A fault prediction method and system for a bidirectional flow membrane filtration device

Abstract

The application provides a fault prediction method and system for a bidirectional flow membrane filtration device, which comprises the following steps: collecting multiple source sensing operation signals and obtaining corrected sensing characteristics; when multiple membrane elements synchronously appear abnormal, determining factors and generating prior constraint parameters according to the determination results; adjusting the threshold value according to the prior constraint parameters; identifying the pollution type based on the hydraulic impedance parameters and the hydraulic impedance threshold value; obtaining the integrity damage result of each membrane element according to the decay time constant and the damage threshold value; and fusing the pollution type result and the integrity damage result to output the fault prediction result. According to the application, the abnormal causes are actively verified through isolation operation, and the prior constraint is generated accordingly. The determination standards of subsequent diagnostic methods are uniformly regulated by the prior constraint. Through the active correction of the hydraulic coupling distortion by isolation verification and the synergistic effect of eliminating individual differences in the structural integrity diagnosis, the accuracy and reliability of fault prediction are significantly improved.

Description

Technical Field

[0001] This application relates to the field of wastewater treatment, and more specifically, to a method and system for predicting the failure of a bidirectional flow membrane filtration device. Background Technology

[0002] Membrane filtration systems rely on membrane elements to separate liquid from feed and produce purified water. Various types of fouling, such as filter cake, adsorbents, biofilm, and inorganic scaling, continuously grow on the membrane surface and inside the membrane pores. This fouling can significantly reduce water flux, increase transmembrane pressure difference, and shorten membrane lifespan.

[0003] Existing fault monitoring methods for membrane filtration systems primarily determine whether membrane elements are contaminated or damaged by comparing their operating parameters, such as flow rate and pressure, with preset thresholds or historical normal conditions. However, these methods typically treat each membrane element as an independent entity. When multiple membrane elements simultaneously exhibit abnormalities, existing methods directly conclude that each element has been contaminated or damaged individually. But in parallel membrane systems, the inlet and outlet pipes of each membrane element are interconnected. A change in the state of one membrane element can cause a redistribution of system flow among the membrane elements through hydraulic coupling in the pipes, leading to synchronous changes in the flow rate, pressure, and other operating parameters of the remaining membrane elements. This collective anomaly caused by hydraulic coupling presents signals highly similar to the actual degradation of multiple membrane elements simultaneously, but current technology cannot distinguish between the two. Summary of the Invention

[0004] Based on the problems existing in the prior art, this application provides a fault prediction method and system for a bidirectional flow membrane filtration device.

[0005] A fault prediction method for a bidirectional flow membrane filtration device, applied to a bidirectional flow membrane filtration device capable of switching between forward and reverse operating states, the method comprising:

[0006] Multi-source sensor operation signals corresponding to each membrane element are collected, and the real-time sensor characteristics are corrected according to the characteristic distortion caused by hydraulic coupling to obtain corrected sensor characteristics. The operating status of each membrane element is continuously monitored based on the corrected sensor characteristics. When multiple membrane elements simultaneously exhibit abnormalities, the abnormality source membrane element is isolated, and based on the state changes of the remaining membrane elements before and after isolation, it is determined whether the abnormality is a false abnormality induced by hydraulic coupling or a real abnormality caused by membrane element deterioration, and prior constraint parameters are generated based on the determination results. The hydraulic impedance threshold is adjusted according to the prior constraint parameters. The pressure step generated by the flow path switching is used as the excitation. The hydraulic impedance parameters of each membrane element are identified in combination with the correction sensing features. The type of fouling is identified based on the hydraulic impedance parameters and the hydraulic impedance threshold to obtain the fouling type result. The damage threshold for structural integrity damage is adjusted according to the prior constraint parameters. Pressure decay signals are collected using the device shutdown or reversing gap, and the decay time constant of each membrane element is extracted from them. The integrity damage results of each membrane element are obtained based on the decay time constant and the damage threshold. By combining the pollution type results and the integrity damage results, a fault prediction result is output.

[0007] A fault prediction system for a bidirectional flow membrane filtration device, applicable to a bidirectional flow membrane filtration device capable of switching between forward and reverse operating states, the system comprising: The sensing correction unit is used to collect multi-source sensing operation signals corresponding to each membrane element, and correct the real-time sensing features according to the characteristic distortion caused by hydraulic coupling to obtain corrected sensing features. Based on the corrected sensing features, the operating status of each membrane element is continuously monitored. The prior constraint unit is used to isolate the anomaly source membrane element when multiple membrane elements simultaneously exhibit anomalies, and based on the state changes of the remaining membrane elements before and after isolation, determine whether the anomaly is a false anomaly induced by hydraulic coupling or a real anomaly caused by membrane element deterioration, and generate prior constraint parameters based on the determination result. The first determination unit is used to adjust the hydraulic impedance threshold according to the prior constraint parameters, use the pressure step generated by the flow path switching as the excitation, identify the hydraulic impedance parameters of each membrane element in combination with the correction sensing features, identify the pollution type based on the hydraulic impedance parameters and the hydraulic impedance threshold, and obtain the pollution type result. The second determination unit is used to adjust the damage threshold of structural integrity damage according to the prior constraint parameters, collect pressure decay signals using device shutdown or reversing gaps and extract the decay time constant of each membrane element from them, and obtain the integrity damage result of each membrane element according to the decay time constant and the damage threshold. The fusion prediction unit is used to fuse the pollution type results and the integrity damage results to output the fault prediction results.

[0008] Beneficial Effects: This application proposes a fault prediction method and system for a bidirectional flow membrane filtration device. By actively verifying the cause of anomalies through isolation operation and generating prior constraints accordingly, the judgment criteria of subsequent diagnostic methods are uniformly controlled by these prior constraints. This forms a complete technical closed loop of isolation verification, prior constraints, global control, and precise diagnosis, avoiding the dilemma of multiple false alarms caused by hydraulic coupling interference, which leads to chaotic operation and maintenance decisions. At the same time, the synergistic effect of active correction of hydraulic coupling distortion by isolation verification and elimination of individual differences by horizontal comparison in structural integrity diagnosis significantly improves the accuracy and reliability of fault prediction. This enables the system to stably output reliable fault prediction results even under hydraulic coupling interference scenarios, providing a systematic technical solution for the refined operation and maintenance of parallel membrane systems.

[0009] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the fault prediction method of this application; Figure 2 This is a schematic diagram illustrating the principle of the fault prediction method of this application; Figure 3 This is a schematic diagram of the cross-validation process of this application; Figure 4 This is a schematic diagram of the process for obtaining the first result of this application; Figure 5 This is a schematic diagram of the fault prediction system module of this application.

[0012] Reference numerals in the attached figures: 1-Correction sensing unit; 2-Priority constraint unit; 3-First decision unit; 4-Second decision unit; 5-Fusion prediction unit. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0014] This application proposes a fault prediction method for a bidirectional flow membrane filtration device, and a flowchart of the fault prediction method is attached. Figure 1 As shown in the attached diagram, the principle is as follows: Figure 2 As shown, the specific solution is as follows: A fault prediction method for a bidirectional flow membrane filtration device, applied to a bidirectional flow membrane filtration device capable of switching between forward and reverse operating states, the method comprising: 101. Collect multi-source sensor operation signals corresponding to each membrane element, and correct the real-time sensor characteristics based on the characteristic distortion caused by hydraulic coupling to obtain corrected sensor characteristics. Continuously monitor the operating status of each membrane element based on the corrected sensor characteristics. 102. When multiple membrane elements simultaneously exhibit abnormalities, isolate the membrane element that caused the abnormality and determine whether the abnormality is a false abnormality induced by hydraulic coupling or a real abnormality caused by membrane element deterioration based on the state changes of the remaining membrane elements before and after isolation. Generate prior constraint parameters based on the determination results. 103. Adjust the hydraulic impedance threshold according to the prior constraint parameters, use the pressure step generated by the flow path switching as the excitation, combine the correction sensing characteristics to identify the hydraulic impedance parameters of each membrane element, identify the fouling type based on the hydraulic impedance parameters and hydraulic impedance threshold, and obtain the fouling type result. 104. Adjust the damage threshold of structural integrity damage according to the prior constraint parameters, collect pressure decay signals by using the device shutdown or reversing gap and extract the decay time constant of each membrane element from it, and obtain the integrity damage results of each membrane element according to the decay time constant and the damage threshold. 105. Combine the pollution type results and integrity damage results to output the fault prediction results.

[0015] This embodiment provides a fault prediction method for a bidirectional flow membrane filtration device, applicable to a bidirectional flow membrane filtration device capable of switching between forward and reverse operating states. The core flow path structure of this device includes a feed inlet pipe, a concentrate outlet pipe, a permeate outlet pipe, a reversing valve assembly, and multiple parallel spiral wound membrane elements. The membrane bag of the spiral wound membrane element consists of two membrane layers and a permeate separator in between. Fluid flows axially in the feed grid, while the permeate flows along a spiral path to the central collection pipe. This structure results in a longer fluid flow path compared to filament membrane elements, and the capacitive and inductive reactance components in the hydraulic impedance parameters are relatively more significant. Therefore, the three-parameter combined fouling type identification method is more applicable to spiral wound membrane elements. Each membrane element is equipped with an independent inlet and outlet valve. By switching the opening and closing states of these valves, any membrane element can be independently isolated from the parallel piping system. A reversing valve assembly is used to switch between the forward and reverse operating states of the device. In the forward operating state, the feed solution flows into the membrane element from the inlet, undergoes membrane separation, and the concentrate is discharged from the outlet, while the permeate is collected in the central collection pipe. In the reverse operating state, the fluid flows backward from the outlet of the membrane element, flows along the membrane surface, and then flows out from the original inlet. The hardware structure of this device itself is the foundation for the implementation of the method of this invention.

[0016] In step 101, the multi-source sensor operation signals corresponding to each membrane element are first collected. The multi-source sensor operation signals refer to the raw physical quantity signals collected by various sensors arranged on each membrane element during normal operation of the device. These sensors include at least a fiber optic strain sensor array arranged along the axial direction of the membrane shell and an acoustic emission sensor array arranged on the outer wall of the membrane shell. They may also include pressure sensors, flow sensors, and temperature sensors at the inlet and outlet of each membrane element.

