Intelligent temperature control and heat dissipation method and system for liquid-cooled computer case

Abstract

The application discloses an intelligent temperature control and heat dissipation method and system for a liquid-cooled computer case. The computer case is internally provided with at least a chip and a liquid cooling system for cooling the chip. The liquid cooling system at least comprises a water pump and a cooling fan. The method comprises the following steps: collecting a first feature set of the computer case in real time, wherein the first feature set comprises a core temperature of the chip, a power consumption of the chip and a liquid temperature of cooling liquid in the liquid cooling system; calibrating a thermal resistance parameter in a preset model based on the first feature set, wherein the thermal resistance parameter comprises a calculated fast thermal resistance value or a preset slow thermal resistance value; analyzing and judging a heat load source of the chip by using the fast thermal resistance value and heat capacity, and obtaining a judgment result; calculating and obtaining a residual heat storage capacity of the liquid cooling system according to the liquid temperature and a preset standard cooling liquid temperature; generating a duty cycle set of the water pump and the cooling fan according to the judgment result and the residual heat storage capacity; and inputting corresponding PWM control signals to the water pump and the cooling fan according to the duty cycle set, so as to control the water pump and the cooling fan to perform heat dissipation on the chip.

Description

Technical Field

[0001] This invention relates to the field of intelligent heat dissipation for computers, and in particular discloses an intelligent temperature control and heat dissipation method and system for a liquid-cooled computer chassis. Background Technology

[0002] With the continuous improvement of computer performance, especially with the power consumption of high-end gaming processors and graphics processors exceeding the 300W or even 450W limit, traditional air cooling solutions are gradually approaching their physical limits. Liquid cooling technology, with its high specific heat capacity of coolant, can efficiently transfer heat from the chip core to the radiator and dissipate it into the air, and has become the mainstream cooling solution for high-performance computers, gaming consoles, and workstations.

[0003] Existing liquid cooling temperature control technologies primarily employ temperature feedback-based PID control schemes, which linearly adjust the speeds of the radiator fan and water pump by monitoring the chip temperature in real time. Some improved schemes introduce feedforward load-linked control, adding a load percentage as a feedforward signal to the temperature feedback, and preemptively increasing the speeds of the radiator fan and water pump when a sudden increase in load is detected. However, these schemes still face a series of technical challenges in practical applications.

[0004] First, the inherent thermal inertia of liquid cooling systems leads to fundamental misjudgments in traditional temperature control strategies. Due to their tiny size and extremely low heat capacity, the chip core's temperature can spike within milliseconds; while the coolant, with its larger volume and extremely high heat capacity, experiences a very slow temperature rise. When a user opens a demanding game or loads large software, the load instantly jumps from low to full, with the chip temperature rising by over 25°C within 1 to 2 seconds, while the coolant temperature rises by only about 0.1°C in the same timeframe. Traditional PID control, focusing solely on chip temperature, immediately commands the radiator fan and water pump to run at full speed upon detecting a temperature spike. However, at this point, heat has not yet been transferred to the coolant, the radiator does not actually need to dissipate heat at full speed, and the water pump does not need to operate at full load immediately. This overreaction of accelerating the speed at high temperatures causes the radiator fan and water pump to run at full speed prematurely when there is no real demand. The radiator fan generates sudden noise surges, and the water pump may experience mechanical resonance and abnormal noise due to sudden speed changes, while also wasting energy and causing mechanical wear.

[0005] Secondly, current technologies typically only detect the surface symptom of temperature rise, failing to determine whether the root cause is a sudden increase in chip heat generation or increased thermal resistance in the heat dissipation path. Both exhibit remarkably similar external characteristics: chip temperature rises while water temperature changes lag behind. However, the corresponding handling methods are drastically different. Transient loads require smooth adjustment of radiator fan and water pump speeds to avoid noise fluctuations, while heat dissipation failures necessitate emergency alarms to prevent hardware damage. The inability to distinguish between the two inevitably leads to inaccurate control strategies. During transient periods of drastic load fluctuations, not all power consumption generated by the chip is transferred to the coolant; a significant portion is used for chip self-heating. Within seconds of a sudden load spike, the chip temperature has already risen sharply, but the water temperature has not yet responded. At this point, attempting to inversely calculate the system state using thermodynamic relationships will yield distorted results due to neglecting the chip's own heat storage. This problem is particularly pronounced in scenarios with frequent transient loads, such as game loading, opening large software programs, or bursts of special effects in video editing software, severely impacting the reliability of diagnostics.

[0006] In addition, after long-term use, the thermal grease in a liquid cooling system will gradually dry out, leading to increased thermal resistance; the coolant may slowly evaporate, reducing its heat capacity; and scale may accumulate on the water pump impeller, causing performance degradation. These aging phenomena will alter the thermodynamic characteristics of the system, causing any control strategy based on fixed parameters to gradually become inaccurate and unable to maintain consistent temperature control over a usage period of several months or even years.

[0007] Therefore, there is an urgent need for an intelligent temperature control and heat dissipation method and system that can accurately identify the source of heat load under transient load, compensate for transient effects, adapt to system aging, and perform differentiated and coordinated control of water pumps and radiator fans. This would solve the misjudgment problem caused by thermal inertia, achieve smooth noise transition while ensuring heat dissipation efficiency, and maintain the stability of diagnostic and control accuracy during long-term use. Summary of the Invention

[0008] The purpose of this application is to provide an intelligent temperature control and heat dissipation method and system for liquid-cooled computer chassis. It can accurately identify the source of heat load under transient load, compensate for the transient effect caused by the chip's own heat storage, adaptively calibrate system parameters that drift due to long-term aging, and perform differentiated and coordinated control of water pump and radiator fan to solve the misjudgment problem caused by thermal inertia. It can achieve smooth noise transition while ensuring heat dissipation efficiency, and maintain the stability of control accuracy during long-term use.

[0009] To achieve the above objectives, a first aspect of this application provides an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis. The chassis contains at least a chip and a liquid cooling system for cooling the chip. The liquid cooling system includes at least a water pump and a radiator fan, comprising: S10 collects the first feature set inside the chassis in real time. The first feature set includes the core temperature of the chip, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system. S20 calibrates the thermal resistance parameters in the preset model based on the first feature set, wherein the thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value. S30 uses the rapid thermal resistance and thermal capacity to analyze and judge the source of heat load of the chip and obtain the judgment result; S40 calculates the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature. Based on the judgment result and the remaining heat storage capacity, S50 generates a set of duty cycles for the water pump and the radiator fan; S60 inputs corresponding PWM control signals to the water pump and radiator fan respectively according to the duty cycle, so as to control the water pump and radiator fan to perform heat dissipation on the chip.

[0010] In one embodiment of this application, step S20 includes: S21 calculates the chip's load percentage based on the real-time power consumption; S22 identifies the steady-state load window of the liquid cooling system based on the load percentage, core temperature, and liquid temperature. S23 calculates and obtains the current real-time thermal resistance value based on the steady-state load window and the corresponding thermodynamic relationship of the steady-state load window. S24 compares the real-time thermal resistance value with the old thermal resistance value of the last calibration or the preset basic thermal resistance value, and determines whether the old thermal resistance value or the basic thermal resistance value needs to be updated. S25 determines that the old thermal resistance value or the basic thermal resistance value needs to be updated, and obtains the fast thermal resistance value based on a first-order low-pass filter. S26 If the number of times the fast thermal resistance value is calculated and obtained is greater than a preset threshold or the difference between the fast thermal resistance value and the slow thermal resistance value is greater than a preset thermal resistance difference threshold, the slow thermal resistance value is updated by overwriting the fast thermal resistance value.

[0011] In one embodiment of this application, step S30 includes: S31 calculates the chip temperature change rate between two adjacent core temperature acquisition times and the temperature difference between the core temperature and liquid temperature at the same acquisition time based on multiple continuously acquired core temperatures and liquid temperatures. S32 determines whether the chip's heating state is abnormal based on the temperature change rate and temperature difference. If the chip's heating is normal, it obtains the judgment result of the heat load source when the chip's heating is normal. The judgment result is a continuous steady-state load. When S33 determines that the chip is overheating abnormally, it calls the preset dual-hypothesis prediction model. The dual-hypothesis prediction model includes a first model that assumes that transient impact load causes the core temperature to rise and a second model that assumes that heat dissipation failure causes the core temperature to rise. Based on the first model, the predicted temperature of the chip is calculated and obtained. At the same time, based on the second model, the change in thermal resistance between the chip and the coolant is calculated in reverse. S34 performs unsteady-state compensation and chip thermal capacity adaptation to provide transient compensation for the change in thermal resistance; S35 calculates and obtains statistical characteristics within multiple time windows based on the predicted temperature and the change in thermal resistance after transient compensation; S36 determines the source of heat load when the chip heats up abnormally based on the statistical characteristics.

