Heat dissipation control method and device based on flip-chip bonding

Abstract

The invention discloses a heat dissipation control method and device based on flip-chip bonding. The method comprises the following steps that S1, temperature information in the flip-chip bonding process is collected in real time; s2, pumping a nanofluid coolant through the micro through hole to take away heat generated in the chip bonding process; s3, inputting the collected temperature information into a temperature control regulation module, and adaptively optimizing the flow velocity of the coolant through fuzzy reasoning and a genetic evolution mechanism; s4, continuously receiving temperature feedback of each monitoring point, comparing a preset safety threshold value with the change trend curve in real time, and performing dynamic adaptive adjustment on a temperature control strategy; and S5, after bonding is completed, bottom filling and curing are carried out, and the chip packaging process is completed. According to the method, a high-thermal-conductivity material structure, nanofluid cooling and an intelligent optimization algorithm are fused, efficient heat management and dynamic self-adaptive temperature control in the flip chip packaging process are achieved, and the packaging reliability and the system intelligence level are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor packaging and intelligent thermal management, and in particular to a heat dissipation control method and device based on flip-chip bonding. Background Art

[0002] As integrated circuit technology continues to evolve toward higher density, miniaturization, and higher performance, flip-chip packaging has become a mainstream technology for high-end chip packaging due to its short interconnects, high I / O density, and excellent electrical performance. In actual flip-chip packaging, the bonding process typically requires multiple high-temperature operations, including thermosonics, reflow soldering, and hot pressing. These high-temperature processes effectively promote solder joint wetting and material adhesion, but they also generate significant heat within the chip, solder joints, base, and packaging materials, significantly increasing the complexity and challenges of process thermal management.

[0003] In a complex and changing thermal environment, the temperature of local areas of the chip and surrounding materials may rise sharply, causing uneven thermal expansion of the material, physical deformation, thermal stress concentration, etc. These problems are specifically manifested as:

[0004] Heat is difficult to dissipate efficiently. Existing packaging structures and material systems are limited by their own thermal conductivity, heat dissipation paths, and structural layout. They are often unable to quickly transfer the instantaneous high heat generated during the bonding process to the outside. This can cause local overheating of the chip, solder joints, and package, increasing the risk of material fatigue damage and local failure.

[0005] Thermal stress and failure risk increase. Due to differences in the thermal expansion coefficients of materials such as chips, solder, and package substrates, high-temperature cycling and rapid temperature changes can easily lead to stress concentration, which can further induce reliability defects such as solder joint cracking, chip cracking, package warping, and voids, affecting product life and long-term stability.

[0006] Changes in material properties and interface failure. Under high-temperature conditions, packaging materials such as solder and adhesives may soften, seep, and lose adhesion. Some interfaces may even delaminate, become voided, or lose structural integrity, leading to degraded electrical connections and insufficient solder joint strength, thus affecting the functional stability and safety of the entire chip package.

[0007] Currently, the traditional flip chip thermal management mode depends on passive heat dissipation structure or single high thermal conductivity material, which has problems such as complex structure, response lag, uneven heat flow distribution and limited local heat dissipation capacity. Some solutions try to use thermal interface material, micro-via cooling or diamond substrate with high thermal conductivity material, but due to the limitations of coolant control accuracy, process adaptability and dynamic adjustment capability, it is often difficult to meet the intelligent cooling needs of high integration, high heat load packaging process. In addition, the existing thermal management scheme generally lacks real-time sensing and adaptive control capability of process dynamic heat flow, and cannot dynamically respond and optimize the adjustment according to the actual temperature distribution of the chip, thermal shock event or material state change.

[0008] In recent years, with the continuous integration and development of adaptive control, intelligent optimization algorithm and advanced material process, some researches have begun to focus on improving the heat dissipation and thermal management capability of flip chip bonding process through multi-point temperature monitoring, real-time adjustment of coolant parameters, intelligent optimization control model and other means. However, existing researches are mostly limited to single parameter adjustment, static optimization or experience-driven strategies, and have not yet formed a solution that takes into account multi-objective dynamic response, real-time process adaptation and energy efficiency optimization, especially lacking an integrated dynamic thermal management control architecture for high-power, high-reliability flip chip packaging process.

[0009] Therefore, how to develop a flip bonding heat dissipation control method and device with high responsiveness, high adaptability and multi-objective optimization capability to realize accurate regulation and intelligent management of chip bonding process heat and improve the reliability, product yield and energy efficiency of packaging process has become a key technical problem to be solved in the field. SUMMARY

[0010] One object of the present application is to provide a flip bonding heat dissipation control method and device based on multi-objective adaptive genetic fuzzy optimization. The present application combines diamond high thermal conductivity base, micro-via nanofluid cooling structure and intelligent temperature control algorithm to realize real-time collection of multi-point temperature and dynamic modeling of thermal field in the chip packaging process, and to adaptively regulate the flow rate of coolant through fuzzy reasoning and multi-objective genetic optimization mechanism, effectively dealing with high heat load and transient thermal shock in flip chip bonding process. The method takes temperature error, change rate and energy consumption as targets to dynamically adjust the adjustment parameters, improve the heat dissipation efficiency and temperature control precision. The device end realizes the whole process accurate management of chip area heat through the integration of multi-channel temperature sensing, intelligent adjustment and closed-loop feedback. The present application comprehensively improves the heat dissipation capability and reliability of flip chip packaging process, has the advantages of intelligent response, dynamic optimization, high adaptability and controllable energy consumption, and is especially suitable for advanced semiconductor manufacturing scenarios with high-density integration, high-power packaging and long-term high reliability requirements.

