Water purifier AI intelligent management and control system

Abstract

The application discloses a water purifier AI intelligent management and control system, and belongs to the technical field of intelligent household electrical appliance control, which comprises a water purifier, wherein the water purifier is provided with an inlet and outlet water pressure sensor, an inlet and outlet water total dissolved solid content sensor, a flow sensor, an electric control water pump and a flushing valve actuating mechanism; and the water purifier is further provided with a filter core pollution state reconstruction module, a cleaning strategy reward evaluation module, a dynamic cleaning parameter generation module and a cleaning effect closed loop verification module. The application breaks through the technical bottleneck of traditional fixed threshold control by constructing a full-link intelligent architecture of multi-feature fusion pollution quantification, multi-target strategy evaluation, dynamic parameter generation and closed loop verification, can accurately identify early hidden blockage of the filter core of the water purifier, avoid misjudgment, automatically balance the cleaning effect, water consumption and membrane damage risk, realize self-adaptive cleaning regulation and control of the filter core in the whole life cycle, and has a built-in closed loop iteration mechanism, so that the system performance is gradually improved with use, and the use and maintenance costs are reduced.

Description

Technical Field

[0001] This invention relates to the field of smart home appliance control technology, and in particular to an AI-powered intelligent control system for water purifiers. Background Technology

[0002] With the intelligent iteration of household water purification equipment, AI-driven dynamic control technology has gradually become the focus of industry research and development. At present, mainstream smart water purifiers have realized basic operation status perception and automatic flushing functions. By deploying sensors such as pressure and flow to collect operation data, and combining preset trigger rules to execute standardized cleaning programs, or using preset timed and pressure fixed flushing programs to complete cleaning, the automation level of the equipment has been improved to a certain extent, and the manual maintenance cost for users has been reduced.

[0003] However, existing intelligent control technologies for water purifiers still have many shortcomings in practical applications. First, in terms of filter status sensing, most systems rely on only a limited number of parameters for simple judgment, which makes it difficult to comprehensively and accurately reflect the actual pollution status of the filter. In particular, there is a lack of effective means to identify early hidden blockage problems, which often leads to untimely or excessive cleaning, affecting the service life of the filter. In terms of cleaning strategy formulation, traditional technologies mostly use fixed procedures, which make it difficult to dynamically adjust cleaning parameters according to factors such as the actual degree of pollution of the filter and water quality conditions. This can easily lead to problems such as incomplete cleaning, excessive rinsing and water consumption, or membrane damage, resulting in a significant increase in operating costs. Summary of the Invention

[0004] To address the shortcomings mentioned in the background technology, we propose an AI-powered intelligent control system for water purifiers.

[0005] The technical solution mainly includes: an AI intelligent control system for water purifiers, comprising a water purifier, wherein the water purifier is equipped with inlet and outlet water pressure sensors, inlet and outlet water total dissolved solids content sensors, flow sensors, an electric water pump, and a flushing valve actuator; the water purifier is also equipped with a filter cartridge contamination status reconstruction module, a cleaning strategy reward evaluation module, a dynamic cleaning parameter generation module, and a cleaning effect closed-loop verification module.

[0006] The filter cartridge fouling state reconstruction module is configured to collect data on influent viscosity, flow rate, membrane area, influent-outfluent pressure difference, and total dissolved solids content, and to perform physical constraint fusion processing on the multi-source data to obtain the fouling state index;

[0007] The cleaning strategy reward evaluation module is configured to collect data on flux, water consumption, rated flushing water volume, peak pressure, and burst threshold before and after cleaning, obtain a comprehensive reward value through collaborative calculation, and output it to the dynamic cleaning parameter generation module and the cleaning effect closed-loop verification module.

[0008] The dynamic cleaning parameter generation module receives the pollution index and reward value, and combines them with the calibrated benchmark value, water quality reference value, and pressure difference threshold to generate the optimal cleaning duration and pulse frequency, which are then output to the electric control water pump and flushing valve actuator.

[0009] The cleaning effect closed-loop verification module collects the pollution index and effect judgment threshold before and after cleaning. The verification result is obtained by comparing the degree of pollution improvement with the threshold, which triggers the secondary cleaning process.

