraint refers to the fact that the physical geometric contours of the heat source simulation are known; therefore, during optimization, the dielectric constant perturbation of voxels located outside these contours is forced to zero. The temporal constraint refers to using the three-dimensional dielectric constant distribution solved in the previous time step as the initial guess for the current solution. Based on this, the optimization objective is defined as minimizing the overall difference between the voxel perturbation distribution guessed using the current iteration step, the signal differences calculated according to the relationships in the nonlinear equations (i.e., calculated values), and the measured set of complex signal differences (i.e., measured values). This overall difference is typically defined as the sum of the squares of the differences between the calculated and measured values ​​of all signals, called the residual.

[0073] Finally, the final distribution is solved iteratively and synthesized. Starting from an initial guess (i.e., an initial value with the spatiotemporal constraints applied), a nonlinear optimization algorithm (e.g., the Levenberg-Marquardt algorithm) is used for iterative calculation. In each iteration, the algorithm calculates the residual and its gradient based on the current guessed perturbation distribution and determines how to adjust the guessed perturbation value of each voxel to reduce the residual. The iteration continues until the residual is less than a preset minimum positive number, or the number of iterations reaches a preset upper limit. The dielectric constant perturbation solution for each spatial voxel obtained at this time is the result at the current moment. Arranging the solutions of all voxels according to their spatial positions synthesizes the three-dimensional dynamic distribution of the dielectric constant of the heat source simulation body under dynamic thermal load. This distribution directly reflects the dynamic changes in material properties in space caused by temperature changes.

[0074] In the temperature field generation module, a preliminary three-dimensional temperature distribution is calculated. For each spatial voxel in the dynamic three-dimensional dielectric constant distribution, its current dielectric constant value is read. Subsequently, in the known material dielectric constant-temperature correspondence, the temperature value corresponding to the data point closest to the dielectric constant value is found; if a fitting formula is stored, the dielectric constant value is substituted into the formula to calculate the temperature value. This temperature value is the independent preliminary temperature value of that spatial voxel. After performing this operation on all voxels, a preliminary three-dimensional temperature distribution map composed of the preliminary temperature values ​​of all voxels is formed.

[0075] Spatial continuity correction is performed to obtain an intermediate three-dimensional temperature distribution. Based on the known physical structure of the heat source simulation body (such as the boundary positions of each material layer), adjacent spatial voxel pairs belonging to the same continuous solid material are identified. For each such pair of adjacent spatial voxels, according to the physical laws of solid heat conduction, its temperature gradient should not exceed the reasonable range that the material can support under local heat flux density. The temperature difference of all adjacent voxel pairs in the preliminary three-dimensional temperature distribution is checked. If the absolute value of a certain temperature difference exceeds a preset threshold estimated based on the material's thermal conductivity and voxel size, the temperature value of that pair of voxels is adjusted so that the adjusted temperature difference tends to the threshold. This adjustment process is iterated several times on all adjacent spatial voxel pairs until all temperature differences meet the constraints, thus obtaining an intermediate three-dimensional temperature distribution that is more spatially consistent with the laws of heat conduction.

[0076] Perform temporal smoothing optimization. Continuously read the intermediate 3D temperature distribution data from several previous historical time points (e.g., the last five time points). For each voxel, calculate a short-term temperature trend based on its temperature change sequence within this historical period; for example, calculate the average slope of its temperature change over time. Based on this trend, predict a reasonable temperature range for the voxel at the current time. Compare the current intermediate temperature value of the voxel with the predicted range. If it exceeds the upper or lower limit of the predicted range, adjust it towards that range. The adjustment magnitude is determined based on the inertial characteristics of temperature change in physical heat transfer laws, ensuring that the temperature change within a single time step remains within a reasonable range. This process is performed on all voxels, completing the temporal smoothing.

