Leakage safety closed-loop control method of refrigerating system

Abstract

The invention relates to the technical field of refrigerant leakage plugging, and discloses a leakage safety closed-loop control method for a refrigerating system, which comprises the following steps: S1, multi-source signal monitoring: collecting multi-mode dynamic signals of a compressor shell in real time through a multi-channel sensor array; s2, leakage risk judgment; s3, leakage positioning: constructing a time difference matrix based on the phase difference of the sensor array; s4, performing pressure coupling dynamic plugging, and activating a shape memory alloy plugging device; and S5, gradient temperature control sealing reinforcement is carried out, and the driving current is adjusted according to leakage point temperature feedback. Precise identification and three-dimensional space positioning of the leakage point are achieved, a traditional passive response mode is broken through, the pressure coupling dynamic sealing technology is combined, the plugging force field is generated in real time according to leakage characteristics, physical plugging is conducted on the leakage source at the initial stage of leakage, and the problem that the leakage source cannot be actively intervened in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of refrigerant leakage plugging, and specifically to a leakage safety closed-loop control method for a refrigeration system. Background Technique

[0002] Refrigerant plays a very important role in the refrigeration system. The refrigerant flows through the motor before or after the compressor completes compression, and realizes the heat dissipation effect of the compressor motor through heat exchange. The temperature of the motor winding, that is, the heat dissipation effect, depends on the refrigerant flow rate and the refrigerant temperature. There may be a slow leakage of refrigerant during the operation of the refrigeration system. When the leakage of the refrigerant reaches a certain amount, it will cause the temperature of the compressor motor to be too high, and the insulation will be damaged and fail.

[0003] After retrieval, the patent with the Chinese patent number CN222190012U discloses a refrigerant leakage alarm system, which relates to the technical field of alarm systems, including a chassis. A refrigerator is provided inside the chassis. A detector is installed on the inner bottom wall of the chassis. An air outlet box is fixed at the bottom end of the chassis. An exhaust structure for forcibly discharging the leaked refrigerant is provided inside the air outlet box. A storage battery is provided at the rear end of the chassis, and a controller is installed on one side of the storage battery. When the refrigerant in the refrigerator leaks, an alarm is given through the cooperation of sound and light. At the same time, the controller will control the refrigerator to stop running, and then control the motor to run, driving the fan to rotate, so that the fan guides and discharges the leaked refrigerant in the chassis through the air outlet groove and the guiding groove.

[0004] In the use of existing refrigerant leakage monitoring technologies, there are the following defects. First, the above-mentioned solutions mostly adopt a passive emergency mode, and only realize post-leakage treatment by shutting down and exhausting air, and cannot actively plug the leakage source. For example, during the operation of the refrigeration unit in a cold chain logistics center, when the refrigerant slowly leaks due to the aging of the compressor flange seal ring, although the system can trigger an alarm and stop the compressor from running, the leakage point is still in an open state, and the refrigerant will continue to leak. Second, the forced exhaust strategy instead increases the safety risk in specific scenarios. Taking the data center computer room as an example, when an axial flow fan is used to forcibly exhaust air in a confined space, although the refrigerant concentration inside the cabinet can be reduced, it will cause the refrigerant-containing air flow to diffuse to the adjacent power distribution room through the cable trench, forming a larger low-concentration dangerous area. Based on this, the present invention designs a leakage safety closed-loop control method for a refrigeration system to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a leakage safety closed-loop control method for a refrigeration system, which solves the problems in the background technique.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A leakage safety closed-loop control method for a refrigeration system, comprising the following steps:

[0008] Step S1, multi-source signal monitoring: Collect multi-modal dynamic signals of the compressor housing in real time through a multi-channel sensor array, dynamically divide 8 kHz - 16 kHz as the leakage-sensitive frequency band, and calculate the energy change rate ΔE of this frequency band;

[0009] Step S2, leakage risk determination: When ΔE≥15 dB / s and lasts for 1.5 seconds, extract the time-frequency domain characteristics of the signal, extract the singular entropy value S_entropy of the high-frequency sub-band. If the increase in S_entropy exceeds 30% of the baseline value, trigger a leakage warning;

[0010] Step S3, leakage location: Construct a time difference matrix based on the phase difference of the sensor array, and use the adaptive beamforming algorithm to invert the three-dimensional coordinates of the leakage point;

[0011] Step S4, pressure-coupled dynamic plugging: Activate the shape memory alloy plugging device, and dynamically calculate the sealing force according to the real-time system pressure P_sys and the equivalent area A_leak inverted from the leakage characteristics as:

[0012] F_seal = α·P_sys·A_leak,

[0013] where the dynamic safety factor α = 1.2 + 0.05·(P_sys / P_nominal)^2, and P_nominal is the rated pressure of the system;

[0014] Step S5, gradient temperature control seal strengthening: Adjust the drive current according to the temperature feedback of the leakage point, so that the phase change temperature T_act of the shape memory alloy rises in a gradient of T_act = T_env + ΔT·(1 - e^{-t / τ}), realizing non-linear pressure loading on the sealing contact surface.