[0017] A fiber optic grating strain sensor array is used to sense the local strain distribution on the membrane shell surface. When the stress or strain of the environment around the grating changes, the center wavelength of the reflected light shifts. The strain value at that measuring point is obtained by measuring the wavelength shift. The membrane shell of the spiral wound membrane element is a cylindrical pressure vessel. The strain sensor senses the circumferential and axial strain of the membrane shell, indirectly reflecting the expansion or contraction state of the membrane bag inside the membrane element. An acoustic emission sensor array is used to sense the transient elastic wave signals released by the membrane element material under stress. In spiral wound membrane elements, when the membrane sheet delaminates from the support layer, the membrane bag bonding interface cracks, releasing elastic waves and generating sudden acoustic emission events. When the membrane bag bonding seam opens and leaks, the fluid passing through the crack generates turbulent noise, which manifests as a continuous acoustic emission signal. Pressure sensors are used to collect the fluid pressure at the inlet and outlet of each membrane element, and flow sensors are used to collect the permeate flow rate of each membrane element. All sensors maintain synchronization of the acquisition timing through a hardware clock synchronization module to ensure that the signals are aligned in time during subsequent multi-source data fusion.

[0018] In some embodiments, the fiber optic strain sensor array is arranged equidistantly along the axial direction of the membrane housing, with eight to twelve measurement points corresponding to each membrane element, and a sampling frequency of not less than 100 Hz; an acoustic emission sensor is arranged at the inlet, middle and outlet of each membrane element housing, with a sampling frequency of not less than 1 MHz; a pressure sensor is arranged at both the inlet and outlet ends of each membrane element to collect the transmembrane pressure difference; and a flow sensor is arranged on the permeate side of each membrane element to collect the permeate flow rate of each membrane element.

[0019] In a multi-membrane element parallel system, the inlet and outlet pipes of each membrane element are interconnected through a common main pipe, and there is an inherent hydraulic coupling relationship between the membrane elements. When multiple membrane elements are running in parallel, the total system flow is distributed inversely proportional to the resistance of each branch, and the sum of the flow rates of each branch equals the total flow rate. When the state of a membrane element changes, for example, if the resistance of a membrane increases due to fouling, the decrease in the flow rate of that branch will release a portion of the total flow rate. This released flow rate will automatically be distributed to the other branches with lower resistance, causing passive changes in the flux and transmembrane pressure difference of the remaining membrane elements. This change in the sensing signal caused by hydraulic coupling is not a real change in the state of the other membrane elements themselves, but a false abnormal signal that will be superimposed on the real state signal of the membrane element itself, forming characteristic distortion. Therefore, in step 101, it is necessary to correct the real-time sensing characteristics based on the characteristic distortion caused by hydraulic coupling. The specific implementation of the correction involves performing periodic, rotating isolation verification. Each membrane element is isolated briefly using its own independent valve control assembly according to a preset time period. During this isolation, the fluid path of the isolated membrane element is completely cut off from the parallel piping system, unaffected by the operating status of other membrane elements or the hydraulic coupling effect of the parallel flow field. The sensor signals acquired at this time accurately reflect the individual state of the membrane element under the current influent water quality and operating conditions, and these signals are used as the reference sensing characteristics of the membrane element. After obtaining the reference sensing characteristics of each membrane element, the sensor signals acquired during real-time parallel operation are compared with these reference sensing characteristics to calculate the difference. This difference represents the characteristic distortion introduced by the hydraulic coupling of the parallel piping. Then, based on this distortion, the real-time sensor signals are reverse-corrected by subtracting the distortion component from the real-time signal to obtain the corrected sensing characteristics. This periodic rotation of isolation verification and real-time correction can continuously track the slow changes in the individual benchmarks of each membrane element as the influent water quality and operating conditions change, ensuring that the corrected sensing characteristics always accurately reflect the state of each membrane element and effectively eliminate false signal interference introduced by parallel hydraulic coupling.

[0020] In step 102, the operating status of each membrane element is continuously monitored based on the corrected sensing characteristics. Monitoring can be performed by comparing the corrected sensing characteristics of each membrane element with its individual baseline, or by comparing the corrected sensing characteristics of each membrane element across the membrane group. When multiple membrane elements are detected to be simultaneously abnormal, an isolation and discrimination process is initiated. Simultaneous abnormality of multiple membrane elements refers to at least two membrane elements exhibiting deviations in their corrected sensing characteristics that exceed the normal range within a similar time window.

[0021] Such synchronized anomalies can have two completely different physical causes. The first is that multiple membrane elements do indeed experience real fouling or damage simultaneously. This is a genuine anomaly caused by the degradation of the membrane elements themselves, and its physical cause is an irreversible change in the state of each membrane element, requiring individual treatment. The second is that the state change of one or a few membrane elements is transmitted to other membrane elements through the hydraulic coupling effect of parallel pipelines, causing passive changes in the flux and transmembrane pressure difference of those other membrane elements. These membrane elements themselves have not undergone real degradation; this is a pseudo-anomaly induced by hydraulic coupling. In current technology, both types of anomalies are indiscriminately attributed to membrane fouling, making it difficult for maintenance personnel to accurately determine whether to treat all membrane elements or only individual elements when faced with synchronized anomalies, often resulting in over-maintenance or under-maintenance.

[0022] To differentiate between these two causes, step 102 employs active isolation verification. Isolation operations typically require the sensing signals of the remaining membrane elements to complete a hydraulic transient response and enter a new steady state. This process usually takes between five and fifteen minutes, depending on the system piping volume and the membrane element specifications. Determining whether the anomaly has subsided should be done after the isolation operation is completed and the system has entered a new steady state, rather than at the moment of isolation, to avoid misinterpreting transient hydraulic shocks as an abnormal condition.

[0023] Upon detecting a cluster anomaly, one membrane element is first selected as the source of the anomaly. This selection can be based on the membrane element deviating most from the group average (as it is most likely the original fault source triggering flow redistribution), the membrane element with the largest deviation from its individual baseline, or the membrane element whose deviation occurred earliest. After selecting the source membrane element, it is temporarily isolated from the parallel piping system using its corresponding independent valve control assembly. Specifically, the inlet and outlet valves of this membrane element are closed to completely cut off its fluid connection to the parallel piping system. During isolation, the total system flow will be redistributed to the remaining parallel membrane elements, but these elements are no longer affected by the hydraulic coupling of the isolated element. The calibration sensor characteristics of the remaining membrane elements are continuously monitored during isolation. If the anomalies of the remaining membrane elements synchronously subside after the isolation operation, it indicates that the anomalies were caused by the isolated membrane element through hydraulic coupling, and are considered false anomalies. In this case, only the isolated source membrane element needs maintenance; the other membrane elements do not require treatment, and their anomalies will automatically disappear after hydraulic rebalancing. If the abnormal state of the remaining membrane elements persists after the isolation operation, it indicates that these membrane elements themselves do indeed have real degradation, belonging to the category of real degradation of multiple membrane elements, and each abnormal membrane element needs to be investigated and dealt with one by one. Based on the above judgment results, a priori constraint parameters are generated. These parameters characterize whether the physical cause of the current group anomaly is a pseudo-anomaly induced by hydraulic coupling or a real anomaly caused by membrane element degradation.

[0024] Furthermore, the selection of anomaly source membrane elements can also be based on the membrane element that deviates the most from the individual benchmark, i.e., the membrane element whose corrected sensing characteristics deviate the most from its own benchmark value; or it can be based on the membrane element that occurs earliest, i.e., the membrane element that first shows an abnormal signal. These three selection methods can be used individually, or the membrane element with the highest overall score can be selected after comprehensive evaluation.

[0025] In step 103, the hydraulic impedance threshold is first adjusted based on the prior constraint parameters generated in step 102. The hydraulic impedance threshold is a benchmark value used to determine whether the hydraulic impedance parameters of the membrane element are abnormal; its level determines the sensitivity of fouling type identification. When the prior constraint parameters indicate that the current anomaly is a false anomaly induced by hydraulic coupling, the hydraulic impedance threshold is increased to make the fouling type identification standard more stringent, avoiding misjudging false signals caused by hydraulic coupling as real fouling. When the prior constraint parameters indicate that the current anomaly is a real anomaly caused by membrane element degradation, the hydraulic impedance threshold is maintained or decreased to maintain or increase the sensitivity of the identification standard, ensuring that real fouling can be accurately detected.

[0026] Then, using the pressure step generated by the flow path switching as excitation, the hydraulic impedance parameters of each membrane element are identified by combining the calibration sensor characteristics. The source of the excitation signal is the pressure step naturally generated in the pipeline when the directional valve group performs the flow path switching operation. During normal operation of the device, the directional valve group will periodically perform flow path switching, and each switching will generate a pressure step in the pipeline lasting tens to hundreds of milliseconds. This pressure step is an inherent operating event of normal operation of the device, which does not require additional excitation or any additional excitation equipment, realizing online excitation with zero hardware cost. At the moment when the pressure step occurs, the pressure response waveforms at both ends of each membrane element are acquired at a high-frequency sampling rate. This waveform is the dynamic response of the membrane element to the pressure step excitation and contains all the information of the hydraulic characteristics of the membrane element.

[0027] System identification of the acquired pressure response waveforms allows for the extraction of equivalent hydraulic impedance parameters for each membrane element in the frequency domain. System identification is a method of establishing a system model using the mathematical relationship between input and output signals. The input signal is the pressure step generated by the switching of the directional valve, and the output signal is the pressure response waveform across each membrane element. Through mathematical processing, the membrane element can be equivalently represented as a second-order fluid system model composed of three parameters: equivalent resistance, equivalent capacitive reactance, and equivalent inductive reactance.

[0028] Equivalent resistance reflects the steady-state flow resistance of fluid passing through the membrane element and is directly related to the thickness of the filter cake layer on the membrane surface. Equivalent capacitive reactance reflects the dynamic response characteristics of the fluid storage and release capacity inside the membrane element and is directly related to the clogging state of the micropores in the membrane separation layer. Equivalent inductive reactance reflects the response characteristics of the membrane element to changes in fluid flow inertia and is directly related to the state of the adsorption layer on the membrane surface. After obtaining the three impedance parameters of each membrane element, these three parameters are compared with the hydraulic impedance threshold adjusted by prior constraints. In practical implementation, time-domain identification methods such as least squares recursive algorithms or subspace identification methods can be used, or frequency-domain identification methods such as transfer function fitting methods can be used. The membrane element is modeled as a second-order fluid system model, the input signal is the pressure step generated by the switching of the reversing valve, and the output signal is the pressure response waveform at both ends of each membrane element. The identification results are the values ​​of the three parameters: equivalent resistance, equivalent capacitive reactance, and equivalent inductive reactance.