[0010] Preferably, the filter cartridge contamination state reconstruction module includes a fluid parameter acquisition unit, a water quality parameter acquisition unit, and a contamination state fusion calculation unit:

[0011] The fluid parameter acquisition unit is configured to collect data on inlet water viscosity, inlet water flow rate, and inlet-outlet water pressure difference of the filter element according to the filter element operation monitoring requirements, and extract fluid resistance characteristics.

[0012] The water quality parameter acquisition unit is configured to collect total dissolved solids content data of the filter cartridge inlet and outlet water, calculate the ratio of total dissolved solids content of inlet and outlet water, and extract membrane retention efficiency characteristics.

[0013] The pollution state fusion calculation unit is configured to acquire fluid resistance characteristics, membrane retention efficiency characteristics, and effective membrane area parameters of the filter element. Through fusion calculation, the filter element pollution state index is obtained, which quantifies the degree of internal blockage of the filter element.

[0014] Preferably, the pollution state fusion calculation unit includes a fluid resistance characteristic normalization subunit, a retention efficiency characteristic conversion subunit, and a pollution index output subunit:

[0015] The fluid resistance characteristic normalization subunit is configured to obtain parameters such as influent viscosity, influent flow rate, effective membrane area of ​​filter element, and influent-outfluent pressure difference, and to obtain the normalized fluid resistance value by calculating the ratio of fluid transport flux to filtration resistance.

[0016] The retention efficiency characteristic conversion subunit is configured to obtain the ratio of total dissolved solids content in influent and effluent, and to obtain a standardized retention efficiency value through a square root nonlinear conversion operation.

[0017] The pollution index output sub-unit is configured to obtain the normalized fluid resistance value and the standardized retention efficiency value, and obtain the pollution state index by multiplying the two types of features, thereby realizing the coupled quantification of multi-dimensional pollution features.

[0018] Preferably, the cleaning strategy reward evaluation module includes a flux recovery rate calculation unit, a water consumption cost calculation unit, a membrane damage risk calculation unit, and a multi-objective fusion output unit:

[0019] The flux recovery rate calculation unit is configured to obtain the flux parameters of the filter element before and after cleaning and the rated flux parameters of the new filter element, and to obtain the flux recovery degree index by calculating the ratio of the change in flux before and after cleaning to the rated flux.

[0020] The water consumption cost calculation unit is configured to obtain the parameters of cleaning water consumption and rated flushing water volume, and obtain the cleaning water consumption cost index by calculating the ratio of water consumption to the rated value and then performing a natural exponential nonlinear transformation.

[0021] The membrane damage risk calculation unit is configured to obtain the peak cleaning pressure and the filter element burst pressure threshold parameters, and to obtain the membrane damage risk index by calculating the ratio of the peak pressure to the burst threshold.

[0022] The multi-objective fusion output unit is configured to acquire flux recovery level indicators, cleaning water consumption cost indicators, and membrane damage risk indicators. It obtains a comprehensive reward value through multiplication of multiple indicators, thereby achieving automatic balance of multiple objectives.

[0023] Preferably, the multi-objective fusion output unit includes an indicator standardization subunit, a collaborative computation subunit, and a reward level division subunit:

[0024] The index standardization subunit is configured to perform linear normalization on the flux recovery degree index, the cleaning water consumption cost index, and the membrane damage risk index, so that the three types of indexes are in the same numerical range.

[0025] The collaborative computing subunit is configured to multiply the three standardized indicators to obtain the original reward value, thereby achieving collaborative optimization of multiple objectives.

[0026] The reward level division sub-unit is configured to divide reward levels based on the original reward value, mark high reward value strategies as executable strategies, and mark low reward value strategies as strategies to be optimized.

[0027] Preferably, the dynamic cleaning parameter generation module includes a cleaning duration calculation unit, a pulse frequency calculation unit, and a parameter safety verification unit:

[0028] The cleaning time calculation unit is configured to acquire the filter cartridge contamination status index, the pre-calibrated baseline value of the contamination status of a brand new filter cartridge, the total dissolved solids content of the influent, the reference value of the total dissolved solids content of the local municipal influent, and the basic cleaning time parameters calibrated in the laboratory. The acquired parameters are then used to obtain the optimal cleaning time through a nonlinear adaptive calculation of a logarithmic function.