[0015] Preferably, the multi-source signal monitoring in step S1 includes:

[0016] Step S1.1, collect the full-band voiceprint data of 0 - 20 kHz during the compressor startup stage, and establish the baseline E_baseline of the frequency band energy distribution;

[0017] Step S1.2, calculate the energy fluctuation index of each 1 / 3 octave frequency band in real time as:

[0018] W = |E_current - E_baseline| / E_baseline, where E_current is the instantaneous value of the frequency band energy obtained in real time during the detection stage;

[0019] Step S1.3: When the W mean square deviation in the 8 kHz - 16 kHz frequency band exceeds 0.25, lock this frequency band as the leakage - sensitive frequency band.

[0020] Preferably, the signal time - frequency domain feature extraction in step S2 further includes:

[0021] Step S2.1: Use a learning algorithm to extract features and classify the voiceprint signal, and establish a leakage determination model for identifying different leakage degrees and leakage positions.

[0022] Step S2.2: Combine the operating parameters of the compressor to comprehensively analyze the leakage characteristics.

[0023] Preferably, the implementation of the adaptive beamforming algorithm in step S3 includes:

[0024] Step S3.1: Construct the transfer function matrix of the sensor array.

[0025] Step S3.2: Introduce the prior constraint conditions of the leakage point:

[0026] |x_l| < 0.5L, y_l ∈ [0, H], z_l ∈ [0.2D, 0.8D], where L, H, and D are the length, width, and height of the compressor respectively.

[0027] Step S3.3: Solve the maximum - likelihood sound source position through a constrained optimization algorithm.

[0028] Preferably, the inversion method of the equivalent area A_leak in step S4 includes:

[0029] Step S4.1: Extract the sound pressure level attenuation slope K = ΔSPL / Δt in the leakage - sensitive frequency band, where ΔSPL is the change in the sound pressure level within the time interval Δt.

[0030] Step S4.2: Establish a leakage area mapping relationship according to the aero - acoustic model:

[0031] A_leak = 10^(a·logK + b), where a = 0.78 and b = - 2.15 are empirical coefficients calibrated through CFD simulation.

[0032] Preferably, the parameter setting of the gradient temperature control in step S5 includes:

[0033] Step S5.1: The initial phase - change temperature T_act0 = T_env + 10°C, where T_env represents the base temperature of the environment where the plugging device is located.

[0034] Step S5.2: According to the contact gap d_gap fed back by the displacement sensor of the plugging device, dynamically adjust the temperature - rise threshold ΔT = 5°C·{1 + tanh(2d_gap)}.

[0035] Step S5.3: The response time constant τ = τ0·(P_sys / P_nominal)^(-0.6), where τ0 = 8 s is the reference value.

[0036] Preferably, the method further includes the following steps:

[0037] Step S6: Negative pressure assisted adsorption. After the plugging device contacts the leakage point, start the micro-hole negative pressure adsorption array with annular distribution. The adsorption intensity V_vacuum satisfies: V_vacuum = k·ln{1 + P_sys·(d0 / d)}, where d is the distance between the leakage point and the adsorption port, d0 = 5 mm is the characteristic distance, and k = 0.7 is the refrigerant viscosity correction coefficient.

[0038] Preferably, the method for determining the prior constraint conditions in step S3.2 includes:

[0039] Step S3.2.1: Extract the shell size parameters L, H, D through the compressor structure finite element model;

[0040] Step S3.2.2: Conduct cluster analysis on the historical leakage point coordinates to generate a spatial distribution probability density function;

[0041] Step S3.2.3: Combine the structural strength weak point distribution map to correct the boundary value of the constraint range.

[0042] Preferably, the specific method for dynamically adjusting the temperature rise threshold in step S5.2 is as follows:

[0043] When d_gap ≤ 0.5 mm, use a constant value of ΔT = 5 °C;

[0044] When 0.5 mm < d_gap ≤ 2 mm, ΔT increases according to ΔT = 5 °C·(d_gap / 0.5)^1.2;

[0045] When d_gap > 2 mm, trigger the emergency pressurization mode, and ΔT is forced to increase to 15 °C.

[0046] Preferably, step S1 further includes the following steps:

[0047] Step S1.4: Adopt a self-calibration mechanism to regularly calibrate the sensitivity of the sensor array to ensure the accuracy of the collected voiceprint signals; the calibration period can be dynamically adjusted according to the running time and working environment of the compressor; during the calibration process, a standard sound source generator is used to generate sound signals with known frequencies and sound pressure levels. After the sensor receives the sound signal, compare the output electrical signal with the standard value, and adjust the gain and bias of the sensor through a calibration algorithm until the deviation is within the allowable range.

