A method for testing a seawater desalination membrane pressure difference disturbance

Abstract

The present application relates to the technical field of seawater desalination membrane monitoring, and particularly relates to a seawater desalination membrane differential pressure disturbance testing method. The method comprises: obtaining environmental parameters of a membrane system; applying a composite excitation signal composed of a low-frequency step and a high-frequency pulse partially overlapping in time; collecting a differential pressure response signal and adaptively decomposing it into a low-frequency trend component and a transient component; after normalization of the two components based on the environmental parameters, extracting a first characteristic parameter reflecting boundary layer resistance and a second characteristic parameter reflecting physical structure damage; constructing a health vector, comparing it with a health benchmark space, and comprehensively determining the boundary layer resistance state and the physical damage degree of the membrane system by combining a prediction model. Through double-mode composite excitation and adaptive decoupling analysis, early warning and precise operation and maintenance are realized, and the reliability and intelligent level of membrane system monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of seawater desalination membrane monitoring technology, and in particular to a method for testing differential pressure disturbances in seawater desalination membranes. Background Technology

[0002] In the long-term operation of seawater desalination reverse osmosis membrane systems, conventional process parameters such as feed water pressure, differential pressure, and permeate flow rate are widely used to monitor membrane performance. However, when a significant increase in differential pressure occurs, operators often struggle to accurately determine the root cause. On one hand, the membrane surface may experience increased fluid flow resistance due to inorganic scaling, organic fouling, or biofilm growth, which represents recoverable performance degradation. On the other hand, physical structures such as membrane bags, flow guides, and central tubes may suffer fatigue, creep, or mechanical damage under long-term alternating pressure, and this type of damage is usually irreversible. In existing technologies, pressure monitoring data is a mixed signal reflecting multiple physicochemical changes, making it impossible to separate the increase in boundary layer resistance from pressure response changes caused by structural damage. Furthermore, daily fluctuations in feed water temperature, salinity, and flow rate further interfere with the interpretation of differential pressure signals, making alarm methods based on static thresholds or simple trends prone to false alarms or missed alarms. Summary of the Invention

[0003] To overcome the above shortcomings, this invention provides a method for testing differential pressure disturbance of seawater desalination membranes, aiming to improve the problems of existing technologies that cannot distinguish between boundary layer resistance and structural damage online, and that suffer from large environmental interference, monitoring lag, and easy false alarms and missed alarms.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for testing differential pressure disturbance in seawater desalination membranes, comprising:

[0006] Obtain the environmental parameters of the membrane system under its current operating conditions;

[0007] The pressure at the inlet of the membrane system generates a composite excitation signal, which includes a low-frequency step and a high-frequency pulse, and the two overlap at least partially in time.

[0008] The differential pressure response signal of the membrane system under the action of the composite excitation signal is collected, and the differential pressure response signal is adaptively decomposed into low-frequency trend component and transient component.

[0009] After normalizing the low-frequency trend component and transient component based on the environmental parameters, a first characteristic parameter reflecting the boundary layer resistance of the membrane surface is extracted from the low-frequency trend component, and a second characteristic parameter reflecting the physical structural damage of the membrane element is extracted from the transient component.

[0010] A health vector is constructed using the first and second feature parameters. The health vector is then compared with the health reference space and combined with a prediction model to determine the boundary layer resistance state and physical damage level of the membrane system.

[0011] Furthermore, the environmental parameters include influent temperature, salinity, and basic flow rate.

[0012] Furthermore, the low-frequency step and high-frequency pulse specifically include:

[0013] The low-frequency step refers to the inlet pressure switching from a first pressure value to a second pressure value and maintaining it at the second pressure value for a preset duration, which is used to induce the controlled evolution of the membrane surface boundary layer.

[0014] The high-frequency pulse refers to a periodic pressure fluctuation superimposed during the duration of a low-frequency step, used to excite the mechanical response of the membrane element.

[0015] Furthermore, the frequency of the high-frequency pulse is dynamically adjusted based on the service life of the membrane system.

[0016] Furthermore, the adaptive decomposition step includes:

[0017] By employing morphological component analysis, and constructing a total variation dictionary for low-frequency trend components and a sparse pulse dictionary for transient components, the optimal solutions for the two components are iteratively solved in the sparse domain.

[0018] Furthermore, the normalization process includes:

[0019] Temperature-based viscosity compensation, salinity-based osmotic pressure compensation, and base flow rate normalization are applied to the low-frequency trend components.

[0020] Perform pressure reference calibration on transient components based on the current inlet water pressure;

[0021] The transient components are demodulated and calibrated based on the instantaneous amplitude of the low-frequency step to eliminate the modulation effect of the low-frequency background on the high-frequency response.

[0022] Furthermore, the first characteristic parameter includes at least one of relaxation time, time constant, and equivalent derived index of relaxation characteristics; the second characteristic parameter includes at least one of phase difference offset, amplitude-frequency response change, and equivalent derived index of phase characteristics.

