Apparatus for thermal management of radar, method for thermal management of radar, and medium for storing a program performing thermal management of radar

Abstract

A disclosed embodiment provides a radar thermal management device using a variable shape microchannel, comprising a microchannel including cooling water that absorbs heat generated in a transmitting and receiving unit of a radar; wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the length direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of the heat generated in the transmitting and receiving unit.

Description

Technology Field

[0001] The disclosed embodiments relate to a radar thermal management device using a variable shape microchannel, a radar thermal management method, and a storage medium storing a program for performing radar thermal management. Background Technology

[0003] GaN (Gallium Nitride)-based Active Electronically Scanned Array (AESA) radar transmit / receive modules (TRMs) locally provide 1000–W / cm² 2 It generates a heat flux exceeding the above. A micro-channel heat sink (MCHS) is used to cool this.

[0004] Conventional rectangular channels suffer from degraded cooling performance downstream due to the growth of thermal boundary layers. Although zigzag channels introduced to address this issue offer excellent heat transfer, excessive pressure drop caused by abrupt flow path changes consumes a large amount of pumping power and reduces system efficiency. Furthermore, the channel geometry is uniform despite the uneven heat generation of the chip, making it inefficient.

[0005] While vibration sensors or data-driven AI are used to diagnose cooling system failures, it is difficult to obtain actual failure data due to the nature of defense weapon systems. Furthermore, purely data-driven AI that does not consider physical principles suffers from low reliability; when training data is insufficient, it makes physically impossible predictions (e.g., violating the law of conservation of energy). The problem to be solved

[0007] The purpose of the embodiments disclosed below is to solve the above problems by providing a radar thermal management device using a variable shape microchannel, a radar thermal management method, and a storage medium storing a program for performing radar thermal management.

[0008] Another objective of the embodiment is to provide a radar thermal management device that locally deforms the shape of a flow path according to a heat concentration area, a radar thermal management method, and a storage medium that stores a program for performing radar thermal management.

[0009] Another objective of the embodiment is to provide a radar thermal management device that predicts a system state based on physical information, a radar thermal management method, and a storage medium that stores a program for performing radar thermal management. means of solving the problem

[0011] One disclosed embodiment provides a radar thermal management device using a variable shape microchannel, comprising a microchannel that includes cooling water that absorbs heat generated in a transmitting and receiving unit of a radar, wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the length direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of heat generated in the transmitting and receiving unit.

[0012] The above zigzag shape structure is characterized by including a first pitch and a first amplitude for a first region of the transmitting and receiving unit in which the amount of heat generated is greater than a threshold value based on the heat distribution, and the above zigzag shape structure is characterized by including a second pitch longer than the first pitch and a second amplitude smaller than the first amplitude for a second region of the transmitting and receiving unit in which the amount of heat generated is less than a threshold value based on the heat distribution.

[0013] At least one of the inner or outer edges of the above zigzag-shaped structure is characterized by having a fillet formed therein having curvature.

[0014] The microchannel is characterized by including a secondary flow path that moves the cooling water of the main channel included in the microchannel from the high-pressure section to the low-pressure section.

[0015] The microchannel comprises an inlet and an outlet through which a cooling fluid flows, and the microchannel is characterized by having a divergent structure in which the channel width of the outlet is greater than the channel width of the inlet.

[0016] The radar thermal management device further comprises a control unit that acquires sensor data related to the microchannel, generates physics residual data based on the sensor data, and inputs the physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information.

[0017] The control unit is characterized by inputting the sensor data into a physical model including fluid dynamics and heat transfer governing equations to calculate a theoretical predicted value, and calculating physical residual data corresponding to the difference between the sensor data and the theoretical predicted value.

[0018] A disclosed embodiment provides a radar thermal management method using a variable-shape microchannel, comprising: acquiring sensor data related to a microchannel containing cooling water that absorbs heat generated in a radar transmitting and receiving unit; generating physics residual data based on the sensor data; and inputting the physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information; wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the length direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of heat generated in the transmitting and receiving unit.

[0019] One disclosed embodiment provides a storage medium for storing a computer-executable program, wherein the microchannel includes a cooling water that absorbs heat generated in a radar transceiver, acquires sensor data related to the microchannel, generates physics residual data based on the sensor data, inputs the physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information, and wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the length direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of the heat generated in the transceiver. Effects of the invention

[0021] According to the disclosed embodiment, pressure loss can be minimized by providing an optimized flow path shape that reduces pressure loss in unnecessary sections while maintaining excellent heat transfer performance of the variable shape zigzag channel.

[0022] According to the disclosed embodiment, by providing an algorithm capable of accurately diagnosing the system state based on physical laws, it is possible to overcome data scarcity and perform high-reliability diagnosis. Brief explanation of the drawing

[0024] FIG. 1 discloses embodiments of a radar thermal management device using a variable shape microchannel according to an embodiment. FIG. 2 discloses embodiments of a zigzag shape structure of a microchannel according to an embodiment. FIG. 3 discloses embodiments of a fillet structure having a zigzag shape structure according to an embodiment. FIG. 4 discloses examples of a comparison of stress concentration phenomena based on a channel structure according to an embodiment. FIG. 5 discloses examples of heat transfer and pressure drop optimization comparisons based on fillet radius according to an embodiment. FIG. 6 discloses embodiments of a secondary flow path structure of a microchannel according to an embodiment. FIG. 7 discloses examples of the thermodynamic state of reverse circulation within a secondary flow path according to an embodiment. FIG. 8 discloses embodiments of a divergent structure of a zigzag shape structure according to an embodiment. FIG. 9 discloses one embodiment of a normalized performance comparison based on a channel structure according to an embodiment. FIG. 10 discloses examples of a comparison of heat transfer performance based on channel shape according to an embodiment. FIG. 11 discloses one embodiment of a pressure drop comparison based on channel shape according to an embodiment. FIG. 12 discloses an embodiment of a radar thermal management method using a variable shape microchannel according to an embodiment. Specific details for implementing the invention

[0025] FIG. 1 discloses embodiments of a radar thermal management device using a variable shape microchannel according to an embodiment.

[0026] The radar thermal management device (100) according to the present disclosure can perform thermal management for the radar. In one embodiment, the radar may include an Active Electronically Scanned Array (AESA) radar. However, the thermal management device according to the present disclosure can be applied not only to radar but also to various high-power electronic equipment.