[0048] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0049] 1. In the present invention, through multi-source signal fusion analysis and adaptive positioning algorithms, precise identification and three-dimensional spatial positioning of the leakage point are realized, breaking through the traditional passive response mode. Combining with the pressure-coupled dynamic sealing technology, a plugging force field is generated in real time according to the leakage characteristics, and the leakage source is physically plugged at the initial stage of leakage, effectively blocking the continuous leakage path of the refrigerant and solving the problem that the prior art cannot actively intervene in the leakage source.

[0050] 2. In the present invention, a composite action mechanism of gradient temperature control for enhanced sealing and negative pressure synergistic adsorption is adopted. During the plugging process, the pressure distribution on the contact surface is dynamically adjusted to form a progressive sealing interface. Compared with the traditional forced exhaust scheme, this technical solution avoids the spread of dangerous areas caused by air flow disturbance, and realizes the directional collection of leakage substances through local negative pressure adsorption, reducing the formation probability of secondary risks in the confined space.

[0051] 3. In the present invention, based on the in-depth coupling analysis of the time-frequency domain characteristics of the acoustic signature signal and the equipment operation parameters, a multi-dimensional leakage risk assessment model is constructed. Through fuzzy logic judgment and parameter linkage mechanism, the recognition sensitivity to complex working conditions such as slow-release leakage and intermittent leakage is enhanced, solving the problem that traditional threshold alarms are easily interfered by the environment; the introduction of the leakage equivalent area inversion algorithm and the phase change material drive mechanism enables the plugging device to perform non-linear responses according to system pressure fluctuations, leakage rates, and environmental temperature changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is the step flow chart of the adaptive plugging method of the present invention;

[0053] Figure 2 is the multi-source signal monitoring flow chart of the present invention;

[0054] Figure 3 is the signal time-frequency feature extraction flow chart of the present invention;

[0055] Figure 4 is the adaptive beamforming algorithm of the present invention;

[0056] Figure 5 is the parameter setting diagram of the gradient temperature control of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] Example 1

[0059] Please refer to Figures 1-5 , in the embodiment of the present invention, a leakage safety closed-loop control method for a refrigeration system includes the following steps:

[0060] Step S1, multi-source signal monitoring: real-time collect multi-modal dynamic signals of the compressor housing through a multi-channel sensor array, dynamically divide 8 kHz - 16 kHz as the leakage-sensitive frequency band and calculate the energy change rate ΔE of this frequency band;

[0061] Step S2, leakage risk determination: when ΔE ≥ 15 dB / s and lasts for 1.5 seconds, extract the time-frequency domain characteristics of the signal, extract the singular entropy value S_entropy of the high-frequency sub-band. If the increase in S_entropy exceeds 30% of the baseline value, trigger a leakage warning;

[0062] Step S3, leakage location: construct a time difference matrix based on the phase difference of the sensor array, and use the adaptive beamforming algorithm to invert the three-dimensional coordinates of the leakage point;

[0063] Step S4, pressure-coupled dynamic plugging: activate the shape memory alloy plugging device, and dynamically calculate the sealing force according to the real-time system pressure P_sys and the equivalent area A_leak inverted from the leakage characteristics as:

[0064] F_seal = α·P_sys·A_leak,

[0065] where the dynamic safety factor α = 1.2 + 0.05·(P_sys / P_nominal)^2, and P_nominal is the rated pressure of the system;

[0066] Step S5, gradient temperature control seal strengthening: adjust the drive current according to the temperature feedback of the leakage point, so that the phase change temperature T_act of the shape memory alloy rises linearly as T_act = T_env + ΔT·(1 - e^{-t / τ}) to achieve non-linear pressure loading on the sealing contact surface.

[0067] The multi-source signal monitoring in Step S1 includes:

[0068] Step S1.1, collect the full-band soundprint data of 0 - 20 kHz during the compressor startup phase and establish the baseline E_baseline of the frequency band energy distribution;

[0069] Step S1.2, calculate the energy fluctuation index of each 1 / 3 octave frequency band in real time as:

[0070] W = |E_current - E_baseline| / E_baseline, where E_current is the instantaneous value of the frequency band energy obtained in real time during the detection phase;

[0071] Step S1.3: When the mean square deviation of W in the 8 kHz - 16 kHz frequency band exceeds 0.25, lock this frequency band as the leakage - sensitive frequency band.