[0023] Furthermore, the extraction of the first feature parameter and the second feature parameter is performed using at least one of the following methods:

[0024] Frequency domain filtering is used to separate the low-frequency and high-frequency responses of the differential pressure response signal;

[0025] Time-domain fitting is used to extract the relaxation time of the differential pressure response signal;

[0026] Transfer function analysis is used to extract the phase difference from the differential pressure response signal.

[0027] Furthermore, the step of combining the prediction model for comprehensive determination includes:

[0028] The first feature parameter and the second feature parameter are combined into a multi-dimensional vector, which serves as the health vector of the current state of the membrane system.

[0029] The distance offset between the health vector and the static health baseline space is calculated, and multi-criteria fusion is performed by combining the dynamic residuals output by the prediction model to determine the state of the membrane system.

[0030] The health baseline space is established based on historical characteristic parameters of the membrane system under its initial health state, and is automatically reset and updated after chemical cleaning is performed and the system is running stably.

[0031] Furthermore, it also includes:

[0032] Based on the determined boundary layer resistance state and physical damage level, graded early warning information is output;

[0033] The first level warning indicates adjusting operating parameters to slow down boundary layer accumulation; the second level warning indicates performing chemical cleaning; and the third level warning indicates replacing damaged membrane elements.

[0034] Furthermore, it also includes:

[0035] During the application of the composite excitation signal, the instantaneous pressure peak at the inlet is monitored in real time. If the instantaneous pressure peak exceeds the safe pressure threshold of the membrane element, the excitation signal is immediately terminated and the pressure is released.

[0036] The present invention has the following beneficial effects:

[0037] 1. In this invention, by employing a composite excitation of low-frequency step and high-frequency pulse time overlap, the boundary layer evolution and structural vibration are simultaneously excited, and the low-frequency trend and transient components are obtained through adaptive decomposition, thereby distinguishing online between the increase in boundary layer drag and the damage to the physical structure.

[0038] 2. In this invention, viscosity, osmotic pressure and flow rate compensation are performed on low-frequency trend components by utilizing inlet water temperature, salinity and flow rate, and pressure calibration and demodulation calibration are performed on transient components to eliminate the influence of environmental fluctuations and background modulation. This ensures that characteristic parameters such as relaxation time and phase difference have high signal-to-noise ratio and repeatability, and avoids misjudgment caused by changes in operating conditions.

[0039] 3. In this invention, by constructing the characteristic parameters into a health vector, and combining the Mahalanobis distance statistical deviation and the exponential smoothing prediction residual for multi-criteria fusion judgment, early identification of boundary layer accumulation and sudden damage is achieved; based on the graded early warning, instructions for adjusting parameters, chemical cleaning or replacing membrane elements are output, and the benchmark is automatically retested and updated after maintenance, forming a closed-loop management of diagnosis, maintenance and verification, thereby improving the reliability and intelligence level of membrane system monitoring. Attached Figure Description

[0040] Figure 1 This is a flowchart of a seawater desalination membrane differential pressure disturbance testing method proposed in this invention;

[0041] Figure 2 This is a flowchart of the adaptive decomposition of differential pressure response signal proposed in this invention;

[0042] Figure 3 This is a flowchart of the normalization of the low-frequency trend component and the extraction of the first feature parameter proposed in this invention.

[0043] Figure 4 This is a flowchart of the normalization of transient components and extraction of the second feature parameter proposed in this invention;

[0044] Figure 5 This is a flowchart of the health vector construction and comprehensive judgment process proposed in this invention. Detailed Implementation

[0045] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] This invention provides a method for testing differential pressure disturbance in seawater desalination membranes, such as... Figure 1 As shown, it includes the following steps:

[0047] S100: Obtain environmental parameters of the membrane system under current operating conditions;

[0048] Furthermore, the environmental parameters include influent temperature, salinity, and basic flow rate.

[0049] Specifically, a sensor array deployed at the inlet of the seawater desalination unit collects basic data reflecting the physical environment of the membrane system in real time.

[0050] The inlet water temperature is obtained in real time through a platinum resistance temperature sensor installed in the inlet water pipeline. The inlet water temperature directly affects the dynamic viscosity of seawater, thereby changing the frictional resistance of the water flow through the membrane element mesh.

[0051] Influent salinity is monitored in real time using a conductivity meter. Salinity determines osmotic pressure and is a key factor affecting the effective pressure differential.

[0052] The base flow rate is obtained by measuring the influent flow rate using an electromagnetic or ultrasonic flow meter. The flow rate determines the velocity distribution and Reynolds number within the membrane bag and is the primary kinetic factor contributing to the initial pressure differential. In this embodiment, the base operating flow rate of the membrane system is set to [value missing]. .

[0053] To ensure the accuracy of environmental parameters, the control system performs a 30-second moving average filtering process on the acquired raw analog signals to eliminate high-frequency noise from the sensors. The preset sampling frequency for environmental parameters is 10Hz. When the standard deviation of the collected parameters is less than 0.5% over 10 consecutive sampling periods, the average value is defined as the baseline environmental parameter for the current operating condition. If the monitored inlet water temperature fluctuation exceeds 2°C per hour, or the salinity change exceeds 5%, or the flow rate fluctuation exceeds [a certain value], [further action is taken]. It will issue an environmental instability alarm and suspend the application of subsequent excitation signals until the parameters stabilize again within the preset threshold range.

