Semiconductor power device efficient heat dissipation structure based on micro-channel liquid cooling technology and manufacturing method

Abstract

The invention discloses a semiconductor power device high-efficiency heat dissipation structure based on a micro-channel liquid cooling technology, which comprises a thermoelectric refrigeration module, a phase change material layer, a micro-channel liquid cooling module, a bionic micro-channel structure, an intelligent regulation and control module and a full-packaging module, and is characterized in that the thermoelectric refrigeration module is arranged on the surface of a hot spot area of a power device, and the periphery of the thermoelectric refrigeration module is coated with the phase change material layer; the micro-channel liquid cooling module is embedded into the packaging substrate and matched with the bionic micro-channel structure, the intelligent regulation and control module collects data in real time through the sensor group and dynamically adjusts heat dissipation parameters, and the full-packaging module achieves integrated packaging. The problems that an existing heat dissipation technology is low in efficiency, uneven in temperature distribution, insufficient in intelligent degree and the like are solved, and a more reliable and efficient heat dissipation scheme is provided for high-power-density devices.

Description

Technical Field

[0001] The present invention relates to the technical field of heat dissipation of semiconductor power devices, and more specifically, to an efficient heat dissipation structure and manufacturing method of semiconductor power devices based on microchannel liquid cooling technology. Background Art

[0002] With the rapid development of semiconductor technology, the integration and power density of power devices have been continuously improved, resulting in a significant increase in their heat flux density. The heat dissipation problem has become a key factor restricting the performance and reliability of power devices. Traditional heat dissipation technologies such as air cooling and water cooling have limitations when facing the heat dissipation requirements of high heat flux density, such as low heat dissipation efficiency, large volume, high noise, etc. Therefore, an efficient heat dissipation structure based on microchannel liquid cooling technology has become a research hotspot. Microchannel liquid cooling technology has the advantages of high heat dissipation efficiency, small volume, low noise, etc., and is suitable for high heat flux density occasions. However, the existing microchannel liquid cooling technology still has certain deficiencies, such as unreasonable flow channel design, poor coolant performance, and complex manufacturing process of heat dissipation structures.

[0003] Therefore, the existing heat dissipation technologies have low efficiency, unreasonable flow channel design, poor coolant performance, and complex manufacturing process. Summary of the Invention

[0004] In order to overcome the problems of low efficiency, unreasonable flow channel design, poor coolant performance, and complex manufacturing process of the existing heat dissipation technologies, the present invention designs an efficient heat dissipation structure and manufacturing method of semiconductor power devices based on microchannel liquid cooling technology, which can effectively solve the above technical problems.

[0005] To solve the above technical problems, the technical solution of the present invention is as follows: An efficient heat dissipation structure of semiconductor power devices based on microchannel liquid cooling technology, comprising: A thermoelectric refrigeration module, a phase change material layer, a microchannel liquid cooling module, a bionic microchannel structure, an intelligent control module, and a full encapsulation module; The thermoelectric refrigeration module is disposed on the surface of the hot spot area of the power device, and its output end is connected to the intelligent control module; the phase change material layer is coated around the thermoelectric refrigeration module; the microchannel liquid cooling module is embedded inside the power device packaging substrate, and its flow channel topology structure matches the bionic microchannel structure; the intelligent control module collects temperature and flow data in real time through a sensor group and dynamically adjusts the power of the thermoelectric refrigeration module and the flow rate of the microchannel coolant; the full encapsulation module realizes the integrated encapsulation of the thermoelectric refrigeration module, the phase change material layer, and the microchannel through 3D additive manufacturing.

[0006] Preferably, the bionic microchannel structure includes a non-uniform flow channel network based on fractal topology optimization, with the main flow channel width being 200 - 500 μm and the branch flow channel width being 50 - 150 μm; the fractal topology optimization achieves the Pareto optimum of flow resistance and thermal resistance through a genetic algorithm, and the objective function is to minimize the pressure drop and the temperature rise of the hot spot.

