A liquid cooling pressure self-compensation control method and system based on double UQD check valve

Abstract

The application discloses a liquid cooling pressure self-compensation control method based on a double UQD check valve, which comprises the following steps: collecting pressure signals at both ends of the check valve and performing adaptive denoising; calculating the pressure fluctuation change rate by using a five-point cubic smoothing numerical differential algorithm, and simultaneously identifying characteristic resonance frequency energy by using fast Fourier transform and power spectral density analysis; adopting an energy weighted fusion algorithm to fuse the above two characteristic values, and calculating a normalized fluctuation energy evaluation index; when the index exceeds a preset safety threshold, a central controller calculates a compensation amount based on a variable gain adaptive control algorithm, and sends an instruction to a drive frequency converter of a circulating pump to adjust the pump speed to generate a reverse pressure wave, so as to actively offset the pipeline pressure pulse caused by the check valve action. Through the collection of pressure signals, the extraction of composite features, the fusion of multi-dimensional energy and the execution of dynamic compensation, the application realizes intelligent perception and accurate control of pressure fluctuation, and effectively suppresses the pipeline pressure pulse generated by the check valve action.

Description

Technical Field

[0001] This application relates to the field of industrial fluid transport and process control technology, and more specifically, to a liquid cooling pressure self-compensation control method and system based on a dual UQD check valve. Background Technology

[0002] In the field of industrial fluid transport and process control technology, precision liquid cooling systems are widely used in scenarios with stringent requirements for pressure stability and quiet operation, such as data center servers, new energy battery packs, and high-power lasers. As core components for maintaining coolant flow and heat exchange, the operational stability of circulating pumps and pipeline valves directly affects system reliability.

[0003] Check valves play a crucial role in preventing coolant backflow and ensuring unidirectional system operation. However, during the start-up and shutdown of the circulating pump or flow regulation, and at the moment the check valve core closes rapidly, severe pressure shocks (water hammer effect) and continuous pressure pulsations can easily occur in the pipeline. These pressure fluctuations not only cause significant noise and vibration, interfering with normal equipment operation, but also, over time, can lead to loosening of pipe joints, inaccurate sensor measurements, and even serious malfunctions such as coolant leakage and pump cavitation, severely threatening the service life and safety performance of the liquid cooling system.

[0004] Existing technologies offer various solutions to these problems, but all suffer from insurmountable drawbacks. Passive damping silencers rely on internal air bladders or elastic diaphragms to absorb pressure wave energy; however, their response is slow, they are only effective for pulses of specific frequencies, and their large size occupies installation space, making them difficult to deploy in compact liquid cooling systems. Long-term use leads to aging of internal components, causing damping characteristics to drift, resulting in high maintenance costs. Variable frequency pumps employ simple PID pressure closed-loop control, adjusting pump speed through pressure sensor feedback. However, because the sensor is typically located far from the check valve, the detection signal experiences significant delay. The PID algorithm responds slowly to instantaneous high-frequency pressure spikes generated by the check valve's action, and the integral term is prone to overcompensation and low-frequency oscillations, exacerbating system instability. Specialized check valve structures, such as swashplate or spring-assisted closing types, attempt to slow the valve core's closing speed but sacrifice the check valve's rapid shut-off characteristics, leading to increased reverse leakage or failure to meet emergency shut-off requirements. Their improvement effect is limited, and they lack active adjustment capabilities.

[0005] A deeper technical bottleneck lies in the fact that existing solutions cannot achieve intelligent perception and precise control of pressure fluctuation characteristics: the pressure pulsations triggered by the check valve action contain characteristic resonant components coupled with the pipeline's inherent frequency, which is the main cause of continuous vibration and noise. However, existing methods process the pressure signal as a whole, failing to effectively separate and quantify this key component from the complex waveform, resulting in a lack of targeted control actions; risk assessment relies solely on a single pressure absolute value exceeding the limit, lacking multi-dimensional dynamic indicators that integrate the rate of pressure change (reflecting instantaneous impact intensity) and characteristic resonant energy (reflecting vibration intensity), and the safety threshold remains fixed, failing to adapt to changes in the pump's tolerance to pressure fluctuations under different speeds and system loads; compensation control uses a fixed gain mechanism, which is insufficient in response to slight fluctuations and causes oscillations due to overcompensation during severe impacts, lacking the adaptive capability to dynamically adjust the compensation intensity based on the real-time pressure difference across the check valve.

[0006] The aforementioned defects make it difficult for existing technologies to effectively suppress harmful pressure pulsations caused by the check valve's operation while maintaining its rapid shut-off performance, resulting in severe challenges for liquid cooling systems in terms of quietness, stability, and long service life. Summary of the Invention

[0007] The purpose of this application is to provide a liquid cooling pressure self-compensation control method and system based on dual UQD check valves, which has the advantages of effectively suppressing pressure pulses caused by check valve operation, reducing noise and vibration, and improving the stability and reliability of the liquid cooling system.

[0008] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0009] A self-compensating control method for liquid cooling pressure based on dual UQD check valves includes the following steps:

[0010] S1. Raw signal co-acquisition and preprocessing: The raw fluid pressure signal is acquired using pressure sensors installed at both ends of the dual UQD check valve, and the raw fluid pressure signal is subjected to adaptive noise reduction processing to obtain preprocessed pressure sequence data;

[0011] S2. Composite Feature Extraction and Time-Frequency Analysis: The first derivative of the pressure sequence data is calculated using a five-point cubic smoothing numerical differential algorithm to obtain the pressure fluctuation rate; at the same time, the pressure sequence data is converted from the time domain to the frequency domain using a fast Fourier transform, and the characteristic resonant frequency energy caused by the action at both ends of the double UQD check valve is identified by calculating the power spectral density.

[0012] S3. Multidimensional Energy Fusion and Turbulence Assessment: The central controller acquires the pressure fluctuation rate and characteristic resonant frequency energy, and uses the built-in energy weighted fusion algorithm to calculate the instantaneous pressure fluctuation energy value; wherein, the energy weighted fusion algorithm performs gain processing on the pressure change rate by introducing a nonlinear correction index, and combines the distribution intensity of the characteristic resonant frequency energy in the frequency domain to output a normalized fluctuation energy assessment index that reflects the transient impact intensity of the pipeline;

[0013] S4. Dynamic Compensation Execution and Feedback Closed Loop: When the normalized fluctuation energy assessment index exceeds the preset safety threshold, the central controller sends a compensation control command to the drive inverter of the circulating pump, and uses a variable gain adaptive control algorithm to adjust the motor speed frequency of the circulating pump in real time, thereby generating a reverse pressure wave to offset the pipeline pressure pulse generated by the action of the dual UQD check valve.

[0014] Furthermore, the adaptive denoising process uses the following formula:

[0015]

[0016] in, This is the preprocessed pressure sequence data. Sampling time, The filter order is... For sampling offset index, These are weighting coefficients that are dynamically adjusted based on local variance. This is the original pressure signal at the historical sampling time.

[0017] Furthermore, the specific method for step S2 is as follows:

[0018] S21. Calculate the rate of change of pressure fluctuation using the following five-point cubic smoothing numerical differential algorithm formula;

[0019]

[0020] in, Sampling time The rate of change of pressure fluctuations The sampling time interval of the pressure sensor. and To use the sampling time The sampled values ​​are centrally symmetric distributions of the pressure sequence.

[0021] S22. The pressure sequence data is converted from the time domain to the frequency domain using the following Fast Fourier Transform formula;

[0022]

[0023] in, Discrete frequency points Complex spectral values ​​at that location, This is the nth discrete sampled pressure data after preprocessing. Here is the preset windowing function, and N is the total number of sample points. It is a complex exponential basis function;

[0024] S23. Calculate the power spectral density using the following formula and identify the characteristic resonance frequency energy;

[0025]

[0026] in, The power spectral density is represented at discrete frequency points. The energy intensity of the signal at that location; Where N is the sampling frequency; N is the total number of sample points; Discrete frequency points Complex spectral values ​​at;

[0027]

[0028] in, The characteristic resonant frequency energy, and These represent the lower and upper limits of the characteristic frequency range. For power spectral density, This refers to the frequency resolution.

[0029] Furthermore, the energy-weighted fusion algorithm calculates the normalized fluctuation energy assessment index using the following formula:

[0030]

[0031] in, The normalized fluctuation energy assessment index. Sampling time The rate of change of pressure fluctuations The preset rate reference value, The characteristic resonant frequency energy, The reference value for the preset characteristic resonant frequency energy. and This is the weighted fusion coefficient.