[0011] According to the flip bonding heat dissipation control method of the present application, the following steps are included:

[0012] S1, collecting temperature information in real time during the flip-chip mounting on the heat dissipation base made of diamond material through a sensor;

[0013] S2, pumping nanofluid coolant through the micro-via to make the coolant fully contact with the U-shaped groove area during the bonding process, and take away the heat generated during the chip bonding process;

[0014] S3, inputting the collected temperature information into a temperature control adjustment module, and the temperature control adjustment module adaptively optimizes the flow rate of the coolant according to the temperature information and historical temperature control data through fuzzy reasoning and genetic evolution mechanism;

[0015] S4, continuously receiving temperature feedback from each monitoring point, comparing the preset safety threshold and the change trend curve in real time, and when the temperature of any monitoring point exceeds the limit or the change rate is abnormal, immediately re-optimizing the flow rate of the coolant and the fuzzy control parameters, repeating the process of S3, and dynamically adapting the temperature control strategy;

[0016] S5, after the bonding is completed, the underfilling and curing are implemented, and the chip packaging process is completed.

[0017] Optionally, the sensor includes at least one of an infrared temperature sensor, a thermocouple or a MEMS temperature chip, and the temperature information includes the real-time temperature of the chip and the base area and the calculated temperature change rate.

[0018] Optionally, the micro-via and the U-shaped groove are integrally etched inside the heat dissipation base, and after the nanofluid coolant is pumped in through the micro-via inlet, it flows through the U-shaped groove area, fully contacts with the heat source area at the bottom of the chip, and performs heat transfer and real-time heat dissipation.

[0019] Optionally, S3 specifically includes:

[0020] S31, calculating the calculation temperature error and the temperature change rate from the current temperature and the set target temperature as fuzzy control input variables, using a fuzzy rule base for reasoning, and outputting a coolant adjustment instruction;

[0021] S32, using a multi-objective adaptive genetic algorithm to adaptively optimize the membership function, rule weight and reasoning parameter of the fuzzy control, and the genetic algorithm has a dynamic disturbance mechanism and a global-local adaptive switching capability, and automatically adjusts the evolution strategy when the temperature control demand changes;

[0022] S33, the genetic algorithm regularly evaluates the temperature error, the temperature change rate and the energy consumption as the fitness function, and combines the sliding window backtracking near-period temperature control effect to dynamically evolve and optimize the fuzzy control parameters.

[0023] Optionally, S31 specifically includes:

[0024] S311, fuzzifying the temperature error and the temperature change rate using a triangular membership function, dividing the temperature error into five fuzzy subsets: negative large, negative small, zero, positive small, and positive large; similarly, dividing the temperature change rate into five fuzzy subsets: fast cooling, slow cooling, stable, slow heating, and fast heating, to form fuzzy input variables;

[0025] S312. Based on expert knowledge in the field of flip-chip bonding thermal management, typical process experience, and engineering simulation results, a knowledge base containing multiple If-Then type fuzzy rules is established, including the rule that if the temperature error is positive and the temperature change rate is a fast temperature increase, then the coolant flow rate is maximum;

[0026] S313, based on the fuzzy input variables, perform fuzzy reasoning on all relevant rules, calculate the activated membership and reasoning strength of each rule, and perform weighted aggregation on the reasoning results in the output dimension to obtain an output fuzzy set;

[0027] S314, performing a maximum-union aggregation on the output fuzzy sets of the activated rules, and for all rule reasoning results, within the output variable definition domain, taking the maximum membership of each point as the final membership value of each point, forming a total output fuzzy set, and performing fuzzy reasoning to obtain corresponding fuzzy output results for coolant flow rate regulation, which are also fuzzified using a triangular membership function, including minimum, small, medium, large, and maximum.

[0028] S315. Defuzzify the fuzzy output result using the centroid method to obtain the final coolant flow rate adjustment amount, which is used as the actual adjustment instruction to control the coolant flow rate pump.