[0007] Preferably, the coolant of the microchannel liquid cooling module is nanofluid, which contains the following components: base liquid, nanoparticles and dispersant; its thermal conductivity ≥ 0.8 W / m·K and viscosity ≤ 3.5 mPa·s.

[0008] Preferably, the intelligent regulation module includes: Sensor group: MEMS temperature sensors arranged on the surface of the power device and at the inlet and outlet of the microchannel, and fiber optic flow sensors embedded in the inner wall of the microchannel; Control unit: A thermal load prediction model based on an LSTM neural network, with the input being historical temperature, flow data and the working state of the power device, and the output being the power adjustment coefficient of the thermoelectric refrigeration module and the set value of the microchannel flow rate; Execution unit: A piezoelectric micropump drives the coolant to flow, and a PID controller adjusts the current of the thermoelectric refrigeration module.

[0009] Preferably, the fully encapsulated module adopts a multi-layer heterogeneous material structure: Bottom layer: A diamond - copper composite substrate, on which a bionic microchannel is formed through laser-induced graphitization technology; middle layer: A paraffin / expanded graphite phase change material; top layer: An aluminum nitride ceramic encapsulation layer, integrating the thermoelectric refrigeration module and sensor wires; each layer achieves an interfacial thermal resistance ≤ 1×10⁻ 6 m²·K / W.

[0010] A manufacturing method based on an efficient heat dissipation structure includes the following steps: Processing an initial microchannel on the surface of the diamond - copper composite substrate by femtosecond laser etching; generating a fractal flow channel network using a topology optimization algorithm based on thermal - flow coupling simulation data; using micro - arc machining to correct the channel size to form a non - uniform density flow channel; Depositing a super - hydrophilic coating on the inner wall of the microchannel, and generating a SiO 2 nanowire array by plasma - enhanced chemical vapor deposition; injecting nanofluid into the microchannel through vacuum perfusion; sealing the inlet and outlet with ultraviolet curable glue; Coating indium solder paste on the hot spot area of the power device and mounting the thermoelectric refrigeration module; coating the paraffin / expanded graphite composite material around the thermoelectric refrigeration module by screen printing; achieving the encapsulation of the aluminum nitride ceramic layer and the substrate by pulsed laser bonding; Fabricate the MEMS temperature sensor on the surface of the encapsulation layer: embed the FPGA controller and burn the LSTM prediction algorithm; calibrate the flow rate-voltage relationship of the piezoelectric micropump.

[0011] Preferably, the topology optimization algorithm includes: Establish a thermal-fluid coupling simulation model, and input the heat source distribution of the power device and the initial channel parameters; Iteratively optimize the flow channel branch angle and density through the genetic algorithm, with the objective function being to minimize the flow resistance and thermal resistance; Output the Pareto optimal solution set, and select the flow channel layout with relatively small flow resistance and thermal resistance.

[0012] Preferably, the preparation of the superhydrophilic coating includes: Spray titanium dioxide sol on the inner wall of the microchannel, and form a TiO 2 nanoparticle layer through annealing; Generate surface micro-nano structures through hydrofluoric acid etching; Perform surface modification using octadecyltrichlorosilane; The training of the LSTM prediction algorithm includes: Collect the temperature and flow rate data of the power device under steady state, transient and overload conditions, and construct a time series data set; Generate training samples through the sliding window method; Train the model using the Adam optimizer; The pulse laser bonding process parameters include: The laser wavelength is in the near-infrared band; The pulse width is in the nanosecond range; Energy density gradient control: the energy density in the central region is higher than that in the edge region; After bonding, the airtightness of the interface meets the high-vacuum standard, and the shear strength meets the requirements of structural reliability.

[0013] An electronic device includes a processor and a memory, and the memory stores a program or instruction that can run on the processor. When the program or instruction is executed by the processor, the steps of the above-mentioned manufacturing method are implemented.

