A membrane pollution type intelligent diagnosis and grading cleaning decision method and system

Abstract

The application provides a membrane pollution type intelligent diagnosis and grading cleaning decision method and system, relates to the technical field of water treatment, and the technical scheme is characterized in that the operation pressure and the multi-dimensional data of the water quality of the membrane system are collected in real time, the characteristic indexes representing colloidal, scaling and organic biological compound pollution are calculated by deep fusion, and the contribution of each pollution type is quantified to accurately determine the current dominant pollution situation. This mechanism breaks the limitations of traditional single threshold or black box prediction, and can generate targeted grading cleaning decision instructions according to the dominant pollution type when the transmembrane pressure difference or the normalized water flux triggers the cleaning condition. This not only realizes the transformation from passive response to active and accurate diagnosis and treatment, effectively avoids the high cost and membrane material damage risk caused by blind cleaning, but also significantly improves the intelligent level of membrane system operation and maintenance under complex working conditions and the stability of long-term operation.

Description

Technical Field

[0001] This application relates to the field of water treatment technology, and in particular to a method and system for intelligent diagnosis and graded cleaning decision-making of membrane fouling types. Background Technology

[0002] In the field of water treatment, membrane separation technologies such as ultrafiltration and microfiltration are core processes, but their widespread application is limited by membrane fouling. Existing technologies generally suffer from diagnostic incompetence and a disconnect between decision-making and the coexistence of multiple fouling mechanisms under real-world operating conditions. Current mainstream solutions include alarm cleaning based on fixed transmembrane pressure thresholds, fouling prediction and early warning based on neural networks or time series analysis, and fouling classification methods based on machine learning. However, none of these solutions effectively address the challenge of precise treatment of complex fouling. Threshold-based solutions cannot distinguish between fouling types, leading to indiscriminate cleaning; predictive solutions can only indicate impending performance deterioration but cannot explain the causes and types of fouling; and classification solutions are mostly black-box models, lacking interpretability and failing to characterize the primary and secondary relationships of multiple coexisting fouling factors. Furthermore, most existing technologies are limited to a single dimension of membrane performance parameters, lacking a systematic correlation with fouling drivers such as influent water quality. This results in a break in the chain from condition warnings to precise cleaning decisions. Maintenance personnel either perform ineffective single cleaning or blindly adopt aggressive full-scale chemical cleaning, which is not only costly but also prone to damaging membrane materials. Summary of the Invention

[0003] This application provides a method and system for intelligent diagnosis and graded cleaning decision-making of membrane fouling types to solve one or more technical problems existing in the prior art, and at least provides a beneficial option or creates conditions that can accurately identify the primary and secondary relationships of various fouling types such as colloids, scaling, organic and biological fouling through the fusion analysis of multi-dimensional data on membrane system operating pressure and influent water quality, and generate graded and orderly cleaning decisions accordingly, effectively improving cleaning efficiency and reducing operation and maintenance costs and membrane damage risks.

[0004] On the one hand, this application provides a method for intelligent diagnosis and graded cleaning decision-making of membrane fouling types, including the following steps: Real-time acquisition of membrane system operating pressure data and multi-dimensional influent water quality data; Based on the operating pressure data and multidimensional influent water quality data, characteristic indicators are calculated and obtained. The characteristic indicators include: turbidity gradient characterizing colloidal pollution, scaling risk index characterizing scaling risk, and organic pollution coefficient characterizing organic and biological composite pollution. Based on the aforementioned characteristic indicators, contribution scores for multiple pollution types are calculated to determine the current dominant pollution situation; wherein, the contribution scores for the pollution types include colloidal pollution contribution scores, scaling pollution contribution scores, organic pollution contribution scores, and biological pollution contribution scores; When the transmembrane pressure difference determined based on the operating pressure data, or the standardized permeate flux determined based on the multidimensional influent water quality data, meets the preset cleaning triggering conditions, a graded cleaning decision instruction is generated based on the dominant contamination situation, and the execution unit is controlled to implement the corresponding cleaning procedure.

[0005] Further, the turbidity abrupt change gradient and the contribution of colloidal contamination are calculated, specifically including: The difference between the current average and historical average of influent turbidity is calculated to obtain the turbidity abrupt change gradient; Obtain the instantaneous rate of rise of the transmembrane pressure difference; The product of the turbidity abrupt change gradient and the instantaneous rise rate is calculated, and combined with a preset effective threshold for the turbidity gradient, the colloidal contamination contribution score is calculated through a linear mapping function.

[0006] Further, the scaling risk index and the scaling fouling contribution score are calculated, specifically including: Based on the deviation of the current values ​​of influent pH and conductivity from the baseline values, the scaling risk index (SRI) is calculated, and its expression is: ; in, This represents the current pH value of the influent. This is the baseline value of the influent pH. This represents the current value of electrical conductivity. This is the baseline value for conductivity; This is the pH sensitivity coefficient. The conductivity sensitivity coefficient, The weighting factor is preset. Based on the numerical range of the Scaling Risk Index (SRI), the scaling pollution contribution score is calculated through a preset mapping relationship.

[0007] Further, the organic pollution coefficient, organic pollution contribution score, and biological pollution contribution score are calculated, specifically including: The fluorescence index of the influent was measured online using an ultraviolet fluorescence analyzer. and bioactivity index And combined with the instantaneous rate of rise of the transmembrane pressure difference Calculate the organic pollution coefficient Its expression is: ; in, and The preset weighting coefficients, and ; When the organic pollution coefficient does not exceed a set threshold, the organic pollution contribution score is calculated through a preset mapping relationship; When the organic pollution coefficient is greater than the set threshold, the instantaneous rise rate of the transmembrane pressure difference is linearly fitted using the recursive least squares method to calculate its rate of change. If the rate of change is greater than a preset first threshold, the biological activity is determined to be significant, and a biological activity bonus is calculated; the organic pollution base score is calculated and summed with the biological activity bonus to obtain the biological pollution contribution score; If the rate of change is less than or equal to the first threshold, the contribution of biological pollution is 0.

[0008] Furthermore, the current dominant pollution situation should be identified, specifically including the following situations: (1) If the contribution score of colloidal pollution exceeds the first set threshold, and the difference between the contribution score of colloidal pollution and the contribution score of organic pollution and the contribution score of biological pollution are both greater than the first preset difference, then it is determined that colloidal pollution is dominant. (2) If the contribution score of scaling pollution exceeds the first set threshold, and the difference between the contribution score of scaling pollution and the contribution score of organic pollution and the contribution score of biological pollution are both greater than the first preset difference, then it is determined that scaling pollution is dominant. (3) When the contribution score of organic pollution exceeds the second set threshold, if the contribution score of biological pollution is lower than the third set threshold, it is determined that organic pollution is dominant; if the contribution score of biological pollution exceeds the third set threshold, it is determined that biological pollution is dominant. (4) If the contribution of colloidal pollution or the contribution of scaling pollution exceeds the first set threshold, and the contribution of organic pollution or the contribution of biological pollution exceeds the second set threshold, then it is determined to be a compound pollution.

[0009] Furthermore, a tiered cleaning decision instruction is generated based on the dominant pollution situation, specifically including: Based on the identified dominant pollution situation, determine the cleaning agent, cleaning intensity coefficient, and cleaning method; The concentration of the cleaning agent, the cleaning cycle time, and the soaking time are adjusted based on the cleaning intensity coefficient; wherein the cleaning intensity coefficient has a positive correlation with the pollution contribution score.

[0010] Furthermore, the method includes determining the endpoint of the cleaning process: After the soaking stage, the changes in turbidity, conductivity, or pH of the cleaning solution are measured. If the increase in turbidity of the cleaning solution is greater than 20 NTU, or the increase in conductivity is greater than 500 NTU. If the pH value drops by more than 0.5, the cleaning reaction is considered complete, and the rinsing stage begins.

[0011] Furthermore, the cleaning triggering conditions specifically include: Relative pressure threshold trigger: Current transmembrane pressure differential ≥ Clean baseline pressure differential + 80 kPa; Absolute pressure red line trigger: Current transmembrane pressure difference ≥ 90% of the maximum allowable transmembrane pressure difference of the membrane module; Performance degradation trigger: Current standardized permeate flux ≤ 70% of initial clean flux.

[0012] On the other hand, this application provides an intelligent diagnostic and graded cleaning decision system for membrane fouling types, including: The sensing layer includes a pressure sensor, an online analyzer for turbidity and suspended solids, an ultraviolet fluorescence organic matter analyzer, an electromagnetic flowmeter, and integrated sensors for pH, conductivity, and temperature. The processing and control layer includes an edge computing industrial controller with a built-in diagnostic module based on recursive least squares and a rule engine. The execution layer includes a cleaning agent storage tank assembly, a metering pump, a pneumatic valve assembly, an electric valve assembly, and a pipeline mixer; the cleaning agent storage tank assembly includes tanks for storing acidic, alkaline, or oxidizing cleaning solutions; The edge computing controller outputs control signals based on the diagnostic results to adjust the frequency of the metering pump and the switching state of the valve group in order to execute the composite cleaning program.

[0013] Furthermore, the processing and control layer includes a memory that stores a computer program. When the computer program is executed by a processor, it implements the aforementioned intelligent diagnosis and graded cleaning decision-making method for membrane fouling types.

[0014] The beneficial effects of this application are as follows: This application provides a smart diagnostic and graded cleaning decision-making method for membrane fouling types. This technical solution collects multi-dimensional data on the operating pressure and influent water quality of the membrane system in real time, deeply integrates and calculates characteristic indicators representing colloidal, scaling, and organic biological composite fouling, and further quantifies the contribution of each fouling type to accurately identify the current dominant fouling situation. This mechanism breaks through the limitations of traditional single threshold or black-box prediction, and can generate targeted graded cleaning decision instructions based on the clearly defined dominant fouling type when cleaning conditions are triggered by transmembrane pressure difference or standardized permeate flux. This not only realizes the transformation from passive response to proactive and precise diagnosis, effectively avoiding the high costs and membrane material damage risks caused by blind cleaning, but also significantly improves the intelligence level of membrane system operation and maintenance and the long-term stability of operation under complex conditions. This application also provides a system corresponding to the above method; the beneficial effects of the system are similar to those of the method and will not be elaborated here.

[0015] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0016] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0017] Figure 1 This is a flowchart of the intelligent diagnosis and graded cleaning decision-making method for membrane fouling types provided in this application; Figure 2 This is a structural diagram of the intelligent diagnosis and graded cleaning decision system for membrane fouling types provided in this application; Figure 3 This is a flowchart illustrating the data acquisition and preprocessing process provided in this application; Figure 4 This is a flowchart illustrating the intelligent diagnosis process for membrane fouling types provided in this application. Figure 5 This is a flowchart illustrating the hierarchical cleaning decision generation and execution process provided in this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0020] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] While membrane separation technology has become a key process in water treatment, the decrease in flux and increase in transmembrane pressure caused by membrane fouling remain the core bottlenecks restricting its large-scale application. This not only significantly increases the system's energy consumption and operating costs, but frequent cleaning also shortens membrane life. Therefore, how to accurately diagnose the type of fouling and take targeted cleaning strategies has become the primary challenge facing the operation and maintenance of membrane technology.