[0029] In step 104, the damage threshold for structural integrity damage is first adjusted based on the prior constraint parameters generated in step 102. Structural integrity damage in spiral wound membrane elements refers to delamination between the membrane and the support layer, delamination and leakage at the membrane bag adhesive seam, membrane rupture or wrinkling, and extrusion deformation between the feed grid and the membrane. The damage threshold is the criterion value used to determine whether structural integrity damage exists in the membrane element. When the prior constraint parameters indicate that the current anomaly is a false anomaly induced by hydraulic coupling, the damage threshold is increased to make the judgment standard for integrity damage more stringent, avoiding misjudging transient pressure changes caused by hydraulic coupling as structural damage. When the prior constraint parameters indicate that the current anomaly is a real anomaly caused by membrane element deterioration, the damage threshold is maintained or decreased to ensure that real structural damage can be accurately detected.

[0030] Pressure decay signals are collected during unit shutdowns or reversing intervals, from which the decay time constant of each membrane element is extracted. During normal unit shutdown maintenance or the pressure decay period after a flow path switch by the reversing valve assembly, the inlet and outlet valves of each membrane element are closed to create a closed fluid space inside the membrane element. Then, the pressure decay signal across each membrane element over time is collected at a preset sampling frequency. This decay signal reflects the natural depressurization process of the fluid inside the membrane element under no external pressure. The pressure decay of a healthy membrane element follows a specific pattern, and its decay time constant—the time required for the pressure to drop from the initial value to approximately 36.8%—usually stabilizes within a certain range.

[0031] Curve fitting was performed on the collected pressure decay signals, and the pressure decay time constant of each membrane element was extracted from the fitting results. When the spiral wound membrane element experiences delamination between the membrane sheet and the support layer or separation of the membrane bag adhesive seam, the integrity of its internal fluid channels is compromised, fluid resistance is significantly reduced, the pressure decay rate is accelerated, and the decay time constant is significantly decreased. After obtaining the decay time constant of each membrane element, a horizontal comparison was performed to calculate the deviation ratio between the decay time constant of each membrane element and the group mean of the decay time constants of all membrane elements. When the proportion of a membrane element whose decay time constant is lower than the group mean exceeds the damage threshold, the membrane element is determined to have structural integrity damage. The advantage of horizontal comparison is that it can eliminate the influence of common-mode interference factors such as temperature and fluid properties on the absolute value of the decay time constant, because these common-mode interferences will have the same effect on all membrane elements. Horizontal comparison focuses on the relative differences between membrane elements, which can more accurately identify membrane elements with structural abnormalities and avoid misjudgments caused by water temperature fluctuations due to seasonal changes, which can lead to synchronous drift of the decay time constants of all membrane elements.

[0032] Different types of damage have varying impacts on pressure decay. Leakage caused by delamination of the membrane bag adhesive seam is the most severe, causing the decay time constant to plummet to below 30% of its normal value. Delamination between the membrane and the support layer has a relatively milder impact, with the decay time constant decreasing to 60% to 80% of its normal value. Membrane wrinkles have the least impact, with only slight fluctuations in the decay time constant. The pressure decay period begins when all inlet and outlet valves of the membrane elements are fully closed, at which point the pressure reaches its initial stable value, after which it begins to decrease naturally. When used for pressure decay after flow path switching in a directional valve assembly, the beginning time is the instant after the directional valve completes its action and closes all inlet and outlet valves of the membrane elements. In practice, the pressure decay signal is typically fitted using a first-order exponential decay model, where the pressure equals the ratio of the initial pressure multiplied by the natural constant e to the negative decay time and the decay time constant. By performing nonlinear least-squares fitting on the collected discrete pressure data points, the initial pressure and decay time constant are determined to minimize the fitting error; the decay time constant is the final extracted feature value.

[0033] In step 105, the pollution type result obtained in step 103 and the integrity damage result obtained in step 104 are combined to output the fault prediction result. Fusion refers to the comprehensive analysis and integration of result information from two different diagnostic dimensions to form a complete multi-dimensional fault prediction conclusion.

[0034] When the fouling type result indicates that a certain membrane element is fouled and the integrity damage result indicates that the membrane element has no structural damage, it means that the problem of the membrane element is limited to the fouling level and its physical structure is intact. At this time, a fouling warning and corresponding cleaning strategy suggestions are output for the single membrane element. The cleaning strategy is determined according to the fouling type. For surface filter cake layer fouling, it is recommended to increase the intensity of hydraulic flushing. For membrane pore blockage fouling, it is recommended to use chemical cleaning. For membrane surface adsorption fouling, it is recommended to use surfactant cleaning.

[0035] When the integrity damage result indicates that any membrane element has structural integrity damage, since structural damage is a serious failure mode, its consequences may lead to deterioration of product water quality or complete failure of the membrane element. At this time, an emergency warning for structural damage is issued to prompt the operation and maintenance personnel to immediately stop the machine, inspect and replace the damaged membrane element.

[0036] When multiple membrane elements are simultaneously contaminated and the anomalies in the prior constraint parameters indicate actual anomalies caused by membrane element degradation, it signifies that the entire system is in a degraded state. In this case, a system-level degradation warning is issued, prompting maintenance personnel to conduct a comprehensive assessment and maintenance at the system level. Through this fusion of outputs, maintenance personnel can obtain clear, layered, and targeted fault prediction information, enabling them to make accurate and timely maintenance decisions.

[0037] In some embodiments, obtaining the contamination type result specifically includes: using the contamination type identified based on hydraulic impedance parameters and hydraulic impedance thresholds as the first result; collecting operating data of the device under forward and reverse operating conditions respectively, constructing a bidirectional asymmetric feature based on the operating data, wherein the bidirectional asymmetric feature is the ratio of the performance recovery amount of the device under forward flushing conditions to the performance recovery amount under reverse flushing conditions; identifying the corresponding contamination type according to the bidirectional asymmetric feature to obtain the second result; and cross-validating the contamination types in the first and second results to correct and obtain the contamination type result. The cross-validation process is as follows: Figure 3 As shown.

[0038] In step 103, the fouling type of each membrane element was identified by recognizing its equivalent hydraulic impedance parameters and comparing the three impedance components with the hydraulic impedance threshold adjusted by prior constraints. This identification result is defined as the first result in this embodiment. A specific increase in equivalent resistance indicates surface cake layer fouling; a specific change in equivalent capacitive reactance indicates membrane pore blockage fouling; and a specific change in equivalent inductive reactance indicates membrane surface adsorption fouling. However, the identification of hydraulic impedance parameters relies on the systematic identification of the pressure response waveform. The identification accuracy is affected by signal noise and model order selection, resulting in some uncertainty. Furthermore, when multiple impedance components deviate to similar degrees, relying solely on the relative magnitude of the impedance parameters to determine the dominant fouling type may lead to misjudgment. Therefore, this embodiment further introduces an independent verification method: collecting operating data of the device under forward and reverse flushing conditions to construct a bidirectional asymmetric feature. Based on this feature, the corresponding fouling type is identified as the second result. Then, the first and second results are cross-validated, and the second result is used to correct the first result, ultimately yielding a more reliable fouling type result.

[0039] Forward flushing refers to the flushing operation performed when the device is in forward operation. The fluid flows in the normal filtration direction of the membrane element, from the inlet to the outlet, and the flushing liquid exerts a forward scouring effect on the membrane surface. Backwashing refers to the flushing operation performed after the device switches to reverse operation. The fluid flows in the opposite direction, from the outlet to the inlet, and the flushing liquid exerts a reverse scouring effect on the membrane surface. Operating data is collected under both flushing conditions. This data includes at least the flux values ​​before and after flushing (i.e., the volume of liquid permeating the membrane element per unit time), and may also include the transmembrane pressure difference before and after flushing. Based on this operating data, the performance recovery under forward flushing and backwashing conditions can be calculated separately. The performance recovery is calculated by subtracting the flux value before flushing from the flux value after flushing, and then dividing by the difference between the flux value before flushing and the flux value in the clean state, yielding a normalized recovery percentage. The ratio of the performance recovery under forward flushing to that under backwashing represents the bidirectional asymmetry characteristic. A ratio greater than 1 indicates that forward flushing is more effective than reverse flushing, a ratio less than 1 indicates that reverse flushing is more effective than forward flushing, and a ratio close to 1 indicates that forward and reverse flushing are equally effective.

[0040] Different types of fouling depend on the direction of flushing to varying degrees, and this difference is the physical basis for using bidirectional asymmetry features to identify fouling types. Surface cake fouling refers to the accumulation of pollutants on the membrane surface, forming a loose cake. Since the cake is mainly distributed on the feed side of the membrane element, forward flushing, with the fluid flowing in the original direction, directly acts on and removes the cake layer. Therefore, forward flushing is more effective than reverse flushing, and the bidirectional asymmetry feature is greater than 1. Membrane pore clogging fouling refers to pollutants embedded in the micropores of the membrane separation layer. Regardless of the direction of fluid flow through the micropores, the effect on the clogging material is basically the same. Therefore, forward and reverse flushing are equally effective, and the bidirectional asymmetry feature is close to 1. Membrane adsorption fouling refers to pollutants firmly adhering to the membrane surface through chemical or physical adsorption. Since the adsorption layer is distributed on both sides of the membrane surface, the shear force change caused by the change in fluid direction during reverse flushing may be more conducive to breaking the adsorption interface. Therefore, reverse flushing is more effective than forward flushing, and the bidirectional asymmetry feature is less than 1. Based on the above correspondences, the corresponding fouling type can be identified according to the specific numerical range of the bidirectional asymmetry feature; this identification result is defined as the second result.

[0041] After obtaining the first and second results, cross-validation is performed to correct the contamination type result. The specific implementation of cross-validation is as follows: The contamination type corresponding to the first result is compared with the contamination type corresponding to the second result. If they match, it indicates that the hydraulic impedance parameter identification result and the bidirectional flushing response characteristics corroborate each other, and the confidence level of the contamination type determination result is high. Therefore, the contamination type corresponding to the first result is confirmed as the final contamination type result. If they do not match, for example, the first result indicates surface filter cake layer contamination while the second result indicates membrane pore blockage contamination, it indicates a discrepancy between the conclusions given by the two diagnostic methods. In this case, the system outputs a diagnostic conflict warning, suggesting that maintenance personnel conduct manual verification by combining other information or through offline detection methods. It is important to note that in the priority setting of cross-validation, the first result is used as the primary identification result, and the second result is used as the basis for verification and correction. Therefore, when they are inconsistent, the second result is not forced to replace the first result; instead, a conflict warning is output for maintenance personnel to judge. The reason for this design is that the frequency domain identification method of hydraulic impedance parameters has clear physical mechanism support, while the bidirectional asymmetric feature is an indirect inference, which is greatly affected by flushing conditions such as flushing pressure and flushing flow rate. The reliability of the first result is usually higher than that of the second result.