[0029] Optionally, the S32 specifically includes:

[0030] S321, encoding the triangle membership function parameters and fuzzy rule weights of the fuzzy controller in a chromosome manner to form a search individual of the genetic operation, i.e., the fuzzy control parameters;

[0031] S322, constructing a weighted comprehensive fitness function including multiple indicators such as temperature error, temperature change rate, and energy consumption, and evaluating the temperature control performance of each set of fuzzy control parameters;

[0032] S323, initialize the parameter population randomly or based on the historical optimal solution, and use the basic genetic operations of selection, crossover, and mutation;

[0033] S324, introducing an adaptive dynamic disturbance operator in the population evolution process, performing directional fine tuning or disturbance on the parameter solution of local convergence or slow fitness improvement, and automatically enhancing the mutation probability and amplitude of the genetic algorithm population when the real-time monitoring of the heat flow density or temperature gradient of the chip packaging area exceeds the preset dynamic threshold;

[0034] S325, by analyzing the chip area temperature field distribution and its time sequence change rate, when detecting that the temperature distribution uniformity index or the maximum temperature difference exceeds the engineering experience threshold, the genetic algorithm automatically improves the global search weight, and when the heat field tends to be balanced and the temperature change is slow, it is switched to local fine tuning;

[0035] S326, in each generation evolution iteration, the search individual with the optimal fitness is selected and applied as the real-time working parameter group of the fuzzy controller in the subsequent temperature control process, which is specifically: using the optimal search individual to perform fuzzy reasoning on the collected temperature error and temperature change rate, and directly outputting the adjustment instruction of the coolant flow rate;

[0036] S327, the genetic algorithm adopts a sliding window backtracking strategy matched with the physical time constant of the chip packaging temperature rise and fall, dynamically records the near-cycle temperature control effect, and preferentially inherits the parameter set that performs well in the thermal process stage.

[0037] Optionally, the temperature control adjustment module continuously receives temperature feedback of each monitoring point, compares the preset safety threshold and the change trend curve in real time, and when the temperature of any monitoring point is out of limit or the change rate is abnormal, immediately reoptimizes the coolant flow rate and the fuzzy control parameter, and dynamically and adaptively adjusts the temperature control strategy.

[0038] The heat dissipation control device based on flip-chip bonding according to the embodiment of the application comprises the following structure:

[0039] The heat dissipation base table made of diamond material is provided with a plane area for chip mounting on the surface, and a micro-via and a U-shaped groove structure are integrally processed inside the chip area, the micro-via and the U-shaped groove are through, and the lower base of the base table is also a diamond structure;

[0040] The nanofluid coolant circulation comprises a plurality of pump inlets and a plurality of pump outlets;

[0041] The multi-point temperature acquisition unit is composed of a plurality of high-precision temperature sensors and is distributed in the chip area, the micro-via and the base table part;

[0042] The temperature control adjustment module comprises an intelligent control unit based on a multi-objective adaptive genetic fuzzy optimization algorithm.

[0043] The beneficial effects of the application are:

[0044] The present application greatly improves the heat dissipation and reliability of flip chip packaging process, and also realizes significant breakthroughs in energy saving, system intelligence, adaptive ability and process adaptability, especially suitable for advanced semiconductor manufacturing and high-end electronic packaging scenarios with high density, high heat load and high reliability requirements.

[0045] Through multi-point high-precision temperature acquisition unit and dynamic thermal field real-time modeling, the temperature distribution and change trend in the chip, base and cooling path can be accurately perceived, and the complex dynamic processes such as process thermal shock and non-uniform heating can be captured in time. Compared with the traditional process relying on single-point temperature measurement or fixed value control, the present application significantly enhances the adaptive monitoring ability of multi-variable and dynamic thermal environment, providing data basis for intelligent regulation and control.

[0046] In addition, the present application introduces a multi-objective adaptive genetic fuzzy optimization algorithm, which includes temperature error, change rate, energy consumption and other key performance indicators in the optimization target. It can adaptively adjust the coolant flow rate, circulation mode and regulation strategy under different process stages, thermal load mutations or external environment disturbances, and realize real-time closed-loop optimization of the temperature control system. Compared with traditional experience regulation or single-objective PID control, the present application has higher response speed, global optimization ability and anti-disturbance performance.

[0047] The device end integrates a closed-loop temperature control feedback mechanism, which can periodically or event-triggered update control parameters according to the dynamic changes of temperature acquisition results and regulation effects, realize self-learning and self-adaptation of fuzzy rules and control parameters. This mechanism effectively improves the generalization ability and long-term stability of the temperature control strategy under various conditions such as chip type, material system and process batch, further ensuring the consistency and product yield of the packaging process.

[0048] In summary, the present application not only improves the heat dissipation performance and reliability of flip chip packaging process, but also realizes significant breakthroughs in energy saving, system intelligence, adaptive ability and process adaptability, especially suitable for advanced semiconductor manufacturing and high-end electronic packaging scenarios with high density, high heat load and high reliability requirements. BRIEF DESCRIPTION OF DRAWINGS

[0049] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application. In the drawings:

[0050] Figure 1 A flowchart of a heat dissipation control method based on flip bonding is proposed for the present application;

[0051] Figure 2 A structure diagram of a heat dissipation control device based on flip-chip bonding is provided in the present application.