[0032] Furthermore, in step S4, the preset safety threshold is a dynamic safety threshold boundary, used to dynamically adjust the safety boundary according to the current system operating conditions, thereby achieving adaptive adjustment of the safety threshold. The calculation formula for the dynamic safety threshold boundary is:

[0033]

[0034] in, For dynamic security threshold boundaries; This is the baseline value for the static safety threshold under standard rated operating conditions; This is the frequency correction factor, which indicates the sensitivity to pressure fluctuations at different operating frequencies; This represents the current real-time operating frequency of the circulating pump; This indicates the rated maximum operating frequency of the circulating pump; For load disturbance operators; This represents the current real-time load rate of the circulating pump.

[0035] Furthermore, in step S4, the variable gain adaptive control algorithm calculates the motor frequency adjustment increment finally sent to the drive inverter of the circulating pump using the following formula:

[0036]

[0037] in, The motor frequency adjustment increment sent to the drive inverter of the circulating pump; It is a time-varying adaptive gain operator used to adjust the compensation intensity in real time according to the pipeline pressure status; The normalized fluctuation energy assessment index; The safety threshold boundary is calculated at the current sampling time; The differential feedforward gain constant is used to anticipate changes in the resonant energy. The characteristic resonant frequency energy.

[0038] Furthermore, the time-varying adaptive gain operator is updated and calculated in real time according to the following formula:

[0039]

[0040] in, It is a time-varying adaptive gain operator; This is the preset initial gain constant; This is the preset fluid damping adjustment coefficient; This is the absolute value of the real-time pressure difference across the two ends of the double UQD check valve.

[0041] A liquid cooling pressure self-compensation control system based on dual UQD check valves, including

[0042] Pressure signal acquisition module: used to acquire real-time pressure signals at the inlet and outlet of the dual UQD check valve, the pressure difference between the two ends of the check valve, and the operating frequency signal of the circulating pump, providing raw time-series data for pressure fluctuation analysis;

[0043] Pressure fluctuation feature extraction module: used to filter and denoise the acquired raw pressure signal, separate and extract the characteristic resonance component in the pipeline pressure fluctuation, and calculate the energy of the identified characteristic resonance frequency.

[0044] Normalized energy calculation module: It is used to normalize the total energy of the extracted full-band pressure fluctuation, calculate the normalized fluctuation energy assessment index, and compare the normalized fluctuation energy assessment index with the preset safety threshold boundary to determine the risk level of the current pipeline pressure fluctuation.

[0045] Variable gain adaptive calculation module: It is used to dynamically update the time-varying adaptive gain operator based on the real-time pressure difference between the two ends of the dual UQD check valve and the preset fluid damping adjustment coefficient, and calculate the motor frequency adjustment increment of the circulating pump drive frequency converter by combining the deviation between the normalized fluctuation energy assessment index and the safety threshold and the change trend of characteristic resonant frequency energy.

[0046] Pump and valve coordinated control module: It is used to superimpose the calculated frequency adjustment increment onto the current operating frequency of the circulating pump to obtain the target speed frequency, generate the corresponding control signal and send it to the drive frequency converter, and output a reverse pressure wave by adjusting the speed of the circulating pump to cancel the pressure pulse generated by the action of the check valve in the pipeline.

[0047] In summary, the present invention has the following beneficial effects:

[0048] By acquiring pressure signals, extracting composite features, fusing multidimensional energy, and performing dynamic compensation, intelligent sensing and precise control of pressure fluctuations are achieved, effectively suppressing pipeline pressure pulses generated by check valve operation. It has the advantages of efficiently suppressing pressure pulses caused by check valve operation, reducing noise and vibration, and improving the stability and reliability of the liquid cooling system. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the liquid cooling pressure pipeline based on the dual UQD check valve described in this invention.

[0050] Figure 2 This is a system module diagram of the liquid cooling pressure self-compensation control described in this invention.

[0051] Figure 3 This is a flowchart of the liquid cooling pressure self-compensation control method described in this invention. Detailed Implementation

[0052] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to the figures and specific embodiments.

[0053] like Figures 1 to 3As shown, the present invention proposes a liquid cooling pressure self-compensation control method based on dual UQD check valves, the specific steps of which are as follows:

[0054] S1. Raw signal co-acquisition and preprocessing: The raw fluid pressure signal is acquired by the pressure sensor installed at both ends of the double UQD check valve, and the raw fluid pressure signal is subjected to adaptive noise reduction processing to obtain preprocessed pressure sequence data;

[0055] S2. Composite Feature Extraction and Time-Frequency Analysis: The first derivative of the pressure sequence data is calculated using a five-point cubic smoothing numerical differential algorithm to obtain the rate of pressure fluctuation. At the same time, the pressure sequence data is converted from the time domain to the frequency domain using a fast Fourier transform, and the characteristic resonant frequency energy caused by the action at both ends of the double UQD check valve is identified by calculating the power spectral density.

[0056] S3. Multidimensional Energy Fusion and Turbulence Assessment: The central controller acquires the pressure fluctuation rate and characteristic resonant frequency energy, and uses the built-in energy weighted fusion algorithm to calculate the instantaneous pressure fluctuation energy value. The energy weighted fusion algorithm introduces a nonlinear correction index to perform gain processing on the pressure change rate, and combines the distribution intensity of the characteristic resonant frequency energy in the frequency domain to output a normalized fluctuation energy assessment index that reflects the transient impact intensity of the pipeline.

[0057] S4. Dynamic Compensation Execution and Feedback Closed Loop: When the normalized fluctuation energy assessment index exceeds the preset safety threshold, the central controller sends a compensation control command to the drive inverter of the circulating pump. The variable gain adaptive control algorithm is used to adjust the motor speed frequency of the circulating pump in real time, and the reverse pressure wave is generated to offset the pipeline pressure pulse caused by the action of the double UQD check valve.

[0058] For ease of understanding, the following explains some key terms in this embodiment:

[0059] Dual UQD check valve: This is an ultra-fast dry check valve designed to shut off quickly to prevent backflow of fluid. However, its rapid shut-off characteristic may also cause transient pressure surges in the pipeline.

[0060] Five-point cubic smoothing numerical differentiation algorithm: This algorithm is a numerical method for calculating the derivative of discrete data sequences. By performing a cubic polynomial fit on five consecutive data points and then differentiating from it, noise in the data can be effectively smoothed, while accurately capturing the data change trend, thereby calculating the first derivative of the pressure sequence data, i.e., the rate of change of pressure fluctuation.

[0061] Fast Fourier Transform (FFT): This is an efficient algorithm used to convert time-domain signals into frequency-domain signals. By using FFT, the energy distribution of a signal at different frequencies can be analyzed, thereby identifying characteristic frequency components related to specific physical phenomena (such as check valve operation).

[0062] Power spectral density: Power spectral density represents the energy distribution intensity of a signal at different frequencies. By calculating the power spectral density, the energy magnitude of a specific frequency component can be quantified, thereby identifying the characteristic resonant frequency energy caused by the check valve's operation.

[0063] Characteristic resonant frequency energy: This refers to the energy within a specific frequency range that is excited by the action of the double UQD check valve in the liquid-cooled pipeline and coupled with the pipeline's natural frequency. This energy is the main cause of continuous vibration and noise in the system.

[0064] Energy-weighted fusion algorithm: This algorithm is used to comprehensively process energy information from different dimensions. In this embodiment, it calculates a comprehensive instantaneous pressure fluctuation energy value by performing nonlinear correction gain processing on the pressure fluctuation change rate and combining it with the distribution intensity of characteristic resonant frequency energy in the frequency domain, so as to comprehensively evaluate the transient impact intensity of the pipeline.

[0065] Normalized Fluctuation Energy Assessment Index: This index is the output of an energy-weighted fusion algorithm. It is a normalized comprehensive indicator used to quantify the intensity of transient shocks and vibrations in pipelines. This index reflects the risk level of current pipeline pressure fluctuations.

[0066] Variable gain adaptive control algorithm: This algorithm can dynamically adjust the control gain according to the real-time operating conditions of the system to achieve precise regulation of the speed and frequency of the circulating pump motor. By generating a pressure wave opposite to the pressure pulse, this algorithm aims to effectively counteract the pipeline pressure pulse generated by the check valve operation, thereby maintaining system pressure stability.

