Liquid-cooled server controllable liquid baffle dynamic adjustment method based on genetic algorithm

Abstract

The application discloses a liquid-cooled server controllable liquid baffle dynamic adjustment method based on a genetic algorithm, and relates to the technical field of liquid-cooled service equipment.The liquid-cooled server controllable liquid baffle dynamic adjustment method based on the genetic algorithm comprises the following steps: S1, collecting liquid baffle feature, flow rate state and temperature state data and preprocessing, and constructing a standardized regulation expression data set; S2, evaluating the matching degree of the liquid baffle combination in the heat dissipation efficiency and energy consumption balance based on the data set and dynamically adjusting the opening degree; S3, analyzing angle control execution feedback characteristics and optimizing the stepping motor driving rhythm; S4, comprehensively analyzing the fluctuation amplitude of the matching degree and the feedback characteristics and dynamically correcting the regulation strategy; and S5, judging the structure adaptability based on the angle and the heat distribution mapping and adjusting the consistency constraint weight in the genetic calculation.The problems that the fixed liquid baffle structure flow field mode is statically solidified, high-heat areas are prone to insufficient cooling, low-heat areas are prone to resource waste, and the heat regulation efficiency in dynamic operation is limited are solved.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling service equipment technology, specifically to a method for dynamic adjustment of controllable baffles in liquid-cooled servers based on genetic algorithms. Background Technology

[0002] With the continuous increase in computing density and thermal power density in data centers, liquid cooling technology has gradually become a core temperature control method for high-performance servers. Controllable baffles, as key components for flow field distribution and heat exchange efficiency regulation within the liquid cooling box, directly affect the flow distribution and heat exchange balance of the coolant in different hot zones. Currently, most mainstream liquid cooling baffle adjustment methods rely on stepper motor drives combined with fixed control strategies. Flow field distribution is adjusted through preset angle opening tables or simple closed-loop feedback, using the regional average value of temperature sensors as the primary control basis, thus balancing heat dissipation efficiency and energy consumption control to a certain extent. Some solutions also introduce linkage logic based on flow velocity sensing and temperature threshold triggering, enabling baffle adjustment to correct for uneven heat dissipation or localized overheating, improving the system's adaptability under steady-state conditions.

[0003] For example, the invention patent with announcement number CN116817477B discloses a liquid-cooled air conditioning refrigeration system and its control method. This invention adds a second water circuit for separate water circulation heat exchange in the liquid-cooled air conditioning unit. When the low pressure of the refrigerant circuit is high, the second water circuit is opened. By switching on and off a proportional flow three-way valve, a second water pump and multiple solenoid valves in the second water circuit, the flow direction and flow rate of the self-circulating water in the second water circuit are controlled to reduce the inlet water temperature of the first heat exchanger, thereby reducing the low pressure of the compressor. In other words, this invention achieves low pressure regulation of the compressor, ensuring that the low pressure of the compressor is within the allowable operating range, thereby ensuring that the compressor is in a stable working state for a long time.

[0004] For example, invention patent CN113803908B discloses a liquid cooling source system and its control method. The liquid cooling source system includes a liquid circulation system for temperature control of the load and a temperature preparation system. The liquid circulation system includes a water pump and a water tank, with an electric heating element in the water tank for generating heat. The temperature preparation system includes a compressor, a refrigerant pump, a liquid receiver, a condenser, a fan, a plate heat exchanger, and multiple valves for generating cooling capacity. This liquid cooling source system can generate either cooling or heating capacity, enabling equipment to operate normally at a suitable temperature. Furthermore, this liquid cooling source system features a compact structure, a miniaturized overall design, high efficiency integration, and intelligent control. The connections between various components are simple, facilitating maintenance. It also allows for multi-mode adjustment, making it suitable for a wide range of scenarios, allowing different operating modes to be used for different situations, thus saving energy.

[0005] However, existing baffle control methods in liquid cooling systems generally suffer from insufficient dynamic sensing of the coupling characteristics of the thermal and flow fields inside the liquid cooling tank. Their adjustment process often ignores the interactive effects between multiple parameters such as baffle opening, regional flow velocity distribution, high-temperature sensor response, and control execution delay, lacking a control mechanism capable of adaptively optimizing the combined structure during multi-cycle evolution. Furthermore, traditional control strategies often rely on fixed thresholds and single feedback variables, failing to dynamically adjust control priorities and angle allocation ratios based on factors such as heat load fluctuations, local flow resistance changes, and abnormal sensor fluctuations. This can easily lead to persistent insufficient heat dissipation or excessive energy consumption in certain cooling areas.

[0006] To address the above issues, there is an urgent need for a method for dynamically adjusting the controllable baffle of a liquid-cooled server based on genetic algorithms. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a method for dynamically adjusting controllable baffles in liquid-cooled servers based on genetic algorithms. This method solves the problem that static flow field patterns in fixed baffle structures easily lead to insufficient cooling in high-heat areas and waste of resources in low-heat areas, thus limiting the efficiency of thermal regulation during dynamic operation.

[0009] Technical solution

[0010] To achieve the above objectives, this invention is implemented through the following technical solution: a dynamic adjustment method for controllable baffles in a liquid-cooled server based on a genetic algorithm, comprising: S1, collecting baffle feature data, flow velocity state data, and temperature state data, and preprocessing the collected baffle feature data, flow velocity state data, and temperature state data to construct a standardized control expression dataset; S2, evaluating the matching degree of heat dissipation efficiency and energy consumption balance of the current baffle combination based on the standardized control expression dataset, and dynamically adjusting the baffle opening configuration based on the evaluation results; S3, analyzing the current angle control execution feedback characteristics based on the standardized control expression dataset, and dynamically optimizing the stepper motor drive rhythm based on the analysis results; S4, using the evaluation results of the matching degree of heat dissipation efficiency and energy consumption balance of the baffle combination and the analysis results of the angle control execution feedback characteristics as input, analyzing the fluctuation amplitude of the baffle under different heat load conditions, and dynamically correcting the control strategy based on the analysis results; S5, determining the structural adaptability based on the spatial mapping analysis of the baffle angle and the heat distribution of the cooling area fed back by the stepper motor, and dynamically adjusting the consistency constraint weights in the genetic computation.

[0011] Further, the specific steps for collecting baffle characteristic data, flow velocity status data, and temperature status data are as follows: Collect baffle characteristic data during the baffle control execution process, including: the angle opening of each baffle and the communication feedback delay time; simultaneously calculate and record the average angle opening and angle range of the current upper and lower baffle arrays; collect flow velocity status data during the liquid cooling circuit operation, including: instantaneous flow velocity values ​​and average flow velocity values ​​in the main channel and branch circuits; collect temperature status data during the server operation cycle, including: high temperature abnormal fluctuation amplitude, heat dissipation efficiency, and target area cooling. Rate; simultaneously record the number of high-temperature sensors and the cumulative number of times the temperature of the high-temperature area continuously rises without falling back within multiple cycles, and record the cumulative cycle of continuous temperature rise without falling back in the high-temperature area as the abnormal duration count value; through the main control unit, the temperature status data of each chip area in the liquid cooling structure and the flow rate status data of the main channel and branches are standardized and input into the genetic optimization algorithm population construction logic, and iteratively search based on the fitness function with the objectives of maximizing the cooling rate, improving the heat distribution balance and minimizing the control energy consumption. Through crossover mutation and winner selection operations, the optimal angle combination data under the current cycle is obtained, and the optimal angle solution corresponding to each baffle is obtained.