[0077] The final three-dimensional temperature field dynamic distribution is synthesized. The intermediate three-dimensional temperature distribution, after spatial continuity correction, is fused with the adjusted result after temporal smoothing optimization. Specifically, for each voxel, a weighted average of its value in the intermediate three-dimensional temperature distribution and the suggested value after temporal smoothing optimization is taken, where the weights are allocated according to the material thermal diffusion rate of the region where the voxel is located. The final result is a three-dimensional temperature field dynamic distribution that satisfies both spatial continuity constraints and temporal evolution rationality, providing full spatial coverage and temporal continuity.

[0078] In the performance evaluation and analysis module, for the dynamic distribution of the three-dimensional temperature field at each independent moment, candidate spatial maximum temperature points are identified. Specifically, this includes: calculating the spatial gradient field of the temperature distribution at that moment. For each voxel in the distribution, the temperature difference between it and all directly adjacent voxels (e.g., in the six directions of front, back, left, right, up, and down) is calculated. The maximum absolute value of the temperature difference in each direction is defined as the local spatial temperature gradient value of that voxel. The preset threshold is determined based on the analysis of the thermal conductivity of the heat source simulation material and statistical analysis of historical test data. For example, it can be set to 60% of the theoretical maximum temperature gradient that the heat source simulation may produce under the known maximum steady-state power density. All voxels are traversed, and all voxels whose local spatial temperature gradient values ​​exceed the preset threshold are selected. These voxels may form several continuous regions in three-dimensional space. For each such continuous region, the arithmetic mean of the spatial coordinates of all voxels contained therein is calculated, and the spatial point corresponding to this mean is taken as a candidate spatial maximum temperature point representing that region. Multiple candidate points may be generated at one moment.

[0079] The spatial coherence between candidate points and historical trajectories in time is calculated, and a priority selection is made. Specifically, this includes selecting the candidate point most likely to continue the historical hotspot movement trend from multiple candidate points, ensuring the physical smoothness of the trajectory. First, based on the final determined location migration trajectory at a historical moment, the spatial coordinate sequence of the most recent N (e.g., N=5) confirmed points is extracted. The estimated movement direction and speed are calculated using the following method: Let the spatial coordinate vectors of these N historical points be: ;in Let be the point at the most recent moment. Calculate the displacement vector between adjacent points: ;in, Then the estimated unit vector of the current movement direction. The direction of this can be obtained by calculating the weighted average of these displacement vectors, with closer displacements assigned higher weights. The estimated movement rate SS can be taken as the distance between the two most recent historical points divided by the time interval. Next, a spatial sector-shaped prediction region is constructed, using the endpoint of the historical trajectory as the starting point. As the vertex, to predict the direction of movement. With the central axis as the axis, open a vertex angle towards the future. A spatial sector. The radius of the sector. Based on the estimated movement rate Multiply by a preset time window length Determined, that is: ;angle It can adaptively adjust based on the rate of change of historical trajectory direction; when the rate of change is large... Increase, for example: ;in It is a basic angle (such as 30 degrees). It is a proportionality coefficient. It is the standard deviation of the direction angle of the historical displacement vector. Then, for each candidate spatial highest temperature point at the current moment... Calculate its distance to the central axis of the sector (i.e., through the vertex). And the direction is perpendicular distance of the straight line .distance The calculation formula involves vector operations: point To the top The vector is denoted as: Then the perpendicular distance from that point to the central axis is... For vectors Direction vector of the central axis The modulus of the cross product is: ;because Since it is a unit vector, the formula simplifies to: This vertical distance This serves as a quantitative indicator of the coherence between the candidate point and the historical trajectory space; a smaller distance signifies better coherence. Furthermore, it is necessary to check whether the candidate point lies within a sector region (i.e., vector region). and The dot product is positive and less than And the included angle is less than Prioritize selection that simultaneously satisfies the condition of being located within the sector area and The candidate point with the smallest value is selected as the final choice to maintain a smooth trajectory at the current moment. If no point is located within the sector, then only the candidate point with the smallest value is selected. Find the point with the smallest value and record this.