[0067] This embodiment provides a liquid cooling pressure self-compensation control method based on dual UQD check valves, and its specific implementation may include the following aspects:

[0068] For the collaborative acquisition and preprocessing of raw signals, pressure sensors installed at both ends of a dual UQD check valve can be used to acquire raw fluid pressure signals. As one implementation, the pressure sensors can be positioned at the inlet and outlet of the check valve to directly obtain the fluid pressure information flowing through it. After obtaining the raw fluid pressure signal, it needs to be denoised to obtain preprocessed pressure sequence data. For example, a moving average algorithm can be used to smooth the raw signal to reduce the impact of random noise.

[0069] In composite feature extraction and time-frequency analysis, numerical differentiation algorithms can be used to calculate the first derivative of pressure sequence data, thereby obtaining the rate of pressure fluctuation. For example, numerical differentiation methods such as the two-point difference method or the three-point difference method can be used to approximate the first derivative of pressure sequence data. Simultaneously, Fourier transform can be used to convert pressure sequence data from the time domain to the frequency domain. For example, the Discrete Fourier Transform (DFT) can be used to perform frequency domain transformation on the acquired time-domain pressure data to analyze its frequency components. After frequency domain transformation, the characteristic resonant frequency energy caused by the action of the dual UQD check valve can be identified by calculating the energy distribution of the signal. For example, the resonant energy within a specific frequency range can be obtained by simply summing the frequency domain energy within that range.

[0070] In terms of multidimensional energy fusion and disturbance assessment, the central controller acquires the pressure fluctuation rate and characteristic resonant frequency energy. The central controller can periodically acquire this data directly from the signal processing module or sensors via wired or wireless communication interfaces. After acquiring this data, the built-in energy fusion algorithm can be used to calculate the instantaneous pressure fluctuation energy value. For example, a simple linear weighted summation method can be used to combine the pressure fluctuation rate and characteristic resonant frequency energy to obtain a comprehensive energy value. When calculating the instantaneous pressure fluctuation energy value, the pressure change rate can be amplified, and combined with the frequency domain distribution intensity of the characteristic resonant frequency energy, a normalized fluctuation energy assessment index reflecting the transient impact intensity of the pipeline can be output. For example, a fixed gain coefficient can be set to amplify the pressure change rate, and the total energy value of the characteristic resonant frequency can be directly used for fusion, then the assessment index can be obtained through a maximum value normalization method.

[0071] Regarding dynamic compensation execution and feedback closed-loop, when the normalized fluctuation energy assessment index exceeds a preset safety threshold, the central controller sends a compensation control command to the drive inverter of the circulating pump. For example, the safety threshold can be set to a fixed value determined based on empirical data from the system under standard operating conditions. Upon receiving a signal indicating that the assessment index exceeds the threshold, the central controller can generate a simple switching signal or a fixed frequency adjustment command and send it to the drive inverter. After receiving the compensation control command, the motor speed frequency of the circulating pump can be adjusted in real time using a control algorithm. For example, a fixed-gain proportional-integral-derivative (PID) control algorithm can be used to adjust the circulating pump speed based on the deviation between the assessment index and the safety threshold. By adjusting the motor speed frequency of the circulating pump, a reverse pressure wave can be generated to counteract the pipeline pressure pulses generated by the operation of the dual UQD check valves. For example, an additional pulse generator can be installed in the pipeline to generate a pressure wave with an opposite phase when a pressure pulse is detected, thus counteracting the pressure pulse.

[0072] The method proposed in this embodiment accurately senses, extracts multi-dimensional features, and intelligently evaluates pressure fluctuations caused by the operation of dual UQD check valves in the liquid cooling system. Combined with adaptive compensation control, it effectively solves the problems of severe pressure pulsation, high noise and vibration, incomplete risk assessment, and rigid compensation control inherent in traditional solutions. This significantly improves the operational stability of the precision liquid cooling system, reduces noise levels, and extends the service life of system components, ensuring the quiet, stable, and long-life operation of a highly reliable liquid cooling system.

[0073] In some of the solutions described above in this application, adaptive denoising processing is proposed to remove noise and obtain accurate pressure sequence data. However, in this process, a specific adaptive denoising implementation method is needed to ensure that the denoising effect can dynamically adapt to local changes in the signal and avoid noise residue or signal distortion affecting the accuracy of subsequent pressure fluctuation feature extraction.

[0074] In this regard, this application further proposes that the adaptive denoising process in step S1 adopts the following formula:

[0075]

[0076] in, This is the preprocessed pressure sequence data. Sampling time, The filter order is... For sampling offset index, These are weighting coefficients that are dynamically adjusted based on local variance. This is the original pressure signal at the historical sampling time.

[0077] Specifically, the adaptive denoising process is a signal processing method that dynamically adjusts filtering parameters based on the local statistical characteristics of the signal and noise. Unlike traditional fixed-parameter filters, it can better preserve signal details while suppressing noise, and is particularly suitable for non-stationary signals. The formula aims to provide a mathematical model that generates preprocessed pressure sequence data for the current moment by weighted averaging of historical sampling data, where the weighting coefficients are dynamically adjusted to achieve effective denoising of the original fluid pressure signal. Besides the method of dynamically adjusting weighting coefficients based on local variance as described in this application, adaptive denoising can also be achieved through Kalman filtering. Kalman filtering can estimate the signal state and perform denoising in real time based on the system dynamic model and the statistical characteristics of measurement noise; or wavelet thresholding denoising can be used, where the signal is decomposed into different scales through wavelet transform, then the wavelet coefficients are thresholded to remove noise, and finally the signal is reconstructed using inverse wavelet transform.

[0078] The preprocessed pressure sequence data refers to the set of continuous data points that, after adaptive denoising, remove or significantly reduce noise components from the original pressure signal, thereby more accurately reflecting the true changes in fluid pressure. This data serves as input for subsequent composite feature extraction and time-frequency analysis, and its quality directly affects the accuracy and reliability of the subsequent analysis. The sampling time refers to the time point at which discrete data points are acquired during the digitization of the continuous-time signal. It is used to identify the time position of each pressure data point and is the basis for constructing the pressure sequence data and performing time-series analysis. The filter order, in digital filtering, typically refers to the highest degree of the polynomial in the filter's transfer function. It determines the smoothness and response characteristics of the filter to the signal and controls the range and depth of historical data participating in the weighted averaging during denoising. The sampling offset index refers to the relative position or offset of historical sampling data points used when calculating the preprocessed data at the current sampling time, ensuring that historical pressure signal data before or after the current sampling time can be correctly accessed during the weighted averaging process.

[0079] The weighting coefficients, dynamically adjusted based on local variance, are weighting factors that change in real time according to the local statistical characteristics of the signal (such as local variance). Local variance reflects the degree of signal fluctuation within a small time window; a larger variance indicates more drastic signal changes or greater noise, while a smaller variance indicates a more stable signal. This is key to achieving adaptive denoising. When the local variance is large, the weighting coefficients are adjusted accordingly to reduce the impact of noise on the current sampling point; when the local variance is small, the weighting coefficients are adjusted to better preserve signal details. These weighting coefficients can be calculated using a function inversely proportional to the local variance, such as a Gaussian or exponential function whose parameters are related to the local variance; or by using an estimate based on the signal-to-noise power ratio (SNR), dynamically adjusting the weights according to the local SNR. When the SNR is high, the weights are biased towards the original signal; when the SNR is low, the weights are biased towards the smoothed result. The original pressure signal at the historical sampling time refers to the raw pressure data directly collected by the pressure sensor before the current sampling time without any processing. This data serves as the input basis for the adaptive denoising formula, providing the raw data for weighted averaging and local variance calculation.

[0080] Through the above technical solution, the formula, especially the weighting coefficients based on dynamic adjustment of local variance, enables the denoising process to intelligently adapt to the local variation characteristics of the fluid pressure signal. When the pressure signal is relatively stable, the local variance is small, and the weighting coefficients will be adjusted accordingly to provide a smoother output, effectively suppressing background noise. When the pressure signal undergoes rapid transient changes due to factors such as the action of the dual UQD check valve, the local variance increases, and the weighting coefficients will be adjusted to better preserve these key transient features, avoiding excessive signal smoothing or distortion. This dynamic adaptive adjustment ensures higher accuracy of the preprocessed pressure sequence data and effectively preserves important pressure fluctuation details, which is crucial for the composite feature extraction and time-frequency analysis in the subsequent step S2. Compared with filters using fixed parameters, this solution significantly reduces the risk of noise residue or signal distortion, thereby improving the reliability of identifying the rate of pressure fluctuation change and the energy of characteristic resonant frequencies. This provides a more accurate data foundation for multidimensional energy fusion and disorder assessment in step S3, ultimately improving the accuracy and response speed of the entire liquid-cooled pressure self-compensation control method.