[0054] By synchronously collecting influent temperature, salinity, and basic flow rate through a sensor array, and combining moving average filtering with multi-parameter stability criteria, it can automatically identify whether the environment is in a steady state before the excitation test; when any parameter exceeds the threshold, an instability alarm is immediately triggered and the test is suspended, thereby avoiding interference from the non-steady-state environment on the differential pressure response.

[0055] S200: The pressure at the inlet of the membrane system generates a composite excitation signal, which includes a low-frequency step and a high-frequency pulse, and the two overlap at least partially in time.

[0056] Furthermore, the low-frequency step and high-frequency pulse specifically include:

[0057] The low-frequency step refers to the inlet pressure switching from a first pressure value to a second pressure value and maintaining it at the second pressure value for a preset duration, which is used to induce the controlled evolution of the membrane surface boundary layer.

[0058] The high-frequency pulse refers to a periodic pressure fluctuation superimposed during the duration of a low-frequency step, used to excite the mechanical response of the membrane element.

[0059] Furthermore, the frequency of the high-frequency pulse is dynamically adjusted based on the service life of the membrane system.

[0060] Specifically, the variable frequency drive of the high-pressure pump works in conjunction with the high-speed regulating valve at the inlet to change the inlet pressure within a preset time window, generating a composite excitation signal composed of a low-frequency step signal and a high-frequency pulse signal.

[0061] The low-frequency step signal is generated by adjusting the output speed of the high-pressure pump. In this embodiment, the inlet pressure is switched from a first pressure value of 5.0 MPa to a second pressure value of 5.5 MPa within 2 seconds. The second pressure value of 5.5 MPa is maintained for 300 seconds. The purpose of this step pressure is to disrupt the original concentration polarization equilibrium by instantaneously changing the membrane flux, inducing controlled thickness evolution and solute redistribution in the membrane surface boundary layer, thereby providing a dynamic response background for subsequent extraction of boundary layer resistance characteristics.

[0062] A high-frequency pulse signal is superimposed on the duration of a low-frequency step signal. Specifically, this is achieved by using a high-speed regulating valve to perform micro-amplitude reciprocating motion at a specific frequency, generating periodic pressure fluctuations. In this embodiment, the amplitude of the high-frequency pulse is set to 0.05 MPa. To avoid fatigue damage to the membrane element caused by the high-frequency pulse, the amplitude of the high-frequency pulse should be controlled within 5% of the membrane element's burst pressure; in this embodiment, 0.05 MPa is used. The frequency should avoid the membrane module's natural frequency, preferably between 10 Hz and 100 Hz. The amplitude of the low-frequency step should not exceed 80% of the membrane element's maximum withstand pressure, and the step rate should be controlled within 0.1-0.5 MPa / s; in this embodiment, the step amplitude is 0.5 MPa, and the step rate is 0.25 MPa / s. This signal is used to excite microscopic mechanical vibrations of the membrane element and its internal support structure through high-frequency fluid pressure impact. By analyzing the membrane system's following characteristics to this high-frequency disturbance, the presence of physical structural fatigue or damage in the membrane element can be detected.

[0063] The frequency of the high-frequency pulse is dynamically adjusted based on the service life of the membrane system, and the calculation formula is as follows:

[0064] ;

[0065] in, The frequency of the currently applied high-frequency pulse, in Hz; The reference frequency for the initial state of the membrane system is 10Hz in this embodiment; This is a frequency correction factor, which in this embodiment represents a frequency increase of 0.05 Hz for each day of service. This represents the cumulative service days of the membrane system, expressed in days. Considering that the material stiffness and mesh support characteristics of the membrane element will drift as the service time increases, the excitation frequency is adjusted through linear compensation to ensure that the excitation signal is always in the sensitive frequency band that most easily induces the mechanical response of the membrane element.

[0066] While outputting the composite excitation signal, the control system monitors the pressure execution deviation in real time through a pressure sensor array. If the root mean square error between the actual generated pressure waveform and the target composite signal exceeds 3% of the target amplitude, the system will activate automatic gain compensation and adjust the high-pressure pump frequency through closed-loop feedback to ensure that the quality of the excitation signal meets the accuracy requirements of subsequent adaptive decomposition. In this embodiment, the set upper limit threshold for safe pressure is 6.5 MPa. If the instantaneous value of the total pressure at the inlet exceeds 6.5 MPa, the system will forcibly limit the amplitude to protect the membrane element from instantaneous overpressure damage.

[0067] By at least partially overlapping low-frequency steps with high-frequency pulses in time, and actively superimposing periodic pressure fluctuations during the maintenance of the low-frequency steps, the membrane system simultaneously generates controlled boundary layer evolution and micromechanical vibration response. Furthermore, by linearly adjusting the frequency of the high-frequency pulses according to the service time of the membrane system, the drift in stiffness and support characteristics caused by membrane element aging is compensated, ensuring that the excitation frequency is always in the sensitive frequency band, thereby improving the detection sensitivity of the high-frequency response to early structural fatigue or damage. At the same time, by real-time monitoring of pressure waveform deviation and closed-loop compensation, combined with safety pressure threshold limiting, the accuracy of the composite excitation signal and the operational safety of the membrane element are guaranteed.