[0029] The pulse frequency calculation unit is configured to acquire the pressure difference between the filter cartridge inlet and outlet, the pressure difference characteristic threshold calibrated in the laboratory, the pre-calibrated minimum pulse frequency, and the pre-calibrated maximum pulse frequency parameters, and to obtain the optimal pulse frequency by performing an adaptive interval mapping operation using an exponential function on the acquired parameters.

[0030] The parameter safety verification unit is configured to verify the optimal cleaning time and optimal pulse frequency within a safe range, correct parameters that exceed the safe range to within the allowable range, and output the final execution parameters.

[0031] Preferably, the cleaning time calculation unit includes a pollution degree difference calculation subunit, a water quality impact factor calculation subunit, and a time output subunit:

[0032] The pollution level difference calculation subunit is configured to obtain the pollution status baseline value of a brand new filter cartridge and the current pollution status index, and calculate the relative pollution level by the ratio of the difference to the baseline value.

[0033] The water quality impact factor calculation subunit is configured to obtain the total dissolved solids content of the influent and the reference value, and then calculate the water quality impact coefficient by ratio calculation.

[0034] The duration output subunit is configured to obtain the relative pollution level and water quality influence coefficient, and after transformation by the natural logarithm function, multiply it by the basic cleaning duration to obtain the optimal cleaning duration, thereby realizing the joint control of the cleaning duration by the pollution level and water quality parameters.

[0035] Preferably, the cleaning effect closed-loop verification module includes a pollution improvement calculation unit, a threshold comparison unit, and a verification result output unit:

[0036] The pollution improvement calculation unit is configured to obtain the filter element pollution status index before and after cleaning, and to obtain the pollution improvement rate index by calculating the ratio of the change in pollution index to the pollution index before cleaning.

[0037] The threshold comparison unit is configured to obtain the pollution improvement rate index and the cleaning effect judgment threshold, and remove negative values ​​through linear rectification operation to obtain the cleaning effect verification value.

[0038] The verification result output unit is configured to determine the cleaning effect based on the cleaning effect verification value, mark the verification result that meets the requirements as cleaning completed, and trigger a secondary cleaning process for the verification result that does not meet the requirements.

[0039] Preferably, the cleaning effect closed-loop verification module further includes a parameter iteration optimization unit and a strategy feedback adjustment unit;

[0040] The parameter iteration optimization unit is configured to feed back the cleaning effect verification results to the dynamic cleaning parameter generation module, and iteratively update the logarithmic function coefficients for calculating the cleaning duration and the exponential function coefficients for calculating the pulse frequency through the verification data, thereby optimizing the calculation logic of the cleaning duration and pulse frequency.

[0041] The strategy feedback adjustment unit is configured to feed back the cleaning effect verification results to the cleaning strategy reward evaluation module. By iteratively optimizing the indicator weights of multi-objective fusion calculation through verification data, the rationality of subsequent cleaning strategies is improved.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] This invention overcomes the technical bottleneck of traditional water purifiers that rely on fixed threshold judgments and fixed cleaning procedures by using a full-link intelligent architecture that integrates multi-feature fusion and quantification of filter cartridge contamination status, multi-objective collaborative strategy evaluation, dynamic parameter adaptive generation, and closed-loop verification of cleaning effect. Compared with traditional control solutions, this invention can accurately identify early hidden blockage problems of filter cartridges, avoid misjudgments caused by single parameter judgments, and automatically balance three conflicting objectives: cleaning effect, water consumption, and membrane damage risk. It achieves adaptive cleaning control throughout the entire life cycle of the filter cartridge, which not only significantly improves the accuracy of filter cartridge contamination judgment and cleaning efficiency, but also minimizes water waste and membrane mechanical damage risk during the cleaning process, effectively extending the actual service life of the filter cartridge and reducing the operating cost of the water purifier.