[0081] In some of the embodiments described above in this application, composite feature extraction and time-frequency analysis are proposed to extract the pressure fluctuation rate and characteristic resonant frequency energy from preprocessed pressure sequence data. However, in its implementation, the lack of specific calculation methods and algorithm details may lead to insufficient feature extraction accuracy, and the inability to effectively separate and quantify the characteristic resonant components coupled with the pipeline's inherent frequency, thereby affecting the accuracy of subsequent multidimensional energy fusion and disorder assessment.

[0082] In this regard, this application further proposes the following steps for composite feature extraction and time-frequency analysis:

[0083] S21. Calculate the rate of change of pressure fluctuation using the following five-point cubic smoothing numerical differential algorithm formula;

[0084]

[0085] in, Sampling time The rate of change of pressure fluctuations The sampling time interval of the pressure sensor. and To use the sampling time Δt represents the sampled values ​​of the pressure sequence that are centrally symmetrically distributed; Δt represents the sampled values ​​of the pressure sequence that are centrally symmetrically distributed with respect to the sampling time t.

[0086] Specifically, this five-point cubic smoothing numerical differentiation algorithm effectively suppresses the influence of high-frequency noise on the differentiation results by performing local polynomial fitting and differentiation on the preprocessed pressure sequence data, thereby accurately capturing the transient change trend of fluid pressure. In practical applications, in addition to the above formula, a method based on the Savitzky-Golay filter can also be used. This method performs least-squares polynomial fitting on the data within a sliding window, and then performs analytical differentiation on the fitted polynomial to achieve simultaneous processing of smoothing and differentiation; alternatively, the pressure sequence data can be first smoothed by low-pass filtering or moving average, and then numerical differentiation methods such as the central difference method can be used to calculate the pressure fluctuation rate to ensure stable and reliable differentiation results in noisy environments.

[0087] S22. The pressure sequence data is converted from the time domain to the frequency domain using the following Fast Fourier Transform formula:

[0088]

[0089] in, Discrete frequency points Complex spectral values ​​at that location, This is the nth discrete sampled pressure data after preprocessing. Here is the preset windowing function, and N is the total number of sample points. It is a complex exponential basis function;

[0090] Specifically, the Fast Fourier Transform (FFT) is an efficient algorithm used to decompose a time-domain signal into a superposition of different frequency components, thereby revealing the signal's frequency characteristics. In this process, the application of a pre-defined windowing function w(n) is crucial. For example, Hanning, Hamming, or Blackman windows can be used, as these window functions effectively reduce spectral leakage and improve the accuracy of frequency analysis. Besides the standard radix-2 FFT algorithm, when the number of sample points is not a power of 2, the Chirp Z-transform or Goertzel algorithm can also be used to adapt to the analysis needs of different data lengths and specific frequency ranges.

[0091] S23. Calculate the power spectral density using the following formula and identify the characteristic resonance frequency energy:

[0092]

[0093] in, The power spectral density is represented at discrete frequency points. The energy intensity of the signal at that location; Where N is the sampling frequency; N is the total number of sample points; Discrete frequency points Complex spectral values ​​at;

[0094]

[0095] in, The characteristic resonant frequency energy, and These represent the lower and upper limits of the characteristic frequency range. For power spectral density, This refers to the frequency resolution.

[0096] Specifically, the power spectral density (PSD) quantifies the distribution of signal energy in the frequency domain, obtained by appropriately normalizing the squared magnitude of the FFT result. When identifying characteristic resonant frequency energy, one or more characteristic frequency intervals [f_{lower}, f_{upper}] can be pre-defined based on the inherent vibration characteristics of the liquid-cooled piping system and the operating frequency of the dual UQD check valve. Integrating the PSD within this interval quantifies the energy associated with piping resonance. Alternatively, adaptive methods can be used, such as using peak detection algorithms to automatically identify significant resonance peaks in the power spectrum and dynamically determining the integration interval centered on these peak frequencies, thereby more accurately capturing characteristic resonant energy under different operating conditions.

[0097] Through the aforementioned technical solutions, this application introduces a specific five-point cubic smoothing numerical differentiation algorithm, which can accurately and stably calculate the rate of pressure fluctuation, effectively suppressing noise interference and ensuring the reliability of transient impact intensity assessment. Simultaneously, by employing a fast Fourier transform combined with a windowing function, accurate conversion of pressure sequence data from the time domain to the frequency domain is achieved, effectively reducing spectral leakage and laying a solid foundation for subsequent frequency analysis. Based on this, by calculating the power spectral density and accurately identifying the energy within specific characteristic frequency ranges, this application can effectively separate and quantify the characteristic resonant components coupled with the pipeline's inherent frequency from complex pressure waveforms, thus providing a high-precision quantitative index of vibration intensity. The synergistic application of these specific algorithms solves the problems of insufficient feature extraction accuracy and inability to effectively separate and quantify key resonant components in existing technologies, significantly improving the accuracy of subsequent multidimensional energy fusion and disorder assessment. This provides more accurate and reliable input data for the active pressure compensation control of the liquid cooling system, thereby more effectively offsetting the pipeline pressure pulses generated by the action of the dual UQD check valves.

[0098] In some of the above-mentioned schemes in this application, an energy weighted fusion algorithm is proposed to calculate the normalized fluctuation energy assessment index. However, in this process, due to the lack of a specific weighted fusion formula, it is impossible to effectively combine the pressure change rate and characteristic resonance energy for dynamic risk assessment, resulting in insufficient control accuracy and difficulty in adapting to the pressure fluctuation characteristics under different working conditions.

[0099] In response, this application further proposes an energy-weighted fusion algorithm to calculate the normalized fluctuation energy assessment index using the following formula:

[0100]

[0101] in, The normalized fluctuation energy assessment index. Sampling time The rate of change of pressure fluctuations The preset rate reference value, The characteristic resonant frequency energy, The reference value for the preset characteristic resonant frequency energy. and This is the weighted fusion coefficient.

[0102] In this technical solution, the energy weighted fusion algorithm is an algorithm that comprehensively evaluates energy indicators from different sources or with different properties. Its core lies in integrating multiple indicators through weighting to form a unified evaluation result. In this application, the algorithm introduces a nonlinear correction exponent to perform gain processing on the pressure change rate and combines the distribution intensity of energy at the characteristic resonant frequency in the frequency domain to output a normalized fluctuation energy evaluation index that reflects the transient impact intensity of the pipeline.

[0103] The Normalized Fluctuation Energy Assessment Index (NFIA) is a dimensionless index used to comprehensively reflect the transient impact intensity of pipelines, providing a unified and comparable risk assessment standard to facilitate subsequent control decisions. It can be normalized by dividing the original fluctuation energy value by its maximum possible value or a reference value, or by using a Sigmoid or Tanh function to map the index to a specific range (e.g., 0 to 1). In practical engineering applications, to ensure the consistency of the physical dimensions of the NFIA, the above calculation process is performed under a per-unit system. Specifically, after acquiring the pressure fluctuation rate, the central controller normalizes it according to a preset benchmark value, converting it into a dimensionless relative value. This dimensionless per-unit value is then weighted and fused with the characteristic resonant frequency energy, which has also undergone normalization. Through this per-unit preprocessing, even though the energy weighted fusion algorithm formula is mathematically expressed in terms of rate dimensions, the actual values ​​involved in the calculation are dimensionless per-unit values, thus ensuring that the final output NFIA is a strictly dimensionless index.

[0104] The rate of change of pressure fluctuation at a sampling time represents how quickly the fluid pressure changes with time at a specific sampling moment, and is used to quantify the intensity of transient shocks. It can be obtained by numerically differentiating the pressure sequence data, such as using a five-point cubic smoothing numerical differentiation algorithm, or through simple difference calculations, or by smoothing and differentiating the pressure signal using a Kalman filter.

[0105] Characteristic resonance frequency energy represents the energy carried by pipeline resonance induced by the action of the double UQD check valve within a specific frequency range. It is used to quantify vibration intensity and reflect the strength of persistent, destructive pressure pulsations. It can be obtained by performing a Fast Fourier Transform (FFT) on the pressure sequence data to obtain frequency domain information, and then calculating the power spectral density integral for the specific frequency range; or by using time-frequency analysis methods such as wavelet transform or empirical mode decomposition (EMD) to extract the energy of specific frequency components.

[0106] The preset reference value for the characteristic resonant frequency energy is used to normalize the baseline value of the characteristic resonant frequency energy. Its function is to convert the characteristic resonant frequency energy into a dimensionless relative index, which facilitates integration with other indicators. This reference value can be determined by statistically averaging historical data during normal system operation or without pressure pulsation, or by determining the upper limit of the system's resonant energy under specific operating conditions through theoretical calculations or simulations, or by measuring and setting it under controlled conditions through experimental calibration.