[0012] In one embodiment of this application, step S33 is performed in parallel: The real-time liquid temperature, power consumption, and rapid thermal resistance value are substituted into the first model to calculate and obtain the predicted temperature; The real-time liquid temperature, core temperature, power consumption, and rapid thermal resistance value are substituted into the second model to calculate the change in thermal resistance in reverse.

[0013] In one embodiment of this application, in step S33, The formula for calculating the predicted temperature is as follows:

[0014] in, For the predicted temperature of the chip, This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value in the dual-hypothesis prediction model. The formula for calculating the change in thermal resistance in reverse is as follows:

[0015] in, This represents the change in the chip's thermal resistance. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

[0016] In one embodiment of this application, step S34 includes: S341 estimates the thermal capacity of the chip during the cold start and load drop phases of the chassis. S342 calculates the heat storage power of the chip itself by reverse calculation based on the temperature change rate and heat capacity; S343 calculates the effective power delivered by the chip to the coolant based on the power consumption and heat storage power. S344 substitutes the effective power into the second model to replace the power consumption, and calculates the change in thermal resistance after transient compensation.

[0017] In one embodiment of this application, in steps S342-S344, The formula for calculating the heat storage power of the chip itself is as follows:

[0018] in, The chip's heat storage power is calculated in reverse. To obtain the optimal chip heat capacity by smoothing and filtering the multiple heat capacity estimates, the term multiplied by the optimal chip heat capacity is the rate of temperature change. This refers to the minute change in the core temperature of the heat-generating chip. This represents the corresponding minute change over time; The formula for calculating the effective power transferred from the chip to the coolant is as follows:

[0019] in, This refers to the effective power transferred to the coolant during the chip's real-time power consumption. This refers to the chip's real-time power consumption. This refers to the chip's real-time heat storage power. The formula for calculating the change in thermal resistance after transient compensation is as follows:

[0020] in, This represents the change in thermal resistance after transient compensation of the chip. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This represents the chip's real-time effective power. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

[0021] In one embodiment of this application, the multiple time windows include short windows, medium windows, and long windows with different durations. In step S35, the prediction error between the predicted temperature and the core temperature, as well as the corrected change in thermal resistance, are continuously collected within each time window based on a preset fixed frequency. Within the short window, the average prediction error of the first model is calculated based on the collected prediction errors. Within the middle window, the variance of the prediction error is calculated based on the collected prediction errors, and the average thermal resistance change of the second model is calculated based on the collected thermal resistance changes. Within the long window, the collected thermal resistance changes are linearly fitted to obtain the slope of the thermal resistance change trend.

[0022] In one embodiment of this application, the determination result includes at least transient impact load, continuous steady-state load, and heat dissipation failure. In step S50, When the judgment result is a transient impact load, the first pump duty cycle is calculated based on the preset reference pump speed at the start of the transient impact load and the maximum allowable pump speed under the transient impact load, and the first radiator fan duty cycle is calculated based on the liquid temperature, remaining heat storage capacity and preset first PID parameters. When the judgment result is a continuous steady-state load, the power consumption is substituted into the preset pump speed mapping table to obtain the second water pump duty cycle corresponding to the power consumption, and the second radiator fan duty cycle is calculated based on the core temperature and the preset second PID parameters. If the determination result is a heat dissipation failure, the fault type of the chip or liquid cooling system is obtained, and the control corresponding to the fault type is executed on the chip or liquid cooling system.

[0023] A second aspect of this application provides an intelligent temperature control and heat dissipation system for performing the intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis as described in any one of the above claims, comprising: The data acquisition module is used to collect a first feature set inside the chassis in real time. The first feature set includes the core temperature of the chip, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system. The parameter calibration module is used to calibrate the thermal resistance parameters in the preset model based on the first feature set. The thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value. The load analysis module is used to analyze and judge the heat load source of the chip using the rapid thermal resistance value and heat capacity, and obtain the judgment result; The status assessment module is used to calculate the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature. The strategy generation module is used to generate a set of duty cycles for water pumps and radiator fans based on the judgment result and the remaining heat storage capacity. The control execution module is used to input corresponding PWM control signals to the water pump and the radiator fan respectively according to the duty cycle, so as to control the water pump and the radiator fan to perform heat dissipation on the chip.

[0024] By employing the aforementioned technical solution, after collecting chip core temperature, power consumption, and coolant temperature as the first feature set, the steady-state load window in the liquid cooling system is identified. Based on this steady-state load window, the thermal resistance parameter is calibrated in real time, achieving adaptive aging calibration of the parameters in the dual-hypothesis prediction model. This ensures the diagnostic model remains accurate throughout long-term use. Through the parallel operation of the two models in the dual-hypothesis prediction model, when the first model predicts accurately and the thermal resistance change in the second model approaches zero, the chip temperature rise is determined to be caused by transient impact load. Conversely, when the first model predicts inaccurately and the thermal resistance change in the second model increases significantly, the chip temperature rise is determined to be caused by a heat dissipation fault. This approach distinguishes the source of heat load, transforming the ambiguous temperature rise phenomenon into a quantifiable physical diagnosis, thus achieving accurate identification of transient impact load and heat dissipation faults. By estimating the chip's own thermal capacity during cold start and load drop phases, and inversely calculating the chip's heat storage power based on the temperature change rate, a method is used during transient processes: the portion used for chip self-heating is subtracted from the total chip power consumption, and the effective power is used to replace the original power consumption for inverse thermal resistance calculation. This eliminates the false appearance of increased thermal resistance caused by the chip's own heat storage, allowing for accurate determination of whether the thermal resistance has actually increased even during transient processes with drastic load fluctuations, avoiding unnecessary heat dissipation fault alarms. Under transient impact loads, a method is adopted where the coolant temperature, rather than the chip temperature, is used as the PID control target for the radiator fan and pump. This utilizes the liquid cooling system's own heat storage capacity to absorb chip temperature spikes, avoiding the overreaction of the radiator fan running at full speed as the chip temperature rises, achieving thermal inertia compensation and smooth speed transition. Under continuous steady-state loads, a method is adopted where the pump speed is matched according to the chip power consumption, and the PID control of the radiator fan and pump is performed with the chip temperature as the target, ensuring the heat dissipation efficiency of the liquid cooling system.

[0025] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0026] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings: Figure 1 The illustration shows a flowchart of an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to an embodiment of this application. Figure 2 The illustration shows a flowchart of step S34 in an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to an embodiment of this application. Figure 3 This schematic diagram illustrates the structural block diagram of an intelligent temperature-controlled heat dissipation system according to an embodiment of this application; Explanation of icon numbers: 100. Intelligent temperature control and heat dissipation system; 10. Data acquisition module; 20. Parameter calibration module; 30. Load analysis module; 40. Status assessment module; 50. Strategy generation module; 60. Control execution module. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0028] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0029] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0030] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0031] Figure 1 This illustration schematically shows a flow chart of an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to an embodiment of this application. Figure 1 As shown in the figure, this application provides an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis, which may include the following steps: The S10 collects a first feature set inside the chassis in real time. The first feature set includes the core temperature of the chip, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system.

[0032] S20 calibrates the thermal resistance parameters in the preset model based on the first feature set, wherein the thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value.

[0033] S30 uses the rapid thermal resistance and thermal capacity to analyze and determine the source of heat load on the chip and obtain the determination result.

[0034] S40 calculates the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature.

[0035] Based on the judgment result and the remaining heat storage capacity, S50 generates a duty cycle set for the water pump and the radiator fan.

[0036] S60 inputs corresponding PWM (Pulse Width Modulation) control signals to the water pump and radiator fan respectively according to the duty cycle, so as to control the water pump and radiator fan to perform heat dissipation on the chip.