[0072] The signal time - frequency domain feature extraction in Step S2 also includes:

[0073] Step S2.1: Use a learning algorithm to extract features and classify the voiceprint signal, and establish a leakage determination model for identifying different leakage degrees and leakage positions;

[0074] Step S2.2: Combine the operating parameters of the compressor to comprehensively analyze the leakage characteristics;

[0075] The comprehensive analysis includes the following steps:

[0076] Step S2.2.1: Construct a multi - parameter feature vector, integrate the leakage characteristics of the voiceprint signal and the operating parameters of the compressor to construct a feature vector;

[0077] Step S2.2.2: Calculate the weighted sum, calculate the comprehensive evaluation index value through weighted sum, and determine leakage and evaluate the severity according to the threshold;

[0078] Step S2.2.3: Fuzzy logic judgment, adopt the fuzzy logic method, convert the parameters into fuzzy membership degrees, and comprehensively reason and judge the leakage possibility and its degree, including the following specific implementation methods:

[0079] Definition of input variable fuzzification: Define the domain of discourse as [0%, 50%], divided into three fuzzy sets {low, medium, high}, using triangular membership functions with vertices at 15%, 25%, and 35% respectively. The domain of discourse is [0, P_nominal×10%], divided into {stable, abnormal, dangerous}, and the membership function adopts a trapezoidal distribution with turning points set according to the system rated pressure levels. For the axial temperature gradient ΔT, the domain of discourse is [-5℃, +15℃], divided into {negative anomaly, normal, leakage feature};

[0080] Construction of the fuzzy rule base: For the multi - sensor conflict scenario, define arbitration rules: when max(ΔT membership degree, η membership degree)×min(ΔS confidence interval) > 0.7, force an upgrade of the leakage level. Introduce a time - decay factor. When the early warning continues to exceed the time limit without being located, reduce the weight of the misjudgment rule according to λ = 1 - e^(-t / τ), where τ is taken as 1.2 times the system response time;

[0081] The reasoning and decision-making mechanism adopts the Mamdani fuzzy inference method, performs max-min composition operations on the antecedents of the rules, divides the output universe of discourse [0, 1] of the leakage possibility into 5 levels, and obtains the exact value through the centroid method for defuzzification;

[0082] Online update of the knowledge base, design a rule weight self-adjustment algorithm. When the number of successful plugging cases accumulates more than 50 times, optimize the rule confidence parameter according to the gradient descent method, introduce an adversarial training mechanism, and dynamically prune redundant fuzzy rules by injecting the feature vectors of historical false alarm samples;

[0083] Step S2.2.4, Parameter linkage analysis and threshold adjustment, pay attention to the linkage changes between operating parameters, and dynamically adjust the leakage determination threshold.

[0084] The implementation of the adaptive beamforming algorithm in step S3 includes:

[0085] Step S3.1, Construct the transfer function matrix of the sensor array;

[0086] Step S3.2, Introduce the prior constraint conditions of the leakage point:

[0087] |x_l| < 0.5L, y_l ∈ [0, H], z_l ∈ [0.2D, 0.8D], where L, H, and D are the length, width, and height of the compressor respectively;

[0088] Step S3.3, Solve the maximum likelihood sound source position through the constrained optimization algorithm.

[0089] The inversion method of the equivalent area A_leak in step S4 includes:

[0090] Step S4.1, Extract the sound pressure level attenuation slope K = ΔSPL / Δt in the leakage-sensitive frequency band, where ΔSPL is the change in the sound pressure level within the time interval Δt;

[0091] Step S4.2, Establish the leakage area mapping relationship according to the aeroacoustic model:

[0092] A_leak = 10^(a·logK + b), where a = 0.78 and b = -2.15 are the empirical coefficients calibrated through CFD simulation.

[0093] The parameter setting of gradient temperature control in step S5 includes:

[0094] Step S5.1, The initial phase change temperature T_act0 = T_env + 10°C, where T_env represents the base temperature of the environment where the plugging device is located;

[0095] Step S5.2, According to the contact gap d_gap feedback by the plugging device displacement sensor, dynamically adjust the temperature rise threshold ΔT = 5°C·{1 + tanh(2d_gap)};

[0096] Step S5.3, the response time constant τ = τ0·(P_sys / P_nominal)^(-0.6), where τ0 = 8s is the reference value.

[0097] The working principle of the embodiment of the present invention is as follows: The system collects the 0-20kHz broadband vibration signal on the surface of the compressor housing in real time through a distributed sensor array, and uses the 1 / 3 octave band energy analysis algorithm to dynamically identify the leakage-sensitive frequency band, that is, 8kHz-16kHz. This frequency band selection is based on the broadband noise characteristics generated by the turbulent cavitation effect during micro-leakage of the refrigerant. When the energy change rate ΔE≥15dB / s in the sensitive frequency band is detected and lasts for 1.5 seconds, a three-level wavelet decomposition process is triggered. The multi-dimensional cross-verification of the leakage characteristics is realized by the increase in the singular entropy of the high-frequency sub-band exceeding the baseline by 30%. Combining the fuzzy logic correlation analysis of the compressor operating pressure and temperature parameters, the false alarm rate is controlled below 0.3%.

