Computing power cabinet coupling device with active thermal management and safety interlocking function and control method
By integrating a current sampling and processing unit into the liquid-cooled coupling device, feedforward control and local interlock protection based on current transient characteristics are realized, solving the problems of thermal management lag and slow safety response, and improving the operational reliability and safety of the computing cabinet.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN GUANGSHENG INTELLIGENT CONTROL TECHNOLOGY CO LTD
- Filing Date
- 2026-05-22
- Publication Date
- 2026-07-24
AI Technical Summary
Existing liquid-cooled coupling devices suffer from thermal inertia lag in thermal management, making it difficult to respond quickly to load changes in computing nodes. Furthermore, their security protection response is slow, failing to meet millisecond-level active security requirements.
By introducing transient current sensing and local logic processing capabilities into the coupling device, and integrating the current sampling unit, processing unit, flow regulation unit and controllable switching unit, feedforward regulation and millisecond-level interlock protection for faults are achieved.
It enables rapid response to thermal management of computing nodes, shortens thermal management response time to the millisecond level, improves the inherent security level of security protection, and avoids local overheating and security accidents.
Smart Images

Figure CN122458375A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of thermal management technology for computing equipment, specifically to a computing cabinet coupling device and control method with active thermal management and safety interlock functions. Background Technology
[0002] With the continuous increase in computing power density in data centers, liquid cooling technology has become a key means to solve the heat dissipation problem of high-power chips. In liquid-cooled computing cabinets, the coupling device serves as the physical interface between server nodes and cooling distribution units (CDUs), undertaking the dual functions of power transmission and fluid circulation.
[0003] Existing liquid-cooled coupling devices are typically designed as passive connection devices, only providing simple fluid on / off and electrical connection functions. In terms of thermal management, traditional control logic relies on a central controller to collect server outlet temperature or chip temperature and then adjust the flow valve opening accordingly. However, this temperature-feedback-based control method suffers from significant thermal inertia hysteresis: when the computing node load experiences a sudden jump, the chip junction temperature rises rapidly, while the fluid temperature change takes time to be transmitted to the sensor, resulting in a delayed flow regulation response and a risk of localized overheating. Furthermore, in terms of safety protection, existing coupling devices usually rely on system-level software protection strategies. When faults such as loose connectors or short circuits occur, the signal must be uploaded to the central processing unit for judgment before a power-off command is issued, resulting in a long response cycle that fails to meet the millisecond-level active safety protection requirements. Therefore, there is an urgent need for a coupling device capable of predicting thermal demands and possessing local rapid safety response capabilities. Summary of the Invention
[0004] To address the issues of delayed thermal management response and slow safety protection response in existing technologies, this application proposes a computing cabinet coupling device and control method with active thermal management and safety interlock functions. By introducing transient current sensing and local logic processing capabilities at the physical connection node, it achieves feedforward adjustment of thermal management and millisecond-level interlock protection for faults.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: A computing cabinet coupling device with active thermal management and safety interlock functions is disclosed. The device includes: a current sampling unit, a processing unit, a flow regulation unit, a controllable switch unit, and a position detection unit. The current sampling unit, the flow regulation unit, the controllable switch unit, and the position detection unit are respectively connected to the processing unit. The processing unit is configured to: output a feedforward control signal to the flow regulation unit based on the transient current characteristics acquired by the current sampling unit; and control the controllable switch unit to perform a disconnection action in response to an abnormal engagement state detected by the position detection unit.
[0006] In one implementation, the transient current characteristics include the rate of change of current; the processing unit is configured to generate a feedforward control signal based on the rate of change of current to adjust the opening of the flow regulating unit.
[0007] In one implementation, the sampling frequency of the current sampling unit is greater than 10kHz.
[0008] In one implementation, the position detection unit is used to monitor the insertion displacement of the coupling device; the processing unit is configured to control the controllable switch unit to perform a disconnection action when the offset of the insertion displacement exceeds a preset threshold.