[0037] Specifically, the data acquisition module 10 collects a first feature set within the chassis in real time. This first feature set includes the chip's core temperature, the chip's power consumption, and the coolant temperature in the liquid cooling system. The core temperature is the fundamental basis for determining whether overheating has occurred. Power consumption represents the heat generated by the chip per unit time and is a key parameter for predicting future temperature trends. The coolant temperature reflects the current heat-carrying capacity of the coolant and is an important reference for evaluating the operating status of the liquid cooling system. The parameter calibration module 20 calibrates the thermal resistance parameters in the preset model based on the first feature set. These thermal resistance parameters include calculated rapid thermal resistance values ​​or preset slow thermal resistance values. By monitoring the changes in the chip's core temperature, power consumption, and coolant temperature over a long period, and using data from when the system is in a stable operating state, the thermal resistance value of the heat dissipation path is calculated in reverse, achieving dynamic updates of the thermal resistance parameters and ensuring that the model always matches the current actual physical characteristics of the system. The load analysis module 30 analyzes and judges the sources of heat load on the chip using rapid thermal resistance values ​​and heat capacity, and obtains the judgment results. By combining real-time collected temperature and power consumption data, the model calculates and distinguishes whether the chip temperature rise is caused by a sharp increase in the chip's own power consumption or by blockage or aging of the heat dissipation path, preventing effective heat transfer. This identifies whether the current operating condition is a transient impact load or a heat dissipation failure. The status assessment module 40 calculates the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and a preset standard coolant temperature. The standard coolant temperature refers to the target temperature value that the coolant should maintain under the design operating conditions. It is a preset benchmark parameter representing the optimal balance between system heat dissipation efficiency and economy. The difference between the current liquid temperature and the standard coolant temperature directly quantifies how much heat the coolant can still absorb. This indicator is used to determine the buffer margin of the liquid cooling system in response to sudden loads. Based on the judgment result and the remaining heat storage capacity, the duty cycle set of the water pump and radiator fan is generated. Different operating conditions correspond to completely different control logics: under transient impact loads, priority is given to utilizing the liquid's heat storage capacity for smooth heat dissipation; under continuous steady-state loads, the balance between heat dissipation efficiency and energy consumption is pursued; and under heat dissipation failures, a protection mechanism must be triggered to ensure hardware safety. The strategy generation module 50 inputs corresponding PWM control signals to the water pump and radiator fan according to the duty cycle set, so as to control the water pump and radiator fan to perform heat dissipation on the chip. The control execution module 60 adjusts the speed of the water pump and radiator fan through the PWM control signals, and translates the control intention of the strategy into the actual heat dissipation action of the liquid cooling system.

[0038] In some embodiments of this application, step S20 may include: S21 calculates the chip's load percentage based on the real-time power consumption.

[0039] S22 identifies the steady-state load window of the liquid cooling system based on the load percentage, core temperature, and liquid temperature.

[0040] S23 calculates the current real-time thermal resistance value based on the steady-state load window and the corresponding thermodynamic relationship.

[0041] S24 compares the real-time thermal resistance value with the old thermal resistance value from the previous calibration or the preset basic thermal resistance value, and determines whether the old thermal resistance value or the basic thermal resistance value needs to be updated.

[0042] S25 determines that the old thermal resistance value or the basic thermal resistance value needs to be updated, and calculates the fast thermal resistance value based on a first-order low-pass filter.

[0043] S26 If the number of times the fast thermal resistance value is calculated and obtained is greater than a preset threshold or the difference between the fast thermal resistance value and the slow thermal resistance value is greater than a preset thermal resistance difference threshold, the slow thermal resistance value is updated by overwriting the fast thermal resistance value.

[0044] Specifically, the microcontroller calculates the ratio between the power consumption and the rated power consumption value (i.e., the load percentage) based on the chip power consumption collected in real time by the data acquisition module 10 and the chip's rated power consumption value pre-stored in the microcontroller according to the chip manufacturer's specifications. The load percentage is used to determine the chip's current operating intensity, providing a basis for subsequently identifying special operating condition windows. The microcontroller continuously monitors the chip's load percentage, as well as the chip core temperature and coolant temperature collected in real time by the data acquisition module 10, to identify steady-state load windows during chassis operation.

[0045] The steady-state load window is determined as follows: if the chip's load percentage is higher than a preset percentage threshold (e.g., 80%), it indicates that the chip is operating under high load. If the core temperature fluctuation of the chip is lower than a first fluctuation threshold (e.g., 1°C), it indicates that the chip temperature has stabilized. If the coolant temperature fluctuation is lower than a second fluctuation threshold (e.g., 0.5°C), it indicates that the liquid cooling system has reached thermal equilibrium. When all three conditions are met simultaneously and the duration reaches a preset time in the microcontroller, the microcontroller determines that the chip and liquid cooling system within the chassis are in a steady-state load window and triggers calibration of the thermal resistance parameters. The fluctuation amplitude refers to the difference between the maximum and minimum values ​​of multiple measurements of the chip core temperature or coolant temperature within the preset time window, used to determine whether the system is in a stable state. The microcontroller continuously monitors temperature changes using a sliding window; that is, each time the microcontroller completes a new sampling, the window moves forward by one sampling interval, ensuring continuous and real-time monitoring of temperature stability. The fluctuation amplitude is calculated using the temperature data collected within this window as a sample. The significance of identifying the steady-state load window lies in the fact that the formula for calculating the real-time thermal resistance value is strictly valid only when the system reaches thermal equilibrium. Only then can the calculated real-time thermal resistance value truly reflect the current state of the heat dissipation path.

[0046] Within the steady-state load window, the microcontroller collects multiple sample values ​​of the chip's core temperature, coolant temperature, and chip power consumption. It then calculates the average value of each parameter within the steady-state load window based on these sample values, and finally calculates the real-time thermal resistance value using the average value of each parameter and thermodynamic relationships. The formula for calculating the real-time thermal resistance value within the steady-state load window is as follows:

[0047] in, This is the real-time thermal resistance value. This represents the average core temperature within the steady-state load window. This represents the average liquid temperature within the steady-state load window. This represents the average power consumption within the steady-state load window. The significance of this formula is that, under thermal equilibrium, the difference between the chip's core temperature and the coolant temperature equals the chip's power consumption multiplied by the thermal resistance of the heat dissipation path. This thermal resistance is the required real-time thermal resistance value, representing the temperature difference between the chip and the coolant per watt of power consumption.

[0048] The microcontroller retrieves the currently stored thermal resistance value as the old thermal resistance value. During the initial system run, the old thermal resistance value is the factory-preset baseline value. After the system operates normally, the old thermal resistance value becomes the value stored after the last calibration. Subsequently, the microcontroller calculates the relative rate of change between the real-time thermal resistance value and the old thermal resistance value in real time, using the following formula:

[0049] in, The relative rate of change This is the real-time thermal resistance value. This is the old thermal resistance value. The relative rate of change represents the degree of change in the thermal resistance of the heat dissipation path relative to the system's historical baseline. After obtaining the relative rate of change, the microcontroller compares it with the maximum and minimum values ​​of a preset rate of change range. If the relative rate of change is less than the minimum value of the range, such as 3%, it indicates that the change in the thermal resistance of the heat dissipation path is minimal and can be considered as measurement noise or normal fluctuation, thus determining that the thermal resistance parameter does not need to be updated. If the relative rate of change is greater than the maximum value of the range, such as 15%, it indicates that the thermal resistance of the heat dissipation path has changed significantly, which means that the heat dissipation system may have a serious malfunction (such as severely dried-out thermal paste or blocked cold plates). In this case, the microcontroller determines that this situation triggers a heat dissipation system alarm, prompting the user to check and maintain it. Specifically, the microcontroller outputs an alarm signal through the system interface, such as controlling the LEDs on the front panel of the chassis or the motherboard to change color or adjust the flashing mode, allowing the user to observe the chassis to be aware of the issue. If the relative rate of change is within the rate of change range, it indicates that the thermal resistance has changed normally and slowly, such as due to natural aging of the thermal paste, thus determining that the thermal resistance parameter needs to be updated. When it is determined that the thermal resistance parameter needs to be updated, the microcontroller uses a first-order low-pass filter algorithm to calculate the new fast thermal resistance value. The calculation formula is as follows:

[0050] in, For rapid thermal resistance values, This is the thermal resistance filter coefficient. This is the real-time thermal resistance value. This is the old thermal resistance value.