[0068] S300: Collects the differential pressure response signal of the membrane system under the action of the composite excitation signal, and adaptively decomposes the differential pressure response signal into low-frequency trend components and transient components;

[0069] Furthermore, such as Figure 2 As shown, the adaptive decomposition steps include:

[0070] By employing morphological component analysis, and constructing a total variation dictionary for low-frequency trend components and a sparse pulse dictionary for transient components, the optimal solutions for the two components are iteratively solved in the sparse domain.

[0071] Specifically, during the application of the composite excitation signal, pressure data is synchronously acquired by high-precision pressure transmitters located at the inlet and concentrate ends of the membrane system. The difference between the inlet and concentrate pressures is calculated to obtain the time-varying differential pressure response signal. To fully capture the dynamic characteristics caused by the high-frequency pulse, the sampling frequency of the differential pressure signal in this embodiment is set to 100Hz to ensure that the sampling theorem is satisfied and the characteristic frequency band of the high-frequency response is covered.

[0072] After acquiring the raw differential pressure response signal, morphological component analysis was used for signal separation. Utilizing the sparsity differences of different characteristic signals in a specific transform domain, component separation was achieved through iterative calculation. For the low-frequency trend component, a total variation dictionary was constructed. The total variation dictionary leverages the first-order differential sparsity of the signal to effectively capture the quasi-static changes caused by low-frequency step excitation and the slow relaxation process generated by the evolution of the membrane surface boundary layer in the differential pressure signal. For the transient component, a sparse pulse dictionary was constructed. This dictionary, based on wavelet basis functions or discrete cosine basis functions, aims to characterize pressure oscillations caused by high-frequency pulse excitation and transient disturbances in the internal structure of the membrane element due to mechanical response.

[0073] Signal decomposition is achieved by minimizing the total variation norm of the low-frequency trend component and the L1 norm of the transient component, while constraining the reconstruction error between the sum of the two decomposed components and the original observed signal. This is solved using a synchronous iterative shrinking threshold algorithm. During the solution process, the control system employs the synchronous iterative shrinking threshold algorithm for alternating iterations. In this embodiment, the maximum number of iterations is set to 200, and the convergence threshold is [value missing]. The iteration stops when the rate of change of residual energy between two adjacent iterations is lower than the threshold.

[0074] In this embodiment, the regularization parameters are set to 0.15 and 0.1, the maximum number of iterations is 200, and the convergence threshold is... By minimizing the variational norm of the low-frequency component and the vector norm of the transient component, while ensuring that the sum of the two decomposed components is as close as possible to the original observation signal, each component exhibits the most significant sparsity characteristics under its corresponding feature dictionary.

[0075] The low-frequency trend component obtained from the final decomposition is represented by a step response curve reflecting changes in boundary layer drag, which eliminates the interference of high-frequency fluctuations. The transient component obtained from the decomposition is represented by a periodic pulse sequence with zero as the reference, which filters out the influence of background pressure shifts. This adaptive decomposition process provides a clean data source for subsequent extraction of feature parameters reflecting the degree of membrane fouling and the physical damage state.

[0076] By constructing a total variation dictionary to capture low-frequency trend components and a sparse pulse dictionary to capture transient components, and separating the two responses in the iterative solution based on sparsity differences, the spectral aliasing and amplitude interference between the boundary layer relaxation signal caused by low-frequency step and the mechanical vibration signal excited by high-frequency pulse can be effectively eliminated, thereby improving the accuracy and robustness of decoupling diagnosis.

[0077] S400: After normalizing the low-frequency trend component and transient component based on the environmental parameters, extract the first characteristic parameter reflecting the boundary layer resistance of the membrane surface from the low-frequency trend component, and extract the second characteristic parameter reflecting the physical structural damage of the membrane element from the transient component.

[0078] Furthermore, such as Figure 3 and Figure 4 As shown, the normalization process includes the following steps:

[0079] Temperature-based viscosity compensation, salinity-based osmotic pressure compensation, and base flow rate normalization are applied to the low-frequency trend components.

[0080] Perform pressure reference calibration on transient components based on the current inlet water pressure;

[0081] The transient components are demodulated and calibrated based on the instantaneous amplitude of the low-frequency step to eliminate the modulation effect of the low-frequency background on the high-frequency response.

[0082] Specifically, temperature-viscosity compensation is performed for the low-frequency trend component. In this embodiment, the correction formula is:

[0083]

[0084] in, The original low-frequency components were decomposed; 2640 is the viscosity-temperature empirical constant. The inlet water temperature is given in degrees Celsius. In this embodiment, the Arrhenius empirical equation is used to compensate for fluctuations in the dynamic viscosity of the liquid caused by temperature changes, and the pressure difference is uniformly reduced to... Baseline environment.