[0044] In this invention, through a built-in closed-loop iterative optimization mechanism, the pollution quantification model, strategy evaluation logic, and parameter generation algorithm can be continuously calibrated based on the effect verification data of each cleaning. As the usage time increases, the system's cleaning adaptability, water-saving capability, and membrane protection performance will gradually improve, forming a self-evolving intelligent control system. Without the need for manual preset cleaning rules and thresholds, it can adapt to water quality conditions in different regions and water usage habits in different usage scenarios, significantly reducing the user's usage threshold and maintenance costs, while providing an scalable technical framework for subsequent function iterations. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating the overall operation of the present invention.

[0046] Figure 2 This is a flowchart illustrating the generation of dynamic cleaning parameters in this invention;

[0047] Figure 3 This is a flowchart of the multi-objective evaluation of the cleaning strategy in this invention;

[0048] Figure 4 This is a flowchart of the closed-loop verification and iterative optimization of the cleaning effect in this invention. Detailed Implementation

[0049] The present invention will now be described in further detail with reference to the accompanying drawings and preferred embodiments.

[0050] Example 1, refer to Figure 1-4As shown, an AI intelligent control system for a water purifier includes a water purifier equipped with inlet and outlet water pressure sensors, inlet and outlet water total dissolved solids content sensors, flow sensors, an electric water pump, and a flushing valve actuator. The water purifier is also equipped with a filter cartridge contamination status reconstruction module, a cleaning strategy reward evaluation module, a dynamic cleaning parameter generation module, and a cleaning effect closed-loop verification module. Through the coordinated linkage of multiple modules, a fully closed-loop intelligent control system is constructed, from contamination status perception, cleaning strategy optimization, dynamic parameter generation to effect closed-loop verification, realizing adaptive cleaning control of the water purifier filter cartridge throughout its entire life cycle.

[0051] In this implementation scheme, when the system is running, the filter cartridge contamination state reconstruction module first collects five types of operating data simultaneously: water inlet viscosity, water inlet flow rate, effective membrane area of ​​the filter cartridge, inlet and outlet water pressure difference, and total dissolved solids content of the inlet and outlet water. The multi-source data is processed collaboratively by a fusion algorithm that embeds fluid dynamics and mass transfer physical constraints, and the quantified dimensionless contamination state index of the filter cartridge is output and synchronized to the other three modules.

[0052] The cleaning strategy reward evaluation module synchronously collects filter cartridge flux, cleaning water consumption, rated flushing water volume, cleaning peak pressure, and filter cartridge burst pressure threshold data before and after cleaning. Through multi-objective collaborative calculation, it completes a balanced evaluation of three types of indicators: flux recovery effect, water consumption cost, and membrane damage risk. It outputs a comprehensive cleaning strategy reward value and pushes it to the dynamic cleaning parameter generation module and the cleaning effect closed-loop verification module.

[0053] The dynamic cleaning parameter generation module connects the pollution state index and the comprehensive reward value, and combines them with three preset parameters: the pre-calibrated new filter cartridge pollution state benchmark value, the reference value of total dissolved solids content in the influent, and the differential pressure characteristic threshold. Through a nonlinear adaptive algorithm, it generates the optimal cleaning time and pulse frequency parameters, and sends them to the electric control water pump and flushing valve actuator to perform the cleaning action.

[0054] The cleaning effect closed-loop verification module collects the pollution state index and cleaning effect judgment threshold before and after cleaning. It obtains the cleaning effect verification result by comparing the degree of pollution improvement with the threshold. If the requirements are met, the cleaning process ends. If the requirements are not met, a second optimization cleaning is triggered. At the same time, the verification result is fed back to the dynamic cleaning parameter generation module and the cleaning strategy reward evaluation module to iteratively optimize the calculation logic of subsequent cleaning strategies.

[0055] This embodiment achieves intelligent control across the entire chain through the orderly operation of four modules, including quantification of pollution status, evaluation of multi-objective strategies, generation of dynamic parameters, and closed-loop verification of effects. This not only improves the accuracy of water purifier filter contamination judgment and cleaning efficiency, but also forms an iterative optimization mechanism. Through continuous feedback on the cleaning effect, the contamination quantification model and strategy evaluation logic can be continuously calibrated, extending the service life of the water purifier filter while reducing water consumption and membrane damage risks during the cleaning process.