[0012] Furthermore, the specific steps for preprocessing the collected baffle feature data, flow velocity status data, and temperature status data to construct a standardized control expression dataset are as follows: For the angle opening and communication feedback delay time during the baffle control execution process, extract the angle change trajectory and response time information of each baffle within the current cycle; eliminate instantaneous abnormal fluctuations using a sliding window aggregation algorithm; and characterize the overall array state using the group average and range values. For the instantaneous flow velocity values ​​and average flow velocity values ​​of the main channel and branch in the liquid cooling circuit, use local anomaly factor analysis to remove data points affected by instantaneous disturbances. For the target area cooling rate and high temperature abnormal fluctuation amplitude in the server cycle temperature status data, combine the temperature slope change trend of multiple cycles and use a time difference smoothing algorithm to process and remove falsely high data caused by sensor response jitter. Simultaneously, perform periodic deduplication and cumulative summation on the number of high temperature sensors and their corresponding abnormal continuous count values. Normalize the baffle feature data, flow velocity status data, and temperature status data after standardization processing to construct a standardized control expression dataset.

[0013] Furthermore, the specific steps for evaluating the matching degree of heat dissipation efficiency and energy consumption balance of the current baffle assembly based on the standardized control expression dataset are as follows: Instantaneous power data and cumulative running time of each stepper motor in the liquid cooling structure during the current operating cycle are collected; the power consumption of the circulating pump is calculated by combining the average flow velocity in the main channel and branches collected by the flow velocity sensor; and a normalized ratio is calculated with the corresponding heat dissipation efficiency to obtain the combined energy consumption value; the average angle opening of the current upper and lower baffle arrays, the average flow velocity in the main channel during the current cycle, and the number of high-temperature sensors are summed, and the result is incremented by one and the logarithm is taken to obtain the comprehensive flow characteristic value; the reciprocal of the baffle angle range value plus one is calculated to obtain the angle balance correction value; the ratio of the target area cooling rate to the combined energy consumption value is calculated to obtain the unit energy consumption temperature difference; the comprehensive flow characteristic value, the angle balance correction value, and the unit energy consumption temperature difference are added together to obtain the baffle control adaptation value.

[0014] Furthermore, the specific steps for dynamically adjusting the baffle opening configuration based on the evaluation results are as follows: Real-time comparison of the current baffle adjustment adaptation value with the control threshold: When the baffle adjustment adaptation value is less than or equal to the control threshold, it is determined that the current baffle combination cannot meet the current heat dissipation load requirements of the hot zone. The main control unit immediately restarts the genetic calculation process: The angle combination search range is expanded through the real-time coprocessor, a new round of rotation commands is issued to the stepper motor, and the sealed through-shaft is re-driven to rotate the target area baffle. Simultaneously, the priority of the area angle change frequency is temporarily increased. When the standard cooling range cannot be reached for two consecutive cycles, a warning command is issued by the safety monitoring unit, directly switching the corresponding baffle to the fully open state. When the baffle adjustment adaptation value is greater than the control threshold, the main control unit sets the current baffle angle configuration to a short-term frozen state, retaining only the micro-amplitude response channel. The real-time coprocessor enters a low-frequency temperature data polling mode, maintaining a weak intervention mechanism for the temperature rise of memory modules, RAID, and PCI. When the monitored temperature remains stable, the controller updates the baffle adjustment command through the communication interface with a delay.

[0015] Furthermore, the specific steps for analyzing the current angle control execution feedback characteristics based on the standardized control expression dataset are as follows: The actual mechanical resistance experienced by the stepper motor during the rotation of the baffle is collected, and the load resistance value of the current stepper motor is obtained by combining it with a torque conversion algorithm; the communication feedback delay time of the current baffle is divided by the current load resistance value of the stepper motor and then one is added to obtain the communication load correction value; the optimal angle solution of the baffle generated by the genetic algorithm in the current cycle, the communication load correction value, and the corresponding angle conversion coefficient are multiplied together to obtain the control execution angle value.

[0016] Furthermore, the specific steps for dynamically optimizing the stepper motor drive rhythm based on the analysis results are as follows: Based on the control execution angle value generated in the current cycle, the controller directly sends an angle adjustment command to the corresponding stepper motor through the communication interface, driving the sealing through shaft to drive the baffle plate to complete precise rotation; when the control execution angle value of the current cycle is greater than that of the previous cycle, it indicates that the current cooling demand is increasing, the controller maintains a high-frequency sampling mode, the real-time coprocessor shortens the temperature data update cycle, and increases the calculation weight of the corresponding area of ​​the baffle plate in the genetic algorithm to ensure that it is selected first in subsequent combinations; when the control execution angle value of the current cycle is less than that of the previous cycle, it indicates that the current cooling adjustment is stabilizing, the controller suspends new angle update requests, the stepper motor maintains the current rotation state, enters the angle locking mode, the sealing through shaft maintains a constant angle structure, reduces mechanical fatigue and reduces energy consumption; during continuous operation, when a specific area is detected to be higher than the safe temperature rise rate threshold for two consecutive detection cycles, the baffle plate of the area is forcibly triggered to switch to the fully open state, forming an emergency flow channel to ensure CPU thermal safety.

[0017] Furthermore, the specific steps for analyzing the fluctuation amplitude of the baffle plate under different heat load conditions, using the evaluation results of the matching degree between heat dissipation efficiency and energy consumption of the baffle plate assembly and the analysis results of the angle control execution feedback characteristics as input, are as follows: Add one to the current baffle plate adjustment adaptation value, take its natural logarithm, and then multiply it by the square of the value obtained by subtracting the control execution angle value from one to obtain the comprehensive adaptation correction value; square the high temperature abnormal fluctuation amplitude, divide it by the value of adding the abnormal duration count value, add one again, and take its natural logarithm to obtain the high temperature fluctuation penalty value; divide the comprehensive adaptation correction value by the high temperature fluctuation penalty value to obtain the control stability evaluation value.

[0018] Furthermore, the specific steps for dynamically correcting the control strategy based on the analysis results are as follows: Real-time comparison of the current control stability evaluation value with the stability evaluation threshold, which includes a first stability threshold and a second stability threshold: When the control stability evaluation value is less than or equal to the second stability threshold, the controller maintains the current angle configuration of the baffle plate unchanged, the control command channel enters a low-frequency refresh state, and the real-time coprocessor reduces the sampling frequency, only performing intermittent temperature sampling on the CPU and GPU regions; When the control stability evaluation value is greater than the second stability threshold and less than or equal to the first stability threshold, the controller maintains the current genetic computation... While maintaining the same rhythm, the stepper motor fine-tuning strategy is activated: the angle of the baffle is dynamically corrected, the coprocessor continues to monitor temperature fluctuations at the current frequency, and continuously evaluates the convergence of local disturbance parameters; when the control stability evaluation value is greater than the first stability threshold, the main control unit immediately restarts the baffle combination genetic optimization process, and prioritizes eliminating schemes with excessive offset angles and lagging hot zone response in the previous cycle. At the same time, it sends a synchronous full angle adjustment command to the stepper motor, triggering the safety monitoring unit in the controller to execute enhanced heat dissipation commands for high-temperature areas, and switches the local baffle to the fully open state to prevent overheating risks.