[0014] A readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the manufacturing method as described above are implemented.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: By integrating a thermoelectric cooling module and a phase change material layer, the present invention directly targets the hot spot area of the power device for efficient heat dissipation, improving the heat dissipation efficiency. The thermoelectric cooling module can quickly respond to temperature changes, while the phase change material layer can absorb or release a large amount of heat during temperature fluctuations, thereby stabilizing the temperature control, which is the direct reason for improving the heat dissipation efficiency. Secondly, the design of the bionic microchannel structure adopts fractal topology optimization to achieve the Pareto optimum of flow resistance and thermal resistance through a genetic algorithm. This non-uniform flow channel network can more effectively distribute the coolant, reduce the flow resistance, and improve the heat conduction efficiency at the same time, solving the problem of unreasonable flow channel design. The microchannel liquid cooling module uses nanofluid as the coolant, and its high thermal conductivity and low viscosity characteristics make the heat dissipation effect better. The selection of nanofluid improves the performance of the coolant and enhances the heat dissipation effect. The fully encapsulated module adopts 3D additive manufacturing and multi-layer heterogeneous material structure, which simplifies the manufacturing process and ensures the reliability and thermal performance of the structure. Through processing and packaging technologies, low thermal resistance connections between layers are achieved, solving the problem of complex manufacturing processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, other implementation drawings can be obtained according to the provided drawings without creative efforts.

[0017] Figure 1 It is a structural diagram of an efficient heat dissipation structure for a semiconductor power device based on microchannel liquid cooling technology; Figure 2 It is a step diagram of a manufacturing method based on the efficient heat dissipation structure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The drawings are only for illustrative purposes and should not be construed as a limitation of this patent; To better illustrate this embodiment, some components in the drawings will be omitted, enlarged, or reduced, which does not represent the size of the actual product; For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0019] The following will further explain the technical solutions of the present invention with reference to the drawings and embodiments. Embodiment 1

[0020] For an efficient heat dissipation structure of a semiconductor power device based on microchannel liquid cooling technology, please refer to Figure 1 , including: Thermoelectric cooling module, phase change material layer, microchannel liquid cooling module, bionic microchannel structure, intelligent control module and full encapsulation module; The thermoelectric cooling module is arranged on the surface of the hot spot area of the power device, and its output end is connected to the intelligent control module; the phase change material layer is coated around the thermoelectric cooling module; the microchannel liquid cooling module is embedded inside the power device packaging substrate, and its flow channel topology structure matches the bionic microchannel structure; the intelligent control module collects temperature and flow rate data in real time through a sensor group, and dynamically adjusts the power of the thermoelectric cooling module and the flow rate of the microchannel coolant; the full encapsulation module realizes the integrated encapsulation of the thermoelectric cooling module, phase change material layer and microchannel through 3D additive manufacturing.

[0021] The thermoelectric cooling module selects bismuth telluride-based thermoelectric materials, with a Seebeck coefficient of 200 μV / K, an electrical conductivity of 1×10 6 S / m, and a thermal conductivity of 1 W / m·K. The thermoelectric cooling module is closely attached to the surface of the hot spot area of the power device to ensure good heat conduction.

[0022] The phase change material layer adopts a composite phase change material of paraffin and expanded graphite. The phase change temperature of paraffin is 60 °C, and the latent heat is 200 J / g. Expanded graphite can effectively improve the thermal conductivity of the phase change material, making its overall thermal conductivity reach 1.5 W / m·K.

[0023] The depth of the microchannel of the microchannel liquid cooling module is 200 μm, the width of the main flow channel is 300 μm, the width of the branch flow channel is 100 μm, the flow channel spacing is 150 μm, the coolant uses nanofluid, the base liquid is deionized water, the nanoparticles are alumina nanoparticles (particle size 50 nm), the dispersant is polyethylene glycol, and the thermal conductivity of the nanofluid is 0.85 W / m·K, and the viscosity is 3 mPa·s.

[0024] The bionic microchannel structure is based on a non-uniform flow channel network optimized by fractal topology, and realizes the Pareto optimal of flow resistance and thermal resistance through a genetic algorithm. The objective function is to minimize the pressure drop and the temperature rise at the hot spot. The width of the main flow channel is 300 μm, and the width of the branch flow channel is 100 μm.