[0027] In one embodiment, the radar thermal management device (100) can maximize cooling efficiency while reducing pressure loss by optimizing the shape of the microchannel (i.e., cooling path) according to the heat distribution of the radar. In one embodiment, the radar thermal management device (100) can detect signs of abnormality in the cooling system at an early stage using a physics-law-based artificial intelligence (AI) prediction model.

[0028] In one embodiment, the radar thermal management device (100) may include a radar (110), a microchannel (120), and a control unit (130). In one embodiment, the radar (110) may be configured separately and mounted on an aircraft without being integrated with the radar thermal management device (100).

[0029] In one embodiment, the microchannel (120) may include cooling water that absorbs heat generated in the transmitting and receiving portion of the radar (110). That is, heat may be absorbed by flowing cooling water through the interior of the microchannel (120). In one embodiment, the microchannel (120) may include an inlet and an outlet through which a cooling fluid flows, and may include a microchannel heatsink that is in thermal contact with a heat-generating element (e.g., radar (110)).

[0030] In one embodiment, the microchannel (120) may include a zigzag shape structure in which the flow direction of the coolant changes along the length direction. In one embodiment, the flow direction of the channel in the zigzag shape structure may change periodically along the length direction. In one embodiment, the pitch or amplitude of the zigzag shape structure may be formed according to the heat distribution of the heat generated in the transmitting and receiving unit. In one embodiment, the pitch, which is the number of bends per unit length of the zigzag shape structure, or the amplitude, which is the depth of the bend, may be formed differently depending on the location of the channel in correspondence with the local heat flux distribution of the heating element. Further details regarding this are described below.

[0031] In one embodiment, the control unit (130) may acquire sensor data associated with at least one of the radar (110) or the microchannel (120). In one embodiment, the sensor data may include at least one of the flow rate, temperature, or pressure data of the microchannel (120). In one embodiment, the sensor data may include the flow rate of the cooling fluid, inlet and outlet temperatures, and inlet and outlet pressures. In one embodiment, the sensor data may be acquired through a sensor unit (not shown).

[0032] In one embodiment, the control unit (130) can generate physics residual data based on sensor data. In one embodiment, the control unit (130) inputs the sensor data into a physics model including fluid dynamics and heat transfer governing equations to calculate a theoretical prediction value, and can calculate physics residual data corresponding to the difference between the sensor data and the theoretical prediction value.

[0033] In one embodiment, the control unit (130) can input physical residual data into a physics-based artificial intelligence prediction model to generate microchannel state information.

[0034] In one embodiment, a residual term of the physics governing equation (conservation of mass, momentum, and energy) may be added to the learning loss function of a physics-based artificial intelligence prediction model as shown in the mathematical formula below.

[0035]

[0036] Here, Loss is the loss function, MSE data is the error (Mean Squared Error) between the value measured by the actual sensor (temperature, pressure, etc.) and the value predicted by the neural network, MSE is a weighting coefficient that balances the importance between data error and physical law error. physics represents the residual indicating the degree to which governing equations (e.g., Navier-Stokes equations, energy equations) are violated, that is, whether physical laws are observed. This allows for physically valid inference to be performed even when sensor data is insufficient or noise is present.

[0037] In one embodiment, the microchannel state information may include at least one of an energy residual or a Hydraulic Health Index (HHI).

[0038] In one embodiment, the energy residual is used to detect sensor failure or external heat inflow by monitoring the imbalance between the input power and the amount of heat removed by the cooling water, and can be expressed as shown in the following mathematical formula.

[0039]

[0040] Here, E res is the energy residual, P electricε is the electrical input power supplied to the heater or chip, m is the mass flow rate of the cooling fluid, and C p is the specific heat of the cooling fluid, T out is the heatsink (microchannel) outlet temperature, T in represents the heatsink inlet temperature. In one embodiment, the energy residual is an indicator representing the discrepancy between the theoretical thermal equilibrium state and the actual system, and a larger energy residual value may indicate system abnormalities (e.g., sensor failure, fouling, etc.). In one embodiment, if the energy residual exceeds a threshold value, it may be diagnosed as a degradation in cooling performance (fouling, hot spot occurrence) or a malfunction of the temperature sensor.

[0041] In one embodiment, the hydrodynamic integrity index may represent the ratio of the actual measured pressure drop to the theoretical pressure drop model. In one embodiment, the hydrodynamic integrity index may be used to determine fouling (blockage) or leakage inside the microchannel.

[0042] In one embodiment, the hydrodynamic integrity index may include the ratio of the predicted pressure drop to the actual measured pressure drop. In one embodiment, if the hydrodynamic integrity index is greater than a threshold value, it may be determined that a hydrodynamic problem, such as channel clogging or pump malfunction, has occurred.

[0043] In one embodiment, the control unit (130) can determine at least one of fouling, leakage, or sensor failure of the system using an artificial neural network that takes physical residuals as input.

[0044] In one embodiment, the physics-based artificial intelligence prediction model may include a Physics-Informed Neural Network (PINN) trained by including the degree of violation of the physics governing equation (Residual Error) as a penalty term of the loss function.

[0045] In one embodiment, the control unit (130) may provide microchannel status information to the user through a display unit (not shown). In one embodiment, the display unit may include a display screen.

[0046] In one embodiment, the radar thermal management device (100) may use data stored in a separate storage / database under the control of a device responsible for data processing. Here, the storage mainly uses a hard disk or SSD to store data, and the database manages structured data and can perform operations such as searching and modifying. At this time, the functions performed by the radar thermal management device (1000) may be performed by the device's processor (e.g., FPGA (Field Programmable Gate Array), DSP (Digital Signal Processor)). In one embodiment, the control unit (130) may include a processor.

[0047] According to the present disclosure, pressure loss can be minimized by providing an optimized flow path shape that reduces pressure loss in unnecessary sections while maintaining excellent heat transfer performance (vortex generation, boundary layer disruption) of the zigzag channel.

[0048] In addition, according to the present disclosure, temperature uniformity can be ensured by locally deforming the shape of the flow path to match the heat concentration area (Hotspot) of the transmitting and receiving unit.

[0049] In addition, according to the present disclosure, by providing an algorithm that can accurately diagnose the system state based on physical laws (e.g., Navier-Stokes, energy equation) even in an environment where failure data is scarce, it is possible to overcome data scarcity and perform high-reliability diagnosis.

[0050] In addition, according to the present disclosure, by forming turbulence mainly in hot spots and reducing flow resistance in the remaining sections, thermal resistance can be reduced by more than 10% under the same pumping power compared to conventional zigzag channels, thereby optimizing thermal fluid performance.