[0019] Furthermore, the specific steps of the spatial mapping analysis of the baffle angle and the heat distribution of the cooling area based on the feedback from the stepper motor, determining the structural adaptability, and dynamically adjusting the consistency constraint weight in the genetic computation are as follows: The baffle angle information fed back by the stepper motor is obtained in real time through the communication interface; the distribution of each baffle in different cooling areas is spatially mapped; combined with the current cycle's heat zone distribution data collected by the real-time processor, the controller calculates the coupling relationship between the opening angle of each baffle and the physical structure, identifying the flow efficiency in the actual flow channel; based on the angle change amplitude and flow field change trend of the previous and current cycles, and in conjunction with the core indicators of baffle adjustment adaptation value, control execution angle value, and adjustment stability evaluation value, it is determined whether there is local structural adjustment lag, angle response deviation, or slow heat dissipation feedback. When the baffle angle fluctuation amplitude of a certain cooling area is too large in the current cycle, it is determined that the structural combination has insufficient adaptability, and the structural consistency constraint weight is increased in the next round of genetic computation. Simultaneously, the regional structural state offset information is recorded as an auxiliary parameter for evolutionary population selection.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) The controllable baffle dynamic adjustment method of liquid-cooled server based on genetic algorithm introduces periodic population evolution and multi-objective fitness evaluation mechanism, so that the baffle opening combination can achieve dynamic optimization matching between heat dissipation efficiency, energy consumption balance and control stability.

[0023] (2) The controllable baffle dynamic adjustment method of the liquid-cooled server based on genetic algorithm introduces local structural state offset information as screening weight in genetic computing, so that the structural adaptability of the cooling area is continuously strengthened in multiple rounds of evolution.

[0024] (3) The controllable baffle dynamic adjustment method of the liquid-cooled server based on genetic algorithm avoids the problem of local overcooling or overheating caused by fixed threshold control by combining the real-time heat dissipation feedback trend during the angle configuration evolution process.

[0025] (4) The controllable baffle dynamic adjustment method of the liquid-cooled server based on genetic algorithm realizes timely correction of control lag and response deviation by comprehensively evaluating the angle change amplitude and flow field change trend of different cycles.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of the controllable baffle dynamic adjustment method for liquid-cooled servers based on genetic algorithms according to the present invention.

[0028] Figure 2 This is a line graph of the regulation stability assessment value involved in this invention;

[0029] Figure 3 This is a diagram showing the internal structure and baffle distribution of the liquid-cooled server involved in this invention.

[0030] Figure 4 This is a model diagram of the controllable liquid baffle involved in the present invention;

[0031] Figure 5 This is a diagram of the internal structure of the controller involved in this invention.

[0032] In the diagram, 1. Upper baffle array; 2. Lower baffle array; 3. CPU; 4. Memory module; 5. RAID; 6. PCI; 7. Controller; 8. Stepper motor; 9. Sealed through shaft; 10. Baffle; 11. Main control unit; 12. Real-time coprocessor; 13. Communication interface; 14. Safety monitoring unit; 15. Storage unit. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Please see Figures 1-5 This invention provides a technical solution: a method for dynamically adjusting a controllable baffle plate in a liquid-cooled server based on a genetic algorithm, comprising: S1, collecting baffle plate feature data, flow rate status data, and temperature status data, and preprocessing the collected baffle plate feature data, flow rate status data, and temperature status data to construct a standardized control expression dataset; S2, evaluating the matching degree of heat dissipation efficiency and energy consumption balance of the current baffle plate combination based on the standardized control expression dataset, and dynamically adjusting the opening configuration of the baffle plate 10 based on the evaluation results; S3, analyzing the current angle control execution feedback characteristics based on the standardized control expression dataset, and dynamically optimizing the driving rhythm of the stepper motor 8 based on the analysis results; S4, using the evaluation results of the matching degree of heat dissipation efficiency and energy consumption balance of the baffle plate combination and the analysis results of the angle control execution feedback characteristics as input, analyzing the fluctuation amplitude of the baffle plate 10 under different heat load conditions, and dynamically correcting the control strategy based on the analysis results; S5, determining the structural adaptability based on the spatial mapping analysis of the baffle plate 10 angle and the heat distribution of the cooling area fed back by the stepper motor 8, and dynamically adjusting the consistency constraint weights in the genetic computation.

[0035] Specifically, the steps for collecting baffle characteristic data, flow rate data, and temperature data are as follows:

[0036] The system collects baffle characteristic data during the control execution process of the baffle 10. This data includes: the real-time angle opening of each baffle 10 under the current control command, and the communication feedback delay time from the control command issuance to the feedback reception. Combined with the physical position identifier of each baffle 10, the system calculates and records the average angle opening and angle range of the upper baffle array 1 and the lower baffle array 2 to characterize the concentration and extreme difference level of the overall opening distribution. The system also collects flow velocity status data during the operation of the liquid cooling circuit. This data includes: the instantaneous flow velocity value of the main channel and branches per unit time, and the periodically statistically calculated average flow velocity value. This data, combined with the flow sensor deployment location marker data, forms a spatial flow velocity distribution record to reflect the differences in flow uniformity across different channels. Finally, the system collects temperature status data during the server's operating cycle. This data includes: the amplitude of abnormal high-temperature fluctuations, the heat dissipation efficiency under different heat dissipation paths, and the current... The cooling rate of the target area is recorded, along with the number of high-temperature sensors and the cumulative number of times the temperature in the high-temperature area continues to rise without falling back within multiple consecutive acquisition cycles. This cumulative number of cycles is defined as the abnormal persistence count value, used to characterize the lag and heat dissipation imbalance risk in the local area during the heat accumulation process. The temperature status data of each chip area in the liquid cooling structure, the flow rate status data of the main channel and branches, and the baffle characteristic data are standardized and then uniformly input into the genetic optimization algorithm population construction logic through the main control unit 11. Iterative search is carried out based on the fitness function with multiple objectives of maximizing the cooling rate, improving the heat distribution balance, and minimizing control energy consumption. In each generation of population update, crossover mutation and winner selection operations are used to generate the next round of candidate solutions. Finally, the optimal angle combination data is selected in the current running cycle to obtain the optimal angle solution corresponding one-to-one with the position of each baffle 10, providing precise adjustment parameters for subsequent execution control.

[0037] In this implementation scheme, by simultaneously collecting and fusing multi-dimensional baffle characteristic data, flow velocity state data and temperature state data, the key parameters reflecting the operating status of the liquid cooling structure are transformed into standardized input information that can be directly used in the optimization calculation of the genetic algorithm. Thus, under the constraint of the multi-objective fitness function, the optimal baffle 10 angle combination in the current cycle is accurately searched, and the dynamic balance optimization between heat dissipation efficiency, flow balance and energy consumption control is achieved.

[0038] Specifically, the collected baffle characteristic data, flow velocity data, and temperature data are preprocessed to construct a standardized control expression dataset. The specific steps are as follows: For the angle opening and communication feedback delay time during the control execution process of the baffle 10, firstly, the continuous angle change curve and corresponding response time record of each baffle 10 in the current cycle are acquired in real time through a high-speed sampling interface. A sliding window aggregation algorithm is used to locally smooth the time series to eliminate short-term fluctuations caused by mechanical rebound and communication transient anomalies. Then, the average angle opening and angle range within the group are used as key indicators to characterize the overall attitude stability and adjustment range of the array. For the instantaneous flow velocity and average flow velocity of the main channel and branches in the liquid cooling circuit, data are collected through a multi-point synchronous flow measurement unit. Local anomaly factor analysis is used to identify and eliminate disturbances caused by liquid injection in the multi-dimensional flow velocity distribution space. Anomalies caused by bubble generation and abrupt changes in the flow field are identified to ensure the continuity and reliability of the flow velocity sequence. For the cooling rate and high-temperature anomaly fluctuation amplitude in the temperature status data of the target area during the server's operation cycle, multi-cycle temperature slope change trend analysis combined with time difference smoothing algorithm is used to weaken the artificially high temperature measurement values ​​caused by sensor thermal inertia and electrical noise. At the same time, the number of high-temperature sensors and their corresponding abnormal continuous count values ​​are deduplicated and summed across cycles to obtain stable feature quantities reflecting the risk of continuous overheating. Finally, the baffle feature data, flow velocity status data and temperature status data processed by the above process are uniformly normalized to convert data with different dimensions and numerical ranges into a dimensionless unified scale, constructing a standardized regulatory expression dataset that can be directly used as input for genetic optimization, providing a highly consistent and comparable data foundation for subsequent multi-objective fitness calculations.