[0056] Example 2, refer to Figure 1-4 As shown, the filter cartridge contamination state reconstruction module includes a fluid parameter acquisition unit, a water quality parameter acquisition unit, and a contamination state fusion calculation unit. The contamination state fusion calculation unit further includes a fluid resistance characteristic normalization subunit, a retention efficiency characteristic conversion subunit, and a contamination index output subunit.

[0057] During system operation, the fluid parameter acquisition unit first collects inlet water flow data by flow sensor and inlet and outlet water pressure difference data by inlet and outlet water pressure sensor according to the water purifier filter cartridge operation monitoring requirements. Based on the inlet water temperature, the inlet water viscosity is obtained by querying the pre-stored physical property table of water and extracting the fluid resistance characteristics that reflect the ease of fluid flow inside the water purifier filter cartridge.

[0058] The water quality parameter acquisition unit collects the total dissolved solids content data of the influent and effluent water of the water purifier filter cartridge through the total dissolved solids content sensor, calculates the ratio of the total dissolved solids content of the influent and effluent water, and extracts the membrane retention efficiency characteristic that reflects the water purifier filter cartridge's ability to retain dissolved solids.

[0059] The pollution status fusion calculation unit takes into account the above two types of features and the fixed parameter of the effective membrane area of ​​the filter element, and obtains the filter element pollution status index through multi-feature dimensionless fusion calculation. This accurately quantifies the degree of blockage inside the water purifier filter element and avoids misjudgment of the degree of pollution caused by a single feature.

[0060] Specifically, the fluid resistance characteristic normalization subunit first obtains parameters such as influent viscosity, influent flow rate, effective membrane area of ​​the filter element, and influent-outfluent pressure difference. The normalized fluid resistance value is then calculated by using the ratio of fluid transport flux to filtration resistance. The calculation formula is as follows:

[0061] ;

[0062] In the formula, R f The normalized fluid resistance value is obtained, where μ is the influent dynamic viscosity, which is obtained by referring to the physical properties table of water based on the influent temperature. Q in The influent flow rate is directly collected by the flow sensor. A mem P represents the effective membrane area of ​​the filter element and is a fixed factory parameter for the filter element. in Pout These are the inlet and outlet water pressures of the filter element, respectively, which are collected by inlet and outlet water pressure sensors.

[0063] Furthermore, the retention efficiency characteristic transformation subunit obtains the ratio of total dissolved solids content in the influent and effluent, and then uses a square root nonlinear transformation operation to obtain the standardized retention efficiency value. The calculation formula is as follows:

[0064] ;

[0065] In the formula, η is the standardized retention efficiency value, ranging from [1, +∞), reflecting the water purifier filter element's ability to retain dissolved pollutants. A higher value indicates a higher retention efficiency for total dissolved solids (TDS) and better membrane performance. A value closer to 1 indicates a lower retention efficiency, and the membrane may be damaged. in TDS out These are the total dissolved solids (TDS) contents of the influent and effluent, respectively, which are collected by the TDS sensor.

[0066] After calculating the normalized fluid resistance value R f After being combined with the standardized retention efficiency value η, the result is output to the pollution index output subunit. The pollution index output subunit performs a product operation on these two types of features to obtain the pollution state index, calculated as follows:

[0067] ;

[0068] In the formula, E is the filter element pollution state index, with a value range of [0, +∞). It is a comprehensive quantitative indicator of the overall pollution level of the water purifier filter element. The higher the value, the more serious the clogging of the water purifier filter element.

[0069] In this embodiment, by quantifying the fusion of fluid dynamics characteristics and water quality characteristics, the internal pollution state of the water purifier filter element is characterized without deviation. The output pollution state index can objectively reflect the actual degree of clogging of the water purifier filter element, providing accurate input parameters for the formulation of subsequent cleaning strategies. Compared with the traditional single pressure difference or total dissolved solids content threshold judgment scheme, the accuracy of pollution degree identification is significantly improved, which can effectively identify early hidden clogging problems and avoid problems such as performance degradation of the water purifier filter element or untimely cleaning caused by judgment lag.

[0070] Example 3, referring to Figure 1-4 As shown, the cleaning strategy reward evaluation module includes a flux recovery rate calculation unit, a water consumption cost calculation unit, a membrane damage risk calculation unit, and a multi-objective fusion output unit. The multi-objective fusion output unit includes an indicator standardization sub-unit, a collaborative operation sub-unit, and a reward level division sub-unit.