[0025] The sensor group of the intelligent control module arranges MEMS temperature sensors on the surface of the power device and at the inlet and outlet of the microchannel, with an accuracy of ±0.1 °C; an optical fiber flow sensor is embedded in the inner wall of the microchannel, with an accuracy of ±2%.

[0026] The control unit is a thermal load prediction model based on an LSTM neural network. The inputs are historical temperature, flow rate data, and the operating state of the power device, and the outputs are the power adjustment coefficient of the thermoelectric cooling module and the set value of the microchannel flow rate. The number of neurons in the hidden layer of the LSTM network is 128. The training dataset is the temperature and flow rate data of the power device under different working conditions, and the training period is 100 epochs.

[0027] The execution unit uses a piezoelectric micropump to drive the coolant to flow. Its maximum flow rate is 10 mL / min, and a PID controller adjusts the current of the thermoelectric cooling module, with an adjustment range of 0 - 5 A.

[0028] The fully encapsulated module adopts a multi-layer heterogeneous material structure. The bottom layer is a diamond - copper composite substrate with a thickness of 1 mm; the middle layer is a paraffin / expanded graphite phase change material with a thickness of 0.5 mm; the top layer is an aluminum nitride ceramic encapsulation layer with a thickness of 0.3 mm. The interface thermal resistance of each layer is achieved to be ≤1×10⁻ 6 m²·K / W.

[0029] The bionic microchannel structure includes a non-uniform flow channel network based on fractal topology optimization. The width of the main flow channel is 200 - 500 μm, and the width of the branch flow channel is 50 - 150 μm. The fractal topology optimization realizes the Pareto optimality of flow resistance and thermal resistance through a genetic algorithm, and the objective function is to minimize the pressure drop and the temperature rise of the hot spot.

[0030] The coolant of the microchannel liquid cooling module is nanofluid, which contains the following components: base liquid, nanoparticles, and dispersant. Its thermal conductivity is ≥0.8 W / m·K, and its viscosity is ≤3.5 mPa·s.

[0031] The intelligent control module includes: Sensor group: MEMS temperature sensors arranged on the surface of the power device and at the inlet and outlet of the microchannel, and fiber optic flow sensors embedded in the inner wall of the microchannel; Control unit: A thermal load prediction model based on an LSTM neural network. The inputs are historical temperature, flow rate data, and the operating state of the power device, and the outputs are the power adjustment coefficient of the thermoelectric cooling module and the set value of the microchannel flow rate; Execution unit: A piezoelectric micropump drives the coolant to flow, and a PID controller adjusts the current of the thermoelectric cooling module.

[0032] The fully encapsulated module adopts a multi-layer heterogeneous material structure: Bottom layer: A diamond - copper composite substrate, on which a bionic microchannel is formed through laser-induced graphitization technology; Middle layer: Paraffin / expanded graphite phase change material; Top layer: Aluminum nitride ceramic encapsulation layer, integrating the thermoelectric cooling module and sensor wires; The interface thermal resistance of each layer is achieved to be ≤1×10⁻ 6m²·K / W.

[0033] The specific implementation process is as follows: The initial microchannels are machined on the surface of the diamond-copper composite substrate by femtosecond laser etching. The laser power is 500 mW, and the scanning speed is 10 mm / s. Based on the thermal-fluid coupling simulation data, a fractal flow channel network is generated by using a topology optimization algorithm, and the micro-arc machining is used to correct the channel size to form non-uniform density flow channels.

[0034] Titanium dioxide sol is sprayed on the inner wall of the microchannels and annealed at 500 °C to form a TiO 2 particle nano-layer; surface micro-nano structures are generated by hydrofluoric acid etching; surface modification is carried out using octadecyltrichlorosilane to prepare a super-hydrophilic coating, and SiO 2 nano-wire arrays are generated by plasma-enhanced chemical vapor deposition. The nanofluid is injected into the microchannels through vacuum perfusion to ensure no air bubble residue; the inlet and outlet are sealed with ultraviolet curable glue.