[0052] Figure 3 A structure diagram of a heat dissipation base of a heat dissipation control device based on flip-chip bonding is provided in the present application.

[0053] Figure 4 A structure diagram of a bonding device of a heat dissipation control device based on flip-chip bonding is provided in the present application. DETAILED DESCRIPTION

[0054] The present application will now be further described in detail with reference to the drawings. These drawings are simplified schematic diagrams and only schematically show the basic structure of the present application, and thus only show the components related to the present application.

[0055] REFERENCE Figure 1 A heat dissipation control method based on flip-chip bonding, comprising the following steps:

[0056] S1, collecting temperature information in real time during the flip-chip bonding process through a sensor when the flip-chip is mounted on a heat dissipation base made of diamond material;

[0057] S2, pumping nanofluid coolant through a micro-via hole so that the coolant is in full contact with the U-shaped groove area during the bonding process and carries away the heat generated during the chip bonding process;

[0058] S3, inputting the collected temperature information into a temperature control adjustment module, and the temperature control adjustment module optimizes the coolant flow rate through fuzzy reasoning and genetic evolution mechanism according to the temperature information and historical temperature control data;

[0059] S4, continuously receiving temperature feedback from each monitoring point, comparing the preset safety threshold and the change trend curve in real time, and when the temperature of any monitoring point exceeds the limit or the change rate is abnormal, immediately re-optimizing the coolant flow rate and fuzzy control parameters, repeating the S3 process, and dynamically adapting the temperature control strategy;

[0060] S5, after the bonding is completed, implementing underfilling and curing to complete the chip packaging process.

[0061] By breaking down the flip-chip package heat dissipation control process into five steps: real-time acquisition of diamond base temperature, microchannel injection of nanofluid coolant, fuzzy genetic intelligent temperature control, adaptive adjustment of temperature anomaly feedback, and bottom fill curing, the heat dissipation control process has engineering-level feasibility. In particular, the combination of a high thermal conductivity base and a dynamic multi-point temperature control algorithm achieves a high degree of synergy between the heat conduction path and the adjustment strategy, effectively improving the process's responsiveness to transient high heat flux and non-uniform thermal fields. This solution builds a complete thermal management chain from temperature sensing, heat extraction, intelligent optimization to process curing, breaking through the technical bottlenecks of uneven heat flux distribution and lagging parameter adjustment in traditional package heat dissipation.

[0062] In this embodiment, the sensor includes at least one of an infrared temperature sensor, a thermocouple, or a MEMS temperature chip, and the temperature information includes the real-time temperature of the chip and the base area, and the calculated temperature change rate.

[0063] In this embodiment, the sensor includes at least one of an infrared temperature sensor, a thermocouple, or a MEMS temperature chip, and is used to collect real-time temperatures at multiple points in the chip and base area. This temperature information includes the real-time temperature values ​​of the chip and base area during the bonding process, as well as the temperature change rate calculated based on the continuous temperature data collected. This provides comprehensive and accurate data support for subsequent heat dissipation regulation and intelligent temperature control strategies.

[0064] In this embodiment, the micro-through holes and U-shaped grooves are formed by integrated etching inside the heat dissipation base. After the nanofluid coolant is pumped in through the micro-through hole inlet, it flows through the U-shaped groove area and fully contacts the heat source area at the bottom of the chip to transfer heat and dissipate heat in real time.

[0065] In this embodiment, the micro-through holes and U-shaped grooves are formed by integrated etching inside the heat dissipation base. After the nanofluid coolant is pumped in through the micro-through hole inlet, it flows through the U-shaped groove area along the microchannel and fully contacts the heat source area at the bottom of the chip, realizing efficient heat transfer and real-time heat dissipation, effectively reducing local temperature rise and improving the overall heat dissipation performance of the packaging structure.

[0066] In this embodiment, S3 specifically includes:

[0067] S31, calculating the temperature error and temperature change rate obtained by comparing the current temperature with the set target temperature as fuzzy control input variables, using the fuzzy rule base for reasoning, and outputting a coolant adjustment instruction;

[0068] S32. Adaptively optimizing the membership function, rule weights, and inference parameters of the fuzzy control using a multi-objective adaptive genetic algorithm, wherein the genetic algorithm has a dynamic perturbation mechanism and global-local adaptive switching capability, and automatically adjusts the evolution strategy when the temperature control demand changes;

[0069] S33, the genetic algorithm regularly evaluates the temperature error, the temperature change rate and the energy consumption as the fitness function, combines the sliding window backtracking near period temperature control effect, and dynamically evolves and optimizes the fuzzy control parameter.

[0070] By taking the temperature error and the temperature change rate as the fuzzy control input, combining the dynamic optimization of the fuzzy controller parameter by the multi-objective adaptive genetic algorithm, the adaptive intelligent adjustment of the coolant flow rate and the circulation mode is realized. The temperature control system can not only respond to the temperature change in real time, but also continuously evolve and optimize the adjustment strategy according to the historical working conditions and the energy consumption. The scheme overcomes the limitations of traditional single parameter setting and static control, has the advantages of fast response speed, strong self-learning ability and high multi-process adaptability, and greatly improves the intelligentization of the chip packaging process and the system reliability.