[0085] Then, traffic normalization is performed:

[0086] ;

[0087] in, This represents the current baseline traffic flow. For reference standard flow, this embodiment takes The index 1.7 is a characteristic index of pressure loss in typical spiral-wound membrane elements under turbulent conditions, which can be calibrated experimentally based on the Reynolds number. It eliminates background pressure drift caused by fluctuations in the influent pump speed or changes in system resistance. As an alternative or supplementary solution to temperature, salinity, and flow rate compensation, multiple linear regression or neural network models can also be used, with environmental parameters as input and pressure drift response signal as output. Through offline training or online adaptive learning, background drift can be subtracted, further improving the accuracy and robustness of the normalization process.

[0088] For transient components, demodulation calibration is performed. Since the high-frequency pulse is superimposed on the low-frequency step, its response amplitude is affected by the instantaneous pressure change of the step. The transient component is demodulated by division by calculating the instantaneous amplitude gain of the low-frequency step. The specific implementation of demodulation calibration is as follows:

[0089] Let the baseline average pressure before the low-frequency step occurs be The average pressure within 1 second prior to the step trigger is taken. The instantaneous average pressure during the low-frequency step period is... The formula for the pressure gain coefficient is defined as follows:

[0090] ;

[0091] Then for the original transient components Perform division demodulation to obtain the calibrated transient component:

[0092] ;

[0093] When a low-frequency step causes the system's average pressure to increase by 10%, that is... Then the amplitude of the corresponding transient component shrinks to its original value. This ensures that the high-frequency response reflects only the mechanical impedance of the membrane element, rather than changes in the background static pressure. In this embodiment, the system records the inlet pressure in real time at a sampling rate of 100Hz, calculates the average pressure every 10 ms, and dynamically updates the data. And perform real-time division calibration on each sampling point of the transient component.

[0094] Furthermore, the first characteristic parameter includes at least one of relaxation time, time constant, and equivalent derived index of relaxation characteristics; the second characteristic parameter includes at least one of phase difference offset, amplitude-frequency response change, and equivalent derived index of phase characteristics.

[0095] Specifically, the first characteristic parameter includes at least one of relaxation time, time constant, and equivalent derived index of relaxation characteristics. Among them, the equivalent derived index of relaxation characteristics includes, but is not limited to, the rise slope of the step response, the half-life (i.e., the time required for the response to reach 50% of the total change), or the integral area of ​​the response curve within a preset time period.

[0096] The second characteristic parameter includes at least one of phase difference offset, amplitude-frequency response change, and equivalent derived index of phase characteristics. The equivalent derived index of phase characteristics includes, but is not limited to, group delay, average phase difference within a specific frequency band, or the inflection frequency of the phase response curve.

[0097] Furthermore, the extraction of the first feature parameter and the second feature parameter is performed using at least one of the following methods:

[0098] Frequency domain filtering is used to separate the low-frequency and high-frequency responses of the differential pressure response signal;

[0099] Time-domain fitting is used to extract the relaxation time of the differential pressure response signal;

[0100] Transfer function analysis is used to extract the phase difference from the differential pressure response signal.

[0101] Specifically, the first characteristic parameter reflects the boundary layer resistance state of the membrane surface. In this embodiment, it is extracted using a time-domain fitting method. The normalized low-frequency trend component is fitted with a first-order dynamic model:

[0102] ;

[0103] In the formula, This represents the low-frequency differential pressure response that evolves over time. To achieve the pressure difference increment after reaching steady state; The first characteristic parameter is the time constant. The physical meaning of this parameter lies in characterizing the time required for the boundary layer solute to redistribute and reach a new equilibrium. When fouling or organic contamination occurs on the membrane surface, changes in the boundary layer thickness and viscoelasticity lead to… The value increases significantly. In this embodiment, the value in the healthy state is... The preset range is 15 to 45 seconds.

[0104] The second characteristic parameter reflects the physical structural damage of the membrane element. In this embodiment, it is extracted using transfer function analysis. The phase response of the system is calculated by performing a Fast Fourier Transform on the input high-frequency pulse signal and the normalized transient component of the output:

[0105] ;

[0106] in, The second characteristic parameter is the phase difference offset. and These represent the frequency domain representations of the output transient signal and the input pulse signal at their characteristic frequencies, respectively. The physical meaning of this parameter lies in reflecting the time-delay characteristics of fluid pressure wave propagation within the membrane bag. If the membrane element experiences mesh displacement, membrane bag damage, or mechanical fatigue, its structural stiffness matrix will change, causing a phase shift in the pressure wave transmission. In this embodiment, the set healthy phase shift deviation threshold is... .

[0107] By applying temperature and viscosity compensation, osmotic pressure compensation, and flow rate normalization to the low-frequency trend components, and performing pressure reference calibration and demodulation based on the division of the low-frequency step instantaneous amplitude for the transient components, the influence of environmental parameter fluctuations and low-frequency background pressure modulation on the high-frequency response can be effectively isolated. This ensures that the normalized low-frequency trend components are determined only by the boundary layer resistance, and the normalized transient components are determined only by the mechanical impedance of the membrane element. Based on this, the relaxation time or equivalent derived index is extracted by fitting a first-order dynamic model in the time domain, and the phase difference or equivalent derived index is extracted by using transfer function analysis. This can improve the signal-to-noise ratio and repeatability of boundary layer resistance and structural damage characteristic parameters.