[0023] Currently, the industry mainly relies on three technical approaches for monitoring and cleaning decisions regarding membrane fouling: first, passive cleaning based on fixed thresholds, which triggers cleaning when parameters such as transmembrane pressure difference exceed preset values; second, data-driven predictive early warning, which uses artificial intelligence algorithms to model operational data to predict flux decline trends; and third, machine learning-based fouling type classification, which identifies single fouling types by analyzing operational parameters or spectral images.

[0024] However, the aforementioned technologies have significant limitations in addressing the complex fouling prevalent in real-world operating conditions. Firstly, membrane fouling is typically caused by the combined effects of multiple contaminants. Traditional threshold methods lack diagnostic capabilities, and most classification models struggle to accurately characterize the primary and secondary relationships of complex fouling, leading to diagnostic results that significantly deviate from reality. Secondly, existing technologies generally suffer from a disconnect between diagnosis and decision-making. Whether it's fixed-process cleaning based on thresholds or predictive solutions that only indicate cleaning timing, neither provides targeted cleaning guidance. Blindly cleaning not only wastes chemicals but also causes irreversible damage to membrane materials. Furthermore, many data-driven models lack interpretability and often overlook the fundamental driving force of influent water quality, making it difficult to accurately attribute causes at the mechanistic level, thus limiting their reliability and generalization ability under complex and variable operating conditions.

[0025] To address the aforementioned issues, this application provides a method and system for intelligent diagnosis and graded cleaning decision-making of membrane fouling types. It constructs an interpretable composite membrane fouling diagnosis and cleaning decision-making system that integrates mechanistic characteristics and data-driven approaches. This system overcomes the limitations of monitoring single operating parameters, collecting multi-dimensional data on membrane system operating pressure and influent water quality in real time. By calculating characteristic indicators with clear biochemical significance, such as turbidity gradient, scaling risk index, and organic fouling coefficient, it accurately quantifies the contribution of colloidal, inorganic, organic, and biological fouling in composite fouling. This intelligently identifies the dominant fouling type and automatically generates targeted graded cleaning decision instructions when cleaning trigger conditions are met, realizing a shift from passive, blind cleaning to precise diagnosis and treatment based on the causes of fouling.

[0026] First, the intelligent diagnosis and graded cleaning decision method for membrane fouling types provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0027] Reference Figure 1The implementation process of the intelligent diagnosis and graded cleaning decision method for membrane fouling types provided in this application embodiment includes, but is not limited to, the following steps.

[0028] Step S110: Real-time acquisition of membrane system operating pressure data and multi-dimensional data of influent water quality.

[0029] It should be noted that operating pressure data refers to the pressure values ​​at various key nodes of the membrane system during the filtration process, including the feed water side, concentrate side, and product water side. These data directly reflect the resistance encountered by water flowing through the membrane and are the most intuitive parameters for judging whether the membrane has become clogged or fouled. Multidimensional feed water quality data refers to a set of various physicochemical indicators in the raw water before it enters the membrane system, reflecting its potential for fouling. This includes not only conventional turbidity (suspended solids content), pH value, and temperature, but also calcium and magnesium ion concentration (hardness), total dissolved solids (TDS), and organic matter content (such as COD).

[0030] In step S110, a comprehensive data capture network is constructed. By synchronously collecting the operating pressure data of the membrane system and multi-dimensional data of the influent water quality in real time, the traditional monitoring is limited to the lag of membrane performance parameters. The monitoring perspective is moved forward to the source of pollution, providing indispensable underlying data support for the subsequent accurate analysis of the causes of compound pollution.

[0031] Step S120: Based on the operating pressure data and multidimensional influent water quality data, calculate and obtain characteristic indicators.

[0032] Among them, the characteristic indicators include: turbidity gradient characterizing colloidal pollution, scaling risk index characterizing scaling risk, and organic pollution coefficient characterizing organic and biological combined pollution.

[0033] The turbidity gradient is a dynamic indicator specifically designed to detect the risk of colloidal or particulate contamination. It is obtained by calculating the difference between the current average turbidity of the influent and the historical average turbidity under normal conditions. Compared to simple instantaneous turbidity values, this gradient more sensitively reflects whether there has been a sudden deterioration in the influent quality (such as sediment impact after heavy rain or illegal industrial wastewater discharge), thus providing early warning of the risk of blockage.

[0034] The scaling risk index typically refers to the Langelier Saturation Index (LSI). It is a dimensionless value calculated based on parameters such as the actual pH, calcium hardness, total alkalinity, temperature, and salinity of the water. This index is used to quantitatively assess whether sparingly soluble salts such as calcium carbonate in water are in a supersaturated state. If the index is positive and high, it indicates a strong tendency for crystallization and precipitation in the water, making it highly susceptible to forming a hard inorganic scale layer on the membrane surface.

[0035] The organic pollution coefficient is a comprehensive indicator that characterizes the tendency of organic matter and biological pollution. Because natural organic matter, proteins, or polysaccharides in water are easily adsorbed onto the membrane surface to form a gel layer and provide nutrients for microbial reproduction (leading to biological slime), this coefficient quantifies the risk level of this complex pollution by analyzing relevant organic indicators in the influent.

[0036] In step S120, a series of key characteristic indicators are calculated based on the collected multidimensional data. Specifically, these include using the turbidity gradient to keenly capture the dynamic changes of colloidal pollution, quantifying the potential threat of inorganic salt precipitation through the scaling risk index, and using the organic pollution coefficient to comprehensively characterize the combined effects of organic matter and microorganisms, thereby interpreting abstract data fluctuations into interpretable pollution signals.

[0037] Step S130: Based on the feature indicators, calculate the contribution scores of multiple pollution types respectively, and determine the current dominant pollution situation.

[0038] The contribution score for pollution type includes contribution score for colloidal pollution, contribution score for scaling pollution, contribution score for organic pollution, and contribution score for biological pollution.

[0039] In the diagnosis of complex contamination, multiple contaminants often coexist. The contribution score is the result of quantifying and scoring the above-mentioned characteristic indicators through an algorithm. It is used to determine whether colloidal, inorganic scale, organic matter, or biological contamination is dominant at the current stage, i.e., which is the main cause of membrane performance degradation.

[0040] In step S130, by introducing a multi-dimensional weighted scoring mechanism, the technical problem of multiple pollutants coexisting under complex working conditions and the difficulty in distinguishing the primary and secondary pollutants is effectively solved. This step calculates the contribution scores of colloidal, scaling, organic and biological pollutants based on the aforementioned characteristic indicators, thereby accurately identifying the dominant pollutant situation that leads to the decline in membrane performance among many interfering factors, and realizing the leap from vague qualitative judgment to accurate quantitative attribution diagnosis.

[0041] Step S140: When the transmembrane pressure difference determined based on the operating pressure data, or the standardized permeate flux determined based on the multidimensional data of the influent water quality, meets the preset cleaning triggering conditions, a graded cleaning decision instruction is generated based on the dominant pollution situation, and the execution unit is controlled to implement the corresponding cleaning procedure.

[0042] Transmembrane pressure (TMP) refers to the net pressure difference required to drive water molecules through the membrane, typically calculated as the difference between the average pressure on the feed side and the pressure on the product side. It is one of the most critical operating parameters in membrane separation technology. When membrane pores are clogged with contaminants or a filter cake layer forms on the membrane surface, water flow resistance increases. To maintain the same product flow rate, the driving pressure must be increased, leading to a significant rise in transmembrane pressure. Therefore, an abnormal increase in TMP is the most direct signal of membrane fouling.

[0043] Standardized permeate flux refers to the water production capacity of a membrane under actual operating temperature and pressure, converted mathematically to standard conditions (usually 25°C). Because water viscosity changes with temperature (viscosity increases at lower temperatures, naturally reducing permeate flow), directly comparing permeate flow across different seasons can lead to misjudgments. Standardization eliminates the interference of environmental factors such as temperature, accurately reflecting whether membrane performance has irreversibly declined due to fouling.

[0044] In step S140, based on the determination of the dominant fouling situation, the system generates targeted, graded cleaning decision instructions, driving the execution unit to implement a cleaning plan with specific agents and processes. This "targeted approach," such as using acid washing for scaling and alkaline washing or oxidant sterilization for organic and biological fouling, effectively avoids the drawbacks of traditional "one-size-fits-all" indiscriminate cleaning. While ensuring efficient recovery of membrane flux, it minimizes the consumption of chemical agents and the risk of membrane material damage.

[0045] In some embodiments of this application, the calculation of turbidity abrupt gradient and colloidal contamination contribution includes the following steps.

[0046] Step S210: Calculate the difference between the current average value and the historical average value of the influent turbidity to obtain the turbidity abrupt change gradient.

[0047] In step S210, the dynamic abrupt change signal of colloidal fouling in the influent water quality is keenly captured. The turbidity abrupt change gradient is obtained by calculating the difference between the current average turbidity of the influent and the average under historical normal conditions. This process can effectively filter out the normal background fluctuations in water quality and accurately identify sudden suspended particles or colloidal load shocks that may cause rapid clogging of the membrane surface (such as sediment inflow after heavy rain), thus providing highly forward-looking source data for subsequent pollution early warning.

[0048] Step S220: Obtain the instantaneous rate of increase of the transmembrane pressure difference.

[0049] In step S220, the instantaneous hydraulic resistance change caused by fouling accumulation in the membrane system is quantified. By obtaining the instantaneous rate of increase in the transmembrane pressure difference, the urgency of membrane pore blockage or filter cake thickening on the membrane surface can be directly reflected. This parameter, as a direct characterization of membrane performance degradation, reveals whether pollutants have substantially increased the resistance of water flow through the membrane, providing crucial pressure-side evidence for determining the actual impact of colloidal fouling on system operation.

[0050] Step S230: Calculate the product of the turbidity abrupt change gradient and the instantaneous rise rate, and combine it with the preset effective threshold of turbidity gradient to calculate the colloidal pollution contribution score through a linear mapping function.

[0051] In step S230, a deep fusion and quantitative assessment of the pollution driving force and system response results are achieved. This is achieved by multiplying the abrupt turbidity gradient, representing the pollution risk on the influent side, with the instantaneous rate of increase representing the change in membrane resistance, and then standardizing the result using a linear mapping function based on a preset effective threshold for the turbidity gradient. This comprehensive algorithm not only gives the diagnostic results clear meaning but also greatly improves the accuracy and interpretability of the system in identifying colloidal-dominated pollution under complex operating conditions, ensuring a scientific basis for subsequent cleaning decisions.

[0052] In some embodiments of this application, the calculation of the scaling risk index and scaling pollution contribution score specifically includes the following steps.

[0053] Step S310: Calculate the scaling risk index SRI based on the deviation of the current values ​​of influent pH and conductivity from the baseline values.