[0071] In this embodiment, the S31 specifically includes:

[0072] S311, the temperature error and the temperature change rate are respectively fuzzy processed by a triangular membership function, the temperature error is divided into five fuzzy subsets of negative large, negative small, zero, positive small and positive large, and the temperature change rate is also divided into five fuzzy subsets of fast cooling, slow cooling, stable, slow heating and fast heating, to form fuzzy input variables;

[0073] S312, according to the expert knowledge in the field of flip chip bonding thermal management, typical process experience and engineering simulation results, a knowledge base containing multiple If-Then type fuzzy rules is established, which contains that the temperature error is positive large and the temperature change rate is fast heating, and then the coolant flow rate is maximum;

[0074] S313, according to the fuzzy input variables, fuzzy reasoning is performed on all related rules, the membership degree and the reasoning strength of each rule being activated are respectively calculated, the reasoning results are weighted and aggregated in the output dimension, and the output fuzzy set is obtained;

[0075] S314, the output fuzzy set of the activated rule is aggregated by the maximum-take-and method, the reasoning results of all rules are aggregated in the output variable definition domain, the maximum membership degree of each point is taken as the final membership degree value of each point, the total output fuzzy set is formed, fuzzy reasoning is performed, the fuzzy output result about the coolant flow rate adjustment which is also divided by the triangular membership function is obtained, including minimum, small, medium, large and maximum;

[0076] S315, the fuzzy output result is de-fuzzied to the final coolant flow rate adjustment amount by the centroid method, which is taken as the actual adjustment instruction to control the coolant flow rate pump.

[0077] The temperature error and temperature change rate are divided into five definite fuzzy subsets by using a triangular membership function, and a plurality of If-Then type fuzzy rules are constructed by combining the expert knowledge in the field of thermal management and process simulation, so as to realize the knowledge-based modeling of the process temperature control condition. The system performs reasoning and aggregation on all rules according to real-time fuzzy input, forms an output fuzzy set about the coolant flow rate adjustment by using the maximum-take method, and de-fuzzifies the unique coolant flow rate adjustment instruction by using the centroid method. The scheme overcomes the roughness of the traditional single temperature control, significantly improves the semantic expression of the temperature signal and the refinement of the adjustment decision, and effectively enhances the self-adaptability and intelligent response capability of the heat dissipation system to the complex process thermal field.

[0078] In the embodiment, the S32 specifically comprises:

[0079] S321, triangular membership function parameters and fuzzy rule weights of the fuzzy controller are encoded in the form of chromosomes to form search individuals for genetic operation, i.e., fuzzy control parameters;

[0080] S322, a weighted comprehensive fitness function including multiple indexes of temperature error, temperature change rate and energy consumption is constructed to evaluate the temperature control performance of each group of fuzzy control parameters;

[0081] S323, the parameter population is initialized randomly or based on the historical optimal solution, and the basic genetic operations of selection, crossover and mutation are adopted;

[0082] S324, a self-adaptive dynamic disturbance operator is introduced in the population evolution process to perform directional fine-tuning or disturbance on the parameter solution with local convergence or slow fitness improvement, and when the real-time monitoring of the heat flow density or temperature gradient of the chip packaging area exceeds the preset dynamic threshold, the mutation probability and amplitude of the genetic algorithm population are automatically enhanced;

[0083] S325, by analyzing the chip area temperature field distribution and its time sequence change rate, when the temperature distribution uniformity index or the maximum temperature difference exceeds the engineering experience threshold, the genetic algorithm automatically increases the global search weight, and when the thermal field tends to be balanced and the temperature changes slowly, the local fine-tuning is switched;

[0084] S326, in each generation evolution iteration, the search individual with the optimal current fitness is selected and applied as the real-time working parameter group of the fuzzy controller to the subsequent temperature control process, which specifically means that the optimal search individual is used for fuzzy reasoning on the collected temperature error and temperature change rate to directly output the adjustment instruction of the coolant flow rate;

[0085] S327, the genetic algorithm adopts a sliding window backtracking strategy matched with the physical time constant of the chip packaging temperature rise and fall, dynamically records the near-cycle temperature control effect, and preferentially inherits the parameter set with excellent performance in the thermal process stage by weighting.

[0086] By introducing a multi-objective adaptive genetic optimization algorithm into the intelligent temperature control system and taking fuzzy control parameters as the search and evolution objects, the system can adaptively optimize in the temperature control parameter space, significantly improving the flexibility and intelligence level of the cooling regulation strategy. Specifically, the system first encodes the triangular membership function parameters and fuzzy rule weights into chromosomes for genetic operations. By constructing a multi-objective weighted fitness function with temperature error, temperature change rate, and energy consumption as the core indicators, the temperature control performance of each set of fuzzy control parameters is comprehensively evaluated, realizing the close coupling of parameters and process thermal response.