[0042] The hardware foundation of this embodiment lies in the bidirectional flow membrane filtration device's ability to switch between forward and reverse operating states. This bidirectional switching function is achieved through a reversing valve assembly, which switches the fluid direction from forward to reverse during flow path switching, enabling the device to operate under both forward and reverse flushing conditions. Therefore, the technical solution of this embodiment does not require additional hardware for cross-validation; all necessary operational data can be collected during the device's existing forward and reverse flushing operations. Simultaneously, prior constraint parameters also apply to the cross-validation process in this embodiment. When the prior constraint parameters indicate that the current anomaly is a false anomaly induced by hydraulic coupling, the reliability of the cross-validation result is relatively higher because the acquisition of the first result has already been performed under conditions of increased hydraulic impedance threshold, resulting in a more stringent identification standard. The system can appropriately reduce its reliance on the consistency of the second result; that is, when the two results are inconsistent, the first result is prioritized over directly outputting a conflict warning. Conversely, when the prior constraint parameters characterize the current anomaly as a real anomaly caused by membrane element degradation, the consistency requirements of cross-validation are more stringent. Only when both parameters are consistent is the contamination type confirmed; otherwise, a conflict warning is output to ensure a more rigorous determination of the contamination type in real degradation scenarios. This interconnected constraint further demonstrates the organic synergistic relationship formed between the various steps of this invention through the prior constraint parameters.

[0043] In some embodiments, the method further includes: adjusting the flow equalization threshold according to prior constraint parameters; acquiring flow distribution data and resistance data of each membrane element based on the calibration sensor signal; determining the current flow distribution balance of the device based on the flow distribution data of each membrane element; when the flow distribution balance is lower than the flow equalization threshold, comparing the resistance change of each membrane element with the flow distribution change of that membrane element to analyze whether the device is experiencing operational degradation and obtain operational degradation results; and integrating the operational degradation results into the fault prediction results.

[0044] In a parallel membrane system, the membrane elements are interconnected via shared piping, and the total system flow is distributed among the branches according to the relative magnitude of their resistances. When the resistance of a membrane element changes, the total system flow automatically redistributes among the branches; this flow redistribution is an inherent physical response of the parallel hydraulic system. However, when this flow redistribution response is disrupted—for example, if a membrane element experiences bypass or internal leakage, causing the flow distribution to deviate from theoretical expectations—it indicates genuine operational degradation within the system, rather than simple single-membrane fouling. This embodiment achieves accurate identification of system-level operational degradation by organically combining two steps: flow distribution balance determination and resistance-flow matching analysis. The identification results are then incorporated into the final fault prediction system, making fault prediction more comprehensive and accurate.

[0045] The diagnostic step first adjusts the flow equalization threshold based on the prior constraint parameters generated in step 102. The flow equalization threshold is a benchmark value used to determine whether the current flow distribution in the system is balanced; its level determines the sensitivity of the system-level operational degradation diagnosis. When the prior constraint parameters indicate that the current anomaly is a false anomaly induced by hydraulic coupling, the flow equalization threshold is increased to make the diagnostic criteria for system-level operational degradation more lenient, avoiding misjudging normal hydraulic redistribution caused by single membrane failure as system-level degradation. When the prior constraint parameters indicate that the current anomaly is a real anomaly caused by membrane element degradation, the flow equalization threshold is maintained or decreased to keep the diagnostic criteria stricter or more stringent, ensuring that system-level anomalies can be accurately detected in real degradation scenarios. The specific value of this threshold can be pre-calibrated based on parameters such as system size, number of membrane elements, and standard deviation of flow distribution under normal operating conditions, or it can be dynamically determined based on the statistical distribution of historical operating data.

[0046] After adjusting the flow equalization threshold, flow distribution data and resistance data for each membrane element are acquired based on the calibrated sensor signals. Flow distribution data refers to the permeate flow rate allocated to each membrane element under current operating conditions. This can be obtained directly by flow sensors placed on the permeate side of each membrane element, or calculated from the total flow rate based on the resistance distribution relationship of each branch. Resistance data refers to the current hydraulic resistance of each membrane element, which can be calculated by dividing the transmembrane pressure difference by the permeate flow rate. In a parallel-operating membrane system, there is an inherent hydraulic coupling relationship between the membrane elements. When the system is in a healthy state, the resistance of each membrane element is similar, the flow distribution is relatively uniform, and the flow rate of each branch satisfies the basic hydraulic relationship of parallel pipelines with the total flow rate. When the resistance of a membrane element increases due to fouling, the flow rate of that branch decreases, and the flow rates of the other branches increase accordingly. This redistribution of flow distribution is essentially a normal hydraulic response to changes in the resistance of a single membrane, and the system as a whole is still in normal operation. Therefore, a decrease in flow distribution balance alone cannot directly determine that the system is deteriorating; further analysis of the matching relationship between resistance changes and flow rate changes is required.

[0047] The current flow distribution balance of the device is determined based on the flow distribution data of each membrane element. Flow distribution balance is a quantitative description of the uniformity of flow distribution among membrane elements. When the resistance of all membrane elements is similar, the total flow of the system is evenly distributed among the branches, and the flow of each membrane element is nearly equal, at which point the flow distribution balance is at its highest level. When the resistance of a membrane element increases, the flow in that branch decreases, while the flow in other branches increases, increasing the flow difference among membrane elements, and the flow distribution balance decreases accordingly. Flow distribution balance can be quantified by the coefficient of variation of the flow of each membrane element, i.e., the ratio of the standard deviation to the mean. The larger the coefficient of variation, the more uneven the flow distribution. Flow distribution balance can also be quantified by flow distribution entropy. The formula for calculating flow distribution entropy is the percentage of flow of each membrane element under a negative summation sign multiplied by the natural logarithm of that percentage. The larger the entropy value, the more uniform the flow distribution. When the flow distribution balance is below the flow balance threshold, it indicates that the current flow distribution of the system has become significantly uneven, triggering the second step of matching analysis.

[0048] When the flow distribution balance is below the flow balance threshold, the resistance change of each membrane element is compared with its own flow distribution change. The resistance change refers to the magnitude of the change in the membrane element's current resistance relative to its historical baseline resistance, which can be calculated by dividing the difference between the current transmembrane pressure difference and the baseline transmembrane pressure difference by the permeate flow rate. The flow distribution change refers to the magnitude of the change in the membrane element's current flow distribution ratio relative to its historical baseline flow distribution ratio. The physical significance of this comparison is to verify whether the resistance and flow changes of each membrane element conform to the hydraulic response law of parallel pipelines. In a parallel pipeline system, when the resistance of a membrane element increases, its flow distribution ratio should decrease accordingly. If the increase in resistance of a membrane element matches the decrease in its flow rate, it indicates that the state change of that membrane element conforms to the normal hydraulic response law, and the system as a whole is still in normal operation. In this case, the current state is determined to be hydraulic redistribution caused by single-membrane fouling. If the resistance change of a membrane element exceeds the range matching the flow distribution change—for example, if the resistance of a membrane element increases but the flow rate decreases by a much smaller margin than theoretically expected—it indicates that the membrane element may have a bypass or internal leakage, with fluid bypassing the normal membrane separation channel. Conversely, if the flow rate of a membrane element decreases but the resistance does not increase as expected, it indicates that the membrane element may be damaged or perforated, leading to an abnormally low fluid resistance. Both of these situations constitute system-level operational degradation, rather than a normal single-membrane fouling response. Through the above resistance-flow matching comparison, normal hydraulic redistribution and true system-level operational degradation can be accurately distinguished, avoiding false alarms caused by judging system-level degradation solely based on a decrease in flow distribution balance, and also avoiding missed alarms caused by the fact that flow distribution balance has not yet significantly deteriorated when system-level degradation occurs. The results of the resistance-flow matching analysis are output as operational degradation results, which include a conclusion on whether the system is currently experiencing operational degradation, and the location information of which membrane elements are specifically affected when operational degradation exists.

[0049] Finally, the operational degradation results are integrated into the fault prediction results. Specifically, this integration involves combining the contamination type results obtained in step 103 and the integrity damage results obtained in step 104 to form a complete fault prediction conclusion. When the operational degradation result indicates operational degradation, a system-level degradation warning is output, alerting maintenance personnel that the system as a whole is in an abnormal operating state, requiring attention to system-level issues while addressing specific membrane element faults. When the operational degradation result indicates that the current state is only hydraulic redistribution caused by single membrane fouling, no system-level warning is triggered; only a fouling warning for the specific membrane element is output, thus avoiding misjudging normal physical responses as system-level anomalies. Through this integrated output, maintenance personnel can obtain multi-level fault prediction information from single membrane fouling and single membrane structural damage to system-level operational degradation, providing comprehensive decision-making basis for precise operation and maintenance. This diagnostic step, along with the contamination type diagnosis and structural integrity diagnosis, is independent yet complementary, together forming a complete fault prediction system.

[0050] In some embodiments, when the abnormal state of the remaining membrane elements disappears synchronously after the abnormal source membrane element is isolated, the abnormality is determined to be a false abnormality caused by the abnormal source membrane element being hydraulically coupled to the remaining membrane elements through parallel pipelines; when the abnormal state of the remaining membrane elements persists after isolation, the abnormality is determined to be a real abnormality caused by the deterioration of multiple membrane elements.

[0051] In actual operation, when multiple membrane elements are detected to be simultaneously abnormal based on calibration sensor characteristics, the system initiates an isolation and discrimination process. The core operation of this process is to briefly isolate the abnormal source membrane element from the parallel piping system, and then observe the changes in the abnormal state of the remaining membrane elements after isolation. The physical basis of this judgment lies in the hydraulic coupling characteristics of the parallel piping system. Under normal operating conditions, the inlet and outlet pipes of each membrane element are interconnected through a common main pipe. Changes in the operating state of any branch will affect the operating parameters of other branches through pressure transmission and flow redistribution in the piping system. When a membrane element becomes fouled or damaged, causing its resistance to increase, the flow rate of that branch will decrease, and the released flow will be automatically distributed to the other parallel branches, causing passive changes in the flux and pressure difference of the remaining membrane elements. These passively changing signals are highly similar in appearance to the signals generated by the actual degradation of the membrane elements themselves, but their physical causes are completely different, and current technology cannot distinguish between the two.