[0107] Weighted fusion coefficients are used to adjust the proportion of pressure fluctuation rate and characteristic resonant frequency energy in the normalized fluctuation energy assessment index. These coefficients allow for flexible adjustment of the importance of different risk indicators based on actual application scenarios and system characteristics. They can be manually set through expert experience or system debugging, automatically learned and adjusted based on historical data or simulation results through optimization algorithms (such as genetic algorithms and particle swarm optimization), or dynamically adjusted based on the current system operating conditions (such as fluid type, pipe size, and pump speed).

[0108] Through the above technical solution, this application introduces a specific energy weighted fusion formula, which accurately integrates the pressure fluctuation rate and the characteristic resonant frequency energy to generate a normalized fluctuation energy assessment index. The pressure fluctuation rate in this formula undergoes nonlinear gain processing, enabling it to more sensitively reflect transient impact intensity and avoiding the limitations of relying solely on a single absolute value. The characteristic resonant frequency energy is normalized by dividing by a preset reference value, quantifying the vibration intensity, and a weighted fusion coefficient is introduced, allowing adjustment of the weights of the two components according to system requirements, achieving flexible adaptation to different operating conditions. The overall formula combines time-domain and frequency-domain characteristics to generate a comprehensive assessment index that can comprehensively and accurately reflect the transient impact intensity and continuous vibration intensity of the pipeline, thereby solving the problem of inaccurate dynamic risk assessment and providing a precise and reliable basis for subsequent dynamic compensation execution, significantly improving the accuracy and adaptability of liquid-cooled pressure self-compensation control.

[0109] In some of the solutions mentioned above in this application, a preset safety threshold is proposed to determine the risk of pressure fluctuation. However, in this process, the fixed safety threshold cannot adapt to the dynamic changes of the system under different operating frequencies and loads, resulting in inaccurate risk assessment and untimely compensation control, which may lead to misjudgment or insufficient response.

[0110] To address this, this application further proposes that in step S4, a preset safety threshold is a dynamic safety threshold boundary, used to dynamically adjust the safety boundary according to the current system operating conditions, thereby achieving adaptive adjustment of the safety threshold. The calculation formula for the dynamic safety threshold boundary is:

[0111]

[0112] in, For dynamic security threshold boundaries; This is the baseline value for the static safety threshold under standard rated operating conditions; This is the frequency correction factor, which indicates the sensitivity to pressure fluctuations at different operating frequencies; This represents the current real-time operating frequency of the circulating pump; This indicates the rated maximum operating frequency of the circulating pump; For load disturbance operators; This represents the current real-time load rate of the circulating pump.

[0113] Specifically, the dynamic safety threshold boundary refers to an upper limit for judging pressure fluctuation risk that is not fixed but can be adjusted according to the real-time operating parameters of the system. It replaces the traditional single fixed value, allowing the system's tolerance to pressure fluctuations to change flexibly. Besides calculation by formula, this dynamic safety threshold boundary can also be implemented through a pre-established operating condition-threshold mapping table. This mapping table stores recommended safety thresholds for different operating conditions such as operating frequency and load rate based on historical operating data and expert experience. The system queries the threshold corresponding to the current operating condition during runtime. Alternatively, a machine learning model-based approach can be used to train the model to learn the system's actual tolerance to pressure fluctuations under various operating conditions, thereby predicting and outputting the optimal safety threshold for the current operating condition in real time.

[0114] This dynamic safety threshold boundary is used to dynamically adjust the safety boundary based on the current system operating conditions. Its purpose is to ensure the accuracy of pressure fluctuation risk assessment and the timeliness of compensation control, avoiding misjudgments or insufficient responses due to threshold mismatches under different operating conditions. This can be achieved by real-time monitoring of key operating parameters such as the operating frequency of the circulating pump, load rate, fluid temperature, and ambient temperature, using these parameters as inputs to drive the threshold adjustment mechanism. Alternatively, it can be achieved by integrating data from multiple sensors and combining fuzzy logic or expert systems to comprehensively assess the current operating conditions, thereby triggering corresponding safety boundary adjustment strategies.

[0115] Achieving adaptive adjustment of the safety threshold refers to the system's ability to automatically modify its behavior or parameters to achieve a predetermined goal based on changes in the external environment or internal state. In this context, it means that the safety threshold can automatically adapt to changes in system operating conditions without human intervention. Besides formula calculation, online learning or reinforcement learning algorithms can be used to allow the system to continuously optimize the threshold adjustment strategy during operation, maximizing system stability and minimizing compensation energy consumption. Alternatively, a feedback mechanism can be introduced, using the effect of compensation control (e.g., the degree of pressure pulsation suppression) as input to correct the threshold adjustment parameters, thereby achieving a deeper level of adaptability.

[0116] Through the above technical solution, this application overcomes the limitation of traditional fixed safety thresholds that cannot adaptively adjust when system operating conditions change. Based on S1 (raw signal collaborative acquisition and preprocessing), S2 (composite feature extraction and time-frequency analysis), and S3 (multidimensional energy fusion and disorder assessment), the S4 dynamic compensation execution and feedback closed-loop link introduces a dynamic safety threshold boundary, making the judgment of the "normalized fluctuation energy assessment index" more intelligent and precise. This dynamic threshold mechanism, combined with real-time operating parameters such as the operating frequency and load rate of the circulating pump, dynamically adjusts the safety boundary, making pressure fluctuation risk assessment more accurate and avoiding misjudgments caused by threshold mismatch under different operating conditions. For example, when the pump is running at high speed, the threshold will be more stringent, enabling earlier detection of potential risks; at low speed and light load, the threshold may be appropriately relaxed to avoid overcompensation. Because the threshold can adapt to changes in operating conditions in real time, the system can identify the real risks more promptly, thereby triggering compensation control commands to effectively offset the pipeline pressure pulses generated by the action of the dual UQD check valve, avoiding system instability caused by response lag. This makes the entire liquid cooling pressure self-compensation control method more robust and adaptable in the face of changing operating environments, better protecting the liquid cooling system and extending its service life.

[0117] In some of the solutions mentioned above in this application, a dynamic safety threshold boundary is proposed to dynamically adjust the safety boundary according to the current operating conditions of the system. However, in this process, the compensation control algorithm itself lacks the ability to adjust the compensation intensity in real time according to the pipeline pressure state, which may lead to insufficient response or overcompensation when facing pressure fluctuations of different intensities.

[0118] In this regard, this application further proposes that in step S4, the variable gain adaptive control algorithm calculates the motor frequency adjustment increment finally sent to the drive inverter of the circulating pump using the following formula:

[0119]

[0120] in, The motor frequency adjustment increment sent to the drive inverter of the circulating pump; It is a time-varying adaptive gain operator used to adjust the compensation intensity in real time according to the pipeline pressure status; The normalized fluctuation energy assessment index; The safety threshold boundary is calculated at the current sampling time; The differential feedforward gain constant is used to anticipate changes in the resonant energy. The characteristic resonant frequency energy.

[0121] Specifically, the motor frequency adjustment increment refers to the amount by which the operating frequency of the circulating pump needs to be adjusted to counteract pipeline pressure pulses. This increment is added to the current operating frequency of the circulating pump to form a new target frequency, which is then used to control the speed of the circulating pump via the drive inverter. This can be achieved by having a digital signal processor (DSP) or microcontroller (MCU) directly generate a PWM (Pulse Width Modulation) signal based on the output of the control algorithm. This PWM signal, after filtering and amplification, drives the inverter to adjust the motor frequency. Alternatively, it can directly output a frequency value or rate of change, which is received by the inverter's internal control logic and converted into the actual motor drive frequency.

[0122] The time-varying adaptive gain operator is a dynamically changing coefficient that flexibly adjusts the intensity of compensation control based on the real-time pressure status of the pipeline system. When pipeline pressure fluctuates drastically or the system damping characteristics change, the operator can increase or decrease accordingly to ensure the accuracy and stability of the compensation effect and avoid overcompensation or undercompensation. For example, the time-varying adaptive gain operator can be updated in real time based on a preset initial gain constant, fluid damping adjustment coefficient, and the absolute value of the real-time pressure difference across the dual UQD check valves. Alternatively, it can be implemented using a fuzzy logic controller or a neural network model. For instance, a fuzzy logic controller can output a dynamic gain value based on multiple input variables such as pressure fluctuation amplitude and pressure change rate, combined with a preset fuzzy rule base; a neural network model can adaptively adjust the gain parameters by learning historical data.