[0108] S500: Construct a health vector using the first and second feature parameters, compare the health vector with the health reference space, and combine it with the prediction model to make a comprehensive judgment to determine the boundary layer resistance state and physical damage degree of the membrane system.

[0109] Furthermore, such as Figure 5 As shown, the steps for comprehensive judgment based on the prediction model include:

[0110] The first feature parameter and the second feature parameter are combined into a multi-dimensional vector, which serves as the health vector of the current state of the membrane system.

[0111] The distance offset between the health vector and the static health baseline space is calculated, and multi-criteria fusion is performed in combination with the dynamic residuals output by the prediction model to determine the state of the membrane system.

[0112] The health baseline space is established based on historical characteristic parameters of the membrane system under its initial health state, and is automatically reset and updated after chemical cleaning is performed and the system is running stably.

[0113] Specifically, a quantitative assessment of the membrane system's health status is achieved by combining multidimensional vector space analysis with dynamic residual evaluation. The extracted first characteristic parameter reflecting boundary layer drag (time constant) is then used. ) and the second characteristic parameter reflecting physical structural damage (phase difference offset) Combine these elements to construct a multidimensional health vector. .

[0114] The health baseline space is a set of features established by recording historical data of the membrane system in its initial health state or after chemical cleaning. In this embodiment, the baseline space consists of health vectors over 50 consecutive acquisition cycles, and their mean vector is calculated. Covariance Matrix After the system completes chemical cleaning and returns to stable operation for one hour, the control system will automatically trigger a reference space reset procedure. Stable operation refers to the normal operating state where the influent pressure, permeate flow rate, and temperature of the membrane system do not fluctuate within their respective preset thresholds over one hour. This can be determined by monitoring the real-time change rate of key process parameters. The system will be updated using the first 10 sets of test data after cleaning. and This is to eliminate reference drift caused by normal decay of membrane elements.

[0115] In this embodiment, the health baseline space is a multidimensional statistical distribution composed of feature parameters extracted from the membrane system under healthy conditions. Mahalanobis distance is preferably used to measure the deviation of the current health vector from the center of the baseline space. The baseline space is established when the system is initially put into operation, and is automatically reset and updated after each chemical cleaning and confirmation of system recovery. A prediction model is used to generate theoretical health feature values ​​that match the current operating conditions. Preferably, a digital twin model is used to simulate the expected differential pressure response of the healthy membrane module using real-time collected environmental parameters and operating time. Alternatively, a long short-term memory network can be used to train historical health vector sequences to predict the expected feature parameters under the current state. The dynamic residual output by the model is fused with Mahalanobis distance using multiple criteria for determination.

[0116] The system calculates the current health vector. With Health Benchmark Space Center Mahalanobis distance Its formula is This formula measures the statistical distance of the current membrane system state from the healthy center. Mahalanobis distance considers the correlation between various characteristic parameters and can more accurately identify anomalous deviations. In this embodiment, a first-level judgment threshold is set. Second-level judgment threshold .

[0117] The system synchronously runs a state prediction model based on exponential smoothing. In this embodiment, the smoothing coefficient is... Set the value to 0.2 and output the predicted value of the feature parameters at the next time step. Calculate the dynamic residuals between the actual observed values ​​and the predicted values. This formula evaluates the rate of change of the state of a membrane system. If... A sudden increase indicates that the membrane system has been subjected to a momentary shock or sudden physical damage, rather than a normal, slow accumulation of contaminants.

[0118] The system according to and The combined results determine the state of the membrane system. If Furthermore, the main deviation term originates from the first characteristic parameter. This indicates increased boundary layer resistance, suggesting solute polarization or mild scaling. If... And dynamic residuals Exceeding the preset threshold of 0.8, and simultaneously the second feature parameter Significant offset occurs, such as offset exceeding If the damage is not detected, it is considered physical structural damage, indicating that there may be a rupture of the membrane bag or deformation of the spacer.

[0119] By combining the first and second feature parameters into a health vector, and calculating its Mahalanobis distance offset from the health baseline space and the dynamic residual based on the exponential smoothing model, the statistical correlation between features and the rate of state change can be comprehensively utilized for multi-criteria fusion judgment. The Mahalanobis distance eliminates the interference of feature dimensions and correlation on anomaly identification, while the dynamic residual sensitively captures sudden impacts or rapid damage. The two complement each other and can effectively distinguish between the slow accumulation of boundary layer resistance and sudden damage to physical structures, thereby improving the accuracy and robustness of diagnosis.

[0120] Furthermore, it also includes:

[0121] Based on the determined boundary layer resistance state and physical damage level, graded early warning information is output;

[0122] The first level warning indicates adjusting operating parameters to slow down boundary layer accumulation; the second level warning indicates performing chemical cleaning; and the third level warning indicates replacing damaged membrane elements.