[0052] The core principle of isolation operation lies in cutting off the hydraulic connection between the abnormal membrane element and the parallel piping system. Once the abnormal membrane element is isolated, its resistance state no longer affects the pressure distribution and flow allocation of the shared piping. The remaining membrane elements are no longer hydraulically coupled to the isolated element, and their operating parameters will readjust to a state unaffected by this coupling. If the abnormal states of the remaining membrane elements disappear synchronously after isolation—that is, their flux, differential pressure, strain, and other sensing characteristics return to normal ranges—it proves that the abnormalities of these membrane elements did not originate from their own actual degradation, but were caused by the state changes of the isolated membrane element transmitted through hydraulic coupling. This situation is considered a pseudo-anomaly induced by hydraulic coupling, essentially a physical phenomenon caused by the system piping topology, rather than irreversible degradation of the membrane element itself. In this case, the isolated membrane element is the true source of failure, while the remaining membrane elements are merely passively responding to the result of hydraulic redistribution.

[0053] If the abnormal state of the remaining membrane elements persists after the isolation operation—that is, their sensing characteristics such as flux, differential pressure, and strain do not recover after isolation—it proves that the abnormality of these membrane elements is unrelated to the isolated membrane element and is not caused by hydraulic coupling. These membrane elements themselves do exhibit genuine degradation, representing true degradation of multiple membrane elements. In such cases, the isolation operation has eliminated the interference of hydraulic coupling. If the remaining membrane elements still exhibit abnormal states after being separated from the influence of the isolated membrane element, it indicates that the degradation of these membrane elements occurred independently and needs to be investigated and addressed individually. The significance of this determination lies in helping maintenance personnel accurately identify whether a true collective degradation event has occurred in the system, such as a sudden deterioration in the influent water quality leading to simultaneous fouling of multiple membranes, or long-term operation of the system causing simultaneous aging of multiple membranes.

[0054] The key to the above judgment rule lies in the timing of the judgment. Isolation operations do not immediately reveal the disappearance or persistence of anomalies in other membrane elements because the piping system undergoes a hydraulic transient response. After the source membrane element is isolated, the system flow needs to undergo a redistribution process, and the flux and pressure differential of the remaining membrane elements require several minutes to reach a new steady state. Therefore, the judgment should be made after the operating parameters of the remaining membrane elements have entered a new steady state, rather than at the moment of isolation, to avoid misjudging transient hydraulic shocks as abnormal conditions. Specifically, the duration of the isolation operation is typically between five and fifteen minutes, which is sufficient for the system to complete its hydraulic transient response and enter a new steady state.

[0055] The beneficial effect of this judgment rule lies in utilizing the unique isolable valve-controlled hardware structure of multi-membrane element parallel systems. It elevates isolation operation from a simple fault-cutting measure to an active diagnostic verification tool, inferring the systemic cause of group anomalies by observing changes in the abnormal states of other membrane elements before and after isolation. This active verification method requires no additional hardware or complex mathematical calculations; it can accurately determine the cause of anomalies solely through physical isolation and state observation, offering advantages such as simplicity and reliable results.

[0056] In some embodiments, the process of acquiring the corrected sensing features includes: isolating each membrane element for a short period of time according to a preset cycle, collecting the sensing signals of each membrane element in the isolated state without being disturbed by the hydraulic coupling of other membrane elements as reference sensing features; after resuming parallel operation, comparing the real-time sensing signals with the reference sensing features, calculating the feature distortion variable; and correcting the real-time sensing features based on the feature distortion variable.

[0057] In a multi-membrane element parallel system, the inlet and outlet pipes of each membrane element are interconnected through a common main pipe, and there is an inherent hydraulic coupling relationship between the membrane elements. When multiple membrane elements operate in parallel, the total system flow is distributed inversely proportional to the resistance of each branch. Changes in the state of any one membrane element will affect the operating parameters of other membrane elements through pressure transmission and flow redistribution in the common pipe. The physical essence of this hydraulic coupling effect lies in the flow distribution conservation constraint of the parallel pipe system: the sum of the flow rates of each branch equals the total flow rate, and the pressure in the total pipe equals the pressure at the inlet of each branch. When the resistance of one branch changes, the system must readjust the flow distribution of each branch to satisfy the above conservation constraint. This adjustment is achieved through pressure changes in the common pipe. Therefore, the impact of a change in the state of any membrane element in a parallel system is not limited to that membrane element itself but is transmitted through the piping system to all membrane elements operating in parallel. This change in the sensing signal caused by hydraulic coupling does not represent a real change in the state of the other membrane elements themselves, but rather a false abnormal signal superimposed on the actual state signal of the membrane element, forming a characteristic distortion. If the original sensing signal is used directly for monitoring and diagnosis, this characteristic distortion will cause the system to misjudge a normally functioning membrane element as having contamination or damage.

[0058] The calibration process is implemented in three closely linked operational stages. The first stage involves short-term isolation of each membrane element according to a preset cycle, collecting the sensing signals of each element in isolation as reference sensing characteristics. The preset cycle can be determined based on the stability of the system's operating conditions, typically set to perform a complete cycle of isolation calibration every 24 or 48 hours, or it can be performed before and after each backflushing operation. Short-term isolation refers to controlling the isolation time of a single membrane element to a range sufficient for signal acquisition without affecting the overall water production capacity of the system, typically five to fifteen minutes. The isolation operation is performed by closing the corresponding inlet and outlet valves of the membrane element, completely cutting off its fluid connection with the parallel piping system. In the isolated state, the isolated membrane element is completely disconnected from the parallel piping system, does not participate in the distribution of the total system flow, is unaffected by changes in the operating status of other membrane elements, and is unaffected by the hydraulic coupling effect of the parallel flow field. The sensor signal acquired at this point originates entirely from the individual state of the membrane element under the current influent water quality and operating conditions, unaffected by any external hydraulic coupling factors. Therefore, it can serve as the reference sensing characteristic for that membrane element. Sequential isolation refers to performing the above isolation acquisition operation sequentially on all membrane elements operating in parallel within the system until the reference sensing characteristics of each membrane element are updated. This sequential operation design allows each membrane element to complete reference acquisition independently, avoiding mutual interference between membrane elements.

[0059] The second stage involves comparing real-time sensor signals with baseline sensor characteristics after resuming parallel operation to calculate the characteristic distortion. After baseline acquisition of all membrane elements is completed, the isolated membrane elements are restored to parallel operation, allowing them to re-participate in the distribution of the total system flow. Real-time sensor signals from each membrane element are continuously acquired during parallel operation. These real-time signals are then compared item by item with the baseline sensor characteristics acquired during isolation, and the difference is calculated. Since the baseline sensor characteristics are acquired in a state free from hydraulic coupling interference, they represent the true individual state of the membrane element. In contrast, the real-time sensor characteristics are acquired during parallel operation and include both the true individual state of the membrane element and spurious signal components introduced by hydraulic coupling. Therefore, the difference between the real-time sensor characteristics and the baseline sensor characteristics is the characteristic distortion introduced by hydraulic coupling in the parallel pipeline under the current operating conditions. The magnitude of this distortion dynamically changes with operating conditions, reflecting the degree of mutual influence between the membrane elements through hydraulic coupling.

[0060] The third stage involves correcting the real-time sensing characteristics based on characteristic distortions. The specific method is to subtract the characteristic distortions from the real-time sensing characteristics, that is, subtract the spurious signal component introduced by hydraulic coupling from the real-time acquired sensing signal to obtain the corrected sensing characteristics. The corrected sensing characteristics eliminate the interference of parallel hydraulic coupling and can accurately reflect the state of each membrane element at the current moment, which can be used for subsequent anomaly monitoring and fault diagnosis. It is worth noting that this correction process is continuous; the system continuously performs real-time sensing signal acquisition, distortion calculation, and correction operations during normal operation to ensure that the corrected sensing characteristics of each membrane element are always updated.

[0061] The advantage of the above correction mechanism lies in its dynamic adaptability. Changes in influent water quality synchronously affect the sensing signals of all membrane elements, causing the reference and real-time sensing characteristics of each membrane element to drift in the same direction. This common-mode drift is automatically subtracted through reference comparison and difference calculation, without affecting the correction result. Changes in operating conditions also synchronously affect the operating parameters of each membrane element, and are similarly automatically subtracted. The slow aging or performance degradation of each membrane element itself is retained in the successive updates of the reference sensing characteristics and will not be misjudged as a distorted variable in the difference calculation. Only when relative changes occur between membrane elements, i.e., flow redistribution caused by hydraulic coupling, will it be reflected in the difference between the real-time sensing characteristics and the reference sensing characteristics and be corrected. This dynamic adaptive correction mechanism does not rely on preset fixed thresholds or complex mathematical models; it can be achieved simply by comparing the signals of each membrane element in isolated and parallel states, and has the advantages of simple implementation and strong adaptability. Through this periodic rotation of isolation verification and real-time correction, the system can continuously track the slow changes in the individual benchmarks of each membrane element as operating conditions change, ensuring that the corrected sensing characteristics always accurately reflect the state of each membrane element, providing a reliable data foundation for subsequent detection and isolation of group anomalies.

[0062] In some embodiments, when the prior constraint parameters characterize the anomaly as a false anomaly induced by hydraulic coupling, the hydraulic impedance threshold, damage threshold, and flow balance threshold are increased; when the prior constraint parameters characterize the anomaly as a real anomaly caused by membrane element degradation, the hydraulic impedance threshold, damage threshold, and flow balance threshold are maintained or decreased. These three thresholds control the sensitivity of three different diagnostic modules: contamination type identification, structural integrity determination, and system-level operational degradation diagnosis. The level of the threshold determines the strictness of each diagnostic module in determining anomalies; a lower threshold results in more sensitive detection and a higher chance of false positives, while a higher threshold results in a more conservative approach and a lower false positive rate, but also a higher chance of missing real anomalies. Therefore, the threshold settings need to be dynamically adjusted based on the physical causes of the current anomaly to achieve the optimal balance between detection rate and false alarm rate.