[0009] In one embodiment, the device further includes a temperature detection unit connected to the processing unit; the processing unit is further configured to control the controllable switch unit to perform a disconnection action when the temperature of the connection point detected by the temperature detection unit exceeds a temperature threshold, or the current sampled by the current sampling unit exceeds a current safety threshold.
[0010] In one embodiment, the device also includes a housing; the housing is provided with a guide structure for guiding the coupling device to blindly align with the computing cabinet; the housing is provided with a fluid channel and a power contact, which are spatially isolated from each other.
[0011] In one implementation, fluid channels are symmetrically arranged on both sides of the housing, and electrical contacts are arranged in the central area of the housing.
[0012] As one implementation method, the outer layer of the power contact is wrapped with insulating material, which is used to achieve isolation between strong and weak currents.
[0013] Furthermore, the present invention also provides a control method for the above-mentioned computing cabinet coupling device, the method comprising: acquiring current sampling data and mating state data; determining current transient characteristics based on the current sampling data, and generating a feedforward control signal according to the current transient characteristics to adjust the opening degree of the flow regulating unit; and controlling the controllable switching unit to perform a disconnection action in response to the mating state data indicating an abnormal mating state.
[0014] As one implementation method, generating a feedforward control signal based on the transient characteristics of the current includes: generating a pulse width modulation (PWM) signal based on the rate of change of the current in the transient characteristics of the current and sending it to the flow regulation unit; controlling the controllable switching unit to perform a disconnection action includes: performing active power-off protection within a millisecond time.
[0015] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: (1) By integrating a current sampling unit and a processing unit inside the coupling device, the current change rate, which precedes temperature change, is used as a feedforward signal to directly drive the flow regulation unit. Compared with traditional temperature feedback control, this scheme can predict the heating trend and adjust the flow rate in advance at the moment of current jump, effectively solving the problem of thermal inertia lag and realizing a rapid response to instantaneous load fluctuations of computing nodes.
[0016] (2) By setting up a linkage mechanism between the position detection unit and the controllable switch unit, the safety interlock logic is pushed down to the coupler node. When the insertion displacement offset is too large or the current exceeds the limit, the physical power-off can be performed locally within milliseconds without waiting for the instruction of the central controller, avoiding safety accidents caused by communication delay or system crash, and realizing the integrated active safety protection of "perception-decision-execution".
[0017] (3) By using a space folding design, the fluid channel and the power contact are physically isolated and arranged, and the blind insertion alignment is achieved by combining the guide structure. Under the premise of ensuring the safe isolation of water and electricity, the installation accuracy and reliability of the coupling device are improved. Attached Figure Description
[0018] Figure 1 This is a structural block diagram of the computing cabinet coupling device according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the computing cabinet coupling device according to an embodiment of the present invention; Figure 3 This is a flowchart of the control method for the computing cabinet coupling device according to an embodiment of the present invention.
[0019] Explanation of reference numerals in the attached figures: 10. Current sampling unit; 20. Processing unit; 30. Flow regulation unit; 40. Controllable switch unit; 50. Position detection unit; 60. Temperature detection unit; 70. Housing; 71. Fluid channel; 72. Power contact; 73. Guide structure. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Example 1
[0022] like Figure 1 and Figure 2 As shown, this embodiment provides a computing cabinet coupling device with active thermal management and safety interlock functions. The device includes a current sampling unit 10, a processing unit 20, a flow regulation unit 30, a controllable switch unit 40, and a position detection unit 50. The current sampling unit 10, the flow regulation unit 30, the controllable switch unit 40, and the position detection unit 50 are all connected to the processing unit 20.