[0098] Based on the time difference direction finding principle, a transfer function matrix of the sensor array is constructed, and the constrained maximum likelihood estimation algorithm is used to solve the three-dimensional coordinates of the leakage source. By introducing the geometric constraint conditions of the compressor housing (x_l < 0.5L, y_l ∈ [0, H], z_l ∈ [0.2D, 0.8D]), the dimension of the solution space is reduced. Combining the adaptive beamforming technology, a positioning accuracy of 1.5mm is achieved. Compared with the traditional ultrasonic detection technology, the spatial resolution of this positioning method is improved by two orders of magnitude.

[0099] The shape memory alloy actuator dynamically adjusts the contact pressure according to the leakage equivalent area A_leak inverted from the voiceprint and the real-time system pressure P_sys according to the formula F_seal = α·P_sys·A_leak, where the dynamic safety factor α = 1.2 + 0.05·(P_sys / P_nominal)^2 realizes non-linear compensation through pressure square correction, effectively coping with the risk of seal failure caused by system pressure fluctuations.

[0100] The contact gap d_gap is real-time fed back through a displacement sensor, and the hyperbolic tangent function is used to dynamically adjust the temperature rise threshold ΔT = 5℃·{1 + tanh(2d_gap)}, and a gradient temperature rise curve of T_act = T_env + ΔT·(1 - e^(-t / τ)) is constructed. The design of the response time constant τ = 8·(P_sys / P_nominal)^(-0.6) makes the seal establishment time negatively correlated with the system pressure, realizing a fast seal response under the condition of 4.2 seconds @ 10bar.

[0101] Embodiment 2

[0102] Please refer to Figures 1-5 In the embodiment of the present invention, the method further includes the following steps:

[0103] Step S6, negative pressure assisted adsorption. After the plugging device contacts the leakage point, activate the micro-hole negative pressure adsorption array with annular distribution, and the adsorption intensity V_vacuum satisfies: V_vacuum = k·ln{1 + P_sys·(d0 / d)}, where d is the distance between the leakage point and the adsorption port, d0 = 5mm is the characteristic distance, and k = 0.7 is the refrigerant viscosity correction coefficient. The distance between the leakage point and the adsorption port, d0 = 5mm is the characteristic distance, and k = 0.7 is the refrigerant viscosity correction coefficient.

[0104] The determination method of the prior constraint conditions in Step S3.2 includes:

[0105] Step S3.2.1, extract the shell size parameters L, H, D through the compressor structure finite element model;

[0106] Step S3.2.2, perform cluster analysis on the historical leakage point coordinates to generate a spatial distribution probability density function;

[0107] Step S3.2.3, combine the structural strength weak point distribution map to correct the boundary value of the constraint range.

[0108] The specific method for dynamically adjusting the temperature rise threshold in Step S5.2 is:

[0109] When d_gap ≤ 0.5mm, adopt a constant value of ΔT = 5℃;

[0110] When 0.5mm < d_gap ≤ 2mm, ΔT increases according to ΔT = 5℃·(d_gap / 0.5)^{1.2};

[0111] When d_gap > 2mm, trigger the emergency pressurization mode, and ΔT is forced to increase to 15℃.

[0112] Step S1 also includes the following steps:

[0113] Step S1.4, adopt a self-calibration mechanism to regularly calibrate the sensitivity of the sensor array to ensure the accuracy of the collected voiceprint signals; the calibration period can be dynamically adjusted according to the running duration and working environment of the compressor; during the calibration process, a standard sound source generator is used to generate sound signals with known frequencies and sound pressure levels. After the sensor receives the sound signal, compare the output electrical signal with the standard value, and adjust the gain and bias of the sensor through the calibration algorithm until the deviation is within the allowable range.

[0114] The working principle of the embodiment of the present invention is as follows: The leakage determination module integrates wavelet singular entropy analysis and compressor operating parameters to construct a three-level verification mechanism: When the energy change rate ΔE of the sensitive frequency band reaches 15 dB / s and lasts for 1.5 seconds, it triggers the baseline threshold verification of a 30% increase in the singular entropy of the high-frequency subband; a fuzzy logic decision tree is constructed by combining the pressure pulsation coefficient and the temperature change rate to achieve a quantitative evaluation of the leakage probability. The positioning algorithm extracts the shell size parameters L / H / D through the compressor finite element model, generates a spatial probability density map by clustering analysis of historical leakage points, constructs a three-dimensional search domain with the x-axis limited within the range of 0.5L and the z-axis limited within the interval of 0.2D - 0.8D, and uses an improved beamforming algorithm to achieve a positioning accuracy of 1.2 mm.