[0087] During the genetic evolution process, an adaptive dynamic perturbation operator is introduced. When key thermal field characteristics such as chip package heat flux or temperature gradient experience sudden changes, the algorithm automatically enhances the mutation probability and amplitude, allowing the optimization process to quickly escape from local optimal traps. This mechanism ensures that the cooling regulation parameters can quickly respond to environmental dynamic changes in special packaging scenarios such as process thermal shock and transient thermal load changes, effectively improving the global search ability and process robustness of temperature control regulation.

[0088] The algorithm dynamically adjusts the global and local search weights based on the chip area temperature distribution and its change rate. When the temperature distribution uniformity is poor or the maximum temperature difference exceeds the engineering experience threshold, the system automatically increases the global search proportion to achieve rapid large-scale adjustment of the control strategy. When the thermal field tends to be stable and changes slowly, it switches to local fine-tuning to improve the precision and stability of parameter optimization, thereby achieving deep physical coupling between the thermal management strategy and the actual process state.

[0089] After each generation of evolution iteration is completed, the system selects the parameter set with the best fitness as the working parameters of the fuzzy controller for subsequent temperature control reasoning and coolant flow rate adjustment instruction generation. By introducing a sliding window backtracking mechanism matched with the chip package temperature rise and fall physical time constant, the system dynamically records and weights the near-period temperature control effect, allowing parameter sets that perform well in special thermal process stages to be given priority in inheritance and accumulation, further enhancing the system's process self-learning ability and long-term optimization ability.

[0090] Compared with traditional static parameter setting or experience rule-driven temperature control schemes, the present invention overcomes the limitations of single regulation mode response lag and parameter fixation through multi-objective fitness dynamic optimization, adaptive adjustment of search and disturbance weights, and historical performance inheritance, significantly improving the adaptability, thermal field dynamic response speed, and energy efficiency optimization ability of the temperature control system. This method exhibits outstanding advantages in terms of efficient regulation loop, strong process adaptation ability, and high system intelligence level in high-power density, high-heat flux, and complex change flip-chip packaging application scenarios.

[0091] In this embodiment, the temperature control adjustment module continuously receives temperature feedback of each monitoring point, compares the preset safety threshold and the change trend curve in real time, and re-optimizes the coolant flow rate and the fuzzy control parameter when the temperature of any monitoring point exceeds the limit or the change rate is abnormal, so as to dynamically and adaptively adjust the temperature control strategy.

[0092] The temperature control adjustment module can re-optimize the coolant flow rate and the fuzzy control parameter when the temperature of any monitoring point exceeds the limit or the change rate is abnormal, so as to dynamically and adaptively adjust the temperature control strategy, by continuously receiving temperature feedback of each monitoring point and comparing the preset safety threshold and the change trend curve in real time, thereby effectively guaranteeing the thermal environment stability and process safety in the chip packaging process.

[0093] Reference Figures 2-4 The heat dissipation control device based on flip-chip bonding according to the embodiment of the application comprises the following structure:

[0094] The heat dissipation base table made of diamond material is provided with a plane area for chip mounting on the surface, and a micro-via and a U-shaped groove structure in position with the chip area are integrally processed in the inside, the micro-via and the U-shaped groove are through, and the lower base of the base table is also a diamond structure;

[0095] The nanofluid coolant circulation comprises a plurality of pump inlets and a plurality of pump outlets;

[0096] The multi-point temperature acquisition unit is composed of a plurality of high-precision temperature sensors and is distributed in the chip area, the micro-via and the base table part;

[0097] The temperature control adjustment module comprises an intelligent control unit based on a multi-target adaptive genetic fuzzy optimization algorithm.

[0098] By integrating the high-thermal-conductivity diamond heat dissipation base table, the micro-via and U-shaped groove flow channel structure, the nanofluid coolant circulation, the multi-point temperature acquisition unit and the intelligent temperature control adjustment module, a complete flip-chip bonding heat dissipation control system architecture is constructed. The system has closed-loop control capability of heat collection-multi-point monitoring-intelligent decision-making-real-time execution-feedback optimization, supports multi-channel data fusion, dynamic thermal field perception and adaptive cooling adjustment. The system structure is modularized and the function is partitioned clearly, which is convenient for actual integration and operation. Compared with the traditional single heat dissipation and passive temperature control equipment, the system has achieved significant improvement in thermal management accuracy, response speed and high-power process adaptability.

[0099] Embodiment 1:

[0100] In order to verify the feasibility of the present application in practice, the present application is applied to the flip-chip packaging production line of a high-performance server processor chip of an integrated circuit enterprise. The production line adopts high-power-density chips and multi-level bonding processes, and the thermal load is large and the local temperature rise is acute. The traditional heat dissipation scheme is difficult to guarantee the packaging quality and long-term reliability. Each batch of the production line needs to package hundreds of flip-chip packages. The chip area is large, the interconnection density is high, and there is a significant difference in the thermal expansion coefficient between the packaging substrate and the solder material, which easily causes solder joint failure and thermal stress damage.