[0123] Specifically, by calculating the pollution deterioration index and structural damage index The formula for quantifying the degree of deviation from the state is as follows:

[0124] ;

[0125] ;

[0126] in, This is the first feature parameter extracted so far; The mean of the time constants in the reference space; This reflects the multiplication relationship between the membrane surface boundary layer thickness and solute diffusion resistance relative to the initial healthy state; This refers to the second feature parameter extracted so far. The average phase difference in the reference space; The phase shift reflecting the internal mechanical structure of the membrane element in response to pressure waves directly corresponds to changes in structural integrity.

[0127] The specific criteria and response schemes for tiered early warning information are as follows:

[0128] Level 1 warning, operating parameters optimized when... and At this point, the system triggers a level one warning. Significant solute accumulation has occurred at the membrane surface boundary layer, exacerbating concentration polarization, but a stable filter cake layer or scaling has not yet formed. The response at this time is for the control system to automatically increase the variable frequency output of the influent high-pressure pump, increasing the lateral scouring velocity within the membrane bag. In this embodiment, it is recommended to increase the base flow rate from... Temporarily upgraded to This can be achieved by increasing shear force to slow down the further accumulation of the boundary layer.

[0129] Level 2 alert, chemical cleaning to be performed when the following conditions are met. If the normalized steady-state pressure difference increment exceeds 15% of the initial baseline value, the system triggers a second-level warning. A thick fouling layer has formed on the membrane surface, and simply adjusting hydraulic parameters is no longer sufficient to effectively suppress the increase in resistance. The system outputs a chemical cleaning command and suggests a cleaning cycle based on the evolution rate of the first characteristic parameter. In this embodiment, if... If the growth rate exceeds 5% within 24 hours, an early warning message will be issued to prioritize acid washing to remove possible inorganic scale.

[0130] Level 3 warning: Replace the damaged membrane element when the following conditions are met. At any time, regardless Regardless of the level, the system immediately triggers a Level 3 warning. Irreversible mechanical damage has occurred to the membrane element, such as membrane bag rupture, severe misalignment of the separator, or failure of the central tube connection. This damage alters the transmission path of the pressure wave. The system issues the highest priority alarm, requiring operators to shut down the corresponding membrane pressure vessel for offline disassembly and inspection, and to replace the damaged membrane element to prevent concentrate from seeping into the permeate side through the damaged area.

[0131] When multiple warning conditions are met simultaneously, the highest-level warning is issued in the order of priority: Level 3, Level 2, and Level 1. That is, if the Level 3 condition is met, the operation of replacing the damaged membrane element is performed first, regardless of other conditions; otherwise, if the Level 2 condition is met, chemical cleaning is performed; otherwise, the Level 1 warning is issued.

[0132] The warning information is synchronized to the HMI interface in the central control room via the industrial bus. For each type of warning, the system records the environmental parameters and health vector coordinates at the time of triggering. After performing the corresponding maintenance actions, such as adjusting parameters, cleaning, or replacing the membrane, the system will re-execute the test process from steps S200 to S400. If the test results are negative... and Returning to within the baseline threshold, i.e. and If the warning is lifted, the current health status record will be updated automatically.

[0133] By calculating the pollution degradation index and structural damage index, the first and second characteristic parameters are quantified into dimensionless ratios and absolute deviations, respectively. Based on their respective threshold ranges, three levels of early warning are established, enabling targeted response schemes for boundary layer resistance accumulation and physical structural damage of the membrane system within a unified framework. At the same time, a closed-loop logic for post-maintenance retesting and recovery thresholds is set to ensure automatic updates of early warning cancellation and health status records. This achieves a complete closed loop from condition diagnosis to hierarchical operation and maintenance decision-making, improving the accuracy and automation level of membrane system maintenance.

[0134] Furthermore, it also includes:

[0135] During the application of the composite excitation signal, the instantaneous pressure peak at the inlet is monitored in real time. If the instantaneous pressure peak exceeds the safe pressure threshold of the membrane element, the excitation signal is immediately terminated and the pressure is released.

[0136] Specifically, a high-frequency pressure transmitter deployed at the water inlet end is used to capture the instantaneous pressure value at the water inlet end in real time with a scan period of 1ms. Since the composite excitation signal is composed of a low-frequency step and a high-frequency pulse, its pressure peak value exhibits dynamic changes. When the instantaneous pressure exceeds the safety threshold of 6.5MPa, a protection action is immediately triggered.

[0137] In this embodiment, the rated maximum operating pressure of the membrane element is 6.9 MPa. Considering the superposition effect of pressure wave reflection within the membrane housing, the system's safe pressure threshold is set at 6.5 MPa, and the emergency relief threshold is 6.7 MPa. Once a safety determination is triggered, the control system will execute the following deterministic protection actions:

[0138] Upon excitation cutoff, the frequency converter immediately stops the composite excitation algorithm, forcing the control frequency of the high-pressure pump back to the base operating frequency or the set safe speed. In this embodiment, the output pressure of the high-pressure pump drops below 5.0 MPa within 0.5 seconds.