[0023] Specifically, in this embodiment, the coupling device is no longer a simple physical connection interface, but has evolved into an edge intelligent terminal with sensing, decision-making, and execution capabilities. The current sampling unit 10, acting as the device's "sensory nerve," is connected in series or parallel to the main power supply circuit of the coupling device to collect current signals flowing through it in real time. The processing unit 20, acting as the device's "edge brain," can be a chip with data processing capabilities, such as a microcontroller (MCU), digital signal processor (DSP), or field-programmable gate array (FPGA). Its signal input terminal is connected to the output terminal of the current sampling unit 10 to receive and analyze the current signal. The processing unit 20 is configured to output a feedforward control signal to the flow regulation unit 30 based on the transient current characteristics acquired by the current sampling unit 10. This breaks the lag chain of "temperature rise - feedback regulation" in traditional technology. By analyzing transient current characteristics (such as sudden current changes), the processing unit 20 can predict the heating trend of the computing node and actively send instructions to the flow regulation unit 30 before the temperature rises significantly. The flow regulation unit 30 can be an electric regulating valve or a proportional valve, which is installed in the fluid channel of the coupling device. Its control end receives the feedforward control signal from the processing unit 20 and adjusts the flow opening of the coolant in real time, thereby realizing the "prevention" of thermal management.
[0024] Meanwhile, the position detection unit 50 monitors the engagement status of the coupling device. This position detection unit 50 can be a Hall sensor, photoelectric sensor, or displacement sensor, located inside the housing of the coupling device or near the connector, to detect engagement displacement or connection status between the coupling device and the computing cabinet backplane. The processing unit 20 is connected to the position detection unit 50 and, in response to an abnormal engagement status detected by the position detection unit 50, controls the controllable switch unit 40 to perform a disconnection action. The controllable switch unit 40, as the component performing the safety disconnection, can be an electronic switch such as a MOSFET (Metal-Oxide-Semiconductor Field-Effect Transistor), IGBT (Insulated Gate Bipolar Transistor), or solid-state relay, connected in series in the main power supply circuit of the coupling device. When the position detection unit 50 detects excessive engagement displacement, loose connection, or incomplete engagement, the processing unit 20 immediately sends a disconnection command to the controllable switch unit 40, which then physically disconnects the power supply circuit. This localized perception-decision-execution mechanism allows the coupling device to complete fault isolation within milliseconds without waiting for instructions from the central control system, significantly improving the inherent safety level of the system.
[0025] Through the above architecture, the coupling device in this embodiment achieves a deep integration of power transmission, fluid transmission, and intelligent control. The processing unit 20 acquires the current signal and the position signal respectively, and outputs control commands, enabling the device to actively respond to thermal inertia hysteresis problems and respond quickly when physical connection abnormalities occur, thus solving the problems of existing couplers having single functions and passive responses. Example 2
[0026] This embodiment, based on Embodiment 1, elaborates on the feedforward control logic of thermal management. Specifically, the transient current characteristics include the rate of change of current; the processing unit 20 is configured to generate a feedforward control signal based on the rate of change of current to adjust the opening degree of the flow regulating unit 30.
[0027] Its core principle lies in establishing a cross-domain mapping relationship between electricity and heat. During the operation of computing nodes, current signals and temperature signals have significantly different physical characteristics. When the load changes abruptly, such as at the moment AI model training starts, the current will jump within microseconds. However, due to the thermal capacity of the chip and the thermal inertia of the heat dissipation system, the temperature rise usually lags behind by tens of milliseconds or even several seconds. Traditional temperature feedback control can only adjust the flow rate after the temperature rises, which cannot eliminate this lag time. In this embodiment, the processing unit 20 can directly capture the transient characteristics of load changes by monitoring the rate of change of current, that is, the derivative of current with respect to time. The rate of change of current leads the temperature change, and the processing unit 20 generates a feedforward control signal accordingly, driving the flow regulation unit 30 to increase the coolant flow rate before the chip temperature rises significantly. This "predictive" regulation essentially uses the leading nature of electrical signals to compensate for the lag of thermal signals, thereby shortening the response time of thermal management from the traditional "second level" to the "millisecond level" or even the "microsecond level", effectively avoiding the risk of frequency reduction or downtime of computing nodes due to instantaneous overheating.
[0028] To accurately capture the aforementioned microsecond-level transient current characteristics, the sampling frequency of the current sampling unit 10 needs to meet specific requirements. In this embodiment, the sampling frequency of the current sampling unit 10 is greater than 10kHz.