[0080] By combining real-time changes in dynamic heat load, the physical rationality of temperature changes at selected candidate points is verified. Specifically, this includes: obtaining real-time power values ​​applied to the heat source simulator at the current moment and several previous moments; estimating the reasonable range of temperature change at the selected candidate point within the observed time interval based on the material heat capacity characteristics of the local region and the thermal resistance characteristics between the candidate point and the heat sink; for example, a simplified estimation method is to assume that the temperature change at the point is mainly determined by the net heat flow absorbed by its local region, where the net heat flow equals the local heating power minus the power lost through thermal resistance; the local heating power can be inferred from the total power and the current power density spatial pattern; calculating the theoretical temperature change range at the point since the previous moment; and comparing the actual temperature change value read from the temperature field data with the theoretical range. If the actual change value significantly exceeds the theoretical range (e.g., more than twice the range width), the point is considered to be abnormal due to instantaneous errors or noise in the three-dimensional temperature field reconstruction process and is therefore discarded. If a rejection occurs, the process returns to step two, selecting a candidate point with the second-best spatial coherence quantification index and repeating this verification step until a point with a temperature change within a physically reasonable range is found, or all candidate points are eliminated. If all candidate points are eliminated, the current moment is marked as "hotspot uncertainty," and no trajectory point is generated at this moment; its temperature value can be obtained by interpolating data from previous and subsequent moments.

[0081] The final location migration trajectory and temperature change curve are generated by connecting the three-dimensional spatial coordinates of the highest temperature point at each valid moment, as confirmed in the previous steps, in chronological order to form a spatial polygonal line (or a smoothed curve), which is the location migration trajectory. Simultaneously, the temperature values ​​corresponding to these points are extracted in the same chronological order and plotted as a temperature-time curve, which is the temperature change curve. These two curves together constitute the core basis for evaluating radiator performance: the location migration trajectory reveals the spatial movement pattern and range of the heat source that the radiator needs to handle; the temperature change curve directly reflects the radiator's ability to suppress the temperature of dynamically moving heat sources within a safe range, and its peak value, fluctuation amplitude, and stabilization rate are key evaluation indicators.

[0082] The working principle of this invention is as follows: A high-bandwidth power output unit applies electrical power, varying according to a preset time sequence, to a heat source simulator positioned on the heat transfer interface of a radiator to simulate the dynamic heat load transferred across the surface of an actual electronic device package. Simultaneously, a beamforming unit surrounds the heat source simulator and adaptively radiates a probe wave signal into the space containing the heat source simulator, while simultaneously receiving multiple response wave signals scattered by the simulator. The multiple response wave signals are then compared with a preset reference wave signal in the complex domain to obtain a signal difference set. By introducing the geometric contour of the heat source simulator and the historical dielectric constant distribution as constraints, and solving the inverse scattering problem, the heat source simulator is reconstructed. The source simulates the three-dimensional dynamic distribution of dielectric constant under dynamic load. Based on the known material dielectric constant-temperature relationship, this distribution is mapped to a preliminary temperature field voxel by voxel. Then, combined with the spatial continuity constraint of solid heat conduction and the temporal smoothing optimization based on historical data, a three-dimensional temperature field dynamic distribution with full spatial and temporal continuity is generated. Finally, from the continuous temporal data of this temperature field, candidate high-temperature points are screened by local gradient threshold. Combined with the spatial coherence analysis of historical trajectories and real-time power verification, the position migration trajectory and temperature change curve of the highest temperature point in space are identified and extracted, which serve as a direct basis for evaluating the heat dissipation performance of the radiator on the dynamic moving heat source.