[0051] The thermal resistance filter coefficient ranges from 0.2 to 0.4, balancing the weights of new and historical measurements. A larger coefficient results in a faster response to new measurements but may introduce more noise. A smaller coefficient leads to a stronger reliance on historical values, providing better smoothing but a slower response. Therefore, the coefficient is calibrated to 0.2 to 0.4, which allows for rapid tracking of the actual changes in thermal resistance while effectively filtering out noise from single measurements. The updated fast thermal resistance value is stored in the microcontroller's cache for real-time diagnostic calls in subsequent steps. The fast thermal resistance value is characterized by its high update frequency; it may be updated each time a steady-state load window is encountered, sensitively reflecting the current state of the heat dissipation path. Finally, the microcontroller maintains two versions of the thermal resistance parameter: the fast thermal resistance value and a built-in slow thermal resistance value. The slow thermal resistance value is stored in non-volatile memory as a long-term reference for subsequent calibration during chip thermal capacity estimation.

[0052] The slow thermal resistance value update is triggered when either of the following conditions is met: the cumulative number of fast thermal resistance value updates reaches a threshold or the deviation between the fast and slow thermal resistance values ​​exceeds the limit. After the update, the slow thermal resistance value is immediately written to non-volatile memory to ensure that it is not lost after the system is powered off.

[0053] The trigger condition for the cumulative number of fast thermal resistance value updates reaching a threshold is as follows: the microcontroller records the cumulative number of fast thermal resistance value updates since the last update of the slow thermal resistance value. When the cumulative update count reaches a certain preset value, such as 20 times, it indicates that the thermal resistance has undergone multiple gradual changes, and a new reference needs to replace the old one. At this time, the currently stored fast thermal resistance value is replaced with the slow thermal resistance value, and the cumulative update count is reset to zero to update the slow thermal resistance value.

[0054] The specific trigger condition for exceeding the slow thermal resistance deviation limit is as follows: The microcontroller monitors the relative deviation between the current fast and slow thermal resistance values ​​in real time, that is, the percentage of the difference between the two to the slow thermal resistance value. When the relative deviation exceeds a predetermined percentage, it indicates that the fast thermal resistance value has significantly deviated from the original long-term benchmark, and the slow thermal resistance value needs to be updated to match the current state. At this time, the currently stored fast thermal resistance value is replaced with the slow thermal resistance value to update the slow thermal resistance value.

[0055] In some embodiments of this application, step S30 may include: S31 calculates the chip temperature change rate between two adjacent core temperature acquisition times and the temperature difference between the core temperature and the liquid temperature at the same acquisition time based on multiple continuously acquired core temperatures and liquid temperatures.

[0056] S32 determines whether the chip's heating state is abnormal based on the temperature change rate and temperature difference. If the chip's heating is normal, it obtains the judgment result of the heat load source when the chip's heating is normal, and the judgment result is a continuous steady-state load.

[0057] When S33 determines that the chip is overheating abnormally, it calls a preset dual-hypothesis prediction model. The dual-hypothesis prediction model includes a first model that assumes that transient impact load causes the core temperature to rise and a second model that assumes that heat dissipation failure causes the core temperature to rise. Based on the first model, the predicted temperature of the chip is calculated and obtained. At the same time, based on the second model, the change in thermal resistance between the chip and the coolant is calculated in reverse.

[0058] S34 performs unsteady-state compensation and chip thermal capacity adaptation to provide transient compensation for the change in thermal resistance.

[0059] S35 calculates and obtains statistical characteristics within multiple time windows based on the predicted temperature and the change in thermal resistance after transient compensation.

[0060] S36 obtains the judgment result of the heat load source when the chip heats up abnormally based on the statistical characteristics.

[0061] Specifically, the microcontroller first calculates the chip's temperature change rate between two adjacent sampling times, and the difference between the core temperature and the liquid temperature at the same time, based on the real-time core temperature and liquid temperature. The temperature change rate is obtained by dividing the difference between the current temperature and the previous temperature by the sampling interval, reflecting how quickly the chip heats up or cools down; the temperature difference directly reflects the driving force of heat transfer between the chip and the coolant. These two features together constitute the basic indicators for judging the chip's heating state. When the temperature change rate exceeds a preset threshold and the temperature difference simultaneously exceeds a preset threshold, the microcontroller determines that the chip is in an abnormal heating state and triggers a deeper diagnostic process. If neither the temperature change rate nor the temperature difference exceeds the corresponding threshold, the chip is determined to be heating normally. In this case, the judgment result of the heat load source is directly determined to be a continuous steady-state load, indicating that the system is in a thermal equilibrium state and the chip temperature is stable within an acceptable range, requiring no further diagnosis. After determining that the chip is heating abnormally, the microcontroller calls a pre-built dual-hypothesis prediction model. The dual-hypothesis prediction model runs two hypotheses in parallel. The first model assumes that the current anomaly is caused by a transient impact load, that is, the chip power consumption increases sharply while the thermal resistance of the heat dissipation path remains unchanged, and calculates the predicted temperature of the chip accordingly. The second model assumes the current anomaly is caused by a heat dissipation failure, i.e., increased thermal resistance along the heat dissipation path while chip power consumption remains within a normal range. Based on this, it calculates the change in thermal resistance between the chip and the coolant in reverse. For the thermal resistance change calculated in reverse by the second model, the microcontroller performs unsteady-state compensation and chip thermal capacity adaptive processing. It estimates the chip's own thermal capacity during the cold start and load drop phases, calculates the power of the heat stored in the chip itself using the temperature change rate, and then subtracts this portion from the measured power consumption to obtain the effective power actually transferred to the coolant. This effective power replaces the original power consumption to recalculate the thermal resistance change, eliminating the interference of the chip's own heat storage on the reverse calculation results during transient processes. Subsequently, the microcontroller continuously collects the prediction error of the first model and the thermal resistance change after compensation by the second model within multiple preset time windows. The average prediction error within a short window reflects the instantaneous matching degree, which refers to the deviation between the temperature predicted by the first model and the measured temperature over a very short time. In the initial stage of the heating anomaly, when the chip temperature has just begun to spike, the matching degree of the short window can quickly determine whether the first assumption model holds true. A high degree of matching indicates that the current anomaly is likely due to a transient impact load; a low degree of matching suggests a heat dissipation failure or other causes. The variance of the prediction error and the average value of the change in thermal resistance are calculated within the middle window, reflecting short-term stability and the degree of deviation. These two factors together influence the judgment of the fault's nature. If the variance of the prediction error is small (the model remains inaccurate) and the average value of the change in thermal resistance is significantly positive (thermal resistance has indeed increased), it is determined to be a heat dissipation failure. If the variance of the prediction error is small and the average value of the change in thermal resistance approaches zero, it is determined to be a transient impact load. A large variance indicates interference or uncertainty, requiring further observation.Linear fitting is performed on the thermal resistance change sequence within a long window to obtain the slope of the thermal resistance change trend, reflecting whether the thermal resistance change is an instantaneous fluctuation or a persistent fault. Finally, the microcontroller makes a logical decision based on the statistical characteristics collected in each time window. When the average prediction error of the short window is small, the variance of the prediction error of the medium window is small, and the average value of the thermal resistance change is close to zero, it is determined to be a transient impact load. When the average prediction error of the short window is large, the average value of the thermal resistance change of the medium window is positive, and the slope of the trend in the long window is close to zero, it is determined to be a heat dissipation fault. If the statistical characteristics do not meet any of the above conditions, it is determined to be an uncertain state, and the current control strategy remains unchanged.

[0062] In some embodiments of this application, step S33 is performed in parallel: The predicted temperature is calculated by substituting the real-time liquid temperature, power consumption, and rapid thermal resistance value into the first model.

[0063] The real-time liquid temperature, core temperature, power consumption, and rapid thermal resistance value are substituted into the second model to calculate the change in thermal resistance in reverse.