[0051] In addition, according to the present disclosure, by eliminating localized temperature spikes, thermal fatigue of the radar element can be prevented and its lifespan extended, thereby improving reliability.

[0052] In addition, according to the present disclosure, anomalies are detected using physical correlations (such as energy conservation) without the need for vast failure data, so they can be immediately applied to rare defense and space systems.

[0054] FIG. 2 discloses embodiments of a zigzag shape structure of a microchannel according to an embodiment.

[0055] The microchannel (120) according to the present disclosure can perform concentrated cooling by increasing the density (i.e., amplitude and frequency) of the zigzag pattern only at high-heat hotspot locations of the radar (110) (e.g., heat-generating element, chip, etc.), and minimize unnecessary pressure loss by arranging a flow path that is close to a straight line in low-heat areas.

[0056] In one embodiment, unlike a conventional uniform zigzag channel, the pitch and amplitude of the microchannel (120) according to the present disclosure can be designed to be variable based on the thermal map of the chip.

[0057] In one embodiment, for a hotspot, a short pitch shorter than a threshold and an amplitude larger than a threshold may be applied to the microchannel (122). This rapidly induces a change in the direction of the flow, which generates a strong Dean vortex to locally maximize the heat transfer coefficient. In one embodiment, the threshold values ​​applied to the pitch and amplitude may be different from each other.

[0058] In one embodiment, the zigzag shape structure of the microchannel (122) may include a first pitch and a first amplitude for a first region (112) of the transceiver where the amount of heat generated is greater than a threshold value based on the heat distribution. In one embodiment, the flow path section corresponding to the first region (112) where the amount of heat generated by the heat element is greater than or equal to a preset threshold value may have a first pitch. In one embodiment, the first region (112) may include a hotspot region.

[0059] In one embodiment, for the low-heat section, a long pitch longer than the threshold value or an amplitude smaller than the threshold value or a straight section may be applied to the microchannel (124). This allows for minimizing flow resistance and pressure recovery. According to the present disclosure, the hot spot temperature can be effectively controlled while reducing the total pressure drop by 10 to 15% compared to the conventional zigzag.

[0060] In one embodiment, the zigzag shape structure of the microchannel (124) may include a second pitch longer than the first pitch and a second amplitude smaller than the first amplitude for a second region (114) of the transceiver where the amount of heat generated is smaller than the threshold value based on the heat distribution. In one embodiment, the flow path section corresponding to the second region (114) where the amount of heat generated is less than the threshold value may have a second pitch larger than the first pitch or be formed as a straight flow path. In one embodiment, the second region (114) may include a low-resistance region.

[0061] In one embodiment, the heat transfer efficiency in the first region (112) can be increased and the pressure loss in the second region (114) can be reduced.

[0062] In one embodiment, the zigzag shape structure can increase heat transfer efficiency through flow separation and reattachment at the bending section (bend section) and active fluid mixing by secondary flow.

[0064] FIG. 3 discloses embodiments of a fillet structure having a zigzag shape structure according to an embodiment.

[0065] In one embodiment, at least one of the inner or outer edges of the zigzag shape structure may have a fillet formed thereon having curvature. In one embodiment, the edge to which the fillet according to the present disclosure is applied may be referred to as a rounded edge. In one embodiment, by applying a fillet (curvature) to the inner and outer edges of the zigzag bend of the microchannel (120), a dead zone caused by flow separation can be eliminated and pressure loss reduced. In one embodiment, the inner edge may include the inside of the curvature of the zigzag bend (U-shape) of the microchannel (120). In one embodiment, the outer edge may include the outside of the curvature of the zigzag bend (U-shape) of the microchannel (120).

[0066] In one embodiment, the flow within the microchannel may generally belong to the laminar flow region, where the Reynolds number (Re) is 2000 or less. Unlike turbulent flow, mixing between fluid layers is not active in laminar flow, so the thermal boundary layer formed near the wall can become a major resistance factor that hinders heat transfer. In one embodiment, the zigzag shape may be introduced to artificially disturb this laminar flow, thereby destroying the boundary layer and promoting heat transfer. However, the specific shape of the bend can act as a decisive variable determining the behavior of the flow.

[0067] In conventional zigzag channels with sharp corners bent at right angles (90 degrees), the fluid may not follow the wall surface due to inertia, and separation may occur. In other words, conventional sharp corners can exhibit flow separation and dead zones at right-angle corners, where heat transfer may be drastically reduced.

[0068] By applying a fillet to the bend of the microchannel (120) according to the present disclosure, efficient generation and control of the Dean vortex can be secured. According to the present disclosure, a counter-rotating pair can be formed in which the fluid flowing through the curved channel moves the fast fluid at the center of the channel toward the outer wall and the slow fluid at the wall toward the inside due to centrifugal force, rotating in a direction perpendicular to the cross-section of the channel. This is called a Dean vortex, and heat transfer can be dramatically improved by mixing the bulk fluid.

[0069] In one embodiment, in the case of the Dean Number (De), the intensity of the Dean vortex is the Dean Number ( It can be proportional to ). Here, R crepresents the radius of curvature. Conventional sharp edges are Although it is close to implying infinite centrifugal force mathematically, in reality, flow separation and turbulent-like chaotic mixing dominate, so an ordered secondary flow is not formed. On the other hand, a fillet with a suitable radius of curvature according to the present disclosure can stably guide the flow and generate a strong and continuous vortex.

[0070] The rounded zigzag shape structure according to the present disclosure may exhibit more efficient mixing performance than simple curved channels or sharp channels. This may be because the vortex generated at the bend is maintained until the next straight section, continuously keeping the thermal boundary layer thin. In particular, it can be seen that the rounded corners of the zigzag channel form a secondary flow pattern that is advantageous for particle focusing and mixing.

[0071] In one embodiment, pressure drop is as important as heat transfer performance, and if the pressure drop in the microchannel is high, the load on the pump increases, the risk of leakage increases, and the energy efficiency of the entire system may decrease.

[0072] In one embodiment, for the reduction of minor loss, the loss coefficient (k) of a 90-degree meter bend in fluid dynamics is approximately 1.1 to 1.3, while an appropriate curvature ( The loss factor of a Smooth Bend with ) can drop sharply to the level of 0.2 to 0.3.

[0073] In one embodiment, regarding the fluid resistance reduction effect of the fillet, a heatsink with a fillet profile according to the present disclosure can reduce the pressure drop by 25% compared to sharp channels due to the smooth streamlines. On the other hand, as the number of conventional sharp bends and flow velocity increase, the flow resistance can increase non-linearly. Therefore, in high-heat cooling systems requiring high flow rate operation, the fillet can secure energy savings.