[0101] In the present embodiment, the flip-chip bonding-based heat dissipation control device of the present application is installed at each chip packaging station on the production line. The device includes a heat dissipation base made of diamond material, which is internally integrated with micro-via and U-shaped groove structures. A nanofluid coolant circulation system is equipped with multiple pump inlets and outlets, and high-precision temperature sensors are distributed in the chip area, micro-via and base positions to realize real-time temperature acquisition at multiple points. The temperature control and adjustment module of the device is based on a multi-objective adaptive genetic fuzzy optimization algorithm, which optimizes and adjusts multiple indexes such as temperature error, change rate and energy consumption in real time, and dynamically outputs coolant flow rate adjustment instructions.

[0102] In actual operation, the production line collects a set of temperature data every 2 seconds through high-precision temperature sensors, and the system automatically analyzes the thermal field distribution and temperature rise trend of the key areas of each chip packaging. After each chip bonding is completed, the device processes the collected historical and real-time temperature data through the temperature control and adjustment module, updates the fuzzy controller parameters, and realizes optimal closed-loop adjustment of the coolant flow rate. When a sharp rise in local area temperature or exceeding the safety threshold is detected, the system immediately increases the coolant flow rate and resets the fuzzy control parameters to achieve adaptive response to thermal shock. Through the sliding window mechanism, the system can preferentially inherit the control parameters with excellent performance, improving the generalization ability of the temperature control strategy for different chips and batches.

[0103] After packaging is completed, the chip bottom is automatically injected with filling material and solidified, further improving the structural strength and thermal stability. The system records the temperature control curve and cooling strategy of each chip throughout the process, facilitating subsequent big data analysis and process tracing.

[0104] The implementation results show that, before the deployment of the present application, on average, 7.6 out of every 100 chips on the production line fail due to local overheating, resulting in solder joint failure or packaging failure, and the highest bonding temperature rise exceeds the standard value by 12.4°C, and the system energy consumption is relatively high. After the deployment of the present application, the statistics of 30 consecutive days show that the packaging failure rate decreases to 0.8%, the solder joint failure rate decreases to 0.2%, the maximum temperature of the chip bonding area is controlled within ±2.1°C, and the cooling system energy consumption is reduced by more than 15%. Even under peak production and continuous operation conditions, the system can still achieve precise temperature control and heat dissipation under thermal shock through intelligent adjustment of the coolant, improving the chip yield and long-term reliability.

[0105] The above examples verify the practical application effect of the application under high heat load, high integration and complex heat management requirements, not only significantly improve the temperature control accuracy and heat dissipation capacity of the packaging process, but also reduce the frequency of manual intervention and energy consumption, providing an intelligent, reliable and sustainable heat management solution for the high-end chip packaging industry.

[0106] The above describes only the preferred specific embodiments of the application, but the protection scope of the application is not limited thereto, any person skilled in the art within the technical range disclosed by the application, according to the technical scheme and the inventive concept of the application, equivalent replacement or change, should be covered within the protection scope of the application.

Claims

1. A heat dissipation control method based on flip-chip bonding, characterized in that: The steps include: S1. The flip chip is mounted on a heat dissipation base made of diamond material, and the temperature information during the flip chip bonding process is collected in real time through a sensor; S2. Pumping nanofluid coolant through the micro-through hole allows the coolant to fully contact the U-shaped groove area during the bonding process, removing the heat generated during the chip bonding process; S3. Inputting the collected temperature information into a temperature control and adjustment module, which adaptively optimizes the coolant flow rate through fuzzy reasoning and genetic evolution mechanisms based on the temperature information and historical temperature control data; S4: Continuously receive temperature feedback from each monitoring point and compare the preset safety threshold and change trend curve in real time. When the temperature of any monitoring point exceeds the limit or the change rate is abnormal, immediately re-optimize the coolant flow rate and fuzzy control parameters, repeat the S3 process, and dynamically and adaptively adjust the temperature control strategy; S5. After bonding is completed, bottom filling and curing are performed to complete the chip packaging process.

2. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The sensor includes at least one of an infrared temperature sensor, a thermocouple or a MEMS temperature chip, and the temperature information includes the real-time temperature of the chip and the base area, and the calculated temperature change rate.

3. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The micro-through holes and U-shaped grooves are formed by integrated etching inside the heat dissipation base. After the nanofluid coolant is pumped in through the micro-through hole inlet, it flows through the U-shaped groove area and fully contacts the heat source area at the bottom of the chip to transfer heat and dissipate heat in real time.

4. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The S3 specifically includes: S31, calculating the temperature error and temperature change rate obtained by comparing the current temperature with the set target temperature as fuzzy control input variables, using the fuzzy rule base for reasoning, and outputting a coolant adjustment instruction; S32. Adaptively optimizing the membership function, rule weights, and inference parameters of the fuzzy control using a multi-objective adaptive genetic algorithm, wherein the genetic algorithm has a dynamic perturbation mechanism and global-local adaptive switching capability, and automatically adjusts the evolution strategy when the temperature control demand changes; S33. The genetic algorithm regularly evaluates the temperature error, temperature change rate, and energy consumption as fitness functions, and combines the sliding window to trace the recent cycle temperature control effect to perform dynamic evolutionary optimization on the fuzzy control parameters.

5. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The S31 specifically includes: S311, fuzzifying the temperature error and the temperature change rate using a triangular membership function, dividing the temperature error into five fuzzy subsets: negative large, negative small, zero, positive small, and positive large; similarly, dividing the temperature change rate into five fuzzy subsets: fast cooling, slow cooling, stable, slow heating, and fast heating, to form fuzzy input variables; S312. Based on expert knowledge in the field of flip-chip bonding thermal management, typical process experience, and engineering simulation results, a knowledge base containing multiple If-Then type fuzzy rules is established, including the rule that if the temperature error is positive and the temperature change rate is a fast temperature increase, then the coolant flow rate is maximum; S313, based on the fuzzy input variables, perform fuzzy reasoning on all relevant rules, calculate the activated membership and reasoning strength of each rule, and perform weighted aggregation on the reasoning results in the output dimension to obtain an output fuzzy set; S314, performing a maximum-union aggregation on the output fuzzy sets of the activated rules, and for all rule reasoning results, within the output variable definition domain, taking the maximum membership of each point as the final membership value of each point, forming a total output fuzzy set, and performing fuzzy reasoning to obtain corresponding fuzzy output results for coolant flow rate regulation, which are also fuzzified using a triangular membership function, including minimum, small, medium, large, and maximum. S315. Defuzzify the fuzzy output result using the centroid method to obtain the final coolant flow rate adjustment amount, which is used as the actual adjustment instruction to control the coolant flow rate pump.

6. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The S32 specifically includes: S321, encoding the triangle membership function parameters and fuzzy rule weights of the fuzzy controller in a chromosome manner to form a search individual of the genetic operation, i.e., the fuzzy control parameters; S322, constructing a weighted comprehensive fitness function including multiple indicators such as temperature error, temperature change rate, and energy consumption, and evaluating the temperature control performance of each set of fuzzy control parameters; S323, initialize the parameter population randomly or based on the historical optimal solution, and use the basic genetic operations of selection, crossover, and mutation; S324. Introducing an adaptive dynamic perturbation operator during population evolution to fine-tune or perturb parameter solutions that converge locally or improve slowly. When the heat flux density or temperature gradient in the chip packaging area exceeds a preset dynamic threshold, the mutation probability and amplitude of the genetic algorithm population are automatically enhanced. S325. By analyzing the temperature field distribution in the chip area and its time-series change rate, when it is detected that the temperature distribution uniformity index or the maximum temperature difference exceeds the engineering experience threshold, the genetic algorithm automatically increases the global search weight. When the thermal field tends to be balanced and the temperature changes slowly, it switches to local fine-tuning; S326. In each evolutionary iteration, the search individual with the best current fitness is selected and used as the real-time working parameter set of the fuzzy controller in the subsequent temperature control process. Specifically, the best search individual is used to perform fuzzy reasoning on the collected temperature error and temperature change rate, and directly output a coolant flow rate adjustment instruction; S327. The genetic algorithm uses a sliding window backtracking strategy that matches the physical time constant of chip package temperature rise and fall, dynamically records the recent cycle temperature control effect, and gives weighted priority to the parameter sets that perform well in the thermal process stage.

7. The heat dissipation control method based on flip-chip bonding according to claim 1, characterized in that: The temperature control and adjustment module continuously receives temperature feedback from each monitoring point, compares the preset safety threshold and the change trend curve in real time, and immediately re-optimizes the coolant flow rate and fuzzy control parameters when the temperature of any monitoring point exceeds the limit or the change rate is abnormal, and dynamically and adaptively adjusts the temperature control strategy.

8. A heat dissipation control device based on flip-chip bonding, which implements the heat dissipation control method based on flip-chip bonding according to any one of claims 1 to 7, characterized in that: Includes the following structure: The heat dissipation base is made of diamond material, with a flat area for chip mounting on the surface, and micro-through holes and U-shaped groove structures aligned with the chip area are integrated into the interior. The micro-through holes and U-shaped grooves are connected, and the base under the base is also a diamond structure; a nanofluid coolant circulation including a plurality of pump inlets and a plurality of pump outlets; Multi-point temperature acquisition unit, composed of multiple high-precision temperature sensors, distributed in the chip area, micro-through holes, and base parts; The temperature control module includes an intelligent control unit based on a multi-objective adaptive genetic fuzzy optimization algorithm.

Images