[0083] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A dynamic testing system for radiator heat dissipation performance, characterized in that, include: The dynamic thermal load simulation module is used to set a heat source simulation body on the heat transfer interface of the heat sink and apply electrical power to the heat source simulation body according to a preset time sequence to simulate the dynamic thermal load under actual working conditions. The probe wave signal processing module is used to radiate probe wave signals into the space where the heat source simulator is located while applying dynamic thermal load, and simultaneously receive the multi-channel response wave signals generated after being acted upon by the heat source simulator. The dielectric constant reconstruction module reconstructs the three-dimensional dynamic distribution of the dielectric constant of the heat source simulation body under dynamic thermal load by solving the inverse scattering problem based on the difference between the multi-channel response wave signal and a preset reference wave signal. The temperature field generation module maps and generates the three-dimensional temperature field dynamic distribution of the heat source simulation body based on the dynamic distribution of the three-dimensional dielectric constant and the known material dielectric constant-temperature correspondence. The performance evaluation and analysis module identifies and extracts the location migration trajectory and temperature change curve of the highest temperature point in space from the continuous time-series data of the dynamic distribution of the three-dimensional temperature field, which serves as the basis for evaluating the heat dissipation performance of the radiator on the dynamic moving heat source. The identification and extraction of the location migration trajectory and temperature change curve of the highest temperature point in space specifically includes: For the dynamic distribution of the three-dimensional temperature field at each moment, identify all spatial regions that satisfy the local spatial temperature gradient exceeding the preset threshold, and take the geometric center of the spatial region as the candidate point of highest spatial temperature. Based on the historically determined location migration trajectory, the spatial coherence between the candidate spatial highest temperature point and the trajectory in time is calculated, and candidate points that can maintain the smooth movement direction of the trajectory are selected. By combining the real-time changes of dynamic heat load, the physical rationality of temperature changes at the selected candidate points is verified, and outliers caused by instantaneous measurement disturbances are eliminated. Connect the highest spatial temperature point and its corresponding temperature value at each confirmed moment in chronological order to generate the location migration trajectory and temperature change curve, respectively. The calculation of the spatial coherence of the candidate spatial highest temperature point and the trajectory in terms of time sequence specifically includes: Based on the spatial coordinates of the most recent points on the historical migration trajectory, the estimated direction and speed of movement of the highest temperature point at the current moment are calculated. At the end of the historical trajectory As the vertex, to predict the direction of movement. With the central axis as the axis, open a vertex angle towards the future. A spatial sector; the radius of the sector. Based on the estimated movement rate Multiply by the preset time window length Determined, that is: ; For each candidate space highest temperature point at the current moment Calculate its perpendicular distance to the central axis of the sector. ;distance The calculation formula involves vector operations: point To the top The vector is denoted as: Then the perpendicular distance from the corresponding point to the central axis is... For vectors Direction vector of the central axis The modulus of the cross product is: ;because Since it is a unit vector, the formula simplifies to: Vertical distance That is, it serves as a quantitative indicator for measuring the spatial coherence between the corresponding candidate point and the historical trajectory. Check whether the candidate point is located within the sector region, i.e., the vector. and The dot product is positive and less than And the included angle is less than Select the option that simultaneously satisfies the condition of being located within the sector area and The candidate point with the smallest value is selected as the final choice to maintain a smooth trajectory at the current moment; if no point is located within the sector, then only the candidate point with the smallest value is selected. Find the point with the smallest value and record it.

2. The dynamic testing system for heat dissipation performance based on a radiator according to claim 1, characterized in that, The dynamic thermal load simulating actual working conditions specifically includes: Real-time acquisition of electrical parameters of actual electronic devices during operation, and calculation of instantaneous power data stream; Dynamic heat transfer characteristics are corrected for instantaneous power data streams to eliminate power response delay caused by the thermal capacitance of electronic device packages and generate corrected power timing. Based on the corrected power timing and the resistance-temperature characteristics of the heat source simulator, the voltage control sequence for directly driving the heat source simulator is generated. By executing a voltage control sequence through a high-bandwidth power output unit, the heat source simulator generates a spatial dynamic thermal load consistent with the thermal behavior of actual electronic devices.

3. The dynamic testing system for heat dissipation performance based on a radiator according to claim 1, characterized in that, The receiving process of the multi-channel response wave signal is as follows: Based on the real-time power value of the dynamic heat load and the historical spatial temperature distribution, the estimated location of the high-temperature region on the heat source simulator at the current moment is calculated. Based on the estimated location, multiple beamforming units arranged around the target are controlled to adaptively adjust the energy focus of the probe wave signal and direct it towards the spatial region where the estimated location is located for radiation. At the same time, multiple beamforming units synchronously receive response wave signals from different scattering paths of the heat source simulation body, and separate and buffer the response wave signals according to the corresponding spatial paths to form a set of multi-path response wave signals associated with spatial orientation.