[0064] Specifically, upon determining that the chip is overheating abnormally, the microcontroller enters anomaly detection mode, executing calculations based on both a first model and a second model in parallel to distinguish whether the heat load originates from a transient impact load or a thermal failure. The first model assumes the current overheating anomaly is due to a sharp increase in chip power consumption, while the thermal resistance of the heat dissipation path remains normal. In this case, the microcontroller substitutes the real-time coolant temperature, real-time chip power consumption, and updated rapid thermal resistance value collected by the data acquisition module 10 into the formula to calculate the predicted chip temperature. The second model assumes the current overheating anomaly is due to an increase in the thermal resistance of the heat dissipation path, while the chip power consumption maintains a normal variation. In this case, the microcontroller substitutes the real-time chip core temperature, coolant temperature, real-time chip power consumption, and updated rapid thermal resistance value into the formula to calculate the change in thermal resistance of the heat dissipation path. Within each sampling cycle, the microcontroller executes the calculations of both models simultaneously, without waiting for one to complete before executing the other. Both calculation results are temporarily stored in the microcontroller's cache for later use in transient compensation and heat load source determination.

[0065] In some embodiments of this application, in step S33, The formula for calculating the predicted temperature is as follows:

[0066] in, For the predicted temperature of the chip, This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value in the dual-hypothesis prediction model. The formula for calculating the change in thermal resistance in reverse is as follows:

[0067] in, This represents the change in the chip's thermal resistance. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

[0068] Specifically, the predicted temperature calculation formula in the first model is essentially derived from steady-state thermodynamics. In thermal equilibrium, the core temperature of the chip can be considered as the coolant temperature plus the product of power consumption and thermal resistance. The product of power consumption and rapid thermal resistance represents the expected temperature rise of the chip relative to the coolant at the current power consumption. Through this relationship, the microcontroller can calculate the core temperature of the chip assuming normal heat dissipation. The inverse calculation formula for thermal resistance change in the second model is also based on steady-state thermodynamics. Subtracting the real-time rapid thermal resistance value from the current actual total thermal resistance yields the change in thermal resistance relative to the normal preset standard value. A positive change in thermal resistance indicates that the current thermal resistance is greater than the normal value, and heat dissipation efficiency is reduced. A change in thermal resistance approaching zero indicates that the thermal resistance is normal. A negative change in thermal resistance indicates that the current thermal resistance is less than the normal value. This situation may theoretically be caused by measurement noise or transient effects, but can be addressed through filtering in practice. Finally, the calculation results of the first and second models together constitute the judgment criteria for anomaly detection.

[0069] Figure 2 The diagram schematically illustrates step S34 of an intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to an embodiment of this application. Figure 2 As shown, step S34 may include the following steps: S341 estimates the thermal capacity of the chip during the cold start and load drop phases of the chassis.

[0070] S342 calculates the heat storage power of the chip itself by reverse calculation based on the temperature change rate and heat capacity.

[0071] S343 calculates the effective power delivered by the chip to the coolant based on the power consumption and heat storage power.

[0072] S344 substitutes the effective power into the second model to replace the power consumption, and calculates the change in thermal resistance after transient compensation.

[0073] Specifically, the heat capacity of a chip represents the amount of heat it needs to absorb to raise its temperature by 1°C, and is a key parameter for determining heat storage power during transient processes. Because the heat capacity varies between different chip models, and even the same chip may exhibit slight variations across different temperature ranges, the microcontroller needs to estimate the chip's heat capacity in real time during two specific operating phases of the chassis. The first is the cold start phase, which refers to the initial stage when the chassis is powered on for the first time after being powered off, before the chips and liquid cooling system have reached a stable operating state. During this phase, the coolant temperature has not yet risen significantly, and it is approximated that almost all the heat generated by the chip is used for its own heating, with negligible heat transferred to the coolant. At this time, the calculation process for estimating the chip's heat capacity is simplified, specifically, the heat capacity is equal to the ratio of the chip's power consumption to its rate of temperature change. After the chassis undergoes a cold start, the microcontroller continuously monitors the chip power consumption and temperature change rate. When the difference between the coolant temperature and the initial value (i.e., the difference between the current coolant temperature and the initial coolant temperature recorded when the system was first powered on) is less than the preset temperature difference, the microcontroller determines that the system is still in the cold start phase. It then performs smoothing filtering on the calculation results of multiple consecutive sampling points within the phase to obtain the heat capacity of the cold start phase.

[0074] The load drop phase refers to the transition period during which a chip instantly drops from a high-load state to a low-load state. At this time, the chip stops generating heat or generates significantly less heat, and the stored heat begins to transfer to the coolant, causing the chip temperature to drop. In the short period following the load drop, the thermal equilibrium relationship is as follows:

[0075] in, For the core temperature, Liquid temperature This is the slow thermal resistance value. This is an estimate of the heat capacity. This represents the minute change in the core temperature of the heating chip, and is used as the numerator of the rate of temperature change. This represents the corresponding minute change over time. The first term is the heat flow transferred to the coolant through the heat dissipation path, and the second term is the heat released by the chip's own heat storage. Since the chip stops generating heat, the sum of the two terms is zero. After rearranging the terms, the formula for estimating the chip's heat capacity is as follows:

[0076] in, This represents the minute change in the core temperature of the heating chip, and is used as the numerator of the rate of temperature change. The corresponding minute change over time is used as the denominator of the rate of temperature change.

[0077] The microcontroller monitors the chip power consumption in real time. When it detects that the power consumption drops instantly from above the preset first power consumption threshold, such as 80W, to below the preset second power consumption threshold, such as 10W, the microcontroller determines that the system has entered the load drop phase. In the following few seconds, it continuously collects data, uses the above formula to calculate multiple estimated values ​​of the chip's thermal capacity, and performs smoothing filtering to obtain the thermal capacity of the load drop phase.

[0078] The microcontroller calculates the heat storage power of the chip by multiplying the real-time chip temperature change rate by the estimated chip heat capacity. When the chip temperature rises, the heat storage power is positive, indicating that some power consumption is absorbed by the chip itself. When the chip temperature falls, the heat storage power is negative, indicating that the chip releases the stored heat. This calculation converts temperature changes into a quantifiable power value. The microcontroller then subtracts the heat storage power from the chip's real-time power consumption to obtain the effective power actually transferred to the coolant. This effective power excludes the heat consumed by the chip's own temperature rise, accurately reflecting the heat load that needs to be removed by the liquid cooling system. Under steady-state conditions, when the temperature remains constant, the heat storage power is zero, and the effective power equals the actual power consumption. During transient temperature increases, the effective power is less than the actual power consumption.

[0079] Finally, the microcontroller substitutes the calculated effective power into the second model, replaces the original power consumption, and recalculates the change in thermal resistance to obtain the result after transient compensation. During a transient process, if no compensation is performed and the original chip power consumption is used directly to calculate the change in thermal resistance, the compensated change in thermal resistance will be greater than the uncompensated change in thermal resistance because the effective power is less than the power consumption. In other words, the original formula will underestimate the actual change in thermal resistance or produce a false increase. The compensated change in thermal resistance eliminates the interference of the chip's own heat storage on the inverse calculation formula, enabling the system to accurately determine whether the thermal resistance has truly increased, even during transient processes with drastic load fluctuations, avoiding misjudgments of heat dissipation failures due to false increases in thermal resistance.

[0080] In some embodiments of this application, in steps S342-S344, The formula for calculating the heat storage power of the chip itself is as follows:

[0081] in, The chip's heat storage power is calculated in reverse. To obtain the optimal chip heat capacity by smoothing and filtering the multiple heat capacity estimates, the term multiplied by the optimal chip heat capacity is the rate of temperature change. This refers to the minute change in the core temperature of the heat-generating chip. This represents the corresponding minute change over time; The formula for calculating the effective power transferred from the chip to the coolant is as follows:

[0082] in, This refers to the effective power transferred to the coolant during the chip's real-time power consumption. This refers to the chip's real-time power consumption. This refers to the chip's real-time heat storage power. The formula for calculating the change in thermal resistance after transient compensation is as follows:

[0083] in, This represents the change in thermal resistance after transient compensation of the chip. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This represents the chip's real-time effective power. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

[0084] Specifically, during the chip's heating process, besides a portion being transferred to the coolant, another portion of the power consumed is used for its own temperature increase; this power used for temperature increase is called heat storage power. According to the first law of thermodynamics, heat storage power equals the chip's heat capacity multiplied by the rate of temperature change. When the chip temperature rises, i.e., the rate of temperature change is greater than 0, the heat storage power is positive, indicating that some power consumption is absorbed by the chip itself. When the chip temperature falls, i.e., the rate of temperature change is less than 0, the heat storage power is negative, indicating that the chip releases the stored heat. Only after subtracting the heat stored by the chip itself is the remaining heat that truly needs to be removed by the liquid cooling system. Since the chip temperature remains constant under steady-state conditions, the heat storage power is zero, and the effective power equals the actual power consumption. During transient heating, the effective power is less than the actual power consumption. During cooling, the effective power may be negative, meaning the chip releases heat, but the coolant is heated instead. Substituting the calculated effective power into the inverse calculation formula for the change in thermal resistance in the second model, we obtain the change in thermal resistance after transient compensation. During transient processes, if the thermal resistance change is calculated directly using the chip's power consumption without compensation, the actual thermal resistance change will be underestimated or a false increase in thermal resistance will be generated. Therefore, it is necessary to calibrate the thermal resistance change to reflect a more realistic thermal resistance change in the heat dissipation path.