[0071] During system operation, the flux recovery rate calculation unit collects the production water flow data before and after cleaning through the flow sensor, calculates the filter cartridge flux before and after cleaning by combining the fixed parameter of the effective membrane area of ​​the filter cartridge, and then performs a ratio calculation with the rated flux parameter of the new filter cartridge to obtain the flux recovery degree index.

[0072] The water consumption cost calculation unit uses a flow sensor to count the total water consumption of the cleaning process, calculates the ratio with a fixed parameter of rated rinsing water volume, and obtains the cleaning water consumption cost index through natural exponential nonlinear conversion.

[0073] The membrane damage risk calculation unit collects peak pressure data during the cleaning process through pressure sensors, and calculates the ratio with the fixed parameter of filter element burst pressure threshold to obtain the membrane damage risk index.

[0074] The multi-objective fusion output unit connects to the above three types of indicators and obtains a comprehensive reward value through the multiplication of multiple indicators, thereby achieving an automatic balance between the three objectives of flux recovery, water consumption control, and membrane damage protection.

[0075] The formula for calculating the overall reward value is as follows:

[0076] ;

[0077] in, ;

[0078] In the formula, Z is the calculated comprehensive reward value, ranging from (0,1], used to quantitatively evaluate the comprehensive performance of the cleaning strategy, and is the core evaluation index for reinforcement learning to optimize the cleaning strategy. J is the filter cartridge flux. post J represents the filter cartridge flux after cleaning, reflecting the recovery of the filter cartridge's filtration capacity after cleaning. pre The filter element flow rate before cleaning reflects the degree of clogging before cleaning, and is calculated using the filter element flow rate calculation formula, J. new V is the initial standard permeation flux of a brand new filter element, serving as a performance benchmark. cl V represents the water consumption for cleaning, and V represents the total water consumed during this cleaning process. ref P represents the rated flushing water volume, which is the rated water consumption required for cleaning under standard test conditions, and serves as the benchmark value for water consumption evaluation. pu The peak cleaning pressure, collected by a pressure sensor during the cleaning process, represents the maximum pressure value encountered during cleaning. Higher pressures pose a greater risk of mechanical damage to the membrane. bu The burst pressure threshold is the maximum pressure that the filter membrane of a water purifier can withstand. Exceeding this value will cause irreversible damage to the membrane.

[0079] Furthermore, in the formula for calculating the filter cartridge flow rate J, Q represents the current water production flow rate, which is the actual purified water flow rate produced by the filter cartridge. This directly reflects the actual water flow capacity of the water purifier filter cartridge. The lower the flow rate, the more severe the blockage. This flow rate is collected in real time by the flow sensor at the water outlet of the water purifier. A mem This refers to the effective membrane area of ​​the filter element, which is a fixed factory parameter for the filter element.

[0080] In the formula, The flux recovery rate is a value in the range of [0,1), which reflects the degree of flux recovery of the filter element after cleaning. The closer the value is to 1, the closer the flux recovery is to the state of a brand new filter element after cleaning, and the better the cleaning effect.

[0081] This is a water consumption penalty term with a value range of (0,1]. It implements non-linear water consumption penalty through an exponential function: the more water consumption exceeds the rated value, the faster this term decays and the greater the penalty, thus encouraging water-saving cleaning.

[0082] This is a membrane damage protection item, with a value range of [0,1). The closer the peak pressure is to the burst threshold, the closer this item is to 0, and the greater the penalty, thus avoiding high pressure damage to the filter element.

[0083] Among them, the closer the comprehensive reward value Z is to 1, the better the overall performance of the cleaning strategy, achieving the optimal balance in the three dimensions of flux recovery, water saving, and membrane protection. The closer it is to 0, the worse the overall performance of the strategy, which may have problems such as incomplete cleaning, excessive water consumption, or damage to the filter element.

[0084] By multiplying three types of indicators, the system simultaneously measures three conflicting objectives: flux recovery effect, water consumption, and membrane damage risk, thereby automatically achieving multi-objective synergistic optimization.