[0123] The normalized fluctuation energy assessment index is a comprehensive indicator used to quantify the transient impact intensity of pipelines. It fuses the pressure fluctuation rate and the characteristic resonant frequency energy, and then normalizes them, allowing for comparison and assessment of pressure fluctuation risks under different operating conditions on a unified scale. For example, this index can be calculated using an energy-weighted fusion algorithm, which introduces a nonlinear correction index to gain the pressure change rate and combines it with the frequency domain distribution intensity of the characteristic resonant frequency energy. Alternatively, machine learning methods such as Principal Component Analysis (PCA) or Support Vector Machine (SVM) can be used to fuse and reduce the dimensions of the pressure fluctuation rate, characteristic resonant frequency energy, and other relevant parameters to obtain a comprehensive risk assessment index, which is then normalized.

[0124] The safety threshold boundary is a dynamically adjusted upper limit value used to determine whether the current pipeline pressure fluctuation is within a safe range. When the normalized fluctuation energy assessment index exceeds this boundary, it indicates a high risk of pipeline pressure fluctuation, requiring the initiation of compensation control. Its dynamism is reflected in its ability to adaptively adjust according to the current system operating conditions (such as the operating frequency of the circulating pump, load rate, etc.) to meet safety requirements under different operating conditions. For example, the dynamic safety threshold boundary can be calculated using the static safety threshold benchmark value under standard rated operating conditions, combined with a frequency correction coefficient, the current real-time operating frequency of the circulating pump, the rated maximum operating frequency, the load disturbance operator, and the current real-time load rate of the circulating pump. Alternatively, it can be determined using a lookup table or polynomial fitting model built based on historical operating data and expert experience.

[0125] The differential feedforward gain constant is a preset fixed gain parameter used to anticipate changes in the energy of the characteristic resonant frequency. By introducing a differential term, the control system can predict the direction and amplitude of future pressure fluctuations, thereby compensating before the pressure pulse fully forms or reaches its peak, reducing control lag, and improving response speed. This constant can be obtained through system identification experiments and simulation optimization. For example, during the system commissioning phase, different types of excitation signals are applied, the system response is measured, and a mathematical model is established using system identification techniques. Then, the optimal differential feedforward gain constant is determined through simulation optimization to minimize pressure pulsations.

[0126] The characteristic resonant frequency energy refers to the energy intensity within a specific frequency range in the pipeline pressure signal that is related to the operation of the dual UQD check valve or the natural frequency of the pipeline. It is a major cause of persistent vibration and noise; therefore, accurate identification and quantification are crucial for targeted compensation control. For example, this energy can be obtained by performing a Fast Fourier Transform (FFT) on the pressure sequence data to convert it from the time domain to the frequency domain, calculating the power spectral density, and integrating the power spectral density within a preset characteristic frequency range. Alternatively, it can be extracted using time-frequency analysis methods such as wavelet transform or empirical mode decomposition (EMD).

[0127] Through the above technical solution, this application introduces a variable gain adaptive control algorithm, which solves the problem of the lack of ability of compensation control to adjust the compensation intensity in real time according to the pipeline pressure state, thereby avoiding the problems of insufficient response or overcompensation when facing pressure fluctuations of different intensities. Specifically, by calculating the motor frequency adjustment increment, this increment comprehensively considers the time-varying adaptive gain operator, the normalized fluctuation energy assessment index, the safety threshold boundary, the differential feedforward gain constant, and the characteristic resonant frequency energy. The time-varying adaptive gain operator can adjust the compensation intensity in real time according to the pipeline pressure state, ensuring the accuracy of compensation; the normalized fluctuation energy assessment index provides a quantitative indicator of the current pressure fluctuation; the safety threshold boundary serves as a dynamic benchmark, making compensation triggering more flexible; the differential feedforward gain constant can respond in advance to the changing trend of the characteristic resonant frequency energy, effectively reducing control lag; and the direct participation of the characteristic resonant frequency energy ensures that the compensation action can accurately target the key resonant components that cause continuous vibration and noise. This multi-parameter collaborative and dynamically adaptive control strategy enables the system to quickly, accurately, and stably compensate for the transient and high-frequency pressure pulses generated by the action of the dual UQD check valves, significantly improving the operational stability, quietness, and reliability of the liquid cooling system.

[0128] In some of the solutions mentioned above in this application, a time-varying adaptive gain operator is proposed to adjust the compensation intensity in real time according to the pipeline pressure status. However, in the implementation process, without a specific real-time update mechanism, the operator may not be able to accurately respond to the dynamic changes of the system, resulting in inaccurate compensation control and inability to effectively offset pressure fluctuations.

[0129] In response, this application further proposes a time-varying adaptive gain operator that is updated and calculated in real time according to the following formula:

[0130]

[0131] in, It is a time-varying adaptive gain operator; This is the preset initial gain constant; This is the preset fluid damping adjustment coefficient; It is the absolute value of the real-time pressure difference across the two ends of the double UQD check valve.

[0132] Specifically, the time-varying adaptive gain operator is a key parameter in a control system used to dynamically adjust the output strength of the compensation control. It is not a fixed value but adjusts according to the real-time state of the system and changes in the environment to ensure that the compensation control system maintains optimal performance under different operating conditions. This operator can be implemented in several ways. For example, it can be implemented using a lookup table, pre-storing corresponding gain values ​​based on different operating parameters (such as pressure difference, flow rate, etc.), and retrieving the corresponding gain based on the real-time detected operating parameters during runtime. Alternatively, it can be implemented using a fuzzy logic controller, taking the system state variables as fuzzy inputs and outputting a dynamically adjusted gain value through a fuzzy inference mechanism to adapt to nonlinear or uncertain systems.

[0133] The preset initial gain constant is a baseline value for the time-varying adaptive gain operator under system startup or specific stable operating conditions. It provides a stable starting point for dynamic gain adjustment, ensuring that the system can maintain basic compensation capability in the initial stage or when other dynamic parameters are not significant. This constant can be obtained through theoretical calculations or simulation analysis during the system design phase, determined based on the physical model and expected operating characteristics of the liquid cooling system; alternatively, this constant can also be experimentally adjusted and optimized under typical operating conditions during actual system debugging until satisfactory control effects are achieved and then fixed.

[0134] The preset fluid damping adjustment coefficient is a parameter used to quantify the impact of fluid damping characteristics on pressure fluctuations in liquid-cooled piping. Fluid damping reflects the system's ability to dissipate pressure wave energy, and the introduction of this coefficient allows gain adjustment to consider the influence of the fluid medium and piping structure on pressure wave propagation and attenuation. This coefficient can be calculated and calibrated by performing fluid dynamics simulation analysis on the liquid-cooled system piping, combined with parameters such as fluid viscosity, pipe length, diameter, and number of bends; alternatively, the coefficient can also be measured by conducting impulse response experiments on an actual liquid-cooled system, measuring the attenuation characteristics of the pressure wave, and fitting and calibrating based on the experimental data.

[0135] The absolute value of the real-time pressure difference across a dual UQD check valve directly reflects the intensity of fluid impact or its operational status. When a dual UQD check valve operates (e.g., closes), a significant pressure difference is generated across its terminals. The magnitude of this value is a key indicator of the severity of pipeline pressure fluctuations. This value can be obtained by installing high-precision pressure sensors at the inlet and outlet of the dual UQD check valve, acquiring the pressure signals in real time, and then calculating the difference and taking the absolute value through a central controller. Alternatively, in some cases, if the dual UQD check valve integrates a differential pressure sensor, its output differential pressure signal can be directly read and its absolute value processed.

[0136] Through the above technical solution, the time-varying adaptive gain operator is no longer a fixed value, but can be dynamically adjusted according to the real-time operating status of the liquid cooling system. Specifically, the preset initial gain constant provides a stable benchmark, ensuring that the system has basic compensation capabilities under startup or stable operating conditions. The preset fluid damping adjustment coefficient allows the gain adjustment to fully consider the inherent damping characteristics of the fluid in the liquid cooling pipeline, avoiding overcompensation or undercompensation caused by differences in fluid characteristics. More importantly, by introducing the absolute value of the real-time pressure difference across the dual UQD check valves, the gain operator can directly sense and quantify the intensity of the transient pressure shock caused by the action of the dual UQD check valves. When the pressure difference is large, it indicates severe pipeline shock, and the gain operator will increase accordingly, allowing the compensation control command to more strongly counteract the pressure pulse; when the pressure difference is small, the gain operator will decrease, thereby achieving more precise and stable compensation. This real-time, adaptive gain adjustment mechanism significantly improves the response speed and accuracy of compensation control, enabling the reverse pressure wave generated by the circulating pump to more accurately cancel the pressure pulse generated by the action of the dual UQD check valve, thereby effectively suppressing pipeline pressure fluctuations, reducing system vibration and noise, and extending the service life of the liquid cooling system.