[0035] Indium solder paste is coated on the hot spot area of the power device with a thickness of 50 μm, and a thermoelectric refrigeration module is mounted. The paraffin / expanded graphite composite material is coated on the periphery of the thermoelectric refrigeration module by screen printing with a thickness of 0.5 mm. Pulse laser bonding is used to realize the encapsulation of the aluminum nitride ceramic layer and the substrate. The laser wavelength is 1064 nm, the pulse width is 10 ns, and the energy density gradient is controlled: the energy density in the central area is higher than that in the edge area. After bonding, the airtightness of the interface meets the high vacuum standard, and the shear strength meets the requirements of structural reliability.

[0036] The sensor and the control unit are integrated on the surface of the encapsulation layer. An MEMS temperature sensor is fabricated by photolithography, with a size of 1 mm × 1 mm. The FPGA controller is embedded and the LSTM prediction algorithm is burned. The FPGA model is XC7Z020. The training data set of the LSTM prediction algorithm includes the temperature and flow data of the power device under steady-state, transient, and overload conditions. Training samples are generated by the sliding window method, and the model is trained using the Adam optimizer with a learning rate of 0.001.

[0037] The flow rate-voltage relationship of the piezoelectric micropump is calibrated to ensure that it can accurately output the required flow rate under the set voltage. The thermal performance of the entire heat dissipation structure is tested. When the working power of the power device is 100 W, the hot spot temperature is measured to be 85 °C, which is 20 °C lower than that of the traditional heat dissipation structure; the reliability test is carried out. After 1000 thermal cycles (from 20 °C to 100 °C), the performance of the heat dissipation structure has no obvious attenuation. Example 2

[0038] A manufacturing method based on an efficient heat dissipation structure, please refer to Figure 2 , including the following steps: Fabricate initial microchannels on the surface of a diamond - copper composite substrate by femtosecond laser etching; generate a fractal flow channel network using a topology optimization algorithm based on thermal - fluid coupling simulation data; use micro - arc machining to correct the channel dimensions to form channels with non - uniform density; Deposit a super - hydrophilic coating on the inner wall of the microchannels, and generate SiO 2 nanowire arrays by plasma - enhanced chemical vapor deposition; inject nanofluids into the microchannels through vacuum perfusion; seal the inlets and outlets with ultraviolet - curable glue; Apply indium solder paste to the hot spot area of the power device and mount a thermoelectric cooling module; coat the periphery of the thermoelectric cooling module with a paraffin / expanded graphite composite material by screen printing; achieve the encapsulation of the aluminum nitride ceramic layer and the substrate by pulsed laser bonding; Fabricate a MEMS temperature sensor on the surface of the encapsulation layer by photolithography: embed an FPGA controller and burn in the LSTM prediction algorithm; calibrate the flow - voltage relationship of the piezoelectric micropump.

[0039] The topology optimization algorithm includes: Establish a thermal - fluid coupling simulation model, and input the heat source distribution of the power device and the initial channel parameters; Iteratively optimize the channel branch angle and density through a genetic algorithm, with the objective function being to minimize the flow resistance and thermal resistance; Output the Pareto optimal solution set and select the channel layout with relatively small flow resistance and thermal resistance.

[0040] The preparation of the super - hydrophilic coating includes: Spray titanium dioxide sol on the inner wall of the microchannels and form a TiO 2 nanoparticle layer through annealing; Generate surface micro - nano structures through hydrofluoric acid etching; Perform surface modification using octadecyltrichlorosilane; The training of the LSTM prediction algorithm includes: Collect the temperature and flow data of the power device under steady - state, transient, and overload conditions to construct a time - series data set; Generate training samples through the sliding window method; Train the model using an Adam optimizer; The process parameters of the pulsed laser bonding include: The laser wavelength is in the near - infrared band; The pulse width is in the nanosecond range; Energy density gradient control: the energy density in the central region is higher than that in the edge region; After bonding, the airtightness of the interface meets the high - vacuum standard, and the shear strength meets the requirements of structural reliability.