[0075] FIG. 4 discloses examples of a comparison of stress concentration phenomena based on a channel structure according to an embodiment.

[0076] In one embodiment, in the case of a conventional sharp edge, the stress concentration factor (Kt) increases rapidly, and a crack region with a high risk of local failure may appear. On the other hand, in the case of a rounded edge with a fillet according to the present disclosure, stress is distributed over a wide area, ensuring structural stability, which can be used as a factor to determine fatigue life when driving a variable shape.

[0077] In one embodiment, the stress concentration factor (K) at the tip of a sharp notch at a 90-degree angle in elastic theory. t ) can diverge to infinity as the fillet radius (r) approaches 0. This can be referred to as a stress singularity. In one embodiment, in actual materials, plastic deformation occurs and the stress is limited to a finite value, but in MEMS heatsinks made of brittle materials such as silicon (Si), a crack may occur and propagate immediately without plastic deformation and lead to fracture.

[0078] As disclosed in the present disclosure, applying a fillet (R) to the corners allows the stress flow lines to be smoothly distributed, thereby reducing the maximum stress value to within the allowable stress of the material. In one embodiment, the inter-layer stress was reduced to 34.8% of the tensile strength through MCHS structural optimization, which can dramatically extend the structural life.

[0079] The physical information-based neural network (PINN) according to the present disclosure can be trained using the Navier-Stokes equation as the loss function. In this case, the channel shape acts as a boundary condition, and the geometric characteristics of the shape can directly affect the learning performance of the neural network.

[0080] In one embodiment, the physical information-based neural network is based on Gradient Descent, which may require differentiability for all variables. A sharp edge may be a singularity in the coordinate system. Although it can be numerically approximated, the gradient of flow velocity or pressure at this point may change abruptly or tend to diverge to infinity. Such singularities can be a major cause of the loss function spike or the network failing to converge by getting stuck in a specific local minimum during the process of finding a solution.

[0081] For a physical information-based neural network to accurately predict complex flow fields, boundary conditions must be as smooth as possible (C 1 or C 2 It may be advantageous to express it as a continuity function.

[0082] In one embodiment, applying a fillet to a microchannel causes the boundary normal vector to change continuously, making it much easier for the neural network to learn the boundary layer flow. This can increase the learning speed and reduce errors during the inference stage.

[0084] FIG. 5 discloses examples of heat transfer and pressure drop optimization comparisons based on fillet radius according to an embodiment.

[0085] In one embodiment, as fillet reflection increases, the pressure drop continuously decreases, and heat transfer performance can be maximized in a specific range. According to the present disclosure, an optimal curvature range can be set by taking these trade-offs into account.

[0086] In one embodiment, by setting the fillet radius (R) as a continuous design variable, a physical information-based neural network or a topology optimization algorithm can actively search for the optimal balance point (Pareto Frontier) between the hotspot temperature and the pressure drop. In one embodiment, it can be confirmed that a shape that eliminates stagnant regions and smoothly induces flow during the topology optimization process lowers the entropy generation rate and improves overall performance.

[0087] In one embodiment, the size (radius) of the fillet is the hydraulic diameter (D) of the channel. h It can be expressed as a ratio to ).

[0088] In one embodiment, too small fillet (R / D h In the case of <0.2), flow separation cannot be completely suppressed, and the stress concentration relief effect may also be negligible.

[0089] In one embodiment, too large fillet (R / D hIn the case of >1.5), the microchannel becomes serpentine or nearly straight rather than zigzag, and the secondary flow generation effect caused by the abrupt change of direction characteristic of zigzag channels may be reduced.

[0090] In one embodiment, the optimal interval ( In the case of ), the effects of reduced pressure loss and improved heat transfer (increased Nusselt number) can be optimally balanced when the fillet radius is between 0.5 and 1.0 times the channel width. The Performance Evaluation Index (PEC) can be maximized in this range.

[0092] FIG. 6 discloses embodiments of a secondary flow path structure of a microchannel according to an embodiment.

[0093] In one embodiment, the microchannel (120) may include a secondary channel that moves the cooling water of the main channel (201) included in the microchannel (120) from a high-pressure section to a low-pressure section. In one embodiment, the secondary channel may include a micro-slit channel (203) coupled in parallel to the main channel (201) included in the microchannel (120).

[0094] Regarding the flow within the microchannel (120), the boiling phenomenon at the hot spot can be considered as a factor that dynamically changes the resistance characteristics of the flow path.

[0095] In one embodiment, the hydrodynamic resistance (R) of the channel in single-phase laminar flow h ) is the hydraulic diameter (D) according to the Hagen-Poiseuille law as shown in the mathematical formula below. h It is inversely proportional to the fourth power of ) and the viscosity of the fluid ( It can be proportional to ).

[0096]

[0097] In one embodiment, the micro-slit channel (203), which is narrower than the main channel (201), may have higher flow resistance because it has a smaller hydraulic diameter. Therefore, if only the normal steady-state single-phase flow is considered, the cooling water may not flow by bypassing the wide main channel (201) and into the narrow micro-slit channel (203).

[0098] However, the situation according to the present disclosure may be considered as flow into a micro-slit channel (203) under special thermal conditions called a ‘hot spot (e.g., first region (112))’. In one embodiment, under high thermal conditions of the hot spot, the fluid undergoes a rapid temperature rise, and in the case of liquid coolant, as the temperature rises, the viscosity ( ) can be significantly reduced. In one embodiment, thermal creep and viscosity reduction effects can be considered as variables that change flow characteristics within the microchannel. However, a more dominant influence than changes in liquid viscosity may be the rapid increase in two-phase flow resistance due to boiling.

[0099] In one embodiment, the temperature of the hot spot area is the saturation temperature of the cooling water (T sat If ) is exceeded, nucleate boiling may begin. Bubble formation in narrow spaces, such as microchannels, can exhibit completely different behavior compared to macrochannels. In this case, the bubbles rapidly grow into a slug that fills the cross-section of the channel, which can act as a 'vapor piston' that blocks the flow. The pressure drop in the two-phase flow that occurs at this time ( ) can increase from several times to tens of times compared to single-phase flow.