[0039] In this implementation scheme, before the multi-source collected baffle feature data, flow velocity state data, and temperature state data enter the genetic optimization operation, instantaneous outliers caused by mechanical vibration, flow field disturbance, and sensor noise are removed, multi-period trend fluctuations are smoothed, and the dimensions and value ranges of different physical quantities are unified. This results in the construction of a standardized regulatory expression dataset with high stability and high consistency, providing a reliable data foundation for the genetic algorithm to accurately evaluate the performance of the baffle combination under multi-objective conditions and output the optimal angle solution.

[0040] Specifically, based on the standardized control expression dataset, the evaluation of the matching degree between heat dissipation efficiency and energy consumption balance of the current baffle combination is carried out through the following steps: Instantaneous power data and cumulative running time of each stepper motor 8 in the current operating cycle of the liquid cooling structure are collected; the average flow velocity in the main channel and branches collected by the flow velocity sensor is used to calculate the power consumption of the circulating pump, and a normalized ratio is calculated with the corresponding heat dissipation efficiency to obtain the combined energy consumption value; the average angle opening of the current upper baffle array 1 and lower baffle array 2, the average flow velocity of the main channel in the current cycle, and the number of high-temperature sensors are summed, and the result is incremented by one and the logarithm is taken to obtain the comprehensive flow characteristic value; the reciprocal of the angle range of baffle 10 is calculated and incremented by one to obtain the angle balance correction value; the ratio of the cooling rate of the target area to the combined energy consumption value is calculated to obtain the temperature difference per unit energy consumption; the comprehensive flow characteristic value, the angle balance correction value, and the temperature difference per unit energy consumption are added together to obtain the baffle control adaptation value.

[0041] The formula for calculating the adaptive value of the baffle is:

[0042] ;

[0043] In the formula: This represents the average angle opening of the upper baffle array 1 and the lower baffle array 2. It is used to quantify the average rotation angle of all baffles 10 in the current combination. It is a basic parameter for evaluating the overall flow rate and guiding capacity. It comes from the mapping relationship between the angle feedback signal of the stepper motor 8 and the baffle layout structure. It represents the average flow rate of the main channel in the current cycle, which is used to quantify the actual flow rate of the coolant in the main path after the baffle plate combination is adjusted. It is a direct indicator for judging the fluid transport efficiency and is derived from the periodic average value of the real-time readings of the flow rate sensor deployed on the main liquid cooling circuit of the server. This indicates the number of high-temperature sensors, used to count the number of sensors whose temperature exceeds the temperature threshold within the current cycle. It is a direct basis for determining whether there are multiple thermal anomalies, and is a count by comparing the instantaneous temperature value collected by the temperature sensor with a fixed threshold. The angle range of the baffle plate 10 is used to represent the difference between the maximum and minimum opening of the baffle plate 10 in the current combination. It is a structural indicator for judging whether the adjustment is balanced and whether there is unilateral deflection. It is derived from the angle command values ​​of each group of baffle plates 10 issued by the controller 7. It represents the cooling rate of the target area and is used to measure the effect of the current adjustment scheme on the temperature reduction of the high-heat area. It is a core evaluation indicator that reflects the actual cooling capacity and response speed, and is derived from the time-series difference results of two consecutive temperature sensors. This represents the combined energy consumption value, used to assess the energy consumption of the liquid cooling system caused by the current baffle combination. It is an important benchmark for optimizing the energy efficiency ratio and system power consumption control, and is derived from pump load conversion calculations.

[0044] In this implementation plan, the performance of the upper baffle array 1 and the lower baffle array 2 in the current cycle is comprehensively quantified in terms of angle opening balance, flow distribution rationality, abnormal temperature distribution, structural adjustment coordination, and temperature reduction effect. By normalizing and fusion calculation of multiple key parameters such as angle, flow rate, temperature, deviation, and energy consumption, a fitness value that reflects the overall control capability and operational matching degree of the baffle combination is generated. This provides a core evaluation basis for the genetic algorithm to evaluate the merits of different angle configurations and select the optimal control scheme that meets the balance between cooling efficiency and energy consumption during the population evolution process.

[0045] Specifically, the steps for dynamically adjusting the opening configuration of the baffle plate 10 based on the evaluation results are as follows:

[0046] Real-time comparison of the current baffle adjustment adaptation value with the control threshold:

[0047] When the baffle adjustment adaptation value is less than or equal to the control threshold, it indicates that the current baffle combination is insufficient in terms of comprehensive heat dissipation capacity, flow distribution, and temperature fluctuation suppression to support the heat dissipation load requirements of the target hot area. The main control unit 11 immediately triggers a rapid restart of the genetic calculation process, expands the search range of the angle combination through the real-time coprocessor 12, and issues a new round of rotation commands to the stepper motor 8, causing the sealing through shaft 9 to drive the baffle 10 in the target area to adjust its angle. At the same time, the frequency priority of the angle change in the area is increased to high priority to speed up the response. When the detection results of two consecutive cycles do not reach the standard cooling range, that is, when the temperature of the target area drops by no less than the cooling threshold within two consecutive sampling cycles, the safety monitoring unit 14 issues a warning command and directly switches the corresponding baffle 10 to the fully open state, that is, the opening degree of the baffle 10 is ≥95%, to ensure the maximum cross-section of the cold liquid channel.

[0048] When the baffle adjustment value exceeds the control threshold, the main control unit 11 locks the current baffle 10 angle configuration to a short-term frozen state, meaning the baffle 10 angle adjustment frequency is less than or equal to once per cycle, and only responds to minute commands with an amplitude of less than 3 degrees. At this time, only the micro-amplitude response channel is retained, and the real-time coprocessor 12 enters a low-frequency temperature data polling mode to focus on monitoring the temperature rise of the memory module 4, RAID5, and PCI6 areas and maintain weak intervention control. If the temperature remains stable without significant fluctuations within multiple polling cycles, the controller 7 delays the issuance of new baffle 10 adjustment commands through the communication interface 13 to reduce the interference of frequent adjustments on system stability.

[0049] In this implementation scheme, based on the real-time comparison results of the baffle adjustment adaptation value and the control threshold, the current operating status of the baffle combination is dynamically determined, and when the heat dissipation is insufficient, the genetic calculation to expand the angle combination optimization process is quickly triggered to achieve a rapid cooling response to the target hot zone. When the heat dissipation requirements are met, the baffle 10 is configured to be locked in a low-frequency micro-adjustment mode to reduce mechanical wear and energy consumption fluctuations caused by unnecessary frequent adjustments, thereby ensuring the temperature stability of key components while taking into account the response efficiency and operational stability of the liquid cooling system.