[0041] An electronic device includes a processor and a memory. The memory stores programs or instructions that can run on the processor. When the programs or instructions are executed by the processor, the steps of the above-mentioned manufacturing method are implemented.

[0042] A readable storage medium stores programs or instructions thereon. When the programs or instructions are executed by a processor, the steps of the manufacturing method as described above are implemented.

[0043] In a specific implementation, for microchannel processing: Femtosecond laser etching uses a femtosecond laser with a wavelength of 800 nm and a pulse width of 100 fs to perform microchannel processing on the surface of a diamond-composite copper substrate. The laser power density is 1×10¹² W / cm², the scanning speed is 20 mm / s, and an initial microchannel is processed with a depth of 200 μm and a width of 300 μm.

[0044] A thermal-fluid coupling simulation model is established. The heat source distribution of the power device (heat flux density is 100 W / cm²) and the initial channel parameters (main channel width 300 μm, branch channel width 100 μm) are input. The channel branch angle and density are iteratively optimized through a genetic algorithm. The objective function is to minimize the flow resistance and thermal resistance. The population size of the genetic algorithm is set to 100, the number of iterations is 500 generations, the crossover probability is 0.8, and the mutation probability is 0.1. The Pareto optimal solution set is output, and a channel layout with relatively small flow resistance and thermal resistance is selected. The finally determined channel branch angle is 30°, and the branch density is 5 branches per square centimeter.

[0045] According to the optimized channel layout, the size of the initial microchannel is corrected using micro-arc machining technology to form non-uniform density channels. The voltage of the micro-arc machining is 50 V, the current is 1 A, and the machining time is precisely controlled at the millisecond level to ensure that the channel size accuracy reaches ±10 μm.

[0046] Coating preparation and nanofluid perfusion: Titanium dioxide sol is sprayed on the inner wall of the microchannel, and the coating thickness is 10 μm. After annealing treatment at 550 °C for 2 hours, a TiO 2 nanoparticle layer is formed. The coating is etched using a 10% hydrofluoric acid solution to generate surface micro-nano structures. The etching time is 10 minutes. Finally, the surface is modified with octadecyltrichlorosilane to make the coating have superhydrophilicity, and the water contact angle is less than 5°.

[0047] The base fluid is a mixture of ethylene glycol and water (volume ratio 1:1). Copper oxide nanoparticles (particle size 30 nm) with a mass fraction of 2% are added. Then, an appropriate amount of polyethylene glycol is added as a dispersant and ultrasonic dispersion is carried out for 30 minutes to uniformly disperse the nanoparticles in the base fluid, obtaining a nanofluid with a thermal conductivity of 0.82 W / m·K and a viscosity of 3.2 mPa·s. The nanofluid is injected into the microchannel by vacuum perfusion. First, the microchannel is evacuated to a vacuum degree of 10⁻³ Pa, and then the nanofluid is slowly injected to ensure that the fluid fully fills the flow channel and there is no bubble residue. The inlet and outlet are sealed with ultraviolet curable glue, and the curing time is 10 seconds, with the curing strength meeting the usage requirements.

[0048] Module assembly and encapsulation: Indium solder paste with a thickness of 40 μm is coated on the hot spot area of the power device, and the bismuth telluride-based thermoelectric cooling module is mounted on it. An appropriate pressure is applied to ensure good contact. The welding temperature is controlled at about 150 °C, and the welding time is 30 seconds to ensure that the thermoelectric cooling module is closely attached to the surface of the power device, with a contact thermal resistance less than 0.001 m²·K / W.

[0049] The paraffin / expanded graphite composite material is coated on the periphery of the thermoelectric cooling module by screen printing, controlling the coating thickness to be 0.6 mm, and the filling rate of the phase change material is greater than 90% to ensure its effective absorption and release of heat.