[0054] In step S310, the potential risk of inorganic salt crystallization on the membrane surface is precisely quantified from the perspective of chemical equilibrium. This step derives the scaling risk index (SRI) by real-time monitoring of the pH and conductivity of the influent and calculating the deviation of the current value from the historical normal baseline value. Since an increase in pH significantly increases the supersaturation of sparingly soluble salts such as calcium carbonate, and abnormal fluctuations in conductivity reflect the concentration of total dissolved solids in the water, this process can effectively capture the trend of water chemistry evolving towards scaling, providing crucial chemothermodynamic basis for predicting the formation of inorganic scale.

[0055] Step S320: Calculate the scaling pollution contribution score based on the numerical range of the scaling risk index (SRI) through a preset mapping relationship.

[0056] In step S320, the abstract chemical risk index is transformed into a quantifiable pollution assessment indicator, thereby enabling the classification and determination of the severity of scaling. This step converts the calculated SRI (Scaling Risk Index) into a corresponding scaling pollution contribution score using a pre-defined mapping relationship, based on the specific numerical range of the calculated SRI. This mapping mechanism not only intuitively reflects the proportion of inorganic deposition's impact on membrane system performance under current operating conditions, but also provides scientific and standardized data support for subsequent accurate identification of the dominant pollution situation and the development of targeted acid washing or scale inhibition strategies.

[0057] In some embodiments of this application, the scaling risk index SRI is calculated based on the deviation of the current values ​​of influent pH and conductivity from the baseline values. The expression for SRI is as follows: ; in, This represents the current pH value of the influent. This is the baseline value of the influent pH. This represents the current value of electrical conductivity. This is the baseline value for conductivity; This is the pH sensitivity coefficient. The conductivity sensitivity coefficient, This is a preset weighting factor.

[0058] By introducing a weighted calculation mechanism, a scale risk quantification model was constructed that can dynamically reflect the changing trends of water chemical properties. This model accurately captures water quality fluctuations by utilizing the deviation between current values ​​and baseline values. This combination of difference and ratio calculations makes the scale risk index a finely calibrated comprehensive evaluation indicator. It can keenly identify the trend of water bodies evolving towards a supersaturated state from a thermodynamic perspective, effectively eliminate environmental noise interference, and truly reflect the magnitude of the chemical driving force for inorganic matter deposition under current operating conditions, thereby ensuring the robustness and accuracy of the model under different water quality conditions.

[0059] In some embodiments of this application, the calculation of the organic pollution coefficient, organic pollution contribution score, and biological pollution contribution score specifically includes the following steps.

[0060] In step S410, the fluorescence index and biological activity index of the influent are measured online using an ultraviolet fluorescence analyzer, and the organic pollution coefficient is calculated by combining the instantaneous rise rate of the transmembrane pressure difference.

[0061] In step S410, a comprehensive pollution assessment index is constructed that can simultaneously reflect the content of organic matter and the potential for biological activity in the water. This algorithm, which combines the optical properties of water quality with the hydraulic performance response, can effectively distinguish between simple organic matter adsorption and complex pollution accompanied by microbial growth, providing a key preliminary criterion for subsequent accurate identification of pollution types.

[0062] Step S420: When the organic pollution coefficient does not exceed the set threshold, the organic pollution contribution score is calculated through a preset mapping relationship.

[0063] In step S420, for operating conditions where organic matter adsorption is the primary factor and biological contamination is not yet significant, the impact of organic contamination on system performance is quantified. This process ensures that, within the low biological risk range, the diagnostic results accurately reflect the severity of organic matter clogging membrane pores or forming a gel layer, providing a direct basis for subsequent strategies based on alkaline washing or surfactant cleaning.

[0064] Step S430: When the organic pollution coefficient is greater than the set threshold, the instantaneous rise rate of the transmembrane pressure difference is linearly fitted using the recursive least squares method to calculate its rate of change.

[0065] In step S430, the underlying biofouling characteristics in the transmembrane pressure difference variation trend are explored to address potential biofouling problems under high organic loads. Compared to simple instantaneous values, this trend fitting based on historical data can effectively filter out noise interference from hydraulic fluctuations and keenly capture the signal of accelerated resistance increase caused by biofilm thickening, providing mathematical support for judging the activity level of biofouling.

[0066] In step S440, if the rate of change is greater than the preset first threshold, the biological activity is determined to be significant, and the biological activity bonus is calculated; the organic pollution base score is calculated and summed with the biological activity bonus to obtain the biological pollution contribution score.

[0067] In step S440, this superposition calculation method fully considers the fundamental role of organic matter as a nutrient source for microorganisms and the additional resistance contribution brought by biological activity, so that the final contribution score can not only cover the background of organic pollution, but also highlight the explosive power of biological pollution, thereby ensuring that the system can generate comprehensive decision instructions including sterilization and disinfection and deep cleaning when facing complex organic-biological composite pollution.

[0068] In step S450, if the rate of change is less than or equal to the first threshold, the contribution of biological pollution is 0.

[0069] In step S450, a dynamic threshold determination mechanism is introduced to achieve graded quantification of the severity of biocontamination. This step compares the rate of change calculated in the previous step with a preset first threshold: if the rate of change is greater than the first threshold, it indicates that the biological activity on the membrane surface is significantly enhanced, and microorganisms are rapidly proliferating and forming a sticky filter cake layer; the system will calculate a biological activity bonus accordingly. Conversely, if the rate of change is less than or equal to the first threshold, it is considered that biological growth is under control or slow, and the contribution of biocontamination is 0. This step successfully transforms the abstract concept of biological activity into a calculable numerical increment, demonstrating the system's accurate identification capability of the explosive characteristics of biocontamination.

[0070] In some embodiments of this application, in step S410, the fluorescence index of the influent is measured online using an ultraviolet fluorescence analyzer. and bioactivity index And combined with the instantaneous rate of rise of the transmembrane pressure difference Calculate the organic pollution coefficient Its expression is: ;in, and The preset weighting coefficients, and .

[0071] This step aims to construct a comprehensive pollution assessment index (i.e., the organic pollution coefficient) that simultaneously reflects the organic matter content and biological activity potential in water. This index uses an online ultraviolet fluorescence analyzer to acquire the fluorescence index and biological activity index of the influent, representing the dissolved organic matter concentration and microbial metabolic intensity, respectively. It is then combined with the instantaneous rise rate of the transmembrane pressure difference, reflecting the actual change in membrane surface resistance, for comprehensive calculation. This algorithm, which combines the optical properties of water quality with hydraulic performance response, can effectively distinguish between simple organic matter adsorption and complex pollution accompanied by microbial growth, providing a crucial preliminary criterion for subsequent accurate identification of pollution types.

[0072] In some embodiments of this application, step S130, determining the current dominant pollution situation, specifically includes the following situations: (1) If the contribution score of colloidal pollution exceeds the first set threshold, and the difference between it and the contribution scores of organic pollution and biological pollution is greater than the first preset difference, then it is determined that colloidal pollution is dominant. This logic excludes the possibility that other types of pollution may coexist significantly, thus confirming the absolute dominance of colloidal pollution under the current working conditions, and providing a clear basis for subsequent targeted cleaning.

[0073] (2) If the contribution score of scaling pollution exceeds the first set threshold, and the difference between it and the contribution scores of organic pollution and biological pollution is greater than the first preset difference, then it is determined that scaling pollution is dominant, that is, the current pollution is not caused by organic matter or microorganisms, but is mainly caused by scaling. This accurate identification is crucial for guiding the subsequent use of acidic cleaning agents for descaling, and can avoid blindly using oxidants to damage the membrane.

[0074] (3) When the contribution score of organic pollution exceeds the second set threshold, if the contribution score of biological pollution is lower than the third set threshold, it is determined that organic pollution is dominant; if the contribution score of biological pollution exceeds the third set threshold, it is determined that biological pollution is dominant.

[0075] Specifically, when the organic contamination contribution score exceeds the second set threshold, it indicates that organic matter has accumulated in large quantities on the membrane surface. At this point, the system further considers the biological contamination contribution score: if it is below the third set threshold, it indicates weak biological activity, and the contamination mainly consists of inactive organic matter, thus classifying it as organic contamination-dominated; conversely, if the biological contamination contribution score also exceeds the third set threshold, it indicates active microbial reproduction, and the contamination is mainly biological slime, thus classifying it as biological contamination-dominated. This distinction is a key prerequisite for achieving precise dosing of cleaning agents (such as alkaline washing or sterilization).

[0076] (4) If the contribution of colloidal fouling or scaling fouling exceeds a first set threshold, and the contribution of organic fouling or biological fouling exceeds a second set threshold, then it is determined to be a complex fouling. This determination logic reveals that the membrane system is suffering from a dual attack from contaminants of different properties. Identifying such complex fouling is crucial because it means that a single cleaning strategy (such as acid washing or alkaline washing only) will be ineffective, and a more complex multi-step combined cleaning procedure must be initiated, such as first removing the inorganic fouling layer by chemical methods, and then specifically removing the organic or biological fouling layer to ensure the thoroughness and effectiveness of the cleaning.

[0077] In some embodiments of this application, the pollution type contribution score calculation and dominant pollution identification in step S130 can employ a fuzzy logic-based diagnostic system. This scheme utilizes fuzzy set theory to handle uncertainties in diagnosis, mapping characteristic indicators such as the instantaneous rise rate of transmembrane pressure difference into fuzzy linguistic variables through membership functions, and performing logical operations and defuzzification based on a rule base containing multiple fuzzy rules, ultimately outputting a weighted score or directly determining the dominant type. This method does not require a precise mathematical model and can effectively simulate the experience-based judgment of human experts, making it particularly suitable for handling complex pollution diagnosis scenarios with ambiguous boundaries.

[0078] In some embodiments of this application, the weight scoring and discrimination logic in step S130 can also be implemented based on a Bayesian network probabilistic inference model. This scheme sets colloidal contamination and scaling as parent nodes and various sensor characteristic indicators as child nodes, constructing a directed acyclic graph structure and learning the network's conditional probability table parameters based on historical operating data. During real-time diagnosis, the system inputs the collected sensor data as evidence into the network, calculates the posterior probability of each contamination type using a probabilistic inference algorithm, and uses this probability as the contamination contribution score. The type with the highest probability value is determined as the dominant contamination. This scheme provides a rigorous statistical inference mechanism, possesses strong noise resistance, and is robust when some sensor data is missing.

[0079] In some embodiments of this application, the manual feature extraction in step S120 and step S130 can also be replaced by an end-to-end diagnostic method based on deep learning. This approach uses a one-dimensional convolutional neural network or long short-term memory network model, directly taking the preprocessed multi-dimensional raw time-series data as input, eliminating the need for manually designed feature indicators. The model automatically extracts deep abstract features through internal convolutional and pooling layers, and directly outputs the weight scores or dominant type classification results for various types of pollution through fully connected layers. This approach eliminates tedious manual feature engineering, can uncover complex nonlinear correlations that are difficult for the human eye to detect, and is suitable for ultra-large-scale water treatment systems with massive amounts of historical data.