[0085] In some embodiments of this application, the multiple time windows include short windows, medium windows, and long windows with different durations. In step S35, based on a preset fixed frequency, the prediction error between the predicted temperature and the core temperature, as well as the corrected change in thermal resistance, are continuously collected within each time window.

[0086] Within the short window, the average prediction error of the first model is calculated based on the collected prediction errors.

[0087] Within the middle window, the variance of the prediction error is calculated based on the collected prediction errors, and the average thermal resistance change of the second model is calculated based on the collected thermal resistance changes.

[0088] Within the long window, the collected thermal resistance changes are linearly fitted to obtain the slope of the thermal resistance change trend.

[0089] Specifically, the microcontroller continuously collects data at a preset fixed sampling frequency (e.g., 10 times per second, i.e., a sampling interval of 0.1 seconds) and calculates statistical characteristics within three time windows of different lengths. The short window lasts 0.5 seconds and contains 5 sampling points, used to capture instantaneous deviations and reflect the current level of matching. The medium window lasts 3 seconds and contains 30 sampling points, used to assess short-term stability and reflect the amplitude of error fluctuations. The long window lasts 15 seconds and contains 150 sampling points, used to analyze trends and determine whether the thermal resistance change is an instantaneous fluctuation or a persistent fault. At each sampling moment, the microcontroller first calculates the prediction error between the predicted temperature output by the first model and the core temperature; this error reflects the accuracy of the first model's prediction of the current temperature. Simultaneously, the microcontroller acquires the calculated transient compensation thermal resistance change, representing the degree of deviation of the current heat dissipation path thermal resistance from the normal baseline. Within the multiple time windows, the microcontroller stores the latest data in a buffer and removes the oldest data after each new sampling. Within the short window, the microcontroller calculates the average prediction error based on the predicted temperature obtained from five sampling points and the measured chip core temperature. This average prediction error is used to determine the average deviation between the first model's prediction and the measured value near the current moment. A smaller value indicates that the first model's prediction is accurate, and the current temperature change conforms to the assumption of normal thermal resistance and increased power consumption. Within the medium window, the microcontroller calculates the variance of the first model's prediction error within the medium window using a variance formula. Variance reflects the fluctuation of the prediction error; a small variance indicates stable prediction error and the absence of intermittent interference or model mismatch. Simultaneously with the variance calculation, the microcontroller calculates the average value of the thermal resistance change after transient compensation within the medium window. This average value is used to determine whether the thermal resistance has continuously deviated from the normal baseline in the recent period. If the average value is significantly greater than zero and stable, it indicates that the thermal resistance has indeed increased. Within the long window, the microcontroller performs linear fitting sequentially on the thermal resistance changes after transient compensation obtained from 150 sampling points, calculating the slope of the thermal resistance change trend. The linear fitting formula uses the least squares method, as shown below:

[0090] in, The slope of the trend in thermal resistance change. The number of sampling points. For the first The time value corresponding to each sampling point For the first Change in thermal resistance after transient compensation This is the sum of the products of the time value and the corresponding change in thermal resistance at all sampling points, reflecting the covariant relationship between time and the change in thermal resistance. The sum of the time values ​​for all sampling points. This is the sum of the changes in thermal resistance at all sampling points. It is the sum of the squares of the time values ​​at all sampling points.

[0091] This formula uses the least squares method to find an optimally fitted straight line that minimizes the sum of the squared perpendicular distances from all data points to this line. The slope reflects the trend of thermal resistance change over a long window. If the slope approaches zero, it indicates that the thermal resistance change is stable and continuous, rather than a momentary fluctuation. If the slope is positive, it indicates that the thermal resistance is still increasing. If the slope is negative, it indicates that the thermal resistance is recovering.

[0092] Finally, the microcontroller compares the aforementioned statistical characteristics with preset thresholds and performs logical decisions. Each preset threshold is pre-stored in the microcontroller's non-volatile memory based on system characteristics and experimental calibration.

[0093] When the lower limit of the average prediction error threshold is set to 1.0℃, the upper limit of the average prediction error threshold is set to 3.0℃, the prediction error variance threshold is set to 0.5℃², the lower limit of the thermal resistance change threshold is set to 0.02℃ / W, the upper limit of the thermal resistance change threshold is set to 0.08℃ / W, and the slope threshold is set to 0.001℃ / W / s; As an example, if the average prediction error within the short window is 0.8℃, which is less than the lower limit threshold of 1.0℃, and the prediction error variance within the medium window is 0.2℃², which is less than the prediction error variance threshold of 0.5℃², and the average change in thermal resistance within the medium window is 0.01℃ / W, which is less than the lower limit threshold of 0.02℃ / W, then the first model is judged to be accurate and stable. Simultaneously, the change in thermal resistance approaches zero, indicating that the temperature rise is entirely consistent with the expected normal thermal resistance, and the thermal resistance has not actually increased. In this case, the microcontroller determines that the current anomaly is caused by a transient impact load.

[0094] As another example, if the average prediction error within the short window is 4.2℃, which is greater than the upper threshold of 3.0℃ for average prediction error, and the average change in thermal resistance within the medium window is 0.12℃ / W, which is greater than the upper threshold of 0.08℃ / W for change in thermal resistance, and the slope within the long window is 0.0005℃ / W / s, which is less than the slope threshold of 0.001℃ / W / s, then the first model is judged to be severely inaccurate in its prediction, with a significant increase in thermal resistance, and the change in thermal resistance is stable and continuous. This indicates that the thermal resistance of the heat dissipation path has indeed increased in a real way, rather than fluctuating instantaneously. The microcontroller determines that the current anomaly is caused by a heat dissipation failure.

[0095] If the microcontroller detects that the statistical feature does not meet any of the above judgment conditions, the microcontroller concludes that the judgment result is uncertain. At this time, the microcontroller maintains the current control strategy unchanged, does not trigger an alarm, and continues to observe subsequent data.

[0096] The above decision-making logic is based on the fact that although transient impact loads and thermal failures exhibit similar external manifestations (increased chip temperature, delayed water temperature), they differ fundamentally in two dimensions: the accuracy of the first model's prediction and the change in thermal resistance. Under transient impact loads, the thermal resistance is normal, the first model's prediction is accurate, and the change in thermal resistance approaches zero. Under thermal failures, the thermal resistance increases, the first model's prediction is inaccurate, and the change in thermal resistance is positive. By collecting statistical features over multiple time windows, the system can filter out transient noise and make reliable decisions based on stable and continuous evidence, avoiding misjudgments caused by single data fluctuations.

[0097] In some embodiments of this application, the determination result includes at least transient impact load, continuous steady-state load, and heat dissipation failure. In step S50, When the judgment result is a transient impact load, the first pump duty cycle is calculated based on the preset reference pump speed at the start of the transient impact load and the maximum allowable pump speed under the transient impact load, and the first radiator fan duty cycle is calculated based on the liquid temperature, remaining heat storage capacity and preset first PID parameters.

[0098] When the judgment result is a continuous steady-state load, the power consumption is substituted into the preset pump speed mapping table to obtain the second water pump duty cycle corresponding to the power consumption, and the second radiator fan duty cycle is calculated based on the core temperature and the preset second PID parameters.