[0050] The encapsulation of the aluminum nitride ceramic layer and the substrate is realized by pulsed laser bonding technology. The laser wavelength is 1064 nm, the pulse width is 8 ns, and the energy density gradient is controlled: the energy density in the central area is 5 J / cm², and the energy density in the edge area is 3 J / cm² to ensure that the airtightness of the interface after bonding reaches the high-vacuum standard (air pressure below 10⁻ 5 Pa), and the shear strength is greater than 50 MPa, meeting the requirements of structural reliability.

[0051] Integration of the sensor and the control unit: An MEMS temperature sensor is fabricated by photolithography on the surface of the encapsulation layer. A silicon-based material is selected, and a temperature-sensitive element with a size of 800 μm × 800 μm is fabricated through microfabrication technology, with a sensitivity of 0.1 °C and an accuracy of ±0.05 °C.

[0052] Embed an FPGA controller with the model number XC7Z035, and burn the thermal load prediction algorithm based on the LSTM neural network. When training the LSTM prediction algorithm, collect the temperature and flow data of the power device under steady state (power stabilized at 80W), transient state (power rapidly changing between 50W and 150W), and overload condition (power reaching 200W), construct a time series data set, generate training samples through the sliding window method, with the window size of 10 and the step size of 1, use the Adam optimizer to train the model, with the learning rate of 0.002 and the training period of 150 epochs until the model prediction error converges to a satisfactory range.

[0053] Execute unit calibration and overall test: Calibrate the flow - voltage relationship of the piezoelectric micropump, apply different voltage values, measure the corresponding flow output, plot the calibration curve, and ensure that the piezoelectric micropump can stably output at the set flow rate with an error less than ±1% by adjusting the voltage.

[0054] Install the manufactured heat dissipation structure on the semiconductor power device and conduct thermal performance testing. When the working power of the power device is 150W, measure the hot spot temperature to be 90°C, which is 25°C lower than that of the traditional heat dissipation structure; conduct reliability testing. After 1500 thermal cycles (from 15°C to 120°C), the performance of the heat dissipation structure is stable without obvious attenuation, ensuring that it can meet the requirements of long - term stable operation.

[0055] The same or similar reference numerals correspond to the same or similar components; The terms describing the positional relationship in the drawings are for illustrative purposes only and should not be construed as a limitation of this patent; Obviously, the above - mentioned embodiments of the present invention are merely examples for clearly explaining the present invention and are not limitations on the implementation manners of the present invention. For those of ordinary skill in the art, based on the above description, other different forms of changes or modifications can be made. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. An efficient heat dissipation structure for semiconductor power devices based on microchannel liquid cooling technology, characterized in that: include: Thermoelectric cooling module, phase change material layer, microchannel liquid cooling module, bionic microchannel structure, intelligent control module and fully encapsulated module; The thermoelectric cooling module is arranged on the surface of the hot spot area of ​​the power device, and its output end is connected to the intelligent control module; the phase change material layer is coated on the periphery of the thermoelectric cooling module; the microchannel liquid cooling module is embedded in the power device packaging substrate, and its flow channel topology structure matches the bionic microchannel structure; the intelligent control module collects temperature and flow data in real time through a sensor group, and dynamically adjusts the power of the thermoelectric cooling module and the flow rate of the microchannel coolant; the fully encapsulated module realizes the integrated packaging of the thermoelectric cooling module, the phase change material layer and the microchannel through 3D additive manufacturing.

2. The high-efficiency heat dissipation structure according to claim 1, characterized in that: The bionic microchannel structure includes a non-uniform flow channel network based on fractal topology optimization, wherein the width of the main flow channel is 200-500 μm, and the width of the branch flow channel is 50-150 μm; the fractal topology optimization achieves the Pareto optimality of flow resistance and thermal resistance through a genetic algorithm, and the objective function is to minimize the pressure drop and the hot spot temperature rise.

3. The high-efficiency heat dissipation structure according to claim 2, characterized in that: The coolant of the microchannel liquid cooling module is a nanofluid, comprising the following components: a base liquid, nanoparticles and a dispersant; its thermal conductivity is ≥0.8 W / m·K, and its viscosity is ≤3.5 mPa·s.