[0080] In some embodiments of this application, the hierarchical cleaning decision instruction in step S140 specifically includes: Step S510: Based on the identified dominant pollution situation, determine the cleaning agent, cleaning intensity coefficient, and cleaning method.

[0081] In step S510, a logical bridge is established between the contamination diagnosis results and the cleaning plan to achieve precise matching of the cleaning strategy. Through this step, the system can avoid using a one-size-fits-all cleaning mode for all contamination situations, ensuring that a gentle cleaning intensity is maintained when the contamination is slight to protect the membrane module life, while automatically increasing the response level when the contamination is severe or the composition is complex, laying the foundation for subsequent fine-tuning of cleaning parameters.

[0082] Step S520: Adjust the concentration of the cleaning agent, the cleaning cycle time, and the soaking time based on the cleaning intensity coefficient. The cleaning intensity coefficient and the pollution contribution score are positively correlated.

[0083] In step S520, the cleaning intensity coefficient is converted into specific process operation parameters to achieve automated and standardized control of the cleaning process. Since there is a positive correlation between the cleaning intensity coefficient and the contamination contribution, this means that the higher the monitored contamination load and the deeper the contamination, the larger the calculated intensity coefficient will be. This will drive the system to automatically increase the reagent concentration to enhance chemical dissolution capacity, extend the circulation time to ensure that the reagent fully contacts the membrane surface, and increase the soaking time to promote the peeling reaction of stubborn dirt.

[0084] This linkage mechanism ensures that the input of cleaning resources is strictly matched with the actual pollution needs, which not only prevents chemical waste and membrane damage caused by over-cleaning under low pollution conditions, but also avoids incomplete performance recovery caused by insufficient cleaning under high pollution conditions.

[0085] In some embodiments of this application, step S140 further includes determining the endpoint of the cleaning process: Step S610: After the soaking stage is completed, detect the change in turbidity, conductivity or pH value of the cleaning solution.

[0086] In step S610, a real-time monitoring mechanism based on changes in the physicochemical properties of the cleaning solution is constructed to quantitatively assess the actual progress and effectiveness of the chemical cleaning reaction. This step is initiated after the soaking stage, and sensors are used to detect changes in three key indicators in the cleaning solution: turbidity, conductivity, or pH value.

[0087] This design cleverly transforms the invisible fouling process on the membrane surface into a visible fluctuation of water quality parameters: an increase in turbidity directly reflects the extent to which suspended particles and biological slime detach from the membrane surface and enter the solution; an increase in conductivity characterizes the total amount of ions released after the dissolution of insoluble scale or ionic organic matter; and a decrease in pH (especially during alkaline washing) suggests the neutralization reaction of acidic fouling or the consumption of buffer substances. By capturing these subtle chemical signals, the system can overcome the blind reliance on simple time control and provide objective data support for judging whether the cleaning has achieved the expected results.

[0088] Step S620: If the increase in turbidity of the cleaning solution is greater than 20 NTU, or the increase in conductivity is greater than 500 NTU... If the pH value drops by more than 0.5, the cleaning reaction is considered complete, and the rinsing stage begins.

[0089] In step S620, a multi-dimensional set of cleaning endpoint determination criteria is established to ensure that the cleaning process is both thorough and safe and efficient. If any one of these conditions is met, the system determines that the cleaning reaction is sufficient, automatically terminates the soaking, and proceeds to the rinsing stage. This OR-logic determination mechanism greatly improves the system's adaptability and robustness, avoiding misjudgments caused by interference from a single indicator. It prevents incomplete flux recovery due to insufficient cleaning and effectively avoids chemical damage to membrane materials and water waste that may result from over-cleaning, maximizing cleaning benefits.

[0090] In some embodiments of this application, the cleaning triggering conditions in step S140 specifically include: (1) Relative pressure threshold trigger: Current value of transmembrane pressure difference ≥ Clean baseline pressure difference + 80 kPa.

[0091] The core function of the relative pressure threshold triggering mechanism is to capture the gradual increase in resistance caused by fouling accumulation during membrane system operation. This setting is based on the fact that in reverse osmosis or ultrafiltration systems, colloids, microorganisms, and impurity particles gradually precipitate on the membrane surface and form a filter cake layer during long-term operation, resulting in a steady increase in the resistance of water flowing through the membrane. By locking the incremental threshold at 80 kPa, the system can issue an early warning before the fouling deteriorates to an irreversible level, ensuring stable permeate flow under normal operating conditions and preventing unnecessary increases in energy consumption.

[0092] (2) Absolute pressure red line trigger: The current value of the transmembrane pressure difference is ≥ 90% of the maximum allowable transmembrane pressure difference of the membrane module.

[0093] The absolute pressure red line triggering mechanism is designed to build a crucial last line of defense for the safety of membrane modules, preventing catastrophic equipment damage caused by excessive pursuit of water production. When the system encounters sudden severe pollution or failure of the front-end pretreatment, the pressure differential may rise sharply in a short period of time. If operation is continued under these circumstances, the membrane may be compacted or even ruptured.

[0094] Therefore, once this absolute pressure threshold is reached, the system will disregard other performance indicators and forcibly initiate an emergency cleaning or shutdown protection procedure, thereby maximizing the structural integrity and service life of the core membrane element under extreme operating conditions.

[0095] (3) Performance degradation trigger: The current standardized water production flux is ≤ 70% of the initial clean flux.

[0096] The performance degradation triggering mechanism focuses on quantitatively assessing the degree of substantial decline in membrane flux due to fouling and clogging, from the perspective of final permeate efficiency. Since permeate flux directly reflects the treatment efficiency of the membrane system, setting this threshold signifies that membrane pore clogging or surface coverage has reached a point that severely impacts production. By introducing this triggering condition, the system can ensure that, while meeting water quality standards, it consistently maintains the designed production capacity, avoiding the paralysis of the overall water treatment process or water supply shortages caused by excessive membrane performance degradation.

[0097] Secondly, refer to Figure 2 This application provides a hardware architecture for an intelligent diagnosis and graded cleaning decision system for membrane fouling types, which is divided into four core parts: perception layer, control layer, execution layer and auxiliary system.

[0098] The auxiliary system consists of an uninterruptible power supply and instrument protective enclosures, providing power failure protection for the controller and environmental protection such as rain and sun protection for outdoor sensors, ensuring stable operation of the system under harsh conditions. The sensing layer is responsible for collecting key operational data, including TMP sensing units (inlet / outlet pressure sensors) for calculating transmembrane pressure difference, online analyzers for monitoring turbidity, fluorescence, pH, conductivity, etc. of the influent water quality, and flow meter sets for measuring the flow rates of influent, permeate, and concentrate.

[0099] The control layer, acting as the system's brain, is centered around an edge-computing industrial controller equipped with an algorithm engine. This controller processes data uploaded from the perception layer and issues commands. It also includes a human-machine interface unit for local status display and data upload. The execution layer comprises an intelligent dosing unit (acid / alkali / disinfectant tank pump), automatic valve assemblies, and a pretreatment interface (frequency converter control for the dosing pump), directly acting on the bottom membrane filtration system (membrane modules / piping) to achieve precise cleaning and process adjustment.

[0100] In some embodiments of this application, the sensing layer includes a pressure sensor, an online turbidity / suspended solids analyzer, an ultraviolet fluorescence organic matter analyzer, an integrated pH / conductivity / temperature sensor, and an electromagnetic flowmeter. Through the coordinated operation of these hardware components, the system can acquire key indicators in real time such as transmembrane pressure difference, influent colloid content, organic matter and biological activity index, water physicochemical properties, and flow rate, providing accurate and multi-dimensional raw data support for subsequent pollution diagnosis.

[0101] In some embodiments of this application, the processing and control layer includes an edge computing industrial controller for running the aforementioned intelligent diagnosis and graded cleaning decision-making method for membrane fouling types, and has a built-in diagnostic module based on recursive least squares and a rule engine. Further, the processing and control layer includes a memory storing a computer program that, when executed by a processor, implements the aforementioned intelligent diagnosis and graded cleaning decision-making method for membrane fouling types.

[0102] In this way, the processing and control layer can directly run intelligent diagnosis and graded cleaning decision-making methods for membrane fouling types at the device end, achieving millisecond-level real-time response without relying on the cloud.

[0103] In some embodiments of this application, the execution layer includes acidic, alkaline, and oxidizing cleaning agent storage tanks, metering pumps, pneumatic / electric valve assemblies, and pipeline mixers.

[0104] Specifically, once the processing and control layer receives the graded cleaning decision instruction, the execution layer can automatically allocate the corresponding type of cleaning agent, accurately control the concentration through the metering pump, and adjust the cleaning cycle and soaking time in conjunction with the valve and pipeline mixer, so as to achieve efficient, quantitative and safe automated cleaning and maintenance of membrane modules.

[0105] In some embodiments of this application, the edge computing controller outputs control signals based on diagnostic results to adjust the frequency of the metering pump and the switching state of the valve group in order to execute a composite cleaning procedure.

[0106] Specifically, by dynamically adjusting the frequency of the metering pump, the controller can precisely control the concentration and flow rate of cleaning agents such as acids, alkalis, or oxidants, ensuring that the intensity of chemical cleaning is perfectly matched to the current level of contamination. Simultaneously, it can control the on / off status of pneumatic or electric valve assemblies in real time, flexibly switching pipeline flow directions to execute different stages of the process, such as pre-washing, circulating cleaning, or soaking. This highly automated control method not only ensures that the composite cleaning program is implemented precisely according to the predetermined strategy but also effectively avoids errors and delays that may be caused by manual operation, greatly improving the efficiency and safety of membrane system maintenance.

[0107] In some embodiments of this application, the ultraviolet fluorescence organic matter analyzer is used to distinguish between humic acid-like organic matter and protein-like biological metabolites in the influent, and its installation position is parallel to that of the turbidity analyzer on the main influent pipe after the security filter.

[0108] The ultraviolet fluorescence organic matter analyzer, based on the photoluminescence principle, can accurately distinguish between humic acid-like organic matter and protein-like biological metabolites in influent. By analyzing the fluorescence fingerprint at specific wavelengths and combining it with indicators such as fluorescence index (FI) and bioactivity index (BIX), this instrument can not only quantitatively reflect the concentration of organic matter, but also keenly capture the degree of microbial metabolic activity, providing core chemical evidence for systematically identifying complex pollution, whether it is simply organic matter adsorption or accompanied by microbial growth.

[0109] In terms of engineering layout, the analyzer is positioned on the main inlet pipe after the security filter, arranged alongside the turbidity analyzer. Sampling after the security filter ensures that the test data accurately reflects the water quality about to contact the membrane surface, effectively eliminating interference from large particles at the front end. The parallel installation with the turbidity analyzer enables simultaneous online monitoring of indicators such as suspended solids and biochemical indicators such as organic matter. This real-time fusion of multi-dimensional data lays a solid data foundation for the edge computing controller to comprehensively assess the inlet water load and accurately predict the type of membrane fouling.