[0139] The signal is cleared, the high-speed regulating valve stops reciprocating and switches to the fully open state, eliminating the periodic disturbances generated by high-frequency pulses.

[0140] Active relief: If the instantaneous pressure still exceeds the emergency relief threshold of 6.7MPa after the excitation is cut off and the signal is cleared, the system automatically opens the pneumatic quick relief valve located on the outlet side of the high-pressure pump, and directly introduces the excess pressure energy into the return pipeline through the bypass circuit to achieve instantaneous unloading of pressure energy.

[0141] The high-frequency pressure transmitter monitors the instantaneous total pressure at the inlet in real time with a millisecond-level scanning cycle. When the pressure exceeds the safety threshold, the excitation is immediately cut off and the pulse signal is cleared. At the same time, a higher emergency relief threshold is set, and the bypass relief circuit is automatically opened when the pressure is continuously abnormal. This achieves graded safety protection for the dynamic pressure peak generated by the composite excitation signal, effectively preventing the membrane element from suffering irreversible mechanical damage due to instantaneous overpressure, and improving the reliability of the testing process and the safety of the project.

[0142] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for testing differential pressure disturbance in seawater desalination membranes, characterized in that, include: Obtain the environmental parameters of the membrane system under its current operating conditions; The pressure at the inlet of the membrane system generates a composite excitation signal, which includes a low-frequency step and a high-frequency pulse, and the two overlap at least partially in time. The differential pressure response signal of the membrane system under the action of the composite excitation signal is collected, and the differential pressure response signal is adaptively decomposed into low-frequency trend component and transient component; After normalizing the low-frequency trend component and transient component based on the environmental parameters, a first characteristic parameter reflecting the boundary layer resistance of the membrane surface is extracted from the low-frequency trend component, and a second characteristic parameter reflecting the physical structural damage of the membrane element is extracted from the transient component. A health vector is constructed using the first and second feature parameters. The health vector is then compared with the health reference space and combined with a prediction model to determine the boundary layer resistance state and physical damage level of the membrane system.

2. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The environmental parameters include influent temperature, salinity, and basic flow rate.

3. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The low-frequency step and high-frequency pulse specifically include: The low-frequency step refers to the inlet pressure switching from a first pressure value to a second pressure value and maintaining it at the second pressure value for a preset duration, which is used to induce the controlled evolution of the membrane surface boundary layer. The high-frequency pulse refers to a periodic pressure fluctuation superimposed during the duration of a low-frequency step, used to excite the mechanical response of the membrane element. Furthermore, the frequency of the high-frequency pulse is dynamically adjusted based on the service life of the membrane system.

4. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The adaptive decomposition steps include: By employing morphological component analysis, and constructing a total variation dictionary for low-frequency trend components and a sparse pulse dictionary for transient components, the optimal solutions for the two components are iteratively solved in the sparse domain.

5. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The normalization process includes the following steps: Temperature-based viscosity compensation, salinity-based osmotic pressure compensation, and base flow rate normalization are applied to the low-frequency trend components. Perform pressure reference calibration on transient components based on the current inlet water pressure; The transient components are demodulated and calibrated based on the instantaneous amplitude of the low-frequency step to eliminate the modulation effect of the low-frequency background on the high-frequency response.

6. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The first characteristic parameter includes at least one of relaxation time, time constant, and equivalent derived index of relaxation characteristics; the second characteristic parameter includes at least one of phase difference offset, amplitude-frequency response change, and equivalent derived index of phase characteristics.

7. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 6, characterized in that, The first feature parameter and the second feature parameter are extracted using at least one of the following methods: Frequency domain filtering is used to separate the low-frequency and high-frequency responses of the differential pressure response signal; Time-domain fitting is used to extract the relaxation time of the differential pressure response signal; Transfer function analysis is used to extract the phase difference from the differential pressure response signal.

8. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, The steps for comprehensive determination based on the prediction model include: The first feature parameter and the second feature parameter are combined into a multi-dimensional vector, which serves as the health vector of the current state of the membrane system. The distance offset between the health vector and the static health baseline space is calculated, and multi-criteria fusion is performed in combination with the dynamic residuals output by the prediction model to determine the state of the membrane system. The health baseline space is established based on historical characteristic parameters of the membrane system under its initial health state, and is automatically reset and updated after chemical cleaning is performed and the system is running stably.

9. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, Also includes: Based on the determined boundary layer resistance state and physical damage level, graded early warning information is output; The first level warning indicates adjusting operating parameters to slow down boundary layer accumulation; the second level warning indicates performing chemical cleaning; and the third level warning indicates replacing damaged membrane elements.

10. The method for testing differential pressure disturbance of a seawater desalination membrane according to claim 1, characterized in that, Also includes: During the application of the composite excitation signal, the instantaneous pressure peak at the inlet is monitored in real time. If the instantaneous pressure peak exceeds the safe pressure threshold of the membrane element, the excitation signal is immediately terminated and the pressure is released.

Images