4. The high-efficiency heat dissipation structure according to claim 1, characterized in that: The intelligent control module includes: Sensor group: MEMS temperature sensors arranged on the surface of power devices and the inlet and outlet of microchannels, and optical fiber flow sensors embedded in the inner wall of microchannels; Control unit: A heat load prediction model based on an LSTM neural network, with historical temperature, flow data and the working status of the power device as input, and the power adjustment coefficient of the thermoelectric cooling module and the microchannel flow setting value as output; Execution unit: The piezoelectric micro pump drives the coolant to flow, and the PID controller adjusts the current of the thermoelectric cooling module.

5. The efficient heat dissipation structure according to claim 1, characterized in that: The fully encapsulated module adopts a multi-layer heterogeneous material structure: Bottom layer: diamond-copper composite substrate, with bionic microchannels formed by laser-induced graphitization technology; Middle layer: paraffin / expanded graphite phase change material; Top layer: aluminum nitride ceramic packaging layer, integrating the thermoelectric cooling module and sensor wire; each layer is sintered by silver process to achieve interface thermal resistance ≤ 1×10⁻ 6 m²·K / W.

6. The method for manufacturing a high-efficiency heat dissipation structure according to any one of claims 1 to 5, characterized in that: The following steps are involved: Initial microchannels were fabricated on the surface of diamond-copper composite substrates by femtosecond laser etching; Based on the thermal-fluid coupling simulation data, a fractal flow channel network is generated using a topology optimization algorithm; Use micro-arc machining to correct the flow channel size and form a non-uniform density flow channel; Deposit a super-hydrophilic coating on the inner wall of the microchannel and generate SiO2 nanowire arrays by plasma-enhanced chemical vapor deposition; inject nanofluids into the microchannel by vacuum infusion; and seal the inlet and outlet with UV-curable adhesive; Indium solder paste is applied to the hot spot area of ​​the power device to mount the thermoelectric cooling module; paraffin wax / expanded graphite composite material is applied to the periphery of the thermoelectric cooling module by screen printing; pulse laser bonding is used to achieve the encapsulation of the aluminum nitride ceramic layer and the substrate; The MEMS temperature sensor is fabricated by photolithography on the surface of the packaging layer. The FPGA controller is embedded and the LSTM prediction algorithm is burned. The flow-voltage relationship of the piezoelectric micropump is calibrated.

7. The manufacturing method according to claim 6, characterized in that: The topology optimization algorithm includes: Establish a thermal-fluid coupling simulation model, input the heat source distribution of the power device and the initial flow channel parameters; Genetic algorithm is used to iteratively optimize the flow channel branch angle and density, and the objective function is to minimize the flow resistance and thermal resistance; Output the Pareto optimal solution set and select the flow channel layout with smaller flow resistance and thermal resistance.

8. The manufacturing method according to claim 7, characterized in that: The preparation of the super hydrophilic coating comprises: Spray titanium dioxide sol on the inner wall of the microchannel and form a TiO2 nanoparticle layer after annealing; Surface micro-nanostructures are generated by hydrofluoric acid etching; Surface modification was performed using octadecyltrichlorosilane; The training of the LSTM prediction algorithm includes: Collect temperature and flow data of power devices under steady-state, transient and overload conditions to build a time series data set; Generate training samples through sliding window method; Train the model using the Adam optimizer; The pulse laser bonding process parameters include: The laser wavelength is in the near-infrared band; The pulse width is in the nanosecond range; Energy density gradient control: the energy density in the center area is higher than that in the edge area; After bonding, the air tightness of the interface reaches the high vacuum standard, and the shear strength meets the structural reliability requirements.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the manufacturing method according to claims 6 to 8 are implemented.

10. A readable storage medium, characterized in that: The readable storage medium stores programs or instructions, and when the programs or instructions are executed by the processor, the steps of the manufacturing method as described in claims 6-8 are implemented.

Images