[0063] When the prior constraint parameters characterize the current collective anomaly as a pseudo-anomaly induced by hydraulic coupling, the hydraulic impedance threshold, damage threshold, and flow equilibrium threshold are increased. The physical basis for this adjustment is as follows: In the scenario of pseudo-anomalies induced by hydraulic coupling, the essence of the collective anomaly is that the state change of one or a few membrane elements is transmitted to other membrane elements through the hydraulic coupling effect of parallel pipelines, causing passive changes in the flux and pressure difference of other membrane elements. However, these passively changed membrane elements themselves have not undergone actual degradation. In this case, the core risk faced by subsequent diagnostic modules is misjudging the pseudo-signals caused by this hydraulic coupling transmission as real contamination or damage. Increasing the hydraulic impedance threshold means that the judgment criteria of the contamination type identification module are more stringent. Only when the deviation of the hydraulic impedance parameter is large enough will it be judged as contamination, thereby avoiding misjudging slight impedance fluctuations caused by hydraulic redistribution as contamination. Raising the damage threshold means the structural integrity damage assessment module has stricter criteria; only when the attenuation time constant deviates significantly will it be considered structural damage. This avoids misjudging temporary pressure changes caused by hydraulic redistribution as membrane delamination or adhesive seam separation. Raising the flow equalization threshold means the system-level operational degradation diagnosis module has more lenient criteria, increasing the allowable deviation range for system flow distribution uniformity. This avoids misjudging normal hydraulic redistribution caused by single-membrane fouling as system-level operational degradation. The increase in the three thresholds can be the same or different, depending on the sensitivity of each diagnostic module to hydraulic coupling interference. Typically, the hydraulic impedance threshold and flow equalization threshold are increased more significantly, while the damage threshold is increased relatively less, because structural damage assessment has a lower tolerance for misjudgment.

[0064] When the prior constraint parameters characterize the current collective anomaly as a genuine anomaly caused by membrane element degradation, the hydraulic impedance threshold, damage threshold, and flow balance threshold are maintained or lowered. The physical basis for this adjustment direction is as follows: In a genuine anomaly scenario caused by membrane element degradation, the essence of the collective anomaly is that multiple membrane elements have indeed experienced real fouling or damage simultaneously. These membrane elements themselves exhibit irreversible state changes that need to be accurately detected and located. In this case, the core risk faced by subsequent diagnostic modules is missing the detection of genuine fouling or damage. Maintaining or lowering the hydraulic impedance threshold means that the judgment criteria of the fouling type identification module remain unchanged or become more sensitive, ensuring that deviations in impedance parameters from genuine fouling can be accurately identified. Maintaining or lowering the damage threshold means that the judgment criteria of the structural integrity damage assessment module remain unchanged or become more sensitive, ensuring that genuine structural damage can be accurately detected. Maintaining or lowering the flow balance threshold means that the judgment criteria of the system-level operational degradation diagnosis module remain unchanged or become more stringent, ensuring that genuine system-level degradation can be triggered in a timely manner. It's worth noting that in real-world anomaly scenarios, lowering the threshold isn't always necessary; maintaining the original threshold is also a viable option. This is because the signal deviation generated by real anomalies is usually large enough to be detected even without lowering the threshold. Lowering the threshold is suitable for the early stages of degradation, when the signal deviation is not yet significant, and lowering the threshold is necessary to improve detection sensitivity.

[0065] The core value of adjusting the three types of thresholds in the same direction is to ensure that the judgment criteria of each diagnostic module are coordinated and consistent under the guidance of unified prior information. If each module adjusts its threshold independently without being controlled by the same prior constraint, logical contradictions may occur. For example, the pollution type identification module may determine that the pollution is severe, but the system-level diagnostic module may determine that it is operating normally because the threshold has not been adjusted. Such contradictions will prevent maintenance personnel from making accurate decisions. By adjusting the three types of thresholds in the same direction, the judgment criteria of each diagnostic module remain consistent under a unified physical cognitive framework. In the case of false anomalies, the judgment thresholds of each module are raised synchronously to avoid the superposition of multiple false alarms. In the case of real anomalies, the judgment thresholds of each module are lowered synchronously to ensure that real faults are accurately detected. This achieves synergy between the diagnostic steps and makes the output results of the entire fault prediction method more reliable.

[0066] In some embodiments, when the reversing valve group of the bidirectional flow membrane filtration device performs flow path switching, the pressure step generated by the flow path switching is used as the native excitation signal to collect the pressure response waveforms at both ends of each membrane element; system identification is performed on the pressure response waveforms to extract the equivalent hydraulic impedance parameters of each membrane element in the frequency domain; the hydraulic impedance parameters include equivalent resistance, equivalent capacitive reactance and equivalent inductive reactance.

[0067] Specifically, the identification of fouling types based on hydraulic impedance parameters and threshold values ​​includes: when the equivalent resistance exceeds the hydraulic impedance threshold by a greater margin than the equivalent capacitive reactance and equivalent inductive reactance each exceed the hydraulic impedance threshold, it is identified as surface filter cake fouling; when the equivalent capacitive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and equivalent inductive reactance each exceed the hydraulic impedance threshold, it is identified as membrane pore blockage fouling; when the equivalent inductive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and equivalent capacitive reactance each exceed the hydraulic impedance threshold, it is identified as membrane surface adsorption fouling. The fouling type identification process is as follows: Figure 4 As shown.

[0068] In parallel membrane filtration systems, directional valve assemblies are used to switch the device between forward and reverse operating states. When the directional valve assembly performs a flow path switch, the rapid action of the valve generates a steep pressure change in the pipeline, known as a pressure step. This pressure step is a broadband transient signal with a spectral range of several hundred hertz, sufficient to excite the membrane element's dynamic response over a wide frequency band. This pressure step is an inherent operating event during normal operation of the device, occurring naturally with each flow path switch. It requires no additional excitation equipment, consumes no extra energy, and does not interrupt the normal operation of the device. Because this pressure step originates from the device's own operation rather than external equipment, it is called the native excitation signal.

[0069] At the moment of the pressure step, the system acquires the pressure response waveforms at both ends of each membrane element at a preset sampling frequency. Pressure sensors are positioned at the inlet and outlet ends of each membrane element. The inlet pressure sensor is located downstream of the inlet valve of the membrane element, near the membrane housing inlet, while the outlet pressure sensor is located upstream of the outlet valve of the membrane element, near the membrane housing outlet. The difference between the two pressure sensors is the transmembrane pressure difference. The pressure response waveform is a complete time-domain record of the pressure change at each measuring point under pressure step excitation, typically lasting from several seconds to tens of seconds, with a sampling frequency of at least 1 kilohertz to ensure the capture of high-frequency response components. This waveform contains the dynamic response information of the membrane element to pressure excitation, which is represented in the form of impedance.

[0070] System identification is performed on the acquired pressure response waveforms to extract the equivalent hydraulic impedance parameters of each membrane element in the frequency domain. System identification is a method of establishing a mathematical model of a system using its input and output signals. In this embodiment, the input signal is the pressure step generated in the pipeline when the directional valve switches, and the output signal is the pressure response waveform at both ends of each membrane element. Using either time-domain or frequency-domain identification methods, the membrane element can be equivalently represented as a second-order fluid system model. The transfer function of this model is uniquely determined by three parameters: equivalent resistance, equivalent capacitive reactance, and equivalent inductive reactance. Equivalent resistance reflects the steady-state flow resistance of fluid passing through the membrane element, and in the frequency domain, it is expressed as the impedance value at zero frequency, measured in Pascals per cubic meter per second. Equivalent capacitive reactance reflects the dynamic response characteristics of the fluid storage and release capacity within the membrane element, and in the frequency domain, it is expressed as the component of impedance that is inversely proportional to frequency, measured in the reciprocal of Pascals per cubic meter per second. Equivalent inductive impedance reflects the response characteristics of a membrane element to changes in the inertia of fluid flow. In the frequency domain, it is expressed as the component of impedance that is proportional to frequency as it changes with frequency, and the unit is Pascals per second squared per cubic meter.

[0071] The physical meanings of the three parameters mentioned above can be compared to resistance, capacitance, and inductance in a circuit, but they correspond to fluid systems. Equivalent resistance is mainly affected by the thickness of the filter cake layer on the membrane surface; the thicker the filter cake layer, the greater the equivalent resistance. Equivalent capacitive reactance is mainly affected by the micropore state of the membrane separation layer; micropore blockage alters the liquid storage capacity of the membrane element, leading to changes in equivalent capacitive reactance. Equivalent inductive reactance is mainly affected by the adsorption state on the membrane surface; the adsorption layer changes the boundary layer characteristics of the membrane surface, affecting the inertial fluid exchange process, resulting in changes in equivalent inductive reactance. Therefore, these three parameters correspond to three different fouling mechanisms, which is the physical basis for identifying fouling types using a combination of these three parameters.

[0072] After obtaining the equivalent resistance, equivalent capacitive reactance, and equivalent inductive reactance of each membrane element, these three parameters are compared with the hydraulic impedance threshold adjusted for prior constraints. The hydraulic impedance threshold is a common benchmark value for judging whether each parameter is abnormal; when a parameter exceeds this threshold, it indicates a significant deviation. Then, the extent to which each of the three parameters exceeds the threshold is compared. When the equivalent resistance exceeds the hydraulic impedance threshold by a greater margin than the equivalent capacitive reactance and equivalent inductive reactance, it indicates that the main characteristic change of the membrane element is concentrated in the resistance component, while the changes in the dynamic response and inertial response components are relatively small. This matches the physical characteristics of surface cake fouling and is identified as surface cake fouling. When the equivalent capacitive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and equivalent inductive reactance, it indicates that the main characteristic change of the membrane element is concentrated in the dynamic response component, matching the physical characteristics of membrane pore clogging fouling and is identified as membrane pore clogging fouling. When the equivalent inductive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and equivalent capacitive reactance, it indicates that the main characteristic change of the membrane element is concentrated in the inertial response component, which is consistent with the physical characteristics of membrane surface adsorption fouling, and is thus identified as membrane surface adsorption fouling. When none of the three parameters exceed the threshold, it indicates that no obvious fouling characteristics are currently detected in the membrane element. When the exceedances of two or more of the three parameters are similar, it indicates that the membrane element may be in a state of complex fouling, i.e., multiple types of fouling coexist.

[0073] The advantage of this identification method is that it does not rely on preset fixed rules or empirical thresholds. Instead, it determines the dominant pollution type by comparing the relative magnitudes of the exceedances of three parameters under the same threshold, avoiding the problem of different parameters being unable to be directly compared due to their different dimensions. Furthermore, the hydraulic impedance threshold itself has already been adaptively adjusted based on prior constraint parameters in step 103. Therefore, the entire identification process is conducted under the condition of considering the current cause of the anomaly, making it more targeted and reliable.

[0074] In some embodiments, during the normal shutdown of the bidirectional flow membrane filtration device or during the pressure decay period after the reversing valve group performs flow path switching, the inlet and outlet valves of each membrane element are closed, and the pressure decay signal at both ends of each membrane element over time is collected at a preset sampling frequency; the decay signal is curve-fitted to extract the pressure decay time constant of each membrane element; the decay time constants of each membrane element are compared laterally to calculate the deviation ratio of the decay time constant of each membrane element from the group mean; when the proportion of a membrane element whose decay time constant is lower than the group mean exceeds the damage threshold, it is determined that the membrane element has integrity damage.