[0029] Specifically, load current jumps at computing nodes typically occur within extremely short timeframes, such as tens to hundreds of microseconds. If the sampling frequency is too low, the transient details of the current waveform cannot be fully reproduced, leading to distortion in the current change rate calculation or even the loss of critical signals. For comparison, assuming traditional 50Hz power frequency sampling (i.e., a sampling interval of 20ms), for a current surge signal lasting 100 microseconds, this sampling mechanism is highly likely to completely miss the surge during the two sampling intervals, or only capture the steady-state value after the surge, thus failing to calculate the true current change rate and causing feedforward control failure. Therefore, this embodiment sets the sampling frequency to be greater than 10kHz (i.e., a sampling interval less than 100 microseconds), ensuring the capture of the rising and falling edges of the current waveform, providing an accurate basis for differential calculations for the processing unit 20. It should be understood that the specific value of the sampling frequency can be adaptively adjusted according to the load characteristics of the computing node. For example, in scenarios with extremely drastic load fluctuations, the sampling frequency can be further increased to 100kHz or even higher to ensure the fidelity of signal acquisition. Example 3
[0030] This embodiment, based on Embodiment 1, elaborates on the safety interlock protection mechanism. Specifically, the position detection unit 50 is used to monitor the insertion displacement of the coupling device; the processing unit 20 is configured to control the controllable switch unit 40 to perform a disconnection action when the offset of the insertion displacement exceeds a preset threshold.
[0031] Specifically, the position detection unit 50 can employ devices such as Hall sensors, photoelectric sensors, or miniature displacement switches. Taking a Hall sensor as an example, it is installed inside the housing of the coupling device and senses the relative position change between the coupling device and the computing cabinet backplane by detecting changes in magnetic field strength. When the coupling device becomes loose due to vibration, accidental contact, or improper installation, a displacement offset will occur between its actual insertion position and the standard position. The processing unit 20 receives the displacement signal fed back by the position detection unit 50 in real time and compares it with a preset safety threshold. This preset threshold can be set according to the tolerance range of the contact, for example, 0.5mm or 1mm. Once the displacement offset exceeds this threshold, it means that the power contact may have poor contact, posing a risk of increased contact resistance, local overheating, or even arcing and fire. At this time, the processing unit 20 immediately sends a disconnect command to the controllable switch unit 40. The controllable switch unit 40 preferably uses solid-state switching devices such as MOSFETs (Metal-Oxide-Semiconductor Field-Effect Transistors) or IGBTs (Insulated Gate Bipolar Transistors), whose response speed is much faster than that of traditional mechanical relays. By controlling the gate voltage of the MOSFET, the power supply circuit can be cut off within microseconds, thereby achieving physical isolation in the early stages of a fault and preventing the accident from escalating.
[0032] Furthermore, in order to build a multi-dimensional and all-round security defense system, the coupling device in this embodiment also includes a temperature detection unit 60, which is connected to the processing unit 20. The processing unit 20 is also configured to control the controllable switch unit 40 to perform a disconnection action when the temperature of the connection part detected by the temperature detection unit 60 exceeds the temperature threshold, or the current sampled by the current sampling unit 10 exceeds the current safety threshold.
[0033] The temperature detection unit 60 can be a thermistor (NTC / PTC) or thermocouple, which is mounted near the power contact to directly monitor the temperature of critical heat-generating components. The current sampling unit 10 continuously monitors the circuit current. The processing unit 20 has preset temperature and current safety thresholds, forming a triple interlocking logic of "displacement + temperature + current". When any parameter is abnormal, the interlocking mechanism is triggered immediately. For example, when the current sampling unit 10 detects a short-circuit current (the current instantaneously exceeds the safety threshold), the processing unit 20 can identify and drive the controllable switch unit 40 to disconnect within microseconds, with a response speed much faster than traditional fuses or circuit breakers. Similarly, when the temperature detection unit 60 detects an abnormal rise in contact temperature, even if the displacement and current are not exceeded, the processing unit 20 will still perform disconnection protection to prevent thermal runaway caused by a slow increase in contact resistance.