[0110] In some embodiments of this application, a hardware architecture for a smart diagnosis and graded cleaning decision system for membrane fouling types is provided. This architecture consists of four core parts: a perception layer, a processing and control layer, an execution layer, and an auxiliary system. It aims to build a complete closed loop from data acquisition to intelligent decision-making and then to precise execution.

[0111] The sensing layer is responsible for comprehensively capturing the operating status of the membrane system and the quality information of the influent. This layer specifically includes three pressure sensors, two online turbidity and suspended solids analyzers, one ultraviolet fluorescence organic matter analyzer, one integrated pH, conductivity and temperature sensor, and three electromagnetic flow meters.

[0112] The pressure sensor employs a high-precision differential pressure transmitter with a range of 0 to 1.0 MPa and an accuracy of ±0.1% of full scale. The inlet sensor is installed on a straight pipe section at least 5 times the pipe diameter before the membrane housing inlet flange on the membrane module's inlet header, while the outlet sensor is installed on the product water header for accurate calculation of transmembrane pressure differential. Two turbidimeters, using the laser scattering principle with a range of 0.1 to 1000 NTU, are installed on the inlet header after the security filter and before the high-pressure pump, serving as the core data source for turbidity gradient calculation. An ultraviolet fluorescence organic matter analyzer is installed alongside the turbidimeters and shares the sampling flow path for online measurement of fluorescence and bioactivity indices, crucial for distinguishing between organic and biological contaminants. An integrated pH, conductivity, and temperature sensor is installed adjacent to the turbidimeters, providing raw data for calculating water quality anomaly indices. Three electromagnetic flowmeters, with an accuracy of ±0.5%, are installed after the high-pressure pump, on the product water header, and on the concentrate return pipe, respectively, for real-time monitoring of the inlet, product water, and concentrate flow rates.

[0113] The processing and control layer is responsible for running the core algorithms and making decisions. This layer mainly consists of a high-performance edge computing industrial controller, equipped with a multi-core processor and large-capacity memory, installed in the membrane system's field control cabinet, capable of stable operation in environments with temperatures ranging from 0 to 45 degrees Celsius and low dust levels. This controller is responsible for performing data preprocessing, multi-dimensional feature index calculation, and the pollution discrimination rule engine. It can perform recursive least squares fitting, acceleration calculation, and standardized scoring in real time, and locally store at least 90 days of raw and feature data. In addition, this layer also includes a human-machine interface and data service unit, consisting of an industrial touchscreen installed on the main control cabinet door, an industrial switch forming a local area network, and a data upload module, used for local status display, parameter configuration, and uploading diagnostic results to the cloud platform.

[0114] The execution layer is the terminal mechanism that translates decisions into actions. This layer includes an intelligent dosing and cleaning unit, equipped with at least three cleaning agent storage tanks and metering pumps. The acid washing tank stores citric acid or hydrochloric acid for scale removal, the alkaline washing and disinfection tank stores sodium hydroxide and sodium hypochlorite to address organic and biological contamination, and the neutralization and flushing water tank stores reverse osmosis permeable water or softened water. These tanks are located in a separate cleaning chemical dosing room and are connected to the cleaning circulation loop via pump outlet pipes.

[0115] Simultaneously, this layer also includes an automatic process piping valve assembly, composed of pneumatic or electric butterfly valves, including inlet valves, product water valves, concentrate discharge valves, and dedicated cleaning valves. This assembly can automatically switch the pipeline flow direction to execute the cleaning procedure under controller commands. Furthermore, the system is equipped with a pretreatment adjustment unit, which, through a frequency converter control interface connected to the coagulant and scale inhibitor dosing pumps, automatically adjusts the feedforward dosing amount based on diagnostic results.

[0116] To ensure the stable operation of the entire system, this embodiment also includes a comprehensive auxiliary system. This system includes an uninterruptible power supply (UPS) that provides at least 30 minutes of power outage protection for the edge computing controller and critical instruments, preventing data loss or equipment damage due to sudden power outages. Additionally, for sensors installed outdoors, the system is equipped with dedicated instrument protection boxes, providing rain, sun, and freeze protection to ensure the long-term reliability of the sensing layer in various harsh environments.

[0117] In some embodiments of this application, the deployment architecture of the aforementioned processing and control layer can also be implemented using a cloud-edge-device collaborative diagnostic architecture. This alternative distributes computing tasks spatially in layers: the device side is responsible for raw data acquisition and preprocessing; the edge side is responsible for running lightweight diagnostic models, performing millisecond-level real-time monitoring and emergency alarms; the cloud side is responsible for aggregating historical data from multiple sites, using high-performance computing resources to train complex deep learning models or perform long-term trend predictions, and periodically sending optimized diagnostic model parameters to the edge side. This architecture ensures the real-time nature of on-site control while leveraging the big data processing capabilities of the cloud to achieve continuous iteration and optimization of the diagnostic model.

[0118] In some embodiments of this application, the aforementioned human-computer interaction and diagnostic decision-making execution methods can also be implemented based on a mobile-centric interaction and diagnostic solution. This alternative solution involves developing a dedicated smartphone or tablet app that directly connects to the on-site data acquisition unit via a wireless communication protocol. The app integrates a lightweight diagnostic inference engine, capable of independently completing pollution diagnosis and early warning without a fixed industrial control computer on-site, utilizing a simplified rule base or a small machine learning model. Simultaneously, the app serves as the front-end entry point to the cloud platform, allowing maintenance personnel to remotely view diagnostic results, receive alarm push notifications, and manually trigger cleaning commands, thereby achieving highly flexible mobile maintenance management.

[0119] In some embodiments of this application, the data sources and diagnostic criteria of the aforementioned sensing layer can also be implemented using virtual sensing and diagnostic technology based on digital twins. This alternative constructs a high-fidelity digital twin of the membrane system, which includes hydrodynamic, mass transfer, and fouling kinetic models of the membrane process. During the diagnostic process, the system compares sensor readings with the simulated output of the digital twin in real time: when a sensor (such as a pH meter or turbidity meter) malfunctions or drifts, the system uses the digital twin to extrapolate and generate virtual alternative readings based on data from other normal sensors; simultaneously, by analyzing the residuals between the system and the virtual system on key parameters such as transmembrane pressure difference, the system directly infers the fouling distribution and type within the membrane module, thereby achieving highly reliable fault-tolerant diagnosis.

[0120] In some embodiments of this application, reference is made to Figure 3 Based on the hardware architecture of the intelligent diagnosis and graded cleaning decision system for membrane fouling types mentioned above, a real-time data acquisition and preprocessing method is provided. This method aims to ensure the accuracy, continuity and high reliability of the input data, laying a solid foundation for subsequent multi-dimensional feature calculation and fouling type diagnosis.

[0121] System initialization is the first step in data acquisition. After the membrane filtration system completes its startup self-test, the central processing unit automatically executes the initialization procedure. This procedure first uses a device status query command to perform communication checks on all system sensors, confirming that key devices such as the pressure sensor, turbidity meter, pH conductivity sensor, UV fluorescence organic matter analyzer, and turbidity suspended solids online analyzer are all responding normally and have no fault codes.

[0122] Subsequently, the system loads preset operating parameters, including a sampling frequency of 1 Hz by default, the measurement range of each sensor, the Modbus communication address table, filtering algorithm parameters, and safety interlock thresholds, thereby establishing a standardized operating environment.

[0123] At the hardware operation level, the system classifies and processes pressure and water quality data. For pressure data, the system synchronously acquires 4 to 20 mA analog signals of the membrane module inlet pressure, outlet pressure, and product water pressure at a frequency of 1 Hz. After photoelectric isolation and analog-to-digital conversion, these signals are linearly converted according to a preset range, and the transmembrane pressure difference is calculated in real time through fixed-point arithmetic and fast division algorithms.

[0124] For water quality data, the system uses an RS-485 bus and Modbus RTU protocol to poll and collect influent turbidity, pH value, and conductivity. After CRC verification, the system extracts valid measurement values ​​and sensor status information. Simultaneously, the system specifically calculates the instantaneous rise rate of transmembrane pressure differential over the past 5 minutes. and transmembrane pressure difference loss value To monitor the dynamic changes in membrane resistance.

[0125] Take the TMP data from the most recent 5 minutes (collected every 5 seconds) and calculate the rate of change of the difference between the first and last data points: ; Transmembrane pressure difference loss ; The transmembrane pressure difference after cleaning; ;in, This is the value detected by the inlet-side pressure sensor. This is the value detected by the concentrate side pressure sensor. This is the value detected by the pressure sensor on the product water side. Since most systems operate at atmospheric or slightly negative pressure on the product water side, it can be simplified to: In embedded environments, fixed-point arithmetic and fast division algorithms are employed to improve real-time performance. Synchronous monitoring is also included. and If the difference remains abnormal, it indicates sensor drift or pipeline blockage.

[0126] To ensure data quality, the system implements a rigorous validity verification and cleaning mechanism. For water quality parameters, the system sets a reasonableness judgment based on the rate of change. If the difference between the current sampled value and the previous valid value exceeds 20% of the measurement range, it is judged as an erroneous value and discarded. Subsequently, a substitute value is generated using linear interpolation to ensure data continuity. For data loss due to faults or communication interruptions, the system records the missing time period and performs delayed processing or alarms in subsequent calculations.

[0127] In addition, the system pays special attention to the long-term stability of the sensors, has a built-in calibration reminder function, and triggers a "suspicious sensor drift" warning when the reading deviates from the historical benchmark by more than 15% continuously, prompting maintenance personnel to check rather than directly determining it as an abnormal water quality.

[0128] In terms of data caching and storage management, this embodiment employs a circular buffer technique to achieve efficient data read and write. The system pre-defines storage space for 7200 data points, corresponding to 2 hours of data at a 1 Hz sampling frequency, and avoids memory fragmentation by using a circular pointer movement method. Each data point is timestamped with a precision of 1 millisecond, forming a strict time sequence.

[0129] To prevent data loss due to power outages, the system has enabled a power outage protection function. When an abnormal power outage is detected, the system will automatically write the most recent 600 data points (i.e., 10 minutes of data) from the circular buffer to the non-volatile ferroelectric memory to ensure that the historical trend can be restored after power is restored and the integrity of the data is maintained.

[0130] Finally, during the data acquisition process described above, the system pays special attention to handling the inherent timing difference between pressure signals and water quality parameter sampling. A uniform millisecond-level timestamp is applied to each acquired data point via hardware interrupts or a high-precision system clock, ensuring strict alignment of multi-source data on the timeline. Simultaneously, to prevent system stagnation due to communication blockage by a single sensor, when a certain water quality parameter cannot be acquired for several consecutive cycles, the system marks it as "data missing" and fills it with a short-term predicted value based on historical data. This value is clearly marked as an "estimated value" on the human-machine interface for the diagnostic algorithm's reference, thus ensuring the robustness and real-time performance of the entire system.