[0100] In one embodiment, the accelerated pressure drop may represent the pressure loss that occurs when the volume expands by about 1,000 times (under atmospheric pressure in the case of water) and the flow velocity increases rapidly as the liquid undergoes a phase change to a gas. The frictional pressure drop may represent the loss due to friction at the gas-liquid interface and friction between the bubble and the wall.

[0101] Therefore, the moment the hot spot is activated, the main channel (201) can change into a high impedance zone hydrodynamically, even though it is geometrically a wide tube. In this case, the role of the micro-slit channel (203) can be redefined.

[0102] The micro-slit channel (203) according to the present disclosure may function as a manual valve or bypass that operates when a hot spot occurs, rather than as a normal main flow path.

[0103] Under normal conditions (non-hot spot conditions), a single-phase liquid flows within the main channel (201), and since the hydraulic diameter of the main channel (201) is larger than that of the micro-slit channel (203), most of the flow flows through the main channel (203). At this time, the micro-slit channel (230) may be a relatively high-resistance section.

[0104] When a hot spot occurs (activation of the first region (112)), rapid boiling may occur in the main channel (201) of the hot spot area, and a vapor plug may be formed, and this vapor plug may vertically increase the flow resistance of the main channel (201).

[0105] Accordingly, the micro-slit channel (203) adjacent to the hotspot or the adjacent main channel can maintain a relatively low resistance, where boiling has not yet occurred or the vapor quality is low. In this case, similar to the current distribution in a parallel circuit under Ohm's law, the fluid can flow into the micro-slit channel (203) with relatively low resistance (maintaining a liquid state), bypassing the main channel (201) where the resistance has increased.

[0106] In one embodiment, the micro-slit channel (203) may be connected in parallel to the main channel (201) of the micro-channel (120) and may include an inlet through which cooling water from the main channel (201) flows in during a hot spot state and an outlet through which water flows out to the main channel (201).

[0107] In one embodiment, in the case of microchannel boiling heat transfer, Redineg instability may occur in the region where the relationship between pressure drop and flow rate has a negative slope. This may manifest as a phenomenon where, as the flow rate decreases, the pressure drop actually increases, causing the flow rate to decrease even more rapidly and eventually lead to dry-out.

[0108] In one embodiment, the micro-slit channel (203) can function as a pressure equalization device to alleviate such instability. In one embodiment, it can be seen that the transverse connecting channel is suitable for providing a uniform mass flow distribution. When local pressure rises due to steam generation in a hot spot, this pressure can be relieved to an adjacent low-pressure channel through the micro-slit channel (203). This prevents complete flow choking in the main channel (201) and maintains hydrodynamic stability of the entire micro-channel system. Thus, in a hot spot situation, the micro-slit channel (203) can function as a low-resistance relief valve that hydrodynamically accommodates the flow to lower pressure spikes.

[0109] In one embodiment, rapid boiling at a hot spot (e.g., first region (112)) can generate a strong local pressure source. The explosive volume expansion that occurs when the liquid changes phase to gas can exert a force pushing the fluid in both upstream (Inlet) and downstream (Outlet) directions.

[0110] In the case of local overpressure, steam generation within the microchannel (120) can explain the phenomenon of the fluid flowing back toward the inlet plenum. According to the present disclosure, instead of a blocked inlet, a micro-slit channel (203) connected to an adjacent channel may be applied.

[0111] In the case of reverse driving force, the vapor pressure expanding at the hot spot (P hot ) is the pressure of the adjacent cold channel (P cold It can be formed higher than ). If the micro-slit channel (203) is inclined with respect to the flow direction or placed at a specific location, this pressure difference can push the fluid through the slit with a vector having a component opposite to the main flow.

[0112] In the case of forming a recirculation loop, fluid displaced from the hotspot channel (A) can move to an adjacent channel (B) through a slit and then flow back to a point prior to the hotspot in channel A through an upstream pressure equilibrium process. This can consequently form a thermodynamic pumping loop circulating around the hotspot. In one embodiment, the micro-slit channel (203) may include a slit that allows cooling water introduced from the main channel (201) to circulate.

[0113] In terms of heat transfer, this counter-circulation can act as a mechanism to rewet cold liquid to the root of the hot spot, rather than simply pushing the fluid backward. This can dramatically improve the critical heat flux (CHF) by continuously supplying a liquid film to the wall surface, which is prone to drying out due to boiling.

[0114] In one embodiment, even in a single-phase flow region where boiling does not occur, the micro-slit channel (203) can induce a secondary flow to create a flow pattern similar to opposite-direction circulation.

[0115] In the case of flow separation, when the fluid enters the narrow micro-slit channel (203) from the main channel (201) or passes through the micro-slit channel (203), boundary layer separation may occur due to discontinuities in the wall surface.

[0116] In the case of Dean vortices and recirculation, it can be seen that the secondary passage of the microchannel (120) induces flow collision and vortex formation. In particular, secondary flow such as Dean vortices may refer to flow that rotates spirally within the channel cross-section. The velocity vector at the bottom (near the wall) of this vortex may be opposite to the main flow direction, and when observed in a two-dimensional cross-section, this can be represented as local reverse circulation.

[0117] In the case of boundary layer disruption, in a microchannel (120) where laminar flow is dominant, heat transfer efficiency can decrease rapidly as the thermal boundary layer becomes thicker. The opposing circulation component (vortex) induced by the micro-slit channel (203) can act to pump the fluid from the hot wall towards the center of the channel and impingement the cold fluid from the center against the wall. This suppresses the development of the thermal boundary layer and can continuously maintain the thermal entrance length effect of the flow.

[0118] In conclusion, this reverse circulation is not merely a simple backflow, but can renew the boundary layer and mix the fluid.

[0119] The shape of the micro-slit, namely its width, angle, and fillet radius, is an important variable that determines the intensity and location of the counter-circulation described earlier.

[0120] In the case of optimizing the fillet radius, applying a rounded fillet according to the present disclosure instead of a conventional sharp edge can reduce the downstream low-speed region (dead zone) and promote mixing of the hot fluid near the wall and the cold fluid in the center. This allows the micro-slit channel (203) to function not as a simple hole, but as a nozzle that smoothly draws in the flow to maximize circulation.

[0121] In the case of an asymmetric variable structure, the degree of opening or closing of the micro-slit channel (203) or the cross-sectional area of ​​the flow path may be changed according to the thermal load of the hot spot. By adjusting the geometric parameters of the micro-slit channel (203) according to the location and intensity of the hot spot, reverse circulation can be induced to occur precisely at the time required.