[0034] The core advantage of this embodiment lies in its localized proactive protection. Traditional security protection mechanisms typically rely on a central controller or upper-level software system, which involves long signal transmission paths and numerous processing stages, with typical response times in the millisecond or even second range. For transient faults such as arcing and short circuits that occur in microseconds, traditional mechanisms often struggle to effectively cut off power before the incident. This embodiment, however, integrates sensing, decision-making, and execution capabilities within the physical node of the coupling device. The processing unit 20 directly reads sensor data and controls the switch action without going through a communication bus, thus achieving millisecond- or even microsecond-level power-off protection. This hardware-level local interlocking mechanism completely avoids the risks of response lag caused by communication delays, software crashes, or system overload, significantly improving the inherent safety level of the computing cabinet operation. Example 4
[0035] This embodiment, based on Embodiment 1, elaborates on the physical carrier structure of the coupling device. Specifically, the device further includes a housing 70; the housing 70 is provided with a guide structure 73, which is used to guide the coupling device to blindly align with the computing cabinet; the housing 70 is provided with a fluid channel 71 and a power contact 72, which are spatially isolated from each other.
[0036] Specifically, the housing 70, serving as the carrier of the coupling device, is integrally molded from a high-strength insulating material (such as high-performance engineering plastics or precision-cast alloys) to encapsulate the internal electrical components and fluid conduits. The guide structure 73 can be a guide post, guide groove, or mortise and tenon structure located on the outer periphery of the housing 70, adapted to the mating structure on the computing cabinet backplane. During blind insertion operations, the guide structure 73 automatically corrects the insertion angle and position of the coupling device, ensuring simultaneous and precise docking of the fluid interface and electrical interface. For example, the guide structure 73 can employ a tapered guide post design with tolerances controlled at the 0.5mm level, enabling precise insertion and removal even when maintenance personnel cannot directly view the backplane, effectively avoiding connector wear or poor contact caused by misalignment.
[0037] Furthermore, to fundamentally eliminate the risk of short circuits caused by leaks in the liquid cooling system, a strict physical isolation design is adopted inside the housing 70. Fluid channels 71 are symmetrically arranged on both sides of the housing 70, and electrical contacts 72 are located in the central area of the housing 70.
[0038] This spatially folded layout of "left and right water channels, central circuit" constructs a physical safety barrier. The fluid channel 71 circulates coolant and has a self-sealing valve at its interface that automatically closes to prevent leakage when the coupling device is pulled out. The power contact 72, responsible for transmitting high current, is located in the central area, ensuring even force distribution and maintaining sufficient safety distance from the fluid channels 71 on both sides. Even in extreme cases, such as a minor leak in the fluid channel 71, the coolant will flow towards the edge of the casing 70 under gravity or structural guidance, making it unlikely to come into contact with the charged body in the central area, thus achieving an inherently safe "water-electricity separation" design. It should be understood that while this embodiment preferably features a symmetrical layout, in other embodiments, the fluid channel 71 and the power contact 72 can also be isolated by vertical layering or staggered front-to-back separation, as long as physical spatial isolation is satisfied.
[0039] In addition, to further enhance electrical safety, the power contact 72 is wrapped with an insulating material, which is used to isolate strong and weak currents.
[0040] The insulating material is preferably a high-temperature resistant and highly insulating polymer such as polyetheretherketone (PEEK) or polyimide (PI). The power contact 72 typically carries high voltage and high current, belonging to a high-voltage circuit; while the processing unit 20, sensors, and other control circuits belong to a low-voltage circuit. By wrapping the power contact 72 with insulating material, the high-voltage circuit is isolated from the external environment, preventing the risk of electric shock. Furthermore, within the limited space inside the housing 70, electromagnetic interference from the high-voltage circuit to the low-voltage control circuit is effectively blocked, ensuring the stability of signal transmission from the processing unit 20 and various sensors. This structural design enables the coupling device to possess both active thermal management and safety interlocking functions, along with highly reliable physical protection capabilities, adapting to the harsh operating environment of high-density computing cabinets. Example 5
[0041] This embodiment provides a control method for the above-mentioned computing cabinet coupling device, which is applied to, for example... Figure 1 and Figure 2 The processing unit 20 in the coupling device shown. This method protects the control flow of the device from a timing logic perspective, such as... Figure 3 As shown, the specific steps include: Step S100: Acquire current sampling data and mating status data.