[0131] In some embodiments of this application, the instantaneous rate of increase of the transmembrane pressure difference collected in the above steps... Greater than And transmembrane pressure difference loss Greater than Upon arrival, the membrane system immediately enters the pollution development diagnosis state and initiates real-time calculation of characteristic indicators and subsequent pollution type identification processes. If the above conditions are not met, the system only collects water quality data and does not trigger this diagnostic step. This step is triggered every 5 minutes as a periodic task. Before performing specific pollution type identification, the system first determines whether a pollution of concern has occurred. The core criterion is whether there is a continuous and significant increase in transmembrane pressure difference.

[0132] In some embodiments of this application, for four core pollution types, the system calculates four key characteristic indicators in parallel and maps them to a unified pollution contribution interval of 0 to 100 through a standardized function, making them comparable.

[0133] First, calculate the key indicator of colloidal contamination, namely the turbidity gradient. The purpose of this indicator is to quantify a sudden increase in the concentration of colloidal particulate matter in the influent, which is a major precursor to rapid colloidal fouling. The algorithm function is defined as follows: The outlier removal and filtering processes have been applied before calculation, and the unit is NTU / min. Take the turbidity data from the most recent minute and calculate the arithmetic mean of the turbidity data collected every 3 seconds within this time window; Take the forward window data from the 11th to the 10th minute in the past, deliberately avoiding the data from the most recent 10 minutes to prevent ongoing turbidity abrupt changes from contaminating the historical baseline, and calculate the arithmetic mean of the turbidity data within this time window.

[0134] Set the effective threshold for turbidity gradient Only when Only when this occurs is a valid colloidal impact considered to have occurred. The grading criteria are as follows: when... Normal fluctuations at times, continuous monitoring will not trigger alarms; when For mild impacts, record warnings and monitor changes in transmembrane pressure differential; when The incident was of moderate intensity, triggering the colloidal contamination rule and prompting preparation for enhanced chemical pretreatment; when In the event of a severe impact, a high-resolution alarm will be issued and emergency protection procedures may be activated.

[0135] The formula for calculating the contribution of colloidal contamination is as follows: .when That is, when the trigger threshold is just reached, the score is 0; when When the severe impact threshold is reached, the score is approximately 66.67, which means that from a mild impact to a severe impact, the score increases linearly from 0 to approximately 67.

[0136] Second, calculate the key scaling indicator, namely the Scaling Risk Index (SRI). The purpose of this index is to intuitively quantify the immediate risk of inorganic salt scaling on the membrane surface caused by the influent. This index is directly related to the supersaturation degree of scaling ions, rather than simply statistical anomalies deviating from historical averages. The algorithm function is defined as follows: The baseline pH value This refers to the median pH value of the influent over the past 24 hours, the current pH value. This refers to the current pH value of the influent; baseline conductivity value. This refers to the median conductivity of the influent over the past 24 hours, the current conductivity value. This refers to the conductivity value of the current incoming water.

[0137] Specifically, the higher the SRI value, the greater the deviation of the water quality from the normal scaling risk range. The specific grading standard is as follows: when... The fluctuations are normal, the risk of scaling is low, and the system is operating within its historical normal range; when The current level is slightly abnormal; attention is advised as it may indicate an early tendency for scaling. When the condition is significantly abnormal, saturated, and has a high risk of scaling, cleaning intervention should be prepared; when The current condition is severely abnormal, indicating a highly saturated state, and rapid scaling is highly likely. Immediate action is required. The formula for calculating the scaling and fouling contribution is as follows: .when hour, The score increases with the degree of abnormal scaling tendency.

[0138] Third, calculate diagnostic indicators linking organic load and bioactivity. This section employs a two-stage assessment method: first, assessing organic load, and then determining the risk to bioactivity.

[0139] Phase 1: Organic Pollution Coefficient The calculation formula is And satisfy .in, Representing the instantaneous rate of increase of transmembrane pressure difference, it reflects the real-time result of the interaction between pollutants and the membrane. Some organic substances, such as hydrophobic humic acids, are readily adsorbed onto the membrane, thus rapidly increasing the pressure. value.

[0140] Here The fluorescence index is the ratio of the light intensity of a fluorescent substance at 470 nm and 520 nm after irradiating a sample with ultraviolet light of wavelength 370 nm. The bioactivity index is a measure of the ratio of the light intensity of fluorescent substances at 380 nm and 430 nm after a sample is irradiated with ultraviolet light at a wavelength of 310 nm. It is used to reflect recently generated biological organic matter.

[0141] A higher value indicates greater pollution. Its relationship with organic pollution is evaluated as follows: When... The fluctuations were normal, the system was operating stably, the organic load was low, and there was no immediate fouling pressure on the membrane; when At the time, the anomaly was mild; organic matter began to affect the membrane, and the transmembrane pressure differential showed a slow and steady upward trend. This indicates a significant anomaly; organic fouling is accelerating, the rate of increase in transmembrane pressure differential is noticeably faster, and system performance is beginning to deteriorate. A chemical cleaning cycle should be scheduled, and the pretreatment and dosing systems need immediate inspection. This is a severe anomaly; the system is experiencing rapid organic fouling, with a surge in transmembrane pressure and a significant drop in membrane flux. Irreversible fouling is a potential risk, necessitating immediate chemical cleaning or emergency operational intervention. The formula for calculating the organic fouling contribution is as follows: .

[0142] Phase Two: Only when The indicators for this stage are calculated only when a certain organic load exists. The formula for calculating bioactivity is defined as the second derivative of the transmembrane pressure difference. This calculation is based on the most recent 60 reliable transmembrane pressure gradient slope values. A linear fit was then performed on this slope sequence to obtain the rate of change of the slope, expressed in kilopascals per square hour. The 60-minute window was chosen because the processes of microbial metabolism, reproduction, and secretion of extracellular polymers require a certain amount of time, allowing for the detection of acceleration in the early stages of biofilm formation, i.e., within the first few hours. The key coefficients and judgment logic are as follows: when... When the fluctuation is normal, it indicates no acceleration or even deceleration; when A slight anomaly indicates a slight positive acceleration, at the starting point of linear fouling; when A significant anomaly at this time indicates a marked acceleration and worsening of the contamination; when This indicates a severe anomaly, signifying a rapid and uncontrolled acceleration of contamination. The formula for calculating the contribution of biological contamination is as follows: ,in This represents the basic score for organic pollution, indicating the potential biological risks posed by the organic load itself, with a maximum score of 50. This represents a bonus for active organic pollution, when a clear positive acceleration of the transmembrane pressure difference is detected. Extra points are awarded for each instance, up to a maximum of 50 points, which indicates that the microorganisms begin to multiply exponentially, pushing the pollution dynamics from linear to exponential.

[0143] In some embodiments of this application, the calculation and result application of the above-described real-time data acquisition and preprocessing method must strictly follow the following precautions to ensure the accuracy of diagnosis, the stability of the system, and the reliability of decision-making.

[0144] First, all characteristic data used for calculation, including transmembrane pressure difference, turbidity, conductivity, pH value, and fluorescence index, must be time-stamped based on a unified and precise clock source, with all data acquisition frequency set to once every 3 seconds. During calculation, the system must check the time synchronization of each data stream; data with a deviation exceeding 10 seconds should be discarded or time-interpolated for alignment to prevent distortion of characteristic calculations due to data asynchrony. A complete cycle of real-time data calculation is 5 minutes. Furthermore, before reading data from each sensor, its status flag must be checked to verify sensor validity. For sensors reporting faults, undergoing calibration, or with invalid data, the calculation of related indicators should be paused, and the diagnostic conclusion should clearly indicate that the pollution type score is for reference only due to invalid sensor data.

[0145] Second, there are strong constraints on the relationship between organic and biological factors, namely, the contribution of biological pollution. The calculation must be based on the organic pollution coefficient. As a premise. When Regardless of TMP acceleration Both numerical values ​​should be forced. The value is 0 because significant TMP acceleration in the absence of an organic matrix is ​​more likely to be caused by other factors such as mishandling or compaction due to particulate impact rather than biological contamination.

[0146] At the same time, attention should be paid to the weighting adjustment of competitive indicators, when the contribution of colloidal contamination is... Or scale and contamination contribution When the score is abnormally high, for example, exceeding 70 points, the contribution of organic pollution is... and contribution of biological pollution The explanatory power will decrease because high turbidity shocks or severe fouling will dominate the increase in transmembrane pressure gradient, thus masking the signals of organic matter or biological activity. In this case, the system should include a warning in the diagnostic report stating that the current high score of colloidal or fouling contamination may interfere with the accurate assessment of organic or biological contamination.

[0147] Third, the calculation trigger cycle for the above-mentioned real-time data acquisition and preprocessing method is 5 minutes, for TMP acceleration. This metric, which requires a 60-minute window, should employ a recursive update algorithm. Each calculation should reuse most of the previous results, only incrementally calculating new data to reduce the instantaneous CPU load. To ensure the temporary storage and continuity of results, the four contribution scores calculated each time should be stored in a circular buffer. If a calculation fails due to insufficient data, the previous valid result should be used and the data delay should be noted to avoid jumps or interruptions in the diagnostic conclusions.

[0148] In some embodiments of this application, reference is made to Figure 4 , Figure 4This is a flowchart illustrating the intelligent diagnosis process for membrane fouling types provided in this application. After all feature indicators complete one cycle of calculation, the system immediately performs composite fouling weight scoring and dominant type determination, employing an event-triggered mechanism. The system first performs data quality checks, examining the validity and reasonable range of each feature value, and eliminating obviously abnormal calculation results. Only when all input features pass the quality check is the colloidal fouling contribution score then input. Scale contamination contribution Organic pollution contribution and contribution of biological pollution The weighting process is then initiated. These scores are typically calculated from sensor data (such as pressure, flow rate, water conductivity, etc.) using an algorithm model, and the numerical range is usually set from 0 to 100 points. The higher the score, the greater the risk of that type of pollution.

[0149] First, the system executes the highest priority judgment step, which is to check for the existence of serious inorganic contamination risks. The system will then allocate the contribution of colloidal contamination. Contribution of scaling and contamination Each score is compared to the first set threshold of 60. If... ≥60 points or A score of ≥60 indicates that the system may be facing serious blockage or scaling problems, and the process will enter the high-risk treatment branch on the left. Conversely, if both scores are less than 60, it indicates that inorganic pollution is not serious, and the process will jump to the branch on the right to focus on assessing organic and biological pollution.

[0150] Second, when entering the left branch (where there is a high risk of colloids or scaling), the system further introduces organic components. and biological division Conduct a comprehensive analysis. System judgment. or Whether the second set threshold of 40 points has been reached, and at the same time check the dispersion among these four scores. Specifically, if or ≥40 points, and , and , The maximum difference between the two values ​​is less than the third preset difference of 15 points. This means that although the inorganic fraction is high, the organic and biological fractions are also not low, and the values ​​of each indicator are relatively close, without forming an absolute single dominant factor. At this time, the system judges it as "compound pollution dominant". This situation usually indicates that a dense fouling layer mixed with inorganic salts and organic biological slime has formed on the membrane surface.