[0075] The structural integrity of a spiral wound membrane element refers to the degree of integrity of its physical structure. Specifically, this includes whether delamination has occurred between the membrane and the support layer, whether there is adhesive failure or leakage at the seams around the membrane bag, whether the membrane is cracked or wrinkled, and whether there is compression deformation between the feed grid and the membrane. When the membrane element exhibits structural integrity damage as described above, the integrity of its internal fluid channels is disrupted. Fluid can short-circuit through the damaged areas, bypassing the normal membrane separation channels, resulting in a significant reduction in the overall fluid resistance of the membrane element. This reduction in resistance is directly reflected in the pressure decay process, manifesting as a faster pressure decay rate and a smaller decay time constant. Therefore, by measuring the pressure decay time constant of each membrane element and making a lateral comparison, membrane elements with structural integrity damage can be identified.

[0076] There are two opportunities to collect pressure decay signals. The first is during normal system shutdown. After prolonged continuous operation, membrane filtration systems require periodic shutdowns for maintenance. During shutdown, the system is not producing water, and the membrane elements are still filled with liquid, making this an ideal time window for pressure decay testing. The second opportunity is during the pressure decay period after the directional valve assembly performs a flow path switch. When the directional valve assembly performs a flow path switch, the valve action generates pressure fluctuations in the pipeline. After the switch, the pipeline pressure undergoes a natural decay process from fluctuation to stability, which can also serve as a time window for pressure decay testing. Both normal system shutdown and the pressure decay period after the directional valve switch occur during existing idle time during normal system operation. No additional shutdown is required, and water production is not interrupted, achieving zero-time-cost testing.

[0077] Before acquiring pressure decay signals, the inlet and outlet valves of each membrane element need to be closed. The purpose of closing the valves is to create a closed fluid space within each membrane element, preventing fluid exchange between the internal fluid and the external piping system. This ensures that pressure decay is entirely determined by the membrane element's structural characteristics and is unaffected by the external piping system. After the inlet valve is closed, the fluid connection between the membrane element and the inlet pipe is severed; similarly, after the outlet valve is closed, the fluid connection between the membrane element and the outlet pipe is severed. At this point, each membrane element forms an independent, sealed cavity. The fluid within this cavity is under a certain pressure, equal to the hydrostatic pressure of the fluid inside the membrane element at the moment the valve is closed. Because the membrane element contains permeate and concentrate channels, when the inlet and outlet valves are closed, the high-pressure side fluid slowly leaks through the membrane to the low-pressure side, causing the pressure at both ends of the membrane element to gradually decrease. This process is known as natural pressure decay.

[0078] The pressure decay signal across each membrane element is acquired over time at a preset sampling frequency. Pressure sensors are positioned at the inlet and outlet of each membrane element, continuously recording the pressure change over time after the valve is closed. The sampling frequency should be determined based on the pressure decay rate, typically between 10 Hz and 100 Hz, ensuring a complete record of the decay curve's shape while avoiding excessive data volume. The acquisition duration is usually 30 seconds to 2 minutes, sufficient to cover the main decay process from the initial pressure value to near equilibrium. The acquired data is a discrete pressure-time series, with each sampling moment corresponding to a pressure value; these data points collectively constitute the original record of pressure decay.

[0079] Curve fitting is performed on the acquired pressure decay signals to extract the pressure decay time constant of each membrane element. Curve fitting refers to fitting discrete sampling data points to a continuous mathematical function to extract key characteristic parameters that characterize the decay process. Pressure decay signals are typically fitted using a first-order exponential decay model, which assumes that the pressure decay rate is proportional to the degree to which the current pressure deviates from the equilibrium value. The fitting process uses nonlinear least squares to find the model parameters that minimize the fitting error, with the decay time constant being the most important parameter. The decay time constant is defined as the time required for the pressure to decrease from its initial value to 36.8% of the difference between the initial and equilibrium values; it reflects the rate of pressure decay. A smaller decay time constant indicates faster pressure decay and poorer structural integrity of the membrane element. A larger decay time constant indicates slower pressure decay and better structural integrity of the membrane element.

[0080] After obtaining the decay time constants of each membrane element, a horizontal comparison is performed. This horizontal comparison involves statistically analyzing the decay time constants of all membrane elements together, rather than comparing each element's decay time constant to a fixed preset value. The basis for this horizontal comparison is that, under the same operating environment and conditions, the structural characteristics of each membrane element are essentially consistent in the initial state. Even after a period of operation, the decay time constants of each membrane element should remain within a similar range. If the decay time constant of a particular membrane element deviates significantly from that of other membrane elements, it indicates that the membrane element may have structural integrity damage. The specific method for horizontal comparison is to calculate the deviation ratio of each membrane element's decay time constant from the group mean. The group mean is the arithmetic mean of the decay time constants of all membrane elements, representing the overall level of the membrane group. The deviation ratio is the difference between a particular membrane element's decay time constant and the group mean divided by the group mean, reflecting the relative deviation of that membrane element from the group average level. A larger deviation ratio indicates a greater difference between the membrane element's decay time constant and the group mean, and a higher probability of structural integrity damage.

[0081] When the proportion of membrane elements whose decay time constant is lower than the population mean exceeds the damage threshold, the membrane element is determined to have integrity damage. The damage threshold is a pre-set judgment benchmark value used to determine whether the deviation proportion has reached a level sufficient to be considered structural damage. The setting of the damage threshold needs to balance the detection rate and false alarm rate, and is usually set between 20% and 50%, with the specific value depending on the type of membrane element and operating conditions. If the threshold is set too low, normal fluctuations may be misjudged as structural damage, resulting in false alarms. If the threshold is set too high, early structural damage may be missed, delaying maintenance. In step 104, the damage threshold has been adaptively adjusted according to the prior constraint parameters, increasing it in false anomaly scenarios to reduce the risk of false judgment, and maintaining or decreasing it in real degradation scenarios to improve detection sensitivity.

[0082] The advantage of lateral comparison lies in eliminating the influence of common-mode interference on the absolute value of the decay time constant. Temperature changes simultaneously alter the viscosity of the fluid within all membrane elements, thus synchronously affecting the decay time constant of all membrane elements. However, this synchronous drift is offset by changes in the group mean in lateral comparison. Changes in fluid properties also synchronously affect all membrane elements, and are similarly eliminated by lateral comparison. Only when a membrane element suffers structural integrity damage will its decay time constant change independently relative to the group, thus being accurately identified by lateral comparison. Through lateral comparison and deviation ratio calculation, membrane elements with structural integrity damage can be quickly located without shutting down the system or disassembling the membrane elements, providing maintenance personnel with accurate maintenance decision-making basis.

[0083] In some embodiments, the fault prediction results include: when the contamination type result indicates that a single membrane element is contaminated and the integrity damage result indicates that the membrane element has no structural damage, a single membrane contamination warning and a corresponding cleaning strategy suggestion are output; when the integrity damage result indicates that any membrane element has structural integrity damage, an emergency warning for structural damage is output; when multiple membrane elements are contaminated at the same time and the prior constraint parameters characterize the anomaly as a real anomaly caused by membrane element degradation, an overall membrane element degradation warning is output.

[0084] The output of fault prediction results follows preset fusion logic rules. These rules determine the type of warning based on different combinations of contamination type results, integrity damage results, and prior constraint parameters. The first type of output rule addresses the case of a single membrane element being contaminated while its structure remains intact. When the contamination type result indicates contamination in a membrane element but the integrity damage result indicates no structural damage, it means the problem with that membrane element is limited to the contamination level, and its physical structure, such as the bonding state between the membrane and the support layer, the sealing state of the membrane bag joints, and the integrity of the membrane, is intact. In this case, the system outputs a single membrane contamination warning, which includes the specific membrane element's number or location identifier and a corresponding cleaning strategy recommendation. The cleaning strategy recommendation is determined based on the contamination type: for surface cake layer contamination, increased hydraulic flushing intensity is recommended; for membrane pore blockage contamination, chemical cleaning is recommended; and for membrane surface adsorption contamination, surfactant cleaning is recommended. The single membrane contamination warning is a routine maintenance warning, prompting maintenance personnel to perform the corresponding cleaning operation on the membrane element to restore its performance without requiring downtime for maintenance or replacement.

[0085] The second type of output rule addresses structural integrity damage. When the integrity damage result indicates that any membrane element has structural integrity damage, the system outputs an emergency structural damage warning regardless of whether the membrane element is simultaneously fouled. Structural integrity damage includes membrane delamination from the support layer, membrane bag adhesive seam opening and leakage, membrane rupture or wrinkling, and feed grid deformation due to compression. These damages are irreversible physical structural failures and cannot be restored by cleaning. The consequences of structural damage may lead to cross-contamination between the feed solution and permeate, causing deterioration of the permeate quality, or complete failure of the membrane element's separation function. Therefore, the emergency structural damage warning is the highest priority warning, prompting maintenance personnel to immediately shut down the system to inspect and replace the damaged membrane element. This warning has a higher priority than the single membrane fouling warning; that is, when the same membrane element has both fouling and structural damage, the system only outputs an emergency structural damage warning and not a single membrane fouling warning, because the severity and urgency of structural damage are far greater than that of fouling.

[0086] The third type of output rule addresses system-wide degradation. When multiple membrane elements are simultaneously fouled and the prior constraint parameters indicate that the current anomaly is a genuine anomaly caused by membrane element degradation, the system outputs a warning for overall membrane element degradation. In this case, the fouling type result indicates that multiple membrane elements are fouled, and the prior constraint parameters have been verified through isolation to rule out the possibility of hydraulic coupling-related false anomalies, confirming that the fouling is a genuine degradation of each membrane element itself. Simultaneous degradation of multiple membrane elements usually indicates a common cause at the system level, such as a sudden deterioration in influent water quality leading to simultaneous fouling of multiple membranes, or adverse changes in system operating conditions leading to simultaneous aging of multiple membranes. The overall degradation warning prompts maintenance personnel to conduct a comprehensive system-wide assessment and maintenance, rather than focusing solely on individual membrane elements. Upon receiving this warning, maintenance personnel should check for changes in influent water quality, whether the pretreatment system is operating normally, and whether operating parameters deviate from the design range, and formulate corresponding system-level response measures. In the case of overall degradation, if one of the membrane elements also has structural damage, the emergency warning for structural damage will still be issued first, because the severity of structural damage is higher than that of overall degradation.