[0115] The temperature control module implements a hierarchical regulation strategy according to the contact gap d_gap. When the gap ≤ 0.5 mm, it maintains a temperature rise of 5°C; in the interval of 0.5 - 2 mm, the temperature rise threshold is increased according to the exponential relationship of (d_gap / 0.5)^1.2; when it exceeds 2 mm, it triggers an emergency temperature rise mode of 15°C. The negative pressure adsorption unit starts after the sealing contact, generates a dynamic adsorption force through a 200-μm microporous array with an annular distribution. The calculation model of the adsorption intensity V_vacuum introduces a refrigerant viscosity correction coefficient of 0.7, constructs a logarithmic function relationship by combining the system pressure and distance parameters, and can form a negative pressure gradient of 12 kPa at a characteristic distance of 5 mm to achieve the directional capture of the leaked fluid.

[0116] The system integrates acoustic fingerprint features, pressure pulsation, temperature field distribution, and historical leakage data to construct a four-dimensional feature space, assigns feature weights through the entropy weight method, and realizes a quantitative evaluation of the leakage probability. The decision module and the infrared thermal imaging system form a cross-verification mechanism. When there is spatial consistency between the acoustic fingerprint determination result and the thermal imaging temperature difference distribution, it triggers the final sealing instruction.

[0117] Embodiment 3

[0118] A leakage safety closed-loop control method for a refrigeration system is provided. During the compressor startup phase, acoustic fingerprint signals in the full frequency band from 0 kHz to 20 kHz are collected through a distributed acoustic fingerprint sensor array to establish a baseline of the frequency band energy distribution. During real-time monitoring, the energy fluctuation index of each frequency band is calculated with a 1 / 3 octave as a unit. When the mean square deviation of the energy in the frequency band from 8 kHz to 16 kHz exceeds 0.25, this frequency band is dynamically locked as the leakage-sensitive frequency band, and its energy change rate is calculated. When the energy change rate continuously reaches 15 decibels per second for 1.5 seconds, a leakage warning is triggered. The sensor array starts the self-calibration mechanism every 500 hours of operation, uses a standard sound source generator to output a 94-dB, 1-kHz calibration signal, and controls the sensor sensitivity deviation within ±0.5 dB through the least squares algorithm.

[0119] After the warning is triggered, the system performs three-level wavelet decomposition on the voiceprint signal and extracts the singular entropy features of the high-frequency subbands above 16 kHz. When the increase in the singular entropy value exceeds 30% compared to the baseline, a multi-dimensional decision model is constructed by combining the compressor operation parameters: integrating the real-time pressure pulsation coefficient (fluctuation threshold ±7% of the rated pressure), the bearing temperature change rate (monitoring gradient 2°C / minute), and the historical leakage data. The feature weights are assigned by the entropy weight method (voiceprint weight 0.6, pressure weight 0.3, temperature weight 0.1), and the leakage probability is calculated using a fuzzy logic decision tree. When the comprehensive evaluation index exceeds 0.85, it is determined as an effective leakage event, and the false alarm rate is lower than 0.3%.

[0120] Based on the phase difference of the sensor array, a time difference matrix is constructed, and the geometric constraints of the compressor housing are introduced to limit the coordinate range of the leakage point within 50% of the housing length, the full range in the height direction, and the interval of 20% to 80% in the depth direction. Through an improved adaptive beamforming algorithm, combined with the spatial probability density map (kernel radius 3 mm) generated by the historical leakage point clustering analysis, the maximum likelihood sound source position is solved within the constraint domain, achieving a positioning accuracy of 1.2 mm.

[0121] The shape memory alloy plug moves to the leakage point according to the positioning coordinates, and the system pressure is collected in real time to invert the equivalent leakage area. The leakage area is calculated by the slope of the sound pressure level attenuation. When it is detected that the sound pressure level drops by 4 dB within 0.5 s, the equivalent area is mapped based on the pneumoacoustic model, where the empirical coefficients a calibrated by computational fluid dynamics simulation are 0.78 and b is -2.15. The sealing force is calculated according to the dynamic safety factor formula, and the coefficient α is set to 1.2 plus the square of 0.05 times the ratio of the system real-time pressure to the rated pressure, ensuring effective sealing under the condition of pressure fluctuation ±20%.

[0122] After the plug makes contact, the temperature control strategy is adjusted in stages according to the gap value feedback by the displacement sensor: when the gap ≤ 0.5 mm, a 5°C temperature rise is maintained; when the gap is between 0.5 and 2 mm, the temperature rise increases according to the 1.2th power of the ratio of the gap value to 0.5 mm; when the gap exceeds 2 mm, the emergency mode is triggered and the temperature rise is forced to increase to 15°C. The temperature control response time constant τ is designed as 8 s multiplied by the -0.6th power of the ratio of the system pressure to the rated pressure, achieving a rapid sealing response of 4.2 s at 10 bar pressure.

[0123] After the plug makes contact, a micro-porous adsorption array with a circular distribution (pore diameter 200 μm, pore spacing 1.5 mm) is activated, and the adsorption intensity is dynamically adjusted according to the real-time system pressure and the distance to the leakage point. The refrigerant viscosity correction coefficient of 0.7 is introduced into the adsorption force model to generate a negative pressure gradient of 12 kPa at a characteristic distance of 5 mm, effectively capturing the leaked refrigerant.