[0151] If the above-mentioned combined pollution conditions are not met (i.e., the difference is large or the biochemical score is low), the system reverts to single inorganic pollution discrimination. At this point, the system reconfirms... or Does the score ≥ 60, and calculate its difference from the benchmark score? The difference. If ≥60 points and minus If the difference is greater than or equal to the fourth preset difference of 20 points, then other interferences are excluded, and the diagnosis is "scale contamination-dominated"; similarly, if ≥60 points and minus If the difference is greater than or equal to 20 points, the diagnosis is "colloidal contamination dominant". This difference judgment ensures that the determination of the dominant contamination has a significant statistical advantage.

[0152] Third, when the first-level judgment result is negative, the system enters the right-hand branch, focusing on the activity analysis of organic and biological pollutants. At this point, the organic pollution baseline is used. A score of ≥40 is used as the starting threshold, and the score is based on the biopollution activity level. The numerical range is further subdivided into three levels.

[0153] like A score below 40 indicates that although there is an accumulation of organic matter, microorganisms have not yet multiplied in large quantities, and the condition is judged as "organic fouling is dominant but biological activity is not activated". At this time, the main problem is the adsorbed organic matter.

[0154] like A score between 40 and 60 indicates that microorganisms are becoming active, which is classified as "organic fouling dominated with signs of biological activity," suggesting that the initial formation of biofilm should be taken seriously.

[0155] like A score of 60 or higher indicates vigorous microbial metabolism and a significant impact, directly classifying it as "dominantly biological pollution," requiring immediate sterilization and removal measures.

[0156] Regardless of which path is taken, the final conclusion (dominated by complex contamination, dominated by scaling / colloidal contamination, or organic / biological contamination at different stages) will converge on the "diagnosis confirmation" stage. Based on this, the system generates a final diagnostic report and can link with subsequent cleaning control modules to match appropriate chemical cleaning agent formulations or rinsing strategies for different dominant contamination scenarios.

[0157] If a set of values ​​does not fall within any of the contamination judgment ranges defined in the flowchart, the system should classify it as "system operating normally" or "no obvious dominant contamination". In this case, the control logic will not trigger any targeted chemical cleaning or dosing adjustment instructions, and the current maintenance strategy (such as routine backwashing) can be maintained. At the same time, this low score range can also be used as a reverse indicator to confirm that the chemical cleaning has achieved the expected effect or as a baseline health status for the initial operation of a new membrane, thereby avoiding over-maintenance and saving operating costs.

[0158] Fourth, provide the final diagnostic output, which includes the following: The dominant type is determined based on the rules mentioned above. Pollution stages are divided into initial stage (dominant score less than 60), and acceleration stage (dominant score greater than or equal to 60, or TMP acceleration). The severe stage is when the transmembrane pressure differential approaches its limit. A key example of this chain of evidence is high organic loading. Equal to 75, strong transmembrane pressure differential positive acceleration If the condition is as described above, it is considered an active period of biological contamination. The priority order for cleaning strategies is: biological contamination > colloidal or scale > pure organic fouling.

[0159] In some embodiments of this application, a diagnostic confirmation mechanism is provided as a key bridge connecting feature calculation and cleaning execution. When the final diagnostic output simultaneously satisfies three conditions—steady state, confidence level, and trend reinforcement—the system will lock the dominant type, score, and key evidence, generating a diagnostic readiness signal as input for subsequent cleaning decisions. If the dominant type changes during the confirmation period, the confirmation window will be reset and the timer will restart. The steady-state condition requires the dominant contamination type to remain unchanged for three consecutive diagnostic cycles (a total of 15 minutes); the confidence level condition requires the current dominant type's contamination contribution score to be greater than or equal to 50 points; and the trend reinforcement condition requires the dominant type's score to fluctuate by less than 10% or show a continuous upward trend within the last three cycles.

[0160] In some embodiments of this application, a complete set of graded cleaning decision generation and execution steps is provided. Its initiation is determined by both basic conditions and core triggering conditions, forming an "AND" and "OR" logical combination. The basic condition is receiving a diagnostic readiness signal, designed to ensure that cleaning targets identified and stable contamination issues. The core triggering conditions include three sub-conditions, any one of which can trigger the cleaning process: first, a relative pressure threshold trigger, where the current transmembrane pressure difference is greater than or equal to the clean membrane baseline pressure difference plus an 80 kPa current offset, to prevent over-cleaning; second, an absolute pressure red line safety fallback trigger, where the current transmembrane pressure difference is greater than or equal to 0.9 times the maximum permissible transmembrane pressure difference specified by the membrane module manufacturer; and third, a performance degradation trigger, where the current standardized permeate flux is less than or equal to the initial clean flux multiplied by (a 1-30% degradation rate threshold), ensuring timely intervention when operational economics decline.

[0161] When the system simultaneously meets both the basic and core trigger conditions, it will immediately shut down the membrane system's feed water system and pumps and initiate the cleaning process. If the basic conditions are met but the core trigger condition is not, it indicates that fouling is in its initial or accelerated phase. In this case, the system will schedule the cleaning task during the most recent water production interval and prepare the necessary chemicals in advance. If the basic conditions are not met (e.g., the dominant fouling type cannot be identified or the score is below 40), the system will not trigger cleaning but will instead feed back relevant water quality parameters to the pretreatment section of the water treatment process to achieve coordinated optimization across the entire process.

[0162] In some embodiments of this application, reference is made to Figure 5 The system provides a method for selecting a cleaning solution. When both the basic and core triggering conditions meet the cleaning requirements, the system generates a graded and accurate cleaning decision based on the type, severity, stage of development, and historical cleaning results, and coordinates the safe execution of each subsystem. The system incorporates a multi-dimensional cleaning decision matrix, using the diagnostic results confirmed in step 3 as input indexes to match the optimal cleaning strategy.

[0163] First, make a decision on the type of cleaning, which includes the following: When the system inputs information such as the dominant type of scale (biological, colloidal, inorganic, or organic), the stage of contamination (initial, accelerated, or severe), and a complex contamination indicator, it will automatically generate a basic cleaning plan. Specific cleaning types are as follows: If biological contamination is determined to be dominant, a disinfection-type chemical cleaning plan is used, with sodium hypochlorite as the main agent, and the pH value adjusted to between 10 and 11 to enhance the stripping effect; if colloidal or organic contamination is determined to be dominant, an alkaline chemical cleaning plan is used, with sodium hydroxide as the main agent, combined with surfactants or chelating agents; if inorganic scale is determined to be dominant, an acidic chemical cleaning plan is used, with citric acid or hydrochloric acid as the main agent, while paying attention to concentration and contact time control; if complex contamination is determined to be present, a step-by-step complex cleaning is used, typically following the sequence of acid washing followed by alkaline washing or alkaline washing followed by disinfection, with thorough rinsing in between.

[0164] Second, calculate the cleaning intensity, which includes the following: The cleaning intensity is determined by the contribution of pollution, specifically by three factors: cleaning agent concentration, cleaning cycle time, and soaking time, in order to achieve the goal of quantifying the amount of pollution.

[0165] In terms of reagent concentration calculation, cleaning intensity coefficient The calculation formula is ,in This corresponds to a pollution contribution score. For example, when the biological pollution score is 85 points, then... Chemical reagent concentration equal to standard concentration Multiply The standard concentration is set based on the membrane manufacturer's recommended value. The system will prompt you to confirm to prevent damage to the membrane material.

[0166] Regarding the calculation of cleaning cycle time, the formula is as follows: That is, the maximum time shall not exceed 1.5 times the standard time.

[0167] For calculating soaking time, it is crucial for biofilms or heavily soiled organic deposits. The formula for calculation is as follows: .

[0168] In some embodiments of this application, intelligent execution and process monitoring are performed based on the decisions of the above-mentioned cleaning scheme. The cleaning process is not a simple time-series control, but an adaptive control based on real-time feedback.

[0169] First, during the main cleaning phase, add the cleaning agent according to the selected cleaning plan and calculated concentration, and start the cleaning circulation pump. Key monitoring indicators include cleaning solution turbidity, pH value, conductivity, and pressure. If the cleaning solution turbidity rises rapidly and continuously within the first 15 minutes, it indicates a significant stripping effect, and the system can automatically extend the circulation time by 10% to 20%. If the pressure remains unchanged, it may indicate agent failure or misjudgment of the contamination type, and the system will issue a warning.

[0170] Second, decisions are made regarding soaking and secondary cleaning. Based on the dominant contamination type determined above, a differentiated soaking strategy is matched to the system. Soaking is not only a time-based process, but also a dynamic intervention based on the kinetics of chemical reactions.

[0171] For different types of soaking strategies: If biological contamination is the primary concern, sodium hypochlorite or peracetic acid disinfectant should be used. Sodium hypochlorite concentration should be 200-500 ppm, pH adjusted to 10-11 to enhance peeling, and soaking time should be 60-120 minutes. Peracetic acid concentration should be 100-200 ppm, and soaking time should be 30-60 minutes. Its mechanism of action is to oxidize and destroy the microbial cell structure and degrade the EPS adhesive matrix. If colloidal or organic contamination is the primary concern, an alkaline cleaning agent, sodium hydroxide, can be used, which can be combined with 0.1%-0.3% surfactant such as SDS. Sodium hydroxide concentration should be 0.1%-0.5%, pH greater than 12, and soaking time should be 45-90 minutes. Its function is to saponify and hydrolyze organic fouling substances, increasing the solubility and dispersibility of pollutants. If scaling is the primary concern, citric acid (2%-4%) or hydrochloric acid (0.1N-0.5N) should be used. The concentration and time should be controlled so that the pH is less than 3, and the soaking time is 30 to 60 minutes. Citric acid is required for acid-sensitive membranes. Its mechanism of action is that the acid dissolves carbonates, sulfates, and metal oxide scale. If it is a complex contamination, a sequential step-by-step soaking method should be adopted, first acid washing to remove scale and rinse, then alkaline washing to disinfect and remove organic matter or biological substances, or vice versa, with each step of the soaking time shortened by 30%.

[0172] Third, a quantitative evaluation of the soaking effect is conducted. The soaking process is divided into two stages: process monitoring and endpoint determination. During the process monitoring stage, the system records key indicators of the cleaning solution, such as turbidity and conductivity, at 1-minute intervals, calculating the rate of increase or cumulative change of these indicators. For example, for alkaline washing to remove organic contaminants, if the turbidity continuously increases within 30 minutes with a rate greater than 0.5 NTU / min, the stripping reaction is considered to be in progress. In the endpoint determination stage, the endpoint is determined by detecting relevant parameters in the cleaning solution after the soaking time is completed. For alkaline washing to remove organic or colloidal contaminants, the increase in turbidity is greater than 20 NTU or the increase in fluorescence intensity is greater than 30%. For acid washing to remove scale, the increase in conductivity is greater than 500 NTU. If the purpose is disinfection, cleaning, or biological removal, the pH value will decrease by more than 0.5 (for alkalizing sodium hypochlorite).