[0042] Specifically, the processing unit 20 reads the current sampling data output by the current sampling unit 10 and the mating status data output by the position detection unit 50 in real time via an internal bus or GPIO interface. The current sampling data reflects the real-time load current waveform of the computing node and serves as the input source for thermal management feedforward control; the mating status data reflects the physical reliability of the connection between the coupling device and the backplane and serves as the basis for determining safety interlock protection. The acquisition of these two data points is continuous, ensuring that the processing unit 20 always has access to the latest operating status.
[0043] Step S200: Determine the transient characteristics of the current based on the current sampling data, and generate a feedforward control signal based on the transient characteristics of the current to adjust the opening degree of the flow regulating unit 30.
[0044] Specifically, the processing unit 20 performs numerical analysis on the acquired current sampling data, calculating its transient current characteristics, such as the derivative of the current with respect to time, i.e., the rate of change of the current. Since changes in the current signal precede changes in temperature, when the processing unit 20 detects a sudden change in the transient current characteristics, it can immediately generate a feedforward control signal without waiting for feedback from the temperature sensor. This feedforward control signal is transmitted to the flow regulation unit 30 via a drive circuit, driving the flow regulation unit 30 to increase or decrease the valve opening. For example, when the rate of change of the current exceeds a preset threshold, the processing unit 20 generates an instruction to increase the opening, increasing the coolant flow in advance to cope with the upcoming heat load. This step achieves "predictive" adjustment in thermal management, effectively compensating for thermal inertia hysteresis.
[0045] In step S300, in response to the mating state data indicating an abnormal mating state, the controllable switch unit 40 is controlled to perform a disconnection action.
[0046] Specifically, the processing unit 20 compares the acquired mating status data with the preset normal mating range. When the mating status data is abnormal, such as when the displacement exceeds a safety threshold or the mating signal is lost, the processing unit 20 determines that the mating status is abnormal. At this time, the processing unit 20 immediately outputs a cut-off signal to the controllable switch unit 40, controlling the controllable switch unit 40 to physically disconnect the power supply circuit. This step achieves localized fault isolation, and the response speed is limited only by the processing delay of the hardware circuit, which is much faster than traditional software protection logic.
[0047] It is important to emphasize that steps S200 and S300 are not simply sequential in execution logic, but rather performed in parallel or collaboratively. Processing unit 20, as an edge computing node, possesses multi-tasking capabilities. Under normal operating conditions, processing unit 20 continuously executes step S200, dynamically adjusting traffic; simultaneously, the security monitoring logic of step S300 runs in the background. Once step S300 detects an anomaly, its priority is higher than step S200, and processing unit 20 will prioritize the disconnection action, ensuring the immediacy of the security interlock. This parallel and collaborative control logic enables the coupling device to maintain thermal management accuracy while possessing millisecond-level security response capabilities, fully demonstrating the technological advantages of edge intelligent terminals. Example 6
[0048] This embodiment, based on embodiment 5, elaborates on the signal generation and execution timing in the control method. Specifically, a pulse width modulation (PWM) signal is generated based on the current change rate in the transient current characteristics and sent to the flow regulation unit 30.
[0049] Specifically, the processing unit 20 internally stores a mapping table or function model between the current change rate and the PWM duty cycle. A PWM signal is a digital signal that controls the average output voltage or current by adjusting the pulse width, characterized by strong anti-interference capability and high control precision. After calculating the current change rate, the processing unit 20 determines the target PWM duty cycle by looking up a table or performing calculations. For example, when the current change rate is small (e.g., dI / dt < 1A / ms), the processing unit 20 generates a PWM signal with a 30% duty cycle to control the flow regulating unit 30 to maintain a basic opening; when the current change rate is large (e.g., dI / dt > 10A / ms), the processing unit 20 generates a PWM signal with a 90% or even 100% duty cycle to drive the flow regulating unit 30 to quickly open fully to cope with transient thermal shocks. The drive circuit inside the flow regulating unit 30 receives this PWM signal and drives the valve motor or solenoid coil to operate through a power amplifier circuit. Because PWM signals typically have high frequencies (e.g., 1kHz to 10kHz), the flow regulation unit 30 can achieve smooth and continuous flow regulation, avoiding flow fluctuations and mechanical shocks caused by traditional switching control. This direct mapping control based on the rate of change of current eliminates the complex PID calculation process, greatly shortens the response time of the control loop, and enables the thermal management system to keep up with millisecond-level fluctuations in computing load.