[0124] In this embodiment, through the coordination of voiceprint feature recognition, multi-parameter fusion decision-making, high-precision positioning, and intelligent plugging technologies, rapid plugging of micro-leaks (≥0.3 mm aperture) is achieved within 135 seconds during the operation of the compressor, with a plugging success rate exceeding 98.7%, and the system can continuously and stably operate for no less than 2000 hours.

[0125] Working principle: The present invention relates to a refrigerant leakage self-adaptive plugging method based on voiceprint recognition, and its working principle realizes the synergistic effect of leakage detection, positioning, and dynamic plugging through multi-modal signal fusion and closed-loop control. The system uses a distributed voiceprint sensor array to continuously collect the 0-20 kHz broadband vibration signal on the surface of the compressor housing, adopts a dynamic frequency band division algorithm to lock the 8 kHz-16 kHz leakage-sensitive frequency band, and realizes frequency band self-calibration based on the calculation of the 1 / 3 octave energy fluctuation index to ensure the signal acquisition accuracy. When the energy change rate ΔE of the sensitive frequency band reaches 15 dB / s and lasts for 1.5 seconds, multi-scale wavelet decomposition and high-frequency sub-band singular entropy analysis are triggered, combined with the fuzzy logic decision model of the compressor operating pressure and temperature parameters, to construct a multi-dimensional leakage determination mechanism, effectively reducing the false alarm rate.

[0126] During the leakage positioning stage, a time difference matrix is constructed based on the phase difference of the sensor array. The structural parameters of the compressor are extracted through a finite element model and combined with the historical leakage point spatial distribution probability. Geometric constraint conditions are introduced to optimize the beamforming algorithm to achieve millimeter-level three-dimensional coordinate inversion. The actuator calculates the non-linear sealing force using the dynamic safety factor α corrected by the square of the pressure according to the leakage equivalent area A_leak inversed from the voiceprint spectrum and the real-time system pressure P_sys, and drives the shape memory alloy to implement self-adaptive plugging. The gradient temperature control module adjusts the phase change temperature in stages according to the contact gap d_gap, and realizes progressive pressure loading on the sealing surface through an exponential response curve to balance the sealing efficiency and mechanical impact.

[0127] The system further integrates a negative pressure adsorption unit, starts the annular micro-hole array after plugging contact, dynamically adjusts the adsorption intensity based on the viscous characteristics of the refrigerant and the pressure-distance logarithmic model, and forms a collaborative capture mechanism for the leaked fluid. The multi-modal decision-making architecture fuses voiceprint features, thermal imaging data, and historical leakage information, and realizes high-precision determination under complex working conditions through weighted feature space evaluation and threshold dynamic adjustment.

[0128] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirits of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A leakage safety closed-loop control method for a refrigeration system, characterized in that It includes the following steps: Step S1, multi-source signal monitoring: Multimodal dynamic signals of the compressor housing are collected in real time through a multi-channel sensor array. The frequency band of 8 kHz - 16 kHz is dynamically divided as the leakage-sensitive frequency band, and the energy change rate ΔE of this frequency band is calculated; Step S2, leakage risk determination: When ΔE ≥ 15 dB / s and lasts for 1.5 seconds, the time-frequency domain characteristics of the signal are extracted, and the singular entropy value S_entropy of the high-frequency sub-band is extracted. If the increase in S_entropy exceeds 30% of the baseline value, a leakage warning is triggered; Step S3, leakage location: Based on the phase difference of the sensor array, a time difference matrix is constructed, and the three-dimensional coordinates of the leakage point are inverted using the adaptive beamforming algorithm; Step S4, pressure-coupled dynamic plugging: Activate the shape memory alloy plugging device, and according to the real-time system pressure P_sys and the equivalent area A_leak inverted from the leakage characteristics, dynamically calculate the sealing force as: F_seal = α·P_sys·A_leak, where the dynamic safety factor α = 1.2 + 0.05·(P_sys / P_nominal)^2, and P_nominal is the rated pressure of the system; Step S5, gradient temperature control for seal strengthening: Adjust the drive current according to the temperature feedback at the leakage point, so that the phase change temperature T_act of the shape memory alloy rises linearly as T_act = T_env + ΔT·(1 - e^{-t / τ}), realizing non-linear pressure loading on the sealing contact surface.

2. The leakage safety closed-loop control method of a refrigeration system according to claim 1, characterized in that, The multi-source signal monitoring in step S1 includes: Step S1.1, collect the full-band voiceprint data of 0 - 20 kHz during the compressor startup phase, and establish the baseline E_baseline of the frequency band energy distribution; Step S1.2, calculate the energy fluctuation index of each 1 / 3 octave frequency band in real time as: W = |E_current - E_baseline| / E_baseline, where E_current is the instantaneous value of the frequency band energy obtained in real time during the detection phase; Step S1.3, when the mean square deviation of W in the 8 kHz - 16 kHz frequency band exceeds 0.25, lock this frequency band as the leakage-sensitive frequency band.

3. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that, The extraction of the time-frequency domain characteristics of the signal in step S2 also includes: Step S2.1, use the learning algorithm to extract and classify the voiceprint signal, and establish a leakage determination model for identifying different leakage degrees and leakage locations; Step S2.2, combine the operating parameters of the compressor to comprehensively analyze the leakage characteristics; The comprehensive analysis includes the following steps: Step S2.2.1, construct a multi-parameter feature vector, integrate the leakage characteristics of the voiceprint signal and the operating parameters of the compressor to construct a feature vector; Step S2.2.2, weighted summation calculation, calculate the comprehensive evaluation index value through weighted summation, and determine the leakage and evaluate the severity according to the threshold; Step S2.2.3, fuzzy logic judgment, use the fuzzy logic method to convert the parameters into fuzzy membership degrees, and comprehensively reason and judge the leakage possibility and its degree; Step S2.2.4, parameter linkage analysis and threshold adjustment, pay attention to the linkage changes between operating parameters, and dynamically adjust the leakage determination threshold.

4. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that The implementation of the adaptive beamforming algorithm in step S3 includes: Step S3.1: Construct the transfer function matrix of the sensor array; Step S3.2: Introduce the prior constraint conditions of the leakage point: |x_l| < 0.5L, y_l ∈ [0, H], z_l ∈ [0.2D, 0.8D], where L, H, and D are the length, width, and height of the compressor respectively; Step S3.3: Solve the maximum likelihood sound source position through the constrained optimization algorithm.

5. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that The inversion method of the equivalent area A_leak in step S4 includes: Step S4.1: Extract the sound pressure level attenuation slope K = ΔSPL / Δt of the leakage-sensitive frequency band, where ΔSPL is the change in the sound pressure level within the time interval Δt; Step S4.2: Establish the leakage area mapping relationship according to the aeroacoustic model: A_leak = 10^(a·logK + b), where a = 0.78 and b = -2.15 are empirical coefficients calibrated through CFD simulation.

6. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that, The parameter setting of the gradient temperature control in step S5 includes: Step S5.1: The initial phase change temperature T_act0 = T_env + 10°C, where T_env represents the base temperature of the environment where the plugging device is located; Step S5.2: Dynamically adjust the temperature rise threshold ΔT = 5°C·{1 + tanh(2d_gap)} according to the contact gap d_gap feedback by the displacement sensor of the plugging device; Step S5.3: The response time constant τ = τ0·(P_sys / P_nominal)^(-0.6), where τ0 = 8s is the reference value.

7. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that, The method further includes the following steps: Step S6: Negative pressure collaborative adsorption. After the plugging device contacts the leakage point, start the annularly distributed micro-hole negative pressure adsorption array, and the adsorption intensity V_vacuum satisfies: V_vacuum = k·ln{1 + P_sys·(d0 / d)}, where d is the distance between the leakage point and the adsorption port, d0 = 5mm is the characteristic distance, and k = 0.7 is the refrigerant viscosity correction coefficient.

8. A leakage safety closed-loop control method for a refrigeration system according to claim 4, characterized in that, The determination method of the prior constraint conditions in step S3.2 includes: Step S3.2.1: Extract the shell size parameters L, H, and D through the finite element model of the compressor structure; Step S3.2.2: Perform clustering analysis on the historical leakage point coordinates to generate the spatial distribution probability density function; Step S3.2.3: Combine the distribution map of the structural strength weak points to correct the boundary values of the constraint range.

9. A leakage safety closed-loop control method for a refrigeration system according to claim 6, characterized in that The specific method of dynamically adjusting the temperature rise threshold in step S5.2 is: When d_gap ≤ 0.5mm, use the constant value ΔT = 5°C; When 0.5mm < d_gap ≤ 2mm, ΔT increases according to ΔT = 5°C·(d_gap / 0.5)^1.2; When d_gap > 2mm, trigger the emergency pressurization mode, and ΔT is forced to increase to 15°C.

10. A leakage safety closed-loop control method for a refrigeration system according to claim 1, characterized in that, Step S1 further includes the following steps: Step S1.4: Adopt a self-calibration mechanism to regularly calibrate the sensitivity of the sensor array to ensure the accuracy of the collected voiceprint signals; the calibration period can be dynamically adjusted according to the running duration and working environment of the compressor; during the calibration process, a standard sound source generator is used to generate sound signals with known frequencies and sound pressure levels. After the sensor receives the sound signal, the electrical signal output by it is compared with the standard value, and the gain and bias of the sensor are adjusted through a calibration algorithm until the deviation is within the allowable range.

Images