[0122] In one embodiment, the physical information-based neural network can predict the time and location of hotspot occurrence and calculate optimal shape parameters or control signals that allow reverse circulation in the micro-slit channel (203) to contribute most effectively to hotspot cooling. The physical information-based neural network can enable a design that actively utilizes non-linear flow patterns, including reverse flow, beyond conventional intuitive designs.

[0123] The micro-slit channel (203) can function as a dynamic low-resistance zone when a hot spot occurs. Normally, the main channel (201) serves as the main flow path, but when vapor blockage occurs due to boiling at the hot spot and the impedance of the main channel (201) increases rapidly, the slit, which remains in a relatively liquid state, becomes a low-resistance bypass for the fluid, thereby ensuring a continuous supply of cooling water.

[0124] Counter-circulation within the micro-slit channel (203) can occur due to the combined action of pressure waves caused by thermodynamic expansion, the inverse phase pressure difference of the zigzag flow path, and structural secondary flow. This may be an intended local recirculation. This circulating flow can provide a heat transfer benefit by breaking the thermal boundary layer of the hotspot wall and rewetting liquid into the dry patch trapped in bubbles, thereby increasing the critical heat flux (CHF).

[0125] In one embodiment, the secondary flow path may include a link channel (not shown) connecting the first main channel and the second main channel included in the microchannel. In this case, the first main channel and the second main channel may refer to adjacent flow paths in the channel structure. In one embodiment, the secondary flow path may connect adjacent main flow paths to branch off a portion of the fluid.

[0126] In one embodiment, a cross-connection structure of link channels may be applied, in which a micro-slit is placed between zigzag channels arranged in parallel.

[0127] In one embodiment, if a specific channel becomes clogged with bubbles or the pressure increases rapidly due to a local heat load, the fluid may be bypassed to an adjacent low-pressure channel through a cross-connection. This can balance the pressure throughout the channel system and prevent catastrophic flow instability.

[0128] In one embodiment, the temperature uniformity of the entire chip surface can be improved through inter-channel fluid exchange based on transverse mixing.

[0129] According to the present disclosure, by forming a secondary flow path and bypassing the fluid from the high-pressure section to the low-pressure section, the thermal boundary layer can be further broken and mixing can be promoted.

[0131] FIG. 7 discloses examples of the thermodynamic state of reverse circulation within a secondary flow path according to an embodiment.

[0132] In one embodiment, referring to Panel A, the temperature profile of the hotspot can be observed. In this case, the saturation temperature (T sat It can be confirmed that a phase change begins when the temperature rises above )

[0133] Referring to Panel B, the resulting velocity vector field can be verified. T>T sat It can be observed that the local high-pressure region corresponding to the area generates a negative flow velocity vector (reverse flow) that pushes the fluid toward the inlet.

[0134] If the main channel has a zigzag or wave shape, more sophisticated fluid dynamic principles can be at work.

[0135] Fluid flowing through a zigzag channel repeatedly accelerates and decelerates whenever it passes a bend, and a pressure difference may occur between the inner and outer walls of the channel due to centrifugal force. If two adjacent channels are arranged in anti-phase relative to each other (i.e., designed so that the bend (high pressure) of Channel A faces the straight section (low pressure) of Channel B), a strong transverse pressure gradient periodically occurs between the two points separated by a slit ( ) can be formed.

[0136] In this inverse phase arrangement, the pressure of the dispersed phase has an inverse phase relationship with the continuous phase and can induce complex interfacial dynamics. When synthesizing the numerical analysis results, the sign of the transverse pressure difference changes as it proceeds along the channel length direction, and at certain points, such as hotspot areas (e.g., the first region (112)), this pressure gradient can induce a reverse vector component that runs counter to the main flow direction.

[0137] Pressure distribution simulation results confirm that the pressure profiles between adjacent zigzag channels intersect and exhibit a sinusoidal phase difference. Consequently, strong cross-mixing can occur in the slit section, where fluid mixes laterally and sometimes upstream solely due to the pressure difference between the two channels, regardless of the main flow direction. This can provide high heat transfer efficiency that cannot be expected from simple laminar flow.

[0139] FIG. 8 discloses embodiments of a divergent structure of a zigzag shape structure according to an embodiment.

[0140] In one embodiment, the microchannel (120) includes an inlet and an outlet through which a cooling fluid flows, and the microchannel (120) may include a divergent structure in which the channel width of the outlet is greater than the channel width of the inlet.

[0141] Diverging channels have a cross-sectional area in the flow direction (z). It can have a structure where increases. This is a decrease in density in the mass conservation equation ( Increase the area in response to ) It may be a principle that suppresses a rapid increase in flow velocity (u) by ) thereby. If the increase in flow velocity is suppressed, the rate of change of momentum decreases, causing an accelerated pressure drop ( ) can be drastically reduced. Under the same mass flow rate conditions, the total pressure drop of a diverging channel can be measured to be significantly lower compared to a straight channel. This means that more refrigerant can be supplied with the same pump capacity, and the efficiency (COP) of the entire cooling system can be improved.

[0142] In one embodiment, since the microchannel (120) widens as it extends downstream, it may be energetically advantageous for the bubble to expand toward the wider downstream side rather than the narrow upstream side due to the difference in Laplace pressure caused by surface tension and the difference in flow resistance when the bubble grows. This can induce the bubble to naturally be discharged toward the outlet.

[0143] In one embodiment, since the inlet cross-sectional area is designed to be narrow, the liquid flow velocity at the inlet can be relatively fast. This high inertia can act as a strong restoring force that resists the force of bubbles attempting to flow upstream. Even if instantaneous backflow occurs, the fast flow velocity at the narrow inlet can generate a sweep-off effect that pushes the bubbles back downstream.

[0144] In one embodiment, the subcooling number (N sub ) and Phase change number (N pch On the stability map represented by ), divergent channels can maintain stable flow over a much wider area than linear channels.

[0145] In the single-phase flow region, as the channel width increases, the flow velocity decreases, which may result in a slightly lower convective heat transfer coefficient. However, in the two-phase flow (boiling) region, which is the core of high-heat cooling, flow stabilization can suppress the occurrence of dry patches and aid in the rewetting of the liquid film, thereby improving the critical heat flux (CHF). Therefore, divergent channels can be effective for overall performance under high-load operating conditions.

[0146] The zigzag channel structure induces centrifugal force through periodic changes in the flow direction, which can generate a secondary flow called a Dean vortex in the channel cross-section. This secondary flow continuously mixes the cold fluid at the center of the channel with the hot fluid near the walls, thereby hindering the development of a thermal boundary layer and maximizing heat transfer efficiency. Additionally, the centrifugal force pushes the liquid refrigerant toward the channel walls (especially the concave surfaces), which can delay the drying phenomenon that occurs during the boiling crisis (CHF).