[0050] Furthermore, regarding the execution process of the safety interlock, the controllable switch unit 40 performs a disconnection action, including: performing active power-off protection within milliseconds.
[0051] Specifically, traditional power failure protection often relies on the software logic judgment of the upper-level controller. This results in long signal transmission paths and is easily affected by system load, leading to communication delays, typically in the tens or even hundreds of milliseconds range. In this embodiment, however, the processing unit 20 is directly integrated into the coupling device and directly connected to the controllable switching unit 40 via GPIO (General Purpose Input / Output) or a dedicated drive circuit. When step S300 determines that the engagement state is abnormal, or the temperature reported by the temperature detection unit 60 exceeds a threshold, or the current sampling unit 10 detects a short-circuit current, the processing unit 20 immediately triggers an interrupt service routine, pulling down or raising the control level to directly drive the gate of the controllable switching unit 40 (such as a MOSFET). This process does not require a complex communication protocol stack and is only limited by the level transition time of the hardware circuit and the action time of the switching device. For example, after detecting a fault signal, the processing unit 20 can complete the circuit breaking action within 10ms; and in a preferred high-speed hardware circuit configuration, the response time can be shortened to the microsecond level. This millisecond- or even microsecond-level active power-off protection can instantly cut off the power supply in the early stages of faults such as arcing and short circuits, effectively preventing equipment burnout or fire accidents caused by continuous energy injection, and significantly improving the safety of the computing cabinet operation. Example 7
[0052] This embodiment describes in detail the actual operating effect of the above-mentioned computing cabinet coupling device in a specific application scenario. The application scenario is set as a high-density computing cabinet, which deploys multiple high-power server nodes to perform complex tasks such as AI large model training and scientific computing.
[0053] Scenario 1: Sudden increase in server node load.
[0054] During AI model training, computing nodes often switch from standby to full-load operation in a very short time. Specifically, when the model begins loading and performing matrix operations, the current flowing through the coupling device undergoes drastic changes within microseconds. For example, the current sampling unit 10 detects a sudden jump in current from 10A to 50A, a very high rate of change. Because the sampling frequency of the current sampling unit 10 is greater than 10kHz, the processing unit 20 can accurately capture this microsecond-level transient current characteristic and immediately calculate the rate of change.
[0055] Based on the current change rate, processing unit 20 quickly generates a feedforward control signal and sends it to flow regulation unit 30. At this time, the chip temperature of the server node has not yet risen significantly, and traditional temperature feedback control has not yet been activated. However, the coupling device in this embodiment has already "predicted" the upcoming heat load, and the flow regulation unit 30 increases the valve opening the instant it receives the signal, increasing the coolant flow rate from the base flow rate to the peak flow rate. This current-feedforward-based thermal management strategy successfully shortens the thermal management response time from the "second level" of traditional solutions to the "millisecond level," effectively avoiding the problem of instantaneous overheating and frequency reduction caused by thermal inertia hysteresis of the chip, and ensuring the continuity and stability of AI training tasks.
[0056] Scenario 2: The coupler becomes loose due to accidental contact by maintenance personnel.
[0057] During the maintenance of high-density computing cabinets, maintenance personnel may need to operate in confined spaces, posing a risk of accidental contact with the coupling device. Suppose a maintenance personnel accidentally bumps into the housing 70 of the coupling device, causing a slight shift in the mating position. The position detection unit 50 (such as a built-in Hall sensor) monitors the mating displacement in real time. When the detected displacement exceeds a preset safety threshold (e.g., 0.5mm), it immediately sends an abnormal signal to the processing unit 20.