[0099] If the determination result is a heat dissipation failure, the fault type of the chip or liquid cooling system is obtained, and the control corresponding to the fault type is executed on the chip or liquid cooling system.

[0100] Specifically, the microcontroller generates differentiated control strategies for the water pump and radiator fan based on the judgment result and the calculated remaining heat storage capacity. When the judgment result is a transient impact load, the control objective is to smoothly adjust the load while ensuring heat dissipation, avoiding sudden increases in the speed of the radiator fan and water pump and the resulting noise impact. The water pump duty cycle is increased slowly according to a ramp function, calculated as follows:

[0101] in, For the water pump after the transient load begins PWM duty cycle at any given time. The preset reference pump speed, such as 30%, represents the initial speed at the start of transient load. The preset ramp rate, such as 5% / s, controls the smoothness of the speed increase. The duration from the start of the transient load. Set the maximum permissible pump speed under transient load, such as 70%, to avoid entering the resonant speed zone.

[0102] The radiator fan duty cycle is controlled by the coolant temperature using a PID algorithm. The microcontroller first dynamically adjusts the target coolant temperature based on the calculated remaining heat storage capacity. When the remaining heat storage capacity is large, the target coolant temperature can be appropriately increased, allowing the radiator fan to operate at a slower speed, further reducing power consumption and noise. When the remaining heat storage capacity is small or negative, the target coolant temperature decreases, and the radiator fan accelerates in advance to ensure heat dissipation reserves. Subsequently, the microcontroller calculates the coolant temperature control deviation and substitutes it into the PID formula, combining it with preset coolant temperature control PID parameters to calculate the radiator fan duty cycle. The coolant temperature control deviation is the difference between the dynamically adjusted target coolant temperature and the real-time collected coolant temperature. By using coolant temperature as the control target, the radiator fan speed changes slowly with the coolant temperature, avoiding the power waste and noise impact caused by the radiator fan running at full speed as the chip temperature rises in traditional solutions.

[0103] When the judgment result is a continuous steady-state load, the control objective is to ensure heat dissipation efficiency and stabilize the chip's core temperature at the target value. The water pump duty cycle is obtained by querying a preset power consumption-pump speed mapping table based on the current chip power consumption. The radiator fan duty cycle is adjusted using a PID algorithm with the chip core temperature as the control target. The microcontroller calculates the chip temperature control deviation, which is the difference between the core temperature and the target temperature. This deviation value is then substituted into the PID formula, combined with preset chip temperature control PID parameters, to calculate the radiator fan duty cycle.

[0104] When the diagnosis is a heat dissipation fault, the microcontroller first identifies the fault type by integrating multiple diagnostic indicators. If the average change in thermal resistance over a long window consistently exceeds the upper limit threshold and the trend slope approaches zero, it's identified as an abnormally increased thermal resistance fault, typically caused by dried-out thermal grease, buildup on the coolant block, or loose installation. If the deviation between the output PWM duty cycle and the feedback speed of the water pump or radiator fan exceeds a preset threshold, it's identified as an actuator failure fault. Water pump failure manifests as an abnormal rise in liquid temperature without a corresponding increase in chip power consumption; radiator fan failure manifests as a continuous rise in chip temperature while the liquid temperature approaches saturation. If the remaining heat storage capacity remains abnormally low and the coolant temperature fluctuates drastically, it's identified as a coolant insufficiency fault, typically caused by coolant evaporation or pipe leaks.

[0105] Based on the identified fault types, the microcontroller executes differentiated protection actions. The microcontroller distinguishes between minor and severe faults caused by abnormally increased thermal resistance based on the relative rate of change of thermal resistance and a preset threshold. When a minor thermal resistance increase fault is identified, the microcontroller forces the PWM duty cycle of both the water pump and radiator fan to 100%, compensating for the reduced cooling efficiency with maximum flow rate and airflow. Simultaneously, it sends an alarm message to the operating system via lighting control or the system interface, prompting the user to check the thermal paste status and the cleanliness of the cooling block. When a severe thermal resistance increase fault is identified, the microcontroller forces full-speed cooling while sending a frequency reduction request to the BIOS via the PECI or SMBus interface to proactively reduce the chip's real-time power consumption, preventing the temperature from continuing to rise until user maintenance is required. When a water pump or radiator fan failure is identified (actuator failure fault), the microcontroller first attempts to request a reduction in chip power consumption via the system interface. If the chip temperature continues to rise and approaches the critical threshold, it triggers the system emergency shutdown procedure, simultaneously issuing a continuous alarm via buzzer or LED indicator to prevent the chip from burning out due to lack of cooling. When a coolant deficiency is detected, the microcontroller sends an alarm message to the operating system via the light control or the system interface to prompt the user to add coolant or check for leaks in the pipeline. At the same time, it actively limits the chip's maximum operating power consumption to prevent overheating due to insufficient heat dissipation.

[0106] In summary, the intelligent temperature control and heat dissipation method for liquid-cooled computer chassis provided in this application uses the chip core temperature, power consumption, and coolant temperature as a first feature set to identify the steady-state load window in the operation of the liquid cooling system. Based on this steady-state load window, the thermal resistance parameters are calibrated in real time, achieving adaptive aging calibration of the parameters in the dual-hypothesis prediction model and ensuring the diagnostic model remains accurate throughout long-term use. Through the parallel operation of the dual-hypothesis prediction model, when the first model predicts accurately and the change in thermal resistance in the second model approaches zero, the cause of chip temperature rise is determined to be a transient impact load; when the first model predicts inaccurately and the change in thermal resistance in the second model increases significantly, the cause of chip temperature rise is determined to be a heat dissipation failure. This method can distinguish the source of heat load, transforming the ambiguous temperature rise phenomenon into a quantifiable physical diagnosis, and achieving accurate identification of transient impact loads and heat dissipation failures. By estimating the chip's own thermal capacity during cold start and load drop phases, and inversely calculating the chip's heat storage power based on the temperature change rate, a method is used during transient processes. This involves subtracting the portion used for chip self-heating from the total chip power consumption and replacing the original power consumption with the effective power for inverse thermal resistance calculation. This eliminates the false appearance of increased thermal resistance caused by the chip's own heat storage, ensuring accurate determination of actual thermal resistance increase even during transients with drastic load fluctuations, thus avoiding unnecessary heat dissipation fault alarms. Under transient impact loads, the coolant temperature, rather than the chip temperature, is used as the PID control target for the radiator fan and pump. This utilizes the liquid cooling system's own heat storage capacity to absorb chip temperature spikes, preventing the radiator fan from overreacting by running at full speed as the chip temperature rises, achieving thermal inertia compensation and smooth speed transition. Under continuous steady-state loads, the pump speed is matched to the chip power consumption, and the radiator fan and pump are controlled using PID control with chip temperature as the target, ensuring the heat dissipation efficiency of the liquid cooling system.

[0107] In one embodiment, such as Figure 3 As shown, an intelligent temperature control and heat dissipation system 100 is provided, including a data acquisition module 10, a parameter calibration module 20, a load analysis module 30, a state evaluation module 40, a strategy generation module 50, and a control execution module 60, wherein: The data acquisition module 10 is used to acquire a first feature set inside the chassis in real time. The first feature set includes the core temperature of the chip, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system.

[0108] The parameter calibration module 20 is used to calibrate the thermal resistance parameters in the preset model based on the first feature set. The thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value.

[0109] The load analysis module 30 is used to analyze and judge the source of heat load of the chip using the rapid thermal resistance value and heat capacity, and obtain the judgment result.

[0110] The status assessment module 40 is used to calculate the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature.

[0111] The strategy generation module 50 is used to generate a set of duty cycles for water pumps and radiator fans based on the judgment result and the remaining heat storage capacity.

[0112] The control execution module 60 is used to input corresponding PWM control signals to the water pump and the radiator fan respectively according to the duty cycle, so as to control the water pump and the radiator fan to perform heat dissipation on the chip.

[0113] The intelligent temperature control and heat dissipation system 100 includes a processor and a memory. The data acquisition module 10, parameter calibration module 20, load analysis module 30, state evaluation module 40, strategy generation module 50 and control execution module 60 are all stored as program units in the memory. The processor executes the program modules stored in the memory to realize the corresponding functions.