[0087] There is a clear priority relationship among the three types of output rules mentioned above. The highest priority is the emergency warning for structural damage; once any membrane element is determined to have structural integrity damage, the system will output this warning regardless of other diagnostic results. The second highest priority is the warning for overall membrane element degradation; this is output when multiple membrane elements have actual contamination but no structural damage. The lowest priority is the warning for single membrane contamination; this is output when a single membrane element has contamination but no structural damage. This priority design ensures the highest response priority for the most severe failure mode—membrane element structural damage—while providing appropriate warning levels for system-wide degradation and single membrane contamination. Through this hierarchical output, maintenance personnel can quickly determine the severity and urgency of the fault based on the warning level, and accordingly arrange maintenance priorities and resource allocation, significantly improving the efficiency and accuracy of maintenance decisions.

[0088] A fault prediction system for a bidirectional flow membrane filtration device is provided, applicable to a bidirectional flow membrane filtration device capable of switching between forward and reverse operating states. The system's module diagram is shown below. Figure 5 As shown, it includes: The sensing correction unit 1 is used to collect the multi-source sensing operation signals corresponding to each membrane element, and correct the real-time sensing characteristics based on the characteristic distortion caused by hydraulic coupling to obtain the corrected sensing characteristics. Based on the corrected sensing characteristics, the operating status of each membrane element is continuously monitored. The prior constraint unit 2 is used to isolate the abnormal source membrane element when multiple membrane elements simultaneously exhibit abnormalities, and to determine whether the abnormality is a false abnormality induced by hydraulic coupling or a real abnormality caused by membrane element deterioration based on the state changes of the remaining membrane elements before and after isolation, and to generate prior constraint parameters based on the determination results. The first determination unit 3 is used to adjust the hydraulic impedance threshold according to the prior constraint parameters, use the pressure step generated by the flow path switching as the excitation, combine the correction sensing characteristics to identify the hydraulic impedance parameters of each membrane element, identify the type of fouling based on the hydraulic impedance parameters and the hydraulic impedance threshold, and obtain the fouling type result. The second determination unit 4 is used to adjust the damage threshold of structural integrity damage according to the prior constraint parameters, collect pressure decay signals using the device shutdown or reversing gap and extract the decay time constant of each membrane element from it, and obtain the integrity damage result of each membrane element according to the decay time constant and the damage threshold. The fusion prediction unit 5 is used to fuse the contamination type results and the integrity damage results, and output the fault prediction results.

[0089] Those skilled in the art will understand that the components of this application described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage system for execution by the computing system. Alternatively, they can be fabricated as separate integrated circuit components, or multiple components or steps can be fabricated as a single integrated circuit component. Thus, this application is not limited to any particular combination of hardware and software.

[0090] Note that the above description is merely a preferred embodiment and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this application, and the scope of this application is determined by the scope of the appended claims.

[0091] The above disclosures are only a few specific implementation scenarios of this application. However, this application is not limited to these. Any variations that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for predicting the failure of a bidirectional flow membrane filtration device, characterized in that, A bidirectional flow membrane filtration device applicable to switchable forward and reverse operating states, the method comprising: Multi-source sensor operation signals corresponding to each membrane element are collected, and the real-time sensor characteristics are corrected according to the characteristic distortion caused by hydraulic coupling to obtain corrected sensor characteristics. The operating status of each membrane element is continuously monitored based on the corrected sensor characteristics. When multiple membrane elements simultaneously exhibit abnormalities, the abnormality source membrane element is isolated, and based on the state changes of the remaining membrane elements before and after isolation, it is determined whether the abnormality is a false abnormality induced by hydraulic coupling or a real abnormality caused by membrane element deterioration, and prior constraint parameters are generated based on the determination results. The hydraulic impedance threshold is adjusted according to the prior constraint parameters. The pressure step generated by the flow path switching is used as the excitation. The hydraulic impedance parameters of each membrane element are identified in combination with the correction sensing features. The type of fouling is identified based on the hydraulic impedance parameters and the hydraulic impedance threshold to obtain the fouling type result. The damage threshold for structural integrity damage is adjusted according to the prior constraint parameters. Pressure decay signals are collected using the device shutdown or reversing gap, and the decay time constant of each membrane element is extracted from them. The integrity damage results of each membrane element are obtained based on the decay time constant and the damage threshold. By combining the pollution type results and the integrity damage results, a fault prediction result is output.

2. The fault prediction method according to claim 1, characterized in that, The acquisition of the pollution type results specifically includes: The pollution type identified based on the hydraulic impedance parameters and the hydraulic impedance threshold is taken as the first result; the operating data of the device under forward and reverse operating conditions are collected respectively, and a bidirectional asymmetric feature is constructed based on the operating data. The bidirectional asymmetric feature is the ratio of the performance recovery of the device under forward flushing condition to the performance recovery under reverse flushing condition; the corresponding pollution type is identified according to the bidirectional asymmetric feature to obtain the second result; the pollution types in the first result and the second result are cross-validated to obtain the pollution type result.

3. The fault prediction method according to claim 1, characterized in that, The method also includes: The flow balance threshold is adjusted according to the prior constraint parameters; the flow distribution data and resistance data of each membrane element are obtained based on the correction sensor signal; the current flow distribution balance of the device is determined according to the flow distribution data of each membrane element; when the flow distribution balance is lower than the flow balance threshold, the resistance change of each membrane element is compared with the flow distribution change of the membrane element to analyze whether the device is deteriorating and obtain the deterioration result. The operational degradation results are then integrated into the fault prediction results.

4. The fault prediction method according to claim 1, characterized in that, When the abnormal state of the remaining membrane elements disappears synchronously after the abnormal source membrane element is isolated, the abnormality is determined to be a false abnormality caused by the abnormal source membrane element being hydraulically coupled to the remaining membrane elements through parallel pipelines. If the abnormal state of the remaining membrane elements persists after isolation, the abnormality is determined to be a real abnormality caused by the deterioration of multiple membrane elements themselves.

5. The fault prediction method according to claim 1, characterized in that, The process of acquiring the corrected sensing features includes: isolating each membrane element for a short period of time according to a preset cycle, collecting the sensing signals of each membrane element in the isolated state without being affected by the hydraulic coupling of other membrane elements as reference sensing features; after resuming parallel operation, comparing the real-time sensing signals with the reference sensing features, and calculating the feature distortion variable; and correcting the real-time sensing features based on the feature distortion variable.

6. The fault prediction method according to claim 3, characterized in that, When the prior constraint parameters indicate that the anomaly is a pseudo-anomaly induced by hydraulic coupling, the hydraulic impedance threshold, the damage threshold, and the flow balance threshold are increased; when the prior constraint parameters indicate that the anomaly is a real anomaly caused by membrane element degradation, the hydraulic impedance threshold, the damage threshold, and the flow balance threshold are maintained or decreased.

7. The fault prediction method according to claim 1, characterized in that, When the reversing valve group of the bidirectional flow membrane filtration device performs flow path switching, the pressure step generated by the flow path switching is used as the original excitation signal, and the pressure response waveforms at both ends of each membrane element are collected; the pressure response waveforms are subjected to system identification, and the equivalent hydraulic impedance parameters of each membrane element in the frequency domain are extracted; the hydraulic impedance parameters include equivalent resistance, equivalent capacitive reactance and equivalent inductive reactance. The method of identifying the type of contamination based on the hydraulic impedance parameters and hydraulic impedance thresholds includes: when the equivalent resistance exceeds the hydraulic impedance threshold by a greater margin than the equivalent capacitive reactance and the equivalent inductive reactance each exceed the hydraulic impedance threshold, it is identified as surface filter cake layer contamination; when the equivalent capacitive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and the equivalent inductive reactance each exceed the hydraulic impedance threshold, it is identified as membrane pore blockage contamination; when the equivalent inductive reactance exceeds the hydraulic impedance threshold by a greater margin than the equivalent resistance and the equivalent capacitive reactance each exceed the hydraulic impedance threshold, it is identified as membrane surface adsorption contamination.

8. The fault prediction method according to claim 1, characterized in that, During the normal shutdown of the bidirectional flow membrane filtration device or during the pressure decay period after the flow path switching of the reversing valve group, the inlet and outlet valves of each membrane element are closed, and the pressure decay signal at both ends of each membrane element over time is collected at a preset sampling frequency; the decay signal is curve fitted to extract the pressure decay time constant of each membrane element. The decay time constants of each membrane element are compared laterally, and the deviation ratio between the decay time constant of each membrane element and the group mean is calculated. When the proportion of a membrane element whose decay time constant is lower than the group mean exceeds the damage threshold, it is determined that the membrane element has integrity damage.

9. The fault prediction method according to claim 1, characterized in that, The fault prediction results include: when the contamination type result indicates contamination of a single membrane element and the integrity damage result indicates that the membrane element has no structural damage, a single membrane contamination warning and corresponding cleaning strategy suggestions are output; when the integrity damage result indicates that any membrane element has structural integrity damage, an emergency warning for structural damage is output; when multiple membrane elements are contaminated at the same time and the prior constraint parameters characterize the anomaly as a real anomaly caused by membrane element degradation, an overall membrane element degradation warning is output.

10. A fault prediction system for a bidirectional flow membrane filtration device, characterized in that, A bidirectional flow membrane filtration device applicable to switchable forward and reverse operating states, the system includes: The sensing correction unit is used to collect multi-source sensing operation signals corresponding to each membrane element, and correct the real-time sensing features according to the characteristic distortion caused by hydraulic coupling to obtain corrected sensing features. Based on the corrected sensing features, the operating status of each membrane element is continuously monitored. The prior constraint unit is used to isolate the anomaly source membrane element when multiple membrane elements simultaneously exhibit anomalies, and based on the state changes of the remaining membrane elements before and after isolation, determine whether the anomaly is a false anomaly induced by hydraulic coupling or a real anomaly caused by membrane element deterioration, and generate prior constraint parameters based on the determination result. The first determination unit is used to adjust the hydraulic impedance threshold according to the prior constraint parameters, use the pressure step generated by the flow path switching as the excitation, identify the hydraulic impedance parameters of each membrane element in combination with the correction sensing features, identify the pollution type based on the hydraulic impedance parameters and the hydraulic impedance threshold, and obtain the pollution type result. The second determination unit is used to adjust the damage threshold of structural integrity damage according to the prior constraint parameters, collect pressure decay signals using device shutdown or reversing gaps and extract the decay time constant of each membrane element from them, and obtain the integrity damage result of each membrane element according to the decay time constant and the damage threshold. The fusion prediction unit is used to fuse the pollution type results and the integrity damage results to output the fault prediction results.

Images