4. The dynamic testing system for heat dissipation performance based on a radiator according to claim 3, characterized in that, The calculation of the estimated location of the high-temperature region on the heat source simulator at the current moment specifically includes: Based on the real-time power values ​​and the known electrical connection layout applied to the heat source simulator, the spatial pattern of the current power density distribution is identified. Extract the migration paths and evolution rates of historical high-temperature regions that have the strongest spatiotemporal correlation with the currently identified spatial patterns from historical spatial temperature distribution; The migration path is extrapolated based on the evolution rate, and combined with spatial patterns to generate a dynamic spatial probability distribution map representing the probability of high temperature occurrence. From the dynamic spatial probability distribution map, continuous spatial regions with probability values ​​exceeding a preset threshold are extracted, and the geometric center coordinates of the continuous spatial regions are determined as the estimated locations.

5. The dynamic testing system for heat dissipation performance based on a radiator according to claim 1, characterized in that, The reconstructed three-dimensional dynamic distribution of dielectric constant of the heat source simulator under dynamic thermal load specifically includes: The complex domain difference between the multi-channel response wave signal and the preset reference wave signal is performed to obtain a complex signal difference set that contains only the dielectric disturbance information caused by dynamic thermal load. Based on the known spatial path and incident angle of each signal in the complex signal difference set, a set of nonlinear equations is constructed with the dielectric constant perturbation of the discrete voxels of the entire heat source simulation volume as the unknown. The physical structure and geometric contour of the heat source simulation body are introduced as spatial constraints, and the three-dimensional dielectric constant distribution obtained from the previous time step is used as the initial value of the temporal constraints. The solution of the nonlinear equation system is transformed into an iterative optimization process with temporal and spatial constraints. By minimizing the residual between the calculated and measured values ​​of the complex signal difference set and satisfying the spatiotemporal constraints, the dielectric constant perturbation of each spatial voxel in the heat source simulation body at the current moment is iteratively solved, and then the three-dimensional dynamic distribution of dielectric constant is synthesized.

6. The dynamic testing system for heat dissipation performance based on a radiator according to claim 5, characterized in that, The process of constructing the nonlinear equation system is as follows: For each signal in the complex signal difference set, based on the spatial path and incident angle of each signal, the additional phase shift and amplitude attenuation accumulated when the wavefront of the probe signal passes through each spatial voxel are calculated, and the initial contribution coefficients of all voxels under the corresponding signal are generated. Based on the initial contribution coefficients and incorporating the estimated influence of multiple internal scattering paths caused by the geometry of the heat source simulation body, the contribution coefficients of each spatial voxel are iteratively corrected to obtain the final spatial weight coefficients. The difference of each complex signal is represented as the superposition of the perturbation of the dielectric constant of all spatial voxels and the corresponding final spatial weight coefficient, and each signal is established as a nonlinear equation. The nonlinear equations established by collecting all signals in the set of complex signal differences constitute a system of nonlinear equations.

7. The dynamic testing system for heat dissipation performance based on a radiator according to claim 1, characterized in that, The mapping generates a dynamic three-dimensional temperature field distribution of the heat source simulation body, specifically including: For each spatial voxel in the three-dimensional dynamic distribution of dielectric constant, based on the material dielectric constant-temperature correspondence, the preliminary temperature value corresponding to the dielectric constant of the spatial voxel is calculated independently to form a preliminary three-dimensional temperature distribution; Based on the known physical structure of the heat source simulation body, the heat conduction path of the material between adjacent spatial voxels is identified, and the spatial continuity of the preliminary three-dimensional temperature distribution is corrected by the constraint of the heat conduction path to obtain the intermediate three-dimensional temperature distribution. The intermediate three-dimensional temperature distribution is compared with the temperature change trend predicted based on historical time series data, and the temperature values ​​that exceed the prediction range are optimized for time series smoothing based on the physical heat transfer law. By integrating the results of spatial continuity correction and temporal smoothing optimization, a final three-dimensional temperature field dynamic distribution that is continuous in space and time is synthesized.

Images