[0147] In one embodiment, the frequent changes in direction and vortex generation in the straight zigzag channel drastically increase flow resistance, and can cause severe pressure drop and choking phenomena, especially in two-phase flow where steam with a large specific volume flows at high speed. This can cause the pump capacity to be exceeded or amplify the instability of the system.

[0148] As disclosed in the present disclosure, the pressure problem of the aforementioned linear zigzag channel can be solved by applying a divergent structure to the zigzag channel structure and gradually increasing the channel width.

[0149] In one embodiment, the flow path that widens towards the downstream end can accommodate a massive volume of steam generated by boiling. This can reduce pressure loss (shape resistance) due to changes in direction by controlling the average flow velocity of the fluid passing through the zigzag bend to an appropriate level.

[0150] In one embodiment, the flow stability provided by the divergent structure can prevent the complex two-phase flow within the zigzag channel from degenerating into turbulent instability. That is, a thermohydraulic sweet spot can be achieved that offsets the penalty of excessive pressure loss while maintaining the effect of increased heat transfer by the Dean vortex.

[0151] In one embodiment, it can be seen that when a divergent structure is combined with a zigzag channel structure, the Nusselt number can be improved by up to 17% while suppressing the increase in pressure drop. Additionally, rounding the corner shape and adjusting the width can reduce the pressure drop by up to 48%.

[0152] In one embodiment, the diverging structure may be determined based on at least one of the diverging angle, edge shape, pitch, or amplitude of the microchannel (120).

[0153] In one embodiment, the divergence angle, In the case of ), excessive expansion can cause flow separation, which degrades heat transfer performance and creates local hot spots.

[0154] In one embodiment, for the divergence angle, a gradual expansion of 0.5 to 2.0 degrees on one side may be determined as optimization. In one embodiment, the change in width (W) according to the channel length (L) may be determined as shown in the following mathematical formula.

[0155]

[0156] Here, W is the width, Win is the width at the entrance, and z is the position variable. represents the divergence angle.

[0157] In one embodiment, regarding the corner shape and fillet radius, if the bends of the zigzag channel are sharp corners, a stagnation point where the flow velocity at the corner becomes zero and a dead zone in the wake may occur, causing a rapid increase in pressure loss.

[0158] In one embodiment, the inner corner radius (r in ) and outer corner radius (r out A smooth curved flow path can be formed by applying ). r out =r in By maintaining the relationship of +W(z), the channel width can be kept constant even in the bends, or resistance can be reduced by designing the width of the bends to be slightly wider than the straight sections.

[0159] In one embodiment, regarding the pitch and amplitude, the pitch and amplitude of the zigzag pattern can be used to determine the intensity of the secondary flow. Since a pitch that is too short causes an excessive pressure drop, optimization through simulation is required, but generally, a pitch of 5 to 10 times the channel width can be applied as an initial design value.

[0160] For example, as shown in the drawing, a divergent structure in which the channel width (W) increases from the inlet to the outlet can be applied to the microchannel (120) (W out >W in In one embodiment, secondary flow (vortex) can be generated and steam can be expanded at the bends of the zigzag structure. In one embodiment, the bends can be rounded (curvature R) to minimize pressure loss.

[0161] In one embodiment, the inlet zone is the narrowest section (W) of the microchannel (120). in It can represent ). In this case, it can act as a virtual nozzle that induces a high flow rate to promote bubble sweep-off in the supercooled boiling region and prevent backflow.

[0162] In one embodiment, the transition zone may have a repeating zigzag pattern and the channel width may be gradually expanded. As the generated bubbles grow and transform into slug flow, the expanded width can accommodate them to buffer sudden changes in flow velocity. The secondary flow (vortex) at the bends can mix the bubbles and liquid to prevent wall drying.

[0163] In one embodiment, the outlet zone is the widest section (W out It can represent ). In one embodiment, the outlet region may be an annular flow region where the refrigerant has turned into vapor. In one embodiment, the outlet region may have a sufficient flow path cross-sectional area to allow the vapor to escape without choking, thereby minimizing back pressure.

[0164] The microchannel (120) according to the present disclosure is applied with a divergent structure in which the channel width widens toward the outlet, thereby preventing the specific volume from increasing or boiling bubbles from being generated as the fluid is heated, which can prevent the flow velocity from increasing rapidly.

[0165] In one embodiment, the stability and reliability of the system can be maximized by basing the design on a divergent zigzag pattern and selectively placing cross-connecting slits (link channels) at key points where hot spots are expected.

[0167] FIG. 9 discloses one embodiment of a normalized performance comparison based on a channel structure according to an embodiment.

[0168] In one embodiment, it can be seen that the hybrid structure according to the present disclosure (integration of a zigzag channel structure and a divergent structure) demonstrates an optimal balance of performance, which can achieve the highest critical heat flux (CHF) (upgraded by about 60%) while significantly lowering the pressure drop compared to a tubular structure (reduced by about 15%).

[0169] In addition, the hybrid structure according to the present disclosure may have a slightly higher pressure drop compared to a simple divergent structure, but this can be sufficiently offset by the superior improvement in heat transfer performance provided by the zigzag structure.

[0170] In other words, it can be confirmed that the hybrid structure according to the present disclosure has a lower pressure drop and a higher critical heat flux compared to a straight tube type, and is effective in terms of the instability index (pressure oscillation amplitude).

[0172] FIG. 10 discloses examples of a comparison of heat transfer performance based on channel shape according to an embodiment.

[0173] In one embodiment, heat transfer performance (e.g., Nusselt number (Nu)) according to the Reynolds number (Re) for each channel shape can be compared. The comparison subjects may include a rectangular channel, a zigzag channel, and an optimized channel shape to which a variable shape microchannel according to the present disclosure is applied.

[0174] First, in the case of conventional rectangular channels, the Nusselt number showed a tendency to increase slightly as the Reynolds number increased. Specifically, at a Reynolds number of 58.27, the Nusselt number was approximately 5.0, and as the Reynolds number increased to approximately 75.9, it increased to approximately 5.2.

[0175] In addition, the conventional zigzag channel exhibited a higher Nusselt number compared to the rectangular channel within the same range of Reynolds numbers. At a Reynolds number of 58.27, the Nusselt number was approximately 8.2, and it was observed to increase to approximately 8.8 as the Reynolds number increased.