[0050] Specifically, based on the standardized control expression dataset, the specific steps for analyzing the current angle control execution feedback characteristics are as follows: The actual mechanical resistance experienced by the stepper motor 8 during the rotation of the baffle plate 10 is collected, and the load resistance value of the current stepper motor 8 is obtained by combining it with a torque conversion algorithm; the communication feedback delay time of the current baffle plate 10 is divided by the current load resistance value of the stepper motor 8 and then one is added to obtain the communication load correction value; the optimal angle solution of the baffle plate 10 generated by the genetic algorithm in the current cycle, the communication load correction value, and the corresponding angle conversion coefficient are multiplied together to obtain the control execution angle value.

[0051] The formula for calculating the control execution angle value is:

[0052] ;

[0053] In the formula: The optimal angle solution of the baffle plate 10 generated by the genetic algorithm in the current cycle is used to reflect the optimal opening configuration calculated by the genetic algorithm in the current cycle. It is a key adjustment parameter to adapt to the heat load distribution and flow field state. The communication feedback delay time of the baffle plate 10 is used to reflect the response time between the issuance of the control command and the execution feedback. It is an important reference for evaluating the stability of the control link and is derived from the periodic time record of the feedback signal by the communication interface 13. This indicates the current load resistance value of the stepper motor 8, which reflects the resistance encountered when the motor drives the baffle plate 10 to rotate in a liquid-cooled environment. It is an important basis for adjusting the output correction amplitude and is derived from the motor torque feedback signal and resistance calculation. This represents the angle conversion coefficient, ranging from 0.5 to 2. It is used in the control execution angle calculation process to proportionally convert the optimal opening solution output by the genetic algorithm with the actual driving angle of the stepper motor 8, adapting to different baffle structures, installation positions, and the geometric characteristics of the transmission mechanism. Specifically, the calculation first determines the actual angular displacement corresponding to a unit step pulse based on the mechanical structural parameters of the baffle, including shaft length, blade radius, and installation tilt angle. Then, considering the gear ratio of the stepper motor 8 transmission chain, coupling deviation, and transmission clearance, the difference between the ideal and actual displacement is corrected to obtain the structural correction coefficient. Next, based on the coolant flow resistance characteristics in the current operating environment and the force deformation curve of the baffle 10, the angle attenuation rate caused by fluid disturbance is calculated and weighted with the structural correction coefficient to generate an angle conversion ratio adapted to the current hardware and flow conditions. When the baffle 10 has low structural rigidity and significant force deformation... The value is increased accordingly to compensate for insufficient actual angle; conversely, when the structure is stable and the transmission clearance is extremely small, and the angle response is precise, The value is set to a smaller value to maintain the true reproduction of the optimal solution in the physical execution process, thereby ensuring the accuracy and consistency of the control execution angle under different hardware and operating conditions.

[0054] In this implementation scheme, the optimal opening solution of the baffle 10 generated by the genetic algorithm in the previous cycle is dynamically corrected by combining the communication feedback delay time of the baffle 10 and the current load resistance value of the stepper motor 8, thereby calculating the target control angle that can truly reflect the physical execution state. Its function is to transform the theoretical optimal solution into an adjustment command that can be accurately implemented by the actuator, compensating for the impact of communication delay on response time and correcting the deviation of load changes on the actual rotation amplitude, ensuring the angle execution accuracy and control stability of the baffle 10 under different thermal loads and flow field conditions.

[0055] Specifically, the steps for dynamically optimizing the driving rhythm of stepper motor 8 based on the analysis results are as follows:

[0056] Based on the control execution angle value generated in the current cycle, the controller 7 sends the adjustment command to the corresponding stepper motor 8 drive module in real time through the high-speed communication interface 13, driving the sealing through shaft 9 to drive the target area baffle 10 to complete high-precision rotation positioning according to the command angle.

[0057] When the control execution angle value of the current cycle is greater than that of the previous cycle, it is determined that the current cooling demand is increasing. The controller 7 maintains a high-frequency sampling mode while issuing adjustment instructions, and the real-time coprocessor 12 shortens the temperature data update cycle. The heat load information and angle execution characteristics corresponding to the regional baffle 10 are given higher weight in the genetic algorithm calculation process to ensure that they participate in the generation of candidate solutions in subsequent angle combination optimization.

[0058] When the control execution angle value of the current cycle is less than that of the previous cycle, it is determined that the current cooling adjustment process is stabilizing. The controller 7 suspends the new angle update request, and the stepper motor 8 maintains its existing rotation position and enters the angle locking mode. The sealed through shaft 9 maintains a constant opening structure to reduce the frequency of the mechanism's movement, delay the fatigue of mechanical parts, and reduce additional energy consumption.

[0059] During continuous operation, if the temperature monitoring module detects that the temperature rise rate of a certain cooling area exceeds the safety threshold for two consecutive detection cycles, the controller 7 will bypass the genetic calculation and directly trigger the emergency channel strategy, forcibly switching the corresponding area baffle 10 to the fully open state, forming an unobstructed liquid flow channel, and providing immediate heat dissipation protection for the CPU3 and its high heat-sensitive components.

[0060] In this implementation scheme, based on the control execution angle value obtained by real-time calculation, the baffle plate 10 is dynamically driven to achieve precise adjustment and state maintenance at different cooling demand stages. This improves the temperature acquisition frequency and optimizes the calculation priority when the cooling load increases, ensuring the timeliness and targeted nature of the heat dissipation response. When the cooling state tends to be stable, the angle locking mode reduces the wear and energy consumption of mechanical parts. In the emergency situation of excessive temperature rise, it quickly switches to the fully open state to establish an efficient flow path to ensure the thermal safety of CPU3 and key components.

[0061] Specifically, using the evaluation results of the heat dissipation efficiency and energy consumption balance matching degree of the baffle assembly and the analysis results of the angle control execution feedback characteristics as input, the fluctuation amplitude of the baffle 10 under different heat load conditions is analyzed. The specific steps are as follows: add one to the current baffle regulation adaptation value, take the natural logarithm, and then multiply it by the square value after subtracting the control execution angle value to obtain the comprehensive adaptation correction value; take the square of the high temperature abnormal fluctuation amplitude, divide it by the value of adding the abnormal duration count value, add one again, and take the natural logarithm to obtain the high temperature fluctuation penalty value; divide the comprehensive adaptation correction value by the high temperature fluctuation penalty value to obtain the regulation stability evaluation value.

[0062] The formula for calculating the stability assessment value is:

[0063] ;

[0064] In the formula, Indicates the adjustment and adaptation value of the baffle plate; Indicates the control angle value; This indicates the amplitude of abnormal high-temperature fluctuations, reflecting the degree of continuous and drastic fluctuations in temperature readings. It is a characterization of thermal instability and originates from the statistical anomalies in the slope changes of temperature sensor data. This represents the abnormal duration count, used to record the cumulative number of times the temperature in a high-temperature area has risen continuously without falling back over multiple cycles. It serves as an indicator signal for judging the failure trend of the current control scheme and originates from the logic judgment of comparing temperature with threshold.

[0065] In this implementation example, the baffle adjustment adaptation value of Case 1 is set to 1.20, the control execution angle value is 0.15, the high temperature abnormal fluctuation amplitude is 0.10, and the abnormal duration count value is 0.00;

[0066] In Case 2, the baffle adjustment adaptation value was set to 0.80, the control execution angle value was 0.30, the high temperature abnormal fluctuation amplitude was 0.20, and the abnormal duration count value was 1.00.

[0067] In Case 3, the baffle adjustment adaptation value was set to 1.80, the control execution angle value was 0.10, the high temperature abnormal fluctuation amplitude was 0.15, and the abnormal duration count value was 0.50.