[0058] Upon receiving an abnormal signal, processing unit 20 determines that the engagement state is abnormal and immediately triggers the safety interlock logic. At this time, power contact 72 may already show signs of increased contact resistance or slight arcing. Within milliseconds (e.g., within 10ms), processing unit 20 sends a disconnect command to controllable switching unit 40, which (e.g., MOSFET) quickly and physically cuts off the power supply circuit. This process is completed entirely locally within the coupling device, without the need for communication transmission and software judgment by the central controller, thus completely avoiding the risks caused by communication delays. Simultaneously, if temperature detection unit 60 detects an abnormal increase in temperature at the connection point, or current sampling unit 10 detects a short-circuit current, processing unit 20 will also execute power-off protection, constructing a triple safety defense line of "displacement + temperature + current".
[0059] As seen in the above application scenarios, the computing cabinet coupling device provided in this embodiment, by integrating units such as current sampling, processing, flow regulation, and controllable switches, successfully upgrades the traditional passive connection interface into an active management terminal with edge intelligence. In terms of thermal management, a current feedforward mechanism overcomes thermal inertia hysteresis; in terms of security protection, a local interlocking mechanism achieves millisecond-level active power-off. This significantly improves the reliability and security of high-density computing cabinet operation.
[0060] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A computing cabinet coupling device with active thermal management and safety interlock functions, characterized in that, include: The system includes a current sampling unit, a processing unit, a flow regulation unit, a controllable switch unit, and a position detection unit. The current sampling unit, the flow regulation unit, the controllable switch unit, and the position detection unit are respectively connected to the processing unit; The processing unit is configured to: output a feedforward control signal to the flow regulation unit based on the transient current characteristics acquired by the current sampling unit; and control the controllable switch unit to perform a disconnection action in response to an abnormal engagement state detected by the position detection unit.
2. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 1, characterized in that, The transient characteristics of the current include the rate of change of the current; The processing unit is configured to generate the feedforward control signal based on the current change rate to adjust the opening degree of the flow regulation unit.
3. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 2, characterized in that, The sampling frequency of the current sampling unit is greater than 10kHz.
4. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 1, characterized in that, The position detection unit is used to monitor the insertion displacement of the coupling device; The processing unit is configured to control the controllable switch unit to perform a disconnection action when the offset of the interlocking displacement exceeds a preset threshold.
5. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 4, characterized in that, It also includes a temperature detection unit, which is connected to the processing unit; The processing unit is further configured to control the controllable switch unit to perform a disconnection action when the temperature of the connection point detected by the temperature detection unit exceeds the temperature threshold, or when the current sampled by the current sampling unit exceeds the current safety threshold.
6. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 1, characterized in that, It also includes the casing; The housing is provided with a guide structure, which is used to guide the coupling device to be blindly aligned with the computing cabinet. The housing contains a fluid channel and an electrical contact, which are spatially isolated from each other.
7. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 6, characterized in that, The fluid channels are symmetrically arranged on both sides of the housing, and the power contacts are arranged in the central area of the housing.
8. The computing cabinet coupling device with active thermal management and safety interlock functions according to claim 6, characterized in that, The power contact is wrapped with an insulating material, which is used to achieve isolation between strong and weak currents.
9. A control method for the coupling device according to any one of claims 1 to 8, characterized in that, The method includes: Acquire current sampling data and interlocking status data; The transient characteristics of the current are determined based on the current sampling data, and a feedforward control signal is generated according to the transient characteristics of the current to adjust the opening degree of the flow regulation unit. In response to the engagement state data indicating an abnormal engagement state, the controllable switch unit is controlled to perform a disconnection action.
10. The control method according to claim 9, characterized in that, The step of generating a feedforward control signal based on the transient characteristics of the current includes: Based on the rate of change of current in the transient current characteristics, a pulse width modulation (PWM) signal is generated and sent to the flow regulation unit; The controllable switch unit performs a disconnection action, including: Active power-off protection is performed within milliseconds.