[0137] In some of the solutions mentioned above in this application, a liquid cooling pressure self-compensation control method based on dual UQD check valves is proposed to actively counteract pressure pulsation. However, in the process of implementing this method, an efficient system is needed to ensure real-time acquisition, accurate separation of characteristic resonance components, dynamic risk assessment, and adaptive adjustment of compensation intensity, so as to avoid the problems of response lag, insufficient feature recognition, static risk assessment, and rigid control gain in the prior art.

[0138] In response, this application proposes a liquid cooling pressure self-compensation control system based on dual UQD check valves. The system includes a pressure signal acquisition module, a pressure fluctuation feature extraction module, a normalized energy calculation module, a variable gain adaptive calculation module, and a pump-valve coordinated control module.

[0139] The pressure signal acquisition module serves as the sensing layer of the entire control system, responsible for acquiring key physical quantities related to the operating status of the liquid cooling system. Its core function is to provide comprehensive, accurate, and real-time raw data for subsequent pressure fluctuation analysis, ensuring the system can promptly detect pressure changes caused by check valve operation. Specifically, this module can employ high-precision piezoresistive pressure sensors, differential pressure sensors, and Hall effect speed sensors to convert the acquired analog signals into digital signals via an analog-to-digital converter (ADC) and transmit them to the central controller for processing. Alternatively, a miniature pressure sensor array based on MEMS technology can be used, combined with a wireless communication module (such as LoRa or Zigbee) to transmit data to an edge computing unit. The edge computing unit then performs preliminary data aggregation and time synchronization before uploading the data to the central controller. The operating frequency signal of the circulating pump can be obtained by reading the control parameters of the frequency converter or motor driver.

[0140] The pressure fluctuation feature extraction module aims to identify and quantify specific pressure fluctuation features related to the operation of the dual UQD check valve from complex raw pressure signals, particularly key frequency energies that may trigger resonance and sustained vibration. Its role is to transform the raw, mixed signal into physically meaningful feature parameters, providing accurate input for subsequent risk assessment. Specifically, this module can incorporate a digital signal processor (DSP) to denoise the raw pressure signal by executing preset filtering algorithms (such as Kalman filtering and wavelet denoising). Subsequently, it uses Fast Fourier Transform (FFT) or Short-Time Fourier Transform (STFT) to convert the time-domain signal to the frequency domain, and identifies and quantifies the energy within a specific frequency range, i.e., the characteristic resonant frequency energy, by analyzing the power spectral density (PSD). Alternatively, machine learning-based methods can be employed, such as training a convolutional neural network (CNN) model to directly learn and extract feature patterns related to check valve operation from the raw pressure sequence data and output the corresponding characteristic resonant frequency energy.

[0141] The normalized energy calculation module is responsible for comprehensively quantifying the extracted pressure fluctuation characteristics and transforming them into a unified and comparable risk assessment index. Its function is to provide an intuitive and comprehensive assessment of the pipeline's transient impact intensity, enabling the system to determine the risk level of the current pressure fluctuation based on this index, thereby triggering corresponding compensation control. Specifically, this module can be integrated into the central controller, executing a preset energy weighted fusion algorithm to weight and fuse the pressure fluctuation change rate (usually obtained through numerical differentiation) and the characteristic resonant frequency energy. The fused result is then normalized, mapped to a 0-1 interval, forming a normalized fluctuation energy assessment index. Alternatively, a fuzzy logic inference system can be used, taking the pressure fluctuation change rate and characteristic resonant frequency energy as input, and outputting a fuzzy value representing the risk level through fuzzy rules and membership functions. This value is then defuzzified and normalized to obtain the normalized fluctuation energy assessment index.

[0142] The variable gain adaptive calculation module is the core of the system's adaptive compensation control. Its function is to dynamically adjust the intensity of the compensation control based on the current operating state and pressure fluctuations of the liquid cooling system. This ensures that the system provides appropriate compensation under different operating conditions, avoiding overcompensation or undercompensation, thereby improving the robustness and effectiveness of the control. Specifically, this module can calculate and update the time-varying adaptive gain operator in real time using a mathematical model based on the real-time collected pressure difference value across the dual UQD check valves and a preset fluid damping adjustment coefficient. This operator is then combined with the deviation of the normalized fluctuation energy assessment index and the safety threshold, as well as the changing trend of the characteristic resonant frequency energy, to calculate the motor frequency adjustment increment of the circulating pump. Alternatively, a reinforcement learning algorithm can be used to learn the optimal gain operator adjustment strategy under different pressure difference, fluctuation energy, and resonant energy changing trends through continuous interaction with the liquid cooling system environment, thereby calculating the most suitable motor frequency adjustment increment.

[0143] The pump-valve coordinated control module is the execution layer of control commands, responsible for converting the compensation commands generated by the variable gain adaptive calculation module into actual control actions for the circulating pump. Its function is to actively generate a reverse pressure wave with a similar amplitude and opposite direction to the pressure pulse caused by the check valve's action by precisely adjusting the circulating pump's speed, thereby canceling out harmful pressure pulsations. Specifically, this module can be a microcontroller or programmable logic controller (PLC) that receives the motor frequency adjustment increment from the variable gain adaptive calculation module. It adds this increment to the circulating pump's current operating frequency to obtain the target speed frequency, and then sends the control signal to the circulating pump's drive inverter via a digital-to-analog converter (DAC) or pulse width modulation (PWM) signal. The inverter then precisely controls the circulating pump's motor speed. Alternatively, a distributed control system based on industrial Ethernet (such as EtherCAT or Profinet) can be used, with the pump-valve coordinated control module acting as the master station. It sends the target speed frequency command to the intelligent inverter acting as the slave station via a high-speed communication protocol, and the inverter's internal motor controller executes precise speed adjustment.

[0144] Through the aforementioned system, this application effectively addresses the problems of response lag, insufficient feature recognition, static risk assessment, and rigid control gain in existing technologies. Specifically, the pressure signal acquisition module can acquire real-time and comprehensive operating data of the liquid cooling system, laying the foundation for subsequent accurate analysis and avoiding control lag caused by missing or delayed data. The pressure fluctuation feature extraction module can accurately separate and quantify the characteristic resonance components in pipeline pressure fluctuations from complex raw signals, calculating the identified characteristic resonance frequency energy, thereby overcoming the shortcomings of existing technologies in identifying key resonance components and making compensation control more targeted. Based on this, the normalized energy calculation module normalizes the extracted total energy of pressure fluctuations across the entire frequency band, calculates the normalized fluctuation energy assessment index, and compares it with a preset safety threshold boundary to dynamically determine the risk level of the current pipeline pressure fluctuation. This multi-dimensional dynamic risk assessment mechanism is significantly superior to a single static threshold assessment and can better adapt to changes in system operating conditions. Furthermore, the variable gain adaptive calculation module dynamically updates the time-varying adaptive gain operator based on the real-time pressure difference across the dual UQD check valves, combined with a preset fluid damping adjustment coefficient. It also calculates the motor frequency adjustment increment for the circulating pump drive inverter by considering the deviation between the normalized fluctuation energy assessment index and the safety threshold, as well as the changing trend of the characteristic resonant frequency energy. This adaptive gain adjustment mechanism ensures that the compensation intensity can be precisely adjusted according to real-time operating conditions, effectively avoiding overcompensation or undercompensation problems that may occur with fixed gain. Finally, the pump-valve collaborative control module superimposes the calculated frequency adjustment increment onto the current operating frequency of the circulating pump to obtain the target speed frequency and generates a corresponding control signal to send to the drive inverter. By adjusting the circulating pump speed to output a reverse pressure wave, the system can actively and accurately counteract the pipeline pressure pulses generated by the action of the dual UQD check valves. Overall, the system of this application, through the collaborative work between modules, achieves intelligent sensing, accurate assessment, and adaptive compensation of pressure pulsations in the liquid cooling system, significantly improving the operational stability, reliability, and quietness of the liquid cooling system, and effectively extending the service life of the equipment.