[0068] In Case 4, the baffle adjustment adaptation value was set to 0.60, the control execution angle value was 0.55, the high temperature abnormal fluctuation amplitude was 0.25, and the abnormal duration count value was 2.00.

[0069] In Case 5, the baffle adjustment adaptation value was set to 0.30, the control execution angle value was 0.20, the high temperature abnormal fluctuation amplitude was 0.60, and the abnormal duration count value was 1.00.

[0070] In Case 6, the baffle adjustment adaptation value was set to 2.00, the control execution angle value was 0.40, the high temperature abnormal fluctuation amplitude was 0.30, and the abnormal duration count value was 0.50.

[0071] In Case 7, the baffle adjustment adaptation value was set to 1.00, the control execution angle value was 0.35, the high temperature abnormal fluctuation amplitude was 0.50, and the abnormal duration count value was 0.20. The control stability assessment value for each case was calculated, as shown in Table 1.

[0072] It should be noted that some parameters mentioned in this specification (such as baffle adjustment adaptation value, control stability assessment value, abnormal duration count value, etc.) are dimensionless calculation indicators obtained through normalization. The role of such indicators is to provide a relative comparison basis in the control algorithm, rather than an absolute measurement of physical quantities. Therefore, specific units are not marked in the embodiments.

[0073] Table 1. Data Table of Regulation and Stability Assessment Values

[0074]

[0075] like Figure 2 As shown in Table 1, this is a line graph of the regulation stability assessment value provided in this application example. Figure 2 As can be seen, Case 1 has the highest stability assessment value, indicating that it has a high degree of matching between the baffle adjustment adaptation value and the control execution angle value, while the high temperature abnormal fluctuation amplitude is low and there is no abnormal continuous count. This reflects that the baffle combination has strong heat dissipation stability and timely adjustment effect under the current operating cycle, and its angle configuration can be prioritized to ensure the stable operation of the heat load area. Case 5 has the lowest stability assessment value. Although the baffle adjustment adaptation value and the control execution angle value are both at a low level, the high temperature abnormal fluctuation amplitude is large and the abnormal continuous count value is high, resulting in low overall stability. Its priority in the genetic calculation iteration will be automatically reduced, and it will be retained as a secondary adjustment object to reduce resource waste and ineffective adjustment, and improve the overall liquid cooling loop adjustment efficiency. The line graph of the stability assessment value can intuitively present the stability distribution characteristics of each case under different operating conditions. The higher the assessment value, the more suitable the baffle combination is to be used first in dynamic heat load scenarios to achieve precise control and stable heat dissipation.

[0076] Specifically, the steps for dynamically adjusting the control strategy based on the analysis results are as follows:

[0077] The current control stability assessment value is compared with the stability assessment threshold in real time. The stability assessment threshold consists of a first stability threshold and a second stability threshold, which correspond to different control response levels, so as to realize hierarchical management of the liquid-cooled baffle at each stage of operation.

[0078] When the stability assessment value is less than or equal to the second stability threshold, it is determined that the current operating state is in a highly stable range. The controller 7 directly maintains the current angle configuration of the baffle 10 unchanged, and the control command channel switches to low-frequency refresh mode to reduce the number of control signals sent to extend the life of the mechanism. At the same time, the real-time coprocessor 12 actively reduces the sampling frequency and performs intermittent temperature sampling only for the high heat-sensitive areas including the CPU3 and GPU to ensure that the core components still have the safety monitoring capability under low intervention.

[0079] When the stability assessment value is greater than the second stability threshold and less than or equal to the first stability threshold, it indicates that the current operation is in a controllable but slightly fluctuating range. The controller 7 maintains the current genetic calculation rhythm unchanged, and at the same time starts the fine-tuning strategy of the stepper motor 8 to make small dynamic corrections to the angle of the baffle plate 10 based on the temperature and flow rate change trends, so as to avoid the efficiency decrease caused by environmental disturbances and local heat load changes. At this time, the coprocessor maintains a medium frequency to monitor temperature fluctuations and continuously performs convergence evaluation on local disturbance parameters to provide a refined initial solution for the next round of calculation.

[0080] When the stability assessment value exceeds the first stability threshold, it indicates that a high-fluidity state requiring rapid intervention has been entered. The main control unit 11 immediately restarts the baffle combination genetic optimization process, prioritizing the elimination of combination schemes that have excessive offset angles, delayed response, and insufficient cooling in the previous cycle. At the same time, it sends a synchronous full-angle adjustment command to the stepper motor 8 to achieve large-range rapid adjustment. During this process, the controller 7, in conjunction with the safety monitoring unit 14, implements an enhanced heat dissipation strategy for the high-temperature area, directly switching the relevant baffles to the fully open state to form an emergency flow channel that takes effect immediately, preventing the further expansion of the risk of local overheating.

[0081] In this implementation scheme, the adjustment rhythm and response strategy of the baffle plate 10 are dynamically matched based on the graded judgment of the control stability assessment value. By setting the first stability threshold and the second stability threshold, the operating state is divided into three categories: high stability zone, slight fluctuation zone and high fluctuation zone. Different control methods such as low frequency maintenance, slight correction and rapid full adjustment are adopted in different zones. In this way, while ensuring heat dissipation safety, energy consumption control and mechanical life extension are taken into account, ensuring the operating stability and control efficiency of the liquid cooling structure under different heat load conditions.

[0082] Specifically, based on the spatial mapping analysis of the angle of the baffle plate 10 fed back by the stepper motor 8 and the heat distribution in the cooling area, the structural adaptability is determined and the consistency constraint weights are dynamically adjusted in the genetic computation. The specific steps are as follows:

[0083] The controller 7 obtains the angle information of the baffle plate 10 fed back by the stepper motor 8 in real time through the communication interface 13, and performs precise spatial mapping on the physical distribution position of each baffle plate 10 in different cooling areas. It then aligns the baffle plate 10 with the current cycle hot zone distribution data collected by the real-time processor in multiple dimensions. Based on this, the controller 7 calculates the coupling relationship between the opening angle of each baffle plate 10 and the physical structure of the liquid cooling circuit, thereby evaluating its flow efficiency in the actual flow channel and its impact on the local heat conduction path.

[0084] By combining the angle change amplitude of the previous cycle and the current cycle, the trajectory of the liquid flow distribution change, and the trend of flow field disturbance, and by linking them with the core operating indicators such as the baffle adjustment adaptation value, the control execution angle value, and the control stability evaluation value, a comprehensive judgment is made on whether there are problems such as local structural adjustment lag, angle response deviation, and slow heat dissipation feedback.

[0085] When the angle fluctuation of the baffle plate 10 in a certain cooling zone exceeds the stable operating tolerance range during the current cycle, the control logic will determine that the structural combination of that zone is not suitable for the current load, and will increase the weight of that zone in the structural consistency constraints during the subsequent genetic calculation process to prioritize the generation of angle combination schemes with higher stability. At the same time, the controller 7 will archive the structural state offset information of that zone as an auxiliary parameter input for evolutionary population screening and fitness evaluation.

[0086] like Figure 3 The diagram shows the internal structure and baffle distribution of the liquid-cooled server involved in this invention, illustrating the relative layout of the core hardware and liquid cooling control components within the chassis. The upper baffle array 1 and lower baffle array 2 are located in different cooling zones, arranged laterally along the liquid cooling channel. The flow distribution of coolant in different zones is controlled by rotating the array, thus achieving targeted heat dissipation. The CPU 3 is located in the downstream high-heat zone to receive priority cooling when the liquid coolant passes through. Multiple memory modules 4 are distributed in the central cooling area, and their uniform cooling is achieved under the influence of the coolant flow regulated by the upper baffle array 1. The RAID 5 and PCI 6 expansion interface modules are located in the bottom area, with the flow field regulated by the lower baffle array 2 to ensure temperature stability under high load conditions. The overall structure, through the layered arrangement of the upper baffle array 1 and lower baffle array 2, achieves thermal isolation and flow field guidance for different hardware components, facilitating dynamic optimization of the liquid cooling distribution strategy and improving the server's thermal management efficiency and stability under varying operating conditions.