[0176] On the other hand, in the case of the optimized channel according to the present disclosure, it can be seen that the Nusselt number increases from approximately 8.0 to approximately 8.6 as the Reynolds number increases. This value is somewhat lower than that of a zigzag channel but remains significantly higher than that of a rectangular channel. In other words, it can be seen that the optimized channel according to the present disclosure exhibits relatively higher heat transfer performance compared to a rectangular channel.

[0178] FIG. 11 discloses one embodiment of a pressure drop comparison based on channel shape according to an embodiment.

[0179] In one embodiment, the pressure drop (Pa) according to the Reynolds number (Re) for each channel shape can be compared. The subjects of comparison may include a rectangular channel, a zigzag channel, and an optimized channel shape to which a variable shape microchannel according to the present disclosure is applied.

[0180] First, in the case of a conventional rectangular channel, the pressure drop showed a tendency to gradually decrease as the Reynolds number increased. At a Reynolds number of about 58.27, the pressure drop was about 7,000 Pa, and it can be seen that it decreased to about 5,400 Pa as the Reynolds number increased to about 75.9.

[0181] In addition, in the case of a conventional zigzag channel, a relatively high pressure drop compared to a rectangular channel was observed under the same Reynolds number conditions. At a Reynolds number of approximately 58.27, the pressure drop was approximately 7,600 Pa, and it was observed to decrease to approximately 5,900 Pa as the Reynolds number increased.

[0182] On the other hand, in the case of the optimized channel according to the present disclosure, it can be seen that the pressure drop decreases from approximately 7,100 Pa to approximately 5,500 Pa as the Reynolds number increases. This value is lower than that of a conventional zigzag channel and maintained a level somewhat higher than that of a conventional rectangular channel.

[0183] Overall, a tendency for pressure drop to decrease in all channel shapes was observed as the Reynolds number increased, and the optimized channel shape according to the present disclosure exhibits pressure drop characteristics intermediate between conventional rectangular channels and zigzag channels, confirming that the difference in flow resistance characteristics according to channel shape is clearly distinguishable.

[0184] In other words, while a typical zigzag channel exhibits the highest heat transfer and pressure drop, it can be confirmed that the optimized channel according to the present disclosure maintains high heat transfer performance while significantly lowering the pressure drop.

[0186] FIG. 12 discloses an embodiment of a radar thermal management method using a variable shape microchannel according to an embodiment.

[0187] Sensor data related to a microchannel containing cooling water that absorbs heat generated in the transmitting and receiving unit of the radar is acquired (S1010). In one embodiment, the microchannel may include a zigzag shape structure in which the flow direction of the cooling water changes along the length direction. In one embodiment, the pitch or amplitude of the zigzag shape structure may be formed according to the heat distribution of heat generated in the transmitting and receiving unit. For a detailed explanation of this, refer to the details described in FIGS. 1 to 11.

[0188] Physics residual data based on sensor data is generated (S1020). In one embodiment, sensor data is input into a physical model including fluid dynamics and heat transfer governing equations to calculate theoretical prediction values, and physics residual data corresponding to the difference between the sensor data and the theoretical prediction values ​​can be calculated. For a detailed explanation of this, refer to the details described in FIGS. 1 to 11.

[0189] Physical residual data is input into a physics-based artificial intelligence prediction model to generate microchannel state information (S1030). In one embodiment, microchannel state information may be provided to a user. For a detailed explanation of this, refer to the details described in FIGS. 1 to 11. Explanation of the symbols

[0191] 100: Radar thermal management unit 110: Radar 112: Zone 1 114: Area 2 120: Microchannel 122: Microchannel 124: Micro channel 130: Control unit 201: Main Channel 203: Micro-slit channel

Claims

Claim 1 A radar thermal management device using a variable shape microchannel, comprising: a microchannel including cooling water that absorbs heat generated in a transmitting and receiving unit of a radar; wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the longitudinal direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of heat generated in the transmitting and receiving unit, and wherein the microchannel includes a secondary flow path that moves the cooling water of a main channel included in the microchannel from a high-pressure section to a low-pressure section. Claim 2 A radar thermal management device using a variable shape microchannel according to claim 1, wherein the zigzag shape structure includes a first pitch and a first amplitude for a first region of the transmitting and receiving unit in which the amount of heat generated is greater than a threshold value based on the heat distribution, and the zigzag shape structure includes a second pitch longer than the first pitch and a second amplitude smaller than the first amplitude for a second region of the transmitting and receiving unit in which the amount of heat generated is less than a threshold value based on the heat distribution. Claim 3 A radar thermal management device using a variable shape microchannel, wherein at least one of the inner or outer edges of the zigzag shape structure has a fillet formed therein having curvature. Claim 4 delete Claim 5 A radar thermal management device using a variable shape microchannel according to claim 1, wherein the microchannel includes an inlet and an outlet through which a cooling fluid flows, and the microchannel includes a divergent structure in which the channel width of the outlet is greater than the channel width of the inlet. Claim 6 A radar thermal management device using a variable shape microchannel, further comprising: a control unit that acquires sensor data related to the microchannel, generates physics residual data based on the sensor data, and inputs the physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information. Claim 7 A radar thermal management device using a variable shape microchannel, wherein the control unit inputs the sensor data into a physical model including fluid dynamics and heat transfer governing equations to calculate a theoretical predicted value, and calculates physical residual data corresponding to the difference between the sensor data and the theoretical predicted value. Claim 8 A radar thermal management method using a variable shape microchannel, comprising: a step of acquiring sensor data related to a microchannel containing cooling water that absorbs heat generated in a transmitting and receiving unit of a radar; a step of generating physics residual data based on the sensor data; and a step of inputting the physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information; wherein the microchannel includes a zigzag shape structure in which the flow direction of the cooling water changes along the length direction, and the pitch or amplitude of the zigzag shape structure is formed according to the heat distribution of heat generated in the transmitting and receiving unit. Claim 9 A storage medium for storing a computer-executable program that acquires sensor data related to a microchannel containing cooling water that absorbs heat generated in a radar transmitting and receiving unit, generates physics residual data based on said sensor data, inputs said physics residual data into a physics-based artificial intelligence prediction model to generate microchannel state information, said microchannel includes a zigzag shape structure in which the flow direction of said cooling water changes along the length direction, and the pitch or amplitude of said zigzag shape structure is formed according to the heat distribution of said heat generated in said transmitting and receiving unit.

Images