[0145] Throughout the process, the pressure signal acquisition module is responsible for acquiring real-time pressure signals at the inlet and outlet of the dual UQD check valve, the pressure difference across the check valve, and the operating frequency signal of the circulating pump, providing raw time-series data for pressure fluctuation analysis. The pressure fluctuation feature extraction module filters and denoises the acquired raw pressure signals, extracts the characteristic resonance components from the pipeline pressure fluctuations, and calculates the identified characteristic resonance frequency energy. The normalized energy calculation module normalizes the extracted total energy of the full-frequency pressure fluctuations, calculates the normalized fluctuation energy assessment index, and compares the normalized fluctuation energy assessment index with a preset dynamic safety threshold boundary to determine the risk level of the current pipeline pressure fluctuation. The variable gain adaptive calculation module dynamically updates the time-varying adaptive gain operator based on the real-time pressure difference across the dual UQD check valve and a preset fluid damping adjustment coefficient. It then calculates the motor frequency adjustment increment of the circulating pump drive inverter by considering the deviation between the normalized fluctuation energy assessment index and the safety threshold, as well as the changing trend of the characteristic resonance frequency energy. The pump-valve coordinated control module superimposes the calculated frequency adjustment increment onto the current operating frequency of the circulating pump to obtain the target speed frequency, generates a corresponding control signal, and sends it to the drive inverter. By adjusting the circulating pump speed, it outputs a reverse pressure wave to counteract the pressure pulses generated in the pipeline by the check valve's operation. Through this coordinated operation, this method can proactively, accurately, and adaptively counteract the harmful pressure pulsations caused by the operation of the dual UQD check valves, fundamentally improving the quietness, stability, and long-life operation of the high-reliability liquid cooling system.

[0146] In this document, the terms "upper," "lower," "front," "back," "left," "right," "top," "bottom," "inner," "outer," "vertical," and "horizontal," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only used for the clarity of expressing the technical solution and for the convenience of description, and therefore should not be construed as limiting the present invention.

[0147] In this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, which includes not only the elements listed but also other elements not expressly listed.

[0148] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A self-compensating control method for liquid cooling pressure based on dual UQD check valves, characterized in that, Includes the following steps: S1. Raw signal co-acquisition and preprocessing: The raw fluid pressure signal is acquired using pressure sensors installed at both ends of the dual UQD check valve, and the raw fluid pressure signal is subjected to adaptive noise reduction processing to obtain preprocessed pressure sequence data; S2. Composite Feature Extraction and Time-Frequency Analysis: The first derivative of the pressure sequence data is calculated using a five-point cubic smoothing numerical differential algorithm to obtain the pressure fluctuation rate; at the same time, the pressure sequence data is converted from the time domain to the frequency domain using a fast Fourier transform, and the characteristic resonant frequency energy caused by the action at both ends of the double UQD check valve is identified by calculating the power spectral density. S3. Multidimensional Energy Fusion and Turbulence Assessment: The central controller acquires the pressure fluctuation rate and characteristic resonant frequency energy, and uses the built-in energy weighted fusion algorithm to calculate the instantaneous pressure fluctuation energy value; wherein, the energy weighted fusion algorithm performs gain processing on the pressure change rate by introducing a nonlinear correction index, and combines the distribution intensity of the characteristic resonant frequency energy in the frequency domain to output a normalized fluctuation energy assessment index that reflects the transient impact intensity of the pipeline; S4. Dynamic Compensation Execution and Feedback Closed Loop: When the normalized fluctuation energy assessment index exceeds the preset safety threshold, the central controller sends a compensation control command to the drive inverter of the circulating pump, and uses a variable gain adaptive control algorithm to adjust the motor speed frequency of the circulating pump in real time, thereby generating a reverse pressure wave to offset the pipeline pressure pulse generated by the action of the double UQD check valve.

2. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 1, characterized in that, In step S1, the adaptive noise reduction process uses the following formula: in, For preprocessed pressure sequence data, Sampling time, The filter order is... For sampling offset index, These are weighting coefficients that are dynamically adjusted based on local variance. This is the original pressure signal at the historical sampling time.

3. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 2, characterized in that, The specific method for step S2 is as follows: S21. Calculate the rate of change of pressure fluctuation using the following five-point cubic smoothing numerical differential algorithm formula; in, Sampling time The rate of change of pressure fluctuations This is the sampling time interval of the pressure sensor. and To use the sampling time The sampled values ​​are centrally symmetric distributions of the pressure sequence. S22. The pressure sequence data is converted from the time domain to the frequency domain using the following Fast Fourier Transform formula; in, Discrete frequency points Complex spectral values ​​at that location, This is the nth discrete sampled pressure data after preprocessing. Here is the preset windowing function, and N is the total number of sample points. It is a complex exponential basis function; S23. Calculate the power spectral density using the following formula and identify the characteristic resonance frequency energy; in, The power spectral density is represented at discrete frequency points. The energy intensity of the signal at that location; Where N is the sampling frequency; N is the total number of sample points; Discrete frequency points Complex spectral values ​​at; in, The characteristic resonant frequency energy, and These represent the lower and upper limits of the characteristic frequency range. For power spectral density, This refers to the frequency resolution.

4. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 3, characterized in that, The energy-weighted fusion algorithm calculates the normalized fluctuation energy assessment index using the following formula: in, The normalized fluctuation energy assessment index. Sampling time The rate of change of pressure fluctuations The preset rate reference value, The characteristic resonant frequency energy, The reference value for the preset characteristic resonant frequency energy. and This is the weighted fusion coefficient.

5. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 4, characterized in that, In step S4, the preset safety threshold is a dynamic safety threshold boundary, used to dynamically adjust the safety boundary according to the current system operating conditions, thereby achieving adaptive adjustment of the safety threshold. The calculation formula for the dynamic safety threshold boundary is: in, For dynamic security threshold boundaries; This is the baseline value for the static safety threshold under standard rated operating conditions; This is the frequency correction factor, which indicates the sensitivity to pressure fluctuations at different operating frequencies; This represents the current real-time operating frequency of the circulating pump; This indicates the rated maximum operating frequency of the circulating pump; For load disturbance operators; This represents the current real-time load rate of the circulating pump.

6. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 5, characterized in that, In step S4, the variable gain adaptive control algorithm calculates the motor frequency adjustment increment that is finally sent to the drive inverter of the circulating pump using the following formula: in, The motor frequency adjustment increment sent to the drive inverter of the circulating pump; It is a time-varying adaptive gain operator used to adjust the compensation intensity in real time according to the pipeline pressure status; The normalized fluctuation energy assessment index; The safety threshold boundary is calculated at the current sampling time; The differential feedforward gain constant is used to anticipate changes in the resonant energy. The characteristic resonant frequency energy.

7. The liquid cooling pressure self-compensation control method based on dual UQD check valves according to claim 6, characterized in that, The time-varying adaptive gain operator is updated and calculated in real time according to the following formula: in, It is a time-varying adaptive gain operator; This is the preset initial gain constant; This is the preset fluid damping adjustment coefficient; It is the absolute value of the real-time pressure difference across the two ends of the double UQD check valve.

8. A liquid cooling pressure self-compensation control system based on dual UQD check valves, used to implement the liquid cooling pressure self-compensation control method based on dual UQD check valves as described in any one of claims 1-7, characterized in that, include Pressure signal acquisition module: used to acquire real-time pressure signals at the inlet and outlet of the dual UQD check valve, the pressure difference between the two ends of the check valve, and the operating frequency signal of the circulating pump, providing raw time-series data for pressure fluctuation analysis; Pressure fluctuation feature extraction module: used to filter and denoise the acquired raw pressure signal, separate and extract the characteristic resonance component in the pipeline pressure fluctuation, and calculate the energy of the identified characteristic resonance frequency. Normalized energy calculation module: It is used to normalize the total energy of the extracted full-band pressure fluctuation, calculate the normalized fluctuation energy assessment index, and compare the normalized fluctuation energy assessment index with the preset safety threshold boundary to determine the risk level of the current pipeline pressure fluctuation. Variable gain adaptive calculation module: It is used to dynamically update the time-varying adaptive gain operator based on the real-time pressure difference between the two ends of the dual UQD check valve and the preset fluid damping adjustment coefficient, and calculate the motor frequency adjustment increment of the circulating pump drive frequency converter by combining the deviation between the normalized fluctuation energy evaluation index and the safety threshold and the change trend of characteristic resonant frequency energy. Pump and valve coordinated control module: It is used to superimpose the calculated frequency adjustment increment onto the current operating frequency of the circulating pump to obtain the target speed frequency, generate the corresponding control signal and send it to the drive frequency converter, and output a reverse pressure wave by adjusting the speed of the circulating pump to cancel the pressure pulse generated by the action of the check valve in the pipeline.

Images