[0087] like Figure 4 The diagram shows a model of the controllable baffle plate involved in this invention, illustrating the structural composition and key component distribution of a single adjustment unit in a liquid-cooled server. The controller 7 receives adjustment commands from the main control unit 11 and the real-time coprocessor 12, converting them into precise execution signals. The stepper motor is the core driving component for fine adjustment of the baffle plate 10's opening angle, achieving high-resolution rotational control through subdivided stepping. The sealed through-shaft 9 penetrates the wall of the liquid-cooling channel, transmitting the rotational torque of the stepper motor 8 to the baffle plate 10 while ensuring no liquid leakage. The baffle plate 10 body is directly located in the coolant flow path. By changing its rotation angle, the cross-sectional area of ​​the flow channel is adjusted, thereby distributing the liquid flow to different heat zones and achieving dynamic heat dissipation optimization. The overall structure balances sealing performance, mechanical stability, and angle adjustment accuracy, enabling the liquid cooling system to flexibly adjust the coolant flow distribution according to real-time heat load changes during operation.

[0088] like Figure 5The diagram shows the internal structure of the controller involved in this invention, illustrating the hardware composition and functional division of the core control unit during the controllable baffle adjustment process of the liquid-cooled server. The main control unit 11 is the core of the entire controller 7, responsible for parsing externally input adjustment strategy commands, executing genetic algorithm calculations, and issuing angle adjustment commands to the actuators. The real-time coprocessor 12 processes high-frequency acquired temperature data, flow field parameters, and control stability evaluation values, enabling parallel computing and rapid response to ensure the real-time performance of the adjustment process. The communication interface 13 handles data interaction with the liquid-cooling system's sensor network, the stepper motor 8 drive module, and the upper-level management platform, supporting high-speed data transmission and control signal issuance. The safety monitoring unit 14 is responsible for real-time detection of abnormal high temperatures, abnormal flow rates, and mechanical faults during operation, triggering emergency protection measures immediately upon detection of a risk. The storage unit 15 stores operating data, genetic algorithm population information, and configuration parameters, providing data support for long-term optimization and adjustment strategy backtracking. The overall structure ensures that the controller 7 achieves precise, efficient, and traceable dynamic baffle adjustment control under high load and low latency conditions.

[0089] In this implementation scheme, by collecting the baffle angle information fed back by the stepper motor 8 in real time and performing spatial mapping, the opening degree of the baffle in different cooling areas is linked with the physical structure of the liquid cooling circuit and real-time heat zone distribution data for analysis, thereby dynamically identifying the flow efficiency and local flow field adaptability. When the baffle adjustment adaptation value, control execution angle value, and adjustment stability evaluation value are combined to find that the angle fluctuation in a certain area is too large and the adjustment is lagging, the system will increase the structural consistency constraint weight and optimize the angle combination scheme in the genetic calculation, effectively breaking the problem of static solidification of the flow field mode of the fixed baffle structure, avoiding the risk of temperature rise in high-heat areas due to insufficient cooling, and preventing the waste of liquid cooling resources in low-heat areas, thereby significantly improving the thermal control efficiency and response flexibility during dynamic operation.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0091] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm, characterized in that: include: S1. Collect baffle feature data, flow velocity state data, and temperature state data, and preprocess the collected baffle feature data, flow velocity state data, and temperature state data to construct a standardized regulation expression dataset; S2, based on the standardized regulation expression dataset, evaluate the matching degree of the current baffle combination in terms of heat dissipation efficiency and energy consumption balance, and dynamically adjust the opening configuration of the baffle (10) based on the evaluation results; S3, based on the standardized control expression dataset, analyze the current angle control execution feedback characteristics, and dynamically optimize the stepper motor (8) driving rhythm based on the analysis results; S4, taking the evaluation results of the matching degree of heat dissipation efficiency and energy consumption balance of the baffle assembly and the analysis results of angle control execution feedback characteristics as input, analyze the fluctuation range of the baffle (10) under different heat load conditions, and dynamically correct the control strategy based on the analysis results; S5. Based on the spatial mapping analysis of the angle of the baffle plate (10) and the heat distribution of the cooling area fed back by the stepper motor (8), the structural adaptability is determined and the consistency constraint weight is dynamically adjusted in the genetic calculation.

2. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for collecting the baffle plate characteristic data, flow velocity data, and temperature data are as follows: Collect baffle characteristic data during the control execution process of baffle (10). The baffle characteristic data includes: the angle opening of each baffle (10) and the communication feedback delay time. At the same time, calculate and record the average angle opening and the angle range of the current upper baffle array (1) and lower baffle array (2). Collect flow velocity status data during the operation of the liquid cooling circuit. The flow velocity status data includes: instantaneous flow velocity values ​​and average flow velocity values ​​in the main channel and branch channels. Collect temperature status data during the server's operation cycle. The temperature status data includes: the amplitude of abnormal high temperature fluctuations, heat dissipation efficiency, and the cooling rate of the target area. At the same time, record the number of high temperature sensors and the cumulative number of times the temperature of the high temperature area has risen continuously without falling back in multiple cycles. The cumulative period of continuous temperature rise without falling back in the high temperature area is recorded as the abnormal duration count value. The temperature status data of each chip area in the liquid cooling structure and the flow rate status data of the main channel and the branch are standardized and input into the genetic optimization algorithm population construction logic through the main control unit (11). Based on the fitness function with the goal of maximizing the cooling rate, improving the heat distribution balance and minimizing the control energy consumption, iterative search is performed. The optimal angle combination data under the current cycle is obtained through crossover mutation and winner selection operations, and the optimal angle solution corresponding to each baffle (10) is obtained.

3. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for preprocessing the collected baffle feature data, flow rate data, and temperature data to construct a standardized regulatory expression dataset are as follows: For the angle opening and communication feedback delay time during the control execution of the baffle (10), the angle change trajectory and response time information of each baffle (10) in the current cycle are extracted, and the instantaneous abnormal fluctuations are eliminated by the sliding window aggregation algorithm. The overall state of the array is characterized by the average value and the range within the group. For the instantaneous and average flow rates of the main and branch channels in the liquid cooling circuit, the local anomaly factor analysis method is used to remove data points affected by instantaneous disturbances. The target area cooling rate and high temperature anomaly fluctuation amplitude in the temperature status data of the server within the cycle are processed by time difference smoothing algorithm in combination with the temperature slope change trend of multiple cycles to remove false high data caused by sensor response jitter. At the same time, the number of high temperature sensors and their corresponding abnormal continuous count values ​​are periodically deduplicated and cumulatively summed. The baffle feature data, flow rate data, and temperature data after standardization were normalized to construct a standardized regulation expression dataset.

4. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for evaluating the matching degree between heat dissipation efficiency and energy consumption balance of the current baffle assembly based on the standardized regulatory expression dataset are as follows: The instantaneous power data and cumulative running time of each stepper motor (8) of the liquid cooling structure in the current operating cycle are collected. The power consumption of the circulating pump is calculated by combining the average flow velocity in the main channel and branch collected by the flow velocity sensor and the normalized ratio is calculated with the corresponding heat dissipation efficiency to obtain the combined energy consumption value. The average angle opening of the current upper baffle array (1) and lower baffle array (2), the average main channel flow velocity of the current cycle, and the number of high temperature sensors are summed, and one is added to the result and the logarithm is taken to obtain the comprehensive flow characteristic value. Calculate the reciprocal of the angle range of the baffle (10) plus one to obtain the angle balance correction value; Calculate the ratio of the cooling rate of the target area to the combined energy consumption value to obtain the temperature difference per unit energy consumption; The combined flow characteristic value, angle balance correction value, and unit energy consumption temperature difference are added together to obtain the baffle plate adjustment adaptation value.

5. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for dynamically adjusting the baffle opening configuration based on the evaluation results are as follows: Real-time comparison of the current baffle adjustment adaptation value with the control threshold: When the baffle plate adjustment adaptation value is less than or equal to the control threshold, it is determined that the current baffle plate combination cannot meet the heat dissipation load requirements of the current hot zone. The main control unit (11) immediately restarts the genetic calculation process: the angle combination search range is expanded through the real-time coprocessor (12), a new round of rotation command is issued to the stepper motor (8), and the sealing through shaft (9) is driven again to drive the baffle plate (10) in the target area to rotate. At the same time, the priority of the area angle change frequency is temporarily increased. When the standard cooling range cannot be reached for two consecutive cycles, the safety monitoring unit (14) issues a warning command to directly switch the corresponding baffle plate (10) to the fully open state. When the baffle adjustment value is greater than the control threshold, the main control unit (11) sets the current baffle (10) angle configuration to a short-term frozen state, retaining only the micro-amplitude response channel. The real-time coprocessor (12) enters the low-frequency temperature data polling mode, maintaining a weak intervention mechanism for the temperature rise of memory module (4), RAID (5) and PCI (6). When the monitored temperature remains at a stable level, the controller (7) updates the baffle (10) adjustment command through the communication interface (13) with a delay.

6. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for analyzing the feedback features of the current angle control execution based on the standardized control expression dataset are as follows: The actual mechanical resistance experienced by the stepper motor (8) during the process of driving the baffle plate (10) to rotate is collected, and the load resistance value of the current stepper motor (8) is obtained by combining the torque conversion algorithm. The communication load correction value is obtained by dividing the current communication feedback delay time of the current baffle (10) by the current load resistance value of the stepper motor (8) and then adding one. The optimal angle solution of the baffle (10) generated by the genetic algorithm in the current cycle, the communication load correction value, and the corresponding angle conversion coefficient are multiplied together to obtain the control execution angle value.

7. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for dynamically optimizing the driving rhythm of the stepper motor (8) based on the analysis results are as follows: Based on the control execution angle value generated in the current cycle, the controller (7) directly sends an angle adjustment command to the corresponding stepper motor (8) through the communication interface (13), driving the sealing through shaft (9) to drive the baffle plate (10) to complete the precise rotation: When the control execution angle value of the current cycle is greater than that of the previous cycle, it indicates that the current cooling demand is increasing. The controller (7) maintains a high-frequency sampling mode, the real-time coprocessor (12) shortens the temperature data update cycle, and increases the calculation weight of the corresponding area of ​​the baffle (10) in the genetic algorithm to ensure that it is selected first in subsequent combinations. When the current cycle's control execution angle value is less than the previous cycle, it indicates that the current cooling adjustment is stabilizing. The controller (7) postpones the new angle update request, the stepper motor (8) maintains the current rotation state, enters the angle locking mode, and the sealed through shaft (9) maintains a constant angle structure, reducing mechanical fatigue and lowering energy consumption. During continuous operation, when a specific area is detected to be above the safe temperature rise rate threshold for two consecutive detection cycles, the area baffle (10) is forcibly triggered to switch to the fully open state, forming an emergency flow channel to ensure the thermal safety of the CPU (3).

8. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for analyzing the fluctuation range of the baffle (10) under different heat load conditions, using the evaluation results of the matching degree between heat dissipation efficiency and energy consumption of the baffle assembly and the analysis results of the angle control execution feedback characteristics as input, are as follows: Add one to the current baffle adjustment adaptation value, take the natural logarithm, and then multiply it by the square of one minus the control execution angle value to obtain the comprehensive adaptation correction value; The high temperature fluctuation penalty value is obtained by squaring the amplitude of the abnormal high temperature fluctuation, dividing it by one, adding the abnormal duration count value, and then adding one and taking the natural logarithm. Divide the comprehensive adaptation correction value by the high temperature fluctuation penalty value to obtain the regulation stability assessment value; The formula for calculating the stability assessment value is: ； In the formula, Indicates the adjustment and adaptation value of the baffle plate; Indicates the angle value to be controlled; This indicates the magnitude of abnormal fluctuations in high temperature. This indicates the duration of the abnormality count.

9. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for dynamically adjusting the control strategy based on the analysis results are as follows: Real-time comparison of the current regulation stability assessment value with the stability assessment threshold, which includes a first stability threshold and a second stability threshold: When the stability assessment value is less than or equal to the second stability threshold, the controller (7) maintains the current angle configuration of the baffle (10) unchanged, the control command channel enters the low frequency refresh state, and the real-time coprocessor (12) reduces the sampling frequency and only performs intermittent temperature sampling on the CPU (3) and GPU areas. When the stability assessment value is greater than the second stability threshold and less than or equal to the first stability threshold, the controller (7) maintains the current genetic calculation rhythm and activates the stepper motor (8) to fine-tune the strategy: dynamically correct the angle of the baffle (10), the coprocessor continues to monitor the temperature fluctuation at the current frequency, and continuously evaluates the convergence of local disturbance parameters. When the stability assessment value exceeds the first stability threshold, the main control unit (11) immediately restarts the genetic optimization process of the baffle combination and prioritizes eliminating schemes with excessive offset angle and lagging hot zone response in the previous cycle. At the same time, it sends a synchronous full angle adjustment command to the stepper motor (8) to trigger the safety monitoring unit (14) in the controller (7) to execute the enhanced heat dissipation command for the high temperature area and switch the local baffle (10) to the fully open state to prevent overheating risk.

10. The method for dynamic adjustment of controllable baffles in a liquid-cooled server based on a genetic algorithm according to claim 1, characterized in that: The specific steps for determining structural adaptability and dynamically adjusting the consistency constraint weights in the genetic computation based on the spatial mapping analysis of the baffle angle and the heat distribution of the cooling area, which is based on feedback from the stepper motor (8), are as follows: The angle information of the baffle (10) fed back by the stepper motor (8) is obtained in real time through the communication interface (13). The distribution of each baffle (10) in different cooling areas is spatially mapped. Combined with the current cycle hot zone distribution data collected by the real-time processor, the controller (7) calculates the coupling relationship between the opening angle of each baffle (10) and the physical structure, and identifies the flow efficiency in the actual flow channel. Based on the angle change amplitude and flow field change trend of the previous cycle and the current cycle, and in conjunction with the core indicators of baffle plate adjustment adaptation value, control execution angle value and adjustment stability evaluation value, it is determined whether there is a local structural adjustment lag, angle response deviation and slow heat dissipation feedback. When the angle fluctuation amplitude of a certain cooling area baffle (10) is too large in the current cycle, it is determined that the structural combination has insufficient adaptability. In the next round of genetic calculation, the structural consistency constraint weight is increased, and the regional structural state offset information is recorded as an auxiliary parameter for evolutionary population screening.

Images