[0173] Fourth, perform rinsing and recovery, rinsing until the permeate conductivity and pH deviate from the feed water by less than 10% and there is no residual chlorine (for sodium hypochlorite cleaning). After cleaning, the system runs at low pressure and low flux for 15 minutes to record the initial transmembrane pressure difference (TMP). Compare this initial TMP after cleaning with the initial TMP before this fouling cycle and the historical best initial TMP to calculate the membrane flux recovery rate. The calculation formula is: recovery rate = (flux before fouling minus flux after fouling) divided by (flux before fouling minus flux after cleaning) multiplied by 100%. If the recovery rate is less than 85%, the system marks this cleaning as partially effective and records it in the log, providing data for future cleaning optimization. If the recovery rate is greater than 95%, it is marked as highly efficient, and the future cleaning rate for the same type of fouling is slightly reduced. The coefficient enables learning-based energy saving.

[0174] Finally, a report is generated and the knowledge base is updated. Each cleaning process serves as a learning case, and the system automatically generates a structured report and updates the knowledge base. The report includes: a pollution diagnosis summary (including dominant type, score, and key evidence); and the cleaning protocol implemented (including agents, concentrations, time, and intensity coefficients). The system quantifies the cleaning effect (including flux recovery rate, energy consumption, and chemical consumption); it also provides a brief analysis of economic benefits (i.e., an estimate of water and chemical savings compared to fixed-cycle cleaning). The system stores the pollution characteristics, cleaning plan, and recovery effect as a case study.

[0175] The underlying technology involves quantifying and recording each cleaning process and its effectiveness using the TMP recovery rate, which is then used for threshold self-optimization. For example, a reinforcement learning algorithm is employed, with cleaning effectiveness and operating cost as dual objectives, to dynamically adjust the key threshold parameters for determining various contamination scenarios in the rules. , , , For example, automatically lowering the detection threshold for biological contamination during peak algae bloom periods. Adjusting the scaling warning threshold during the high-temperature season This adaptive capability ensures that the system maintains optimal performance over a long period of time.

[0176] In summary, the intelligent diagnosis and graded cleaning decision-making method and system for membrane fouling types provided in this application have the following technical effects.

[0177] This application achieves accurate identification of complex pollution through a multi-parameter fusion diagnostic mechanism. Its core lies in quantitatively assessing the contribution of different pollution types, such as colloidal, inorganic, organic, and biological pollutants, and outputting structured diagnostic results, thereby providing a scientific basis for subsequent decision-making.

[0178] Based on the aforementioned accurate diagnosis, this application establishes a data-driven cleaning decision-making model. By automatically matching targeted, tiered cleaning solutions, it effectively avoids the blind spots of traditional experience-based cleaning, ensuring cleaning effectiveness while maximizing the protection of membrane materials.

[0179] In terms of biofouling control, this application achieves early warning of biofouling by monitoring the acceleration of changes in transmembrane pressure gradient (TMP). This technology can specifically capture the characteristics of the initial stage of biofilm exponential growth, providing a valuable window for early intervention, thus allowing for the use of gentler cleaning methods.

[0180] Furthermore, the technical solution of this application forms an adaptive optimization closed loop. The system can continuously learn through the data closed-loop mechanism, thereby adapting to dynamic factors such as changes in water source and membrane aging, and promoting continuous intelligent operation and maintenance management.

[0181] Finally, this application employs a white-box rule engine, making the diagnostic logic completely transparent and traceable, significantly enhancing the system's reliability and maintainability. This standardized decision output reduces reliance on the professional experience of operations and maintenance personnel, improving the standardization level of operations and maintenance.

[0182] It should be noted that in all specific embodiments of this application, all data processing activities related to user identity or personal characteristics, such as user information, user behavior data, historical data, and location information, will be conducted in accordance with the principles of legality, legitimacy, and necessity. All data collection, use, storage, and processing will be subject to compliance with applicable national and regional laws, regulations, and industry standards, and informed consent from users will be obtained in a clear and explicit manner before processing. For the processing of sensitive personal information, separate consent from users will be obtained through prominent means such as pop-up prompts and independent confirmation pages. If any processing conflicts with laws and regulations, the laws and regulations will prevail, and necessary data processing will only be carried out within the scope permitted by laws and regulations, ensuring that all data-based applications, analyses, and technical implementations are conducted within the scope permitted by laws and regulations.

[0183] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0184] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary skill. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0185] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0186] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0187] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Additionally, computer-readable media can even be paper or other suitable media on which programs can be printed, for example, by optically scanning the paper or other media, then editing, interpreting, or, if necessary, processing it in a suitable manner to obtain the program electronically, and then storing it in computer memory.

[0188] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0189] In the foregoing description of this specification, the reference to terms such as "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations," etc., indicates that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in an embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0190] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0191] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A method for intelligent diagnosis and graded cleaning decision-making based on membrane fouling type, characterized in that, Includes the following steps: Real-time acquisition of membrane system operating pressure data and multi-dimensional influent water quality data; Based on the operating pressure data and multidimensional influent water quality data, characteristic indicators are calculated and obtained. The characteristic indicators include: turbidity gradient characterizing colloidal pollution, scaling risk index characterizing scaling risk, and organic pollution coefficient characterizing organic and biological composite pollution. Based on the aforementioned characteristic indicators, contribution scores for multiple pollution types are calculated to determine the current dominant pollution situation; wherein, the contribution scores for the pollution types include colloidal pollution contribution scores, scaling pollution contribution scores, organic pollution contribution scores, and biological pollution contribution scores; When the transmembrane pressure difference determined based on the operating pressure data, or the standardized permeate flux determined based on the multidimensional influent water quality data, meets the preset cleaning triggering conditions, a graded cleaning decision instruction is generated based on the dominant contamination situation, and the execution unit is controlled to implement the corresponding cleaning procedure.

2. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, The calculation of the turbidity abrupt change gradient and the contribution of colloidal contamination specifically includes: The difference between the current average and historical average of influent turbidity is calculated to obtain the turbidity abrupt change gradient; Obtain the instantaneous rate of rise of the transmembrane pressure difference; The product of the turbidity abrupt change gradient and the instantaneous rise rate is calculated, and combined with a preset effective threshold for the turbidity gradient, the colloidal contamination contribution score is calculated through a linear mapping function.

3. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, The calculation of the scaling risk index and the scaling fouling contribution score specifically includes: Based on the deviation of the current values ​​of influent pH and conductivity from the baseline values, the scaling risk index (SRI) is calculated, and its expression is: ； in, This represents the current pH value of the influent. This is the baseline value of the influent pH. This represents the current value of electrical conductivity. This is the baseline value for conductivity; This is the pH sensitivity coefficient. The conductivity sensitivity coefficient, The weighting factor is preset. Based on the numerical range of the Scaling Risk Index (SRI), the scaling pollution contribution score is calculated through a preset mapping relationship.

4. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, The calculation of the organic pollution coefficient, organic pollution contribution score, and biological pollution contribution score specifically includes: The fluorescence index of the influent was measured online using an ultraviolet fluorescence analyzer. and bioactivity index And combined with the instantaneous rate of rise of the transmembrane pressure difference Calculate the organic pollution coefficient Its expression is: ； in, and The preset weighting coefficients, and ; When the organic pollution coefficient does not exceed a set threshold, the organic pollution contribution score is calculated through a preset mapping relationship; When the organic pollution coefficient is greater than the set threshold, the instantaneous rise rate of the transmembrane pressure difference is linearly fitted using the recursive least squares method to calculate its rate of change. If the rate of change is greater than a preset first threshold, the biological activity is determined to be significant, and a biological activity bonus is calculated; the organic pollution base score is calculated and summed with the biological activity bonus to obtain the biological pollution contribution score; If the rate of change is less than or equal to the first threshold, the contribution of biological pollution is 0.

5. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, To determine the current dominant pollution situation, the following situations must be considered: (1) If the contribution score of colloidal pollution exceeds the first set threshold, and the difference between the contribution score of colloidal pollution and the contribution score of organic pollution and the contribution score of biological pollution are both greater than the first preset difference, then it is determined that colloidal pollution is dominant. (2) If the contribution score of scaling pollution exceeds the first set threshold, and the difference between the contribution score of scaling pollution and the contribution score of organic pollution and the contribution score of biological pollution are both greater than the first preset difference, then it is determined that scaling pollution is dominant. (3) When the contribution score of organic pollution exceeds the second set threshold, if the contribution score of biological pollution is lower than the third set threshold, it is determined that organic pollution is dominant; if the contribution score of biological pollution exceeds the third set threshold, it is determined that biological pollution is dominant. (4) If the contribution of colloidal pollution or the contribution of scaling pollution exceeds the first set threshold, and the contribution of organic pollution or the contribution of biological pollution exceeds the second set threshold, then it is determined to be a compound pollution.

6. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, Based on the dominant pollution situation, a tiered cleaning decision instruction is generated, specifically including: Based on the identified dominant pollution situation, determine the cleaning agent, cleaning intensity coefficient, and cleaning method; The concentration of the cleaning agent, the cleaning cycle time, and the soaking time are adjusted based on the cleaning intensity coefficient; wherein the cleaning intensity coefficient has a positive correlation with the pollution contribution score.

7. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 6, characterized in that, The method includes determining the endpoint of the cleaning process: After the soaking stage, the changes in turbidity, conductivity, or pH of the cleaning solution are measured. If the increase in turbidity of the cleaning solution is greater than 20 NTU, or the increase in conductivity is greater than 500 NTU. If the pH value drops by more than 0.5, the cleaning reaction is considered complete, and the rinsing stage begins.

8. The intelligent diagnosis and graded cleaning decision method for membrane fouling types according to claim 1, characterized in that, The cleaning triggering conditions specifically include: Relative pressure threshold trigger: Current transmembrane pressure differential ≥ Clean baseline pressure differential + 80 kPa; Absolute pressure red line trigger: Current transmembrane pressure difference ≥ 90% of the maximum allowable transmembrane pressure difference of the membrane module; Performance degradation trigger: Current standardized permeate flux ≤ 70% of initial clean flux.

9. A smart diagnostic and graded cleaning decision system for membrane fouling types, characterized in that, include: The sensing layer includes a pressure sensor, an online analyzer for turbidity and suspended solids, an ultraviolet fluorescence organic matter analyzer, an electromagnetic flowmeter, and integrated sensors for pH, conductivity, and temperature. The processing and control layer includes an edge computing industrial controller with a built-in diagnostic module based on recursive least squares and a rule engine. The execution layer includes a cleaning agent storage tank assembly, a metering pump, a pneumatic valve assembly, an electric valve assembly, and a pipeline mixer; the cleaning agent storage tank assembly includes tanks for storing acidic, alkaline, or oxidizing cleaning solutions; The edge computing industrial controller outputs control signals based on the diagnostic results to adjust the frequency of the metering pump and the switching state of the valve group in order to execute the composite cleaning program.

10. The intelligent diagnosis and graded cleaning decision system for membrane fouling types according to claim 9, characterized in that, The processing and control layer includes a memory that stores a computer program. When the computer program is executed by a processor, it implements the intelligent diagnosis and graded cleaning decision method for membrane fouling types as described in any one of claims 1 to 8.

Images