[0114] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and intelligent temperature control and heat dissipation methods for liquid-cooled computer cases can be implemented by adjusting kernel parameters.

[0115] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of implementing the above-described computer program product on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes, and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0117] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes, and / or boxes Figure 1 The function specified in one or more boxes.

[0118] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes, and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0119] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0120] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0121] Computer-readable media include both permanent and non-permanent, removable and non-removable media, where information can be stored by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0122] It should also be noted that 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. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0123] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for intelligent temperature control and heat dissipation of a liquid-cooled computer chassis, wherein the liquid-cooled computer chassis includes a chip and a liquid cooling system for cooling the chip, the liquid cooling system including a water pump and a radiator fan; characterized in that, The intelligent temperature control and heat dissipation method includes: The S10 collects the core temperature of the chip inside the chassis, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system in real time, as the first feature set; S20 calibrates the thermal resistance parameters in the preset model based on the first feature set, wherein the thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value. S30 uses the rapid thermal resistance and thermal capacity to analyze and judge the source of heat load of the chip and obtain the judgment result; S40 calculates the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature. Based on the judgment result and the remaining heat storage capacity, S50 generates a set of duty cycles for the water pump and the radiator fan; S60 inputs corresponding PWM control signals to the water pump and radiator fan respectively according to the duty cycle, so as to control the water pump and radiator fan to perform heat dissipation on the chip.

2. The intelligent temperature control and heat dissipation method for liquid-cooled computer chassis according to claim 1, characterized in that, Step S20 includes: S21 calculates the chip's load percentage based on the real-time power consumption; S22 identifies the steady-state load window of the liquid cooling system based on the load percentage, core temperature, and liquid temperature. S23 calculates and obtains the current real-time thermal resistance value based on the steady-state load window and the corresponding thermodynamic relationship of the steady-state load window. S24 compares the real-time thermal resistance value with the old thermal resistance value of the last calibration or the preset basic thermal resistance value, and determines whether the old thermal resistance value or the basic thermal resistance value needs to be updated. S25 determines that the old thermal resistance value or the basic thermal resistance value needs to be updated, and obtains the fast thermal resistance value based on a first-order low-pass filter. S26 If the number of times the fast thermal resistance value is calculated and obtained is greater than a preset threshold or the difference between the fast thermal resistance value and the slow thermal resistance value is greater than a preset thermal resistance difference threshold, the slow thermal resistance value is updated by overwriting the fast thermal resistance value.

3. The intelligent temperature control and heat dissipation method for liquid-cooled computer chassis according to claim 1, characterized in that, Step S30 includes: S31 calculates the chip temperature change rate between two adjacent core temperature acquisition times and the temperature difference between the core temperature and liquid temperature at the same acquisition time based on multiple continuously acquired core temperatures and liquid temperatures. S32 determines whether the chip's heating state is abnormal based on the temperature change rate and temperature difference. If the chip's heating is normal, it obtains the judgment result of the heat load source when the chip's heating is normal. The judgment result is a continuous steady-state load. When S33 determines that the chip is overheating abnormally, it calls the preset dual-hypothesis prediction model. The dual-hypothesis prediction model includes a first model that assumes that transient impact load causes the core temperature to rise and a second model that assumes that heat dissipation failure causes the core temperature to rise. Based on the first model, the predicted temperature of the chip is calculated and obtained. At the same time, based on the second model, the change in thermal resistance between the chip and the coolant is calculated in reverse. S34 performs unsteady-state compensation and chip thermal capacity adaptation to provide transient compensation for the change in thermal resistance; S35 calculates and obtains statistical characteristics within multiple time windows based on the predicted temperature and the change in thermal resistance after transient compensation; S36 determines the source of heat load when the chip heats up abnormally based on the statistical characteristics.

4. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 3, characterized in that, In step S33, the following steps are executed in parallel: The real-time liquid temperature, power consumption, and rapid thermal resistance value are substituted into the first model to calculate and obtain the predicted temperature; The real-time liquid temperature, core temperature, power consumption, and rapid thermal resistance value are substituted into the second model to calculate the change in thermal resistance in reverse.

5. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 4, characterized in that, In step S33, The formula for calculating the predicted temperature is as follows: in, For the predicted temperature of the chip, This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value in the dual-hypothesis prediction model. The formula for calculating the change in thermal resistance in reverse is as follows: in, This represents the change in the chip's thermal resistance. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This refers to the chip's real-time power consumption. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

6. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 3, characterized in that, Step S34 includes: S341 estimates the thermal capacity of the chip during the cold start and load drop phases of the chassis. S342 calculates the heat storage power of the chip itself by reverse calculation based on the temperature change rate and heat capacity; S343 calculates the effective power delivered by the chip to the coolant based on the power consumption and heat storage power. S344 substitutes the effective power into the second model to replace the power consumption, and calculates the change in thermal resistance after transient compensation.

7. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 6, characterized in that, In steps S342-S344, The formula for calculating the heat storage power of the chip itself is as follows: in, The chip's heat storage power is calculated in reverse. To obtain the optimal chip heat capacity by smoothing and filtering the multiple heat capacity estimates, the term multiplied by the optimal chip heat capacity is the rate of temperature change. This refers to the minute change in the core temperature of the heat-generating chip. This represents the corresponding minute change over time; The formula for calculating the effective power transferred from the chip to the coolant is as follows: in, This refers to the effective power transferred to the coolant during the chip's real-time power consumption. This refers to the chip's real-time power consumption. This refers to the chip's real-time heat storage power. The formula for calculating the change in thermal resistance after transient compensation is as follows: in, This represents the change in thermal resistance after transient compensation of the chip. This refers to the real-time core temperature of the chip. This refers to the real-time temperature of the coolant. This represents the chip's real-time effective power. This represents the rapid thermal resistance value for the dual-hypothesis prediction model.

8. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 3, characterized in that, The multiple time windows include short windows, medium windows, and long windows with different durations. In step S35, based on a preset fixed frequency, the prediction error between the predicted temperature and the core temperature, as well as the corrected change in thermal resistance, are continuously collected within each time window. Within the short window, the average prediction error of the first model is calculated based on the collected prediction errors. Within the middle window, the variance of the prediction error is calculated based on the collected prediction errors, and the average thermal resistance change of the second model is calculated based on the collected thermal resistance changes. Within the long window, the collected thermal resistance changes are linearly fitted to obtain the slope of the thermal resistance change trend.

9. The intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis according to claim 1, characterized in that, The judgment result includes at least transient impact load, continuous steady-state load, and heat dissipation failure. In step S50, When the judgment result is a transient impact load, the first pump duty cycle is calculated based on the preset reference pump speed at the start of the transient impact load and the maximum allowable pump speed under the transient impact load, and the first radiator fan duty cycle is calculated based on the liquid temperature, remaining heat storage capacity and preset first PID parameters. When the judgment result is a continuous steady-state load, the power consumption is substituted into the preset pump speed mapping table to obtain the second water pump duty cycle corresponding to the power consumption, and the second radiator fan duty cycle is calculated based on the core temperature and the preset second PID parameters. If the determination result is a heat dissipation failure, the fault type of the chip or liquid cooling system is obtained, and the control corresponding to the fault type is executed on the chip or liquid cooling system.

10. An intelligent temperature control and heat dissipation system for a liquid-cooled computer chassis, used to execute the intelligent temperature control and heat dissipation method for a liquid-cooled computer chassis as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect a first feature set inside the chassis in real time. The first feature set includes the core temperature of the chip, the power consumption of the chip, and the liquid temperature of the coolant in the liquid cooling system. The parameter calibration module is used to calibrate the thermal resistance parameters in the preset model based on the first feature set. The thermal resistance parameters include the calculated fast thermal resistance value or the preset slow thermal resistance value. The load analysis module is used to analyze and judge the heat load source of the chip using the rapid thermal resistance value and heat capacity, and obtain the judgment result; The status assessment module is used to calculate the remaining heat storage capacity of the liquid cooling system based on the liquid temperature and the preset standard coolant temperature. The strategy generation module is used to generate a set of duty cycles for water pumps and radiator fans based on the judgment result and the remaining heat storage capacity. The control execution module is used to input corresponding PWM control signals to the water pump and the radiator fan respectively according to the duty cycle, so as to control the water pump and the radiator fan to perform heat dissipation on the chip.

Images