[0085] Furthermore, the application of the comprehensive reward value Z is divided into two stages: strategy pre-evaluation and post-cleaning effect evaluation. During strategy pre-evaluation, the post-cleaning flux J... post These are simulation predictions and can be obtained without actually performing the cleaning process.

[0086] Based on historically accumulated pollution state index E, cleaning parameter combinations, and post-cleaning flux correlation datasets, the system pre-trains a pollution-cleaning effect mapping model. When the dynamic cleaning parameter generation module generates multiple sets of candidate cleaning parameters, the cleaning strategy reward evaluation module inputs the current pollution state index E and the candidate parameters. Through the mapping model, it directly simulates and predicts the flux recovery result after the execution of the set of parameters, and then calculates the comprehensive reward value. The parameter combination with the highest reward value is selected as the final execution plan, thus achieving strategy optimization before cleaning.

[0087] During the post-cleaning effect evaluation phase, after the cleaning process is completed, the produced water volume Q is measured using the outlet flow sensor, combined with the effective membrane area A of the filter element. mem The actual flow rate J after cleaning is calculated according to the formula for filter element flow rate J. post The comprehensive reward value Z calculated at this time is used to evaluate the overall performance of this actual cleaning, and serves as feedback data to iteratively optimize the accuracy of the pollution-cleaning effect mapping model, thereby improving the accuracy of subsequent parameter generation and strategy evaluation.

[0088] In this embodiment, a comprehensive quantitative evaluation of the cleaning strategy is achieved through multi-objective collaborative assessment. It can automatically balance three conflicting objectives: cleaning effect, water consumption, and membrane damage risk. This avoids the problem of neglecting one aspect for another caused by a single objective dominating strategy optimization. It can minimize water consumption and membrane damage risk during the cleaning process while ensuring the cleaning effect, thus achieving optimal balance in multiple dimensions.

[0089] Example 4, refer to Figure 1-4 As shown, the dynamic cleaning parameter generation module includes a cleaning duration calculation unit, a pulse frequency calculation unit, and a parameter safety verification unit. The cleaning duration calculation unit includes a pollution degree difference calculation subunit, a water quality impact factor calculation subunit, and a duration output subunit.

[0090] During system operation, the cleaning time calculation unit first obtains five types of parameters: filter element contamination state index E, pre-calibrated new filter element contamination state baseline value, total dissolved solids content data of influent collected by the influent total dissolved solids content sensor, local municipal influent total dissolved solids content reference value, and basic cleaning time. The optimal cleaning time is obtained through nonlinear adaptive calculation of logarithmic function.

[0091] Specifically, the pollution level difference calculation subunit first obtains the baseline value of the pollution state of the new filter cartridge and the current pollution state index E. The relative pollution level is then calculated by the ratio of the difference to the baseline value, as shown in the following formula:

[0092] ;

[0093] In the formula, ΔE represents the calculated relative contamination level, ranging from [0,1]. A larger value indicates a higher level of contamination in the filter element. new The baseline value for the contamination state of a brand new filter element is the baseline value for the contamination state index of a brand new, unused filter element under standard test conditions. It is a fixed constant used as a reference zero point for the degree of contamination.

[0094] The water quality impact factor calculation subunit obtains the total dissolved solids content and reference value of the influent, and calculates the water quality impact coefficient through ratio calculation. The calculation formula is as follows:

[0095] ;

[0096] In the formula, α is the water quality impact coefficient, reflecting the current total dissolved solids content of the influent relative to the local average level, and is used to adjust the cleaning duration. TDS in The total dissolved solids (TDS) content of the influent is obtained from the influent and effluent TDS sensors. ref This is a reference value for the total dissolved solids content of local municipal influent, and is a fixed calibration parameter.

[0097] Furthermore, when α > 1, it indicates that the current total dissolved solids content of the influent is higher than the local average level and the pollutant concentration is higher, so the cleaning time needs to be extended. When α < 1, it indicates that the current total dissolved solids content of the influent is lower than the local average level and the pollutant concentration is lower, so the cleaning time can be appropriately shortened.

[0098] By calculating the ratio of the current total dissolved solids content in the influent to the local average level, the pollution level of the current water quality relative to normal conditions is quantified, enabling the cleaning duration to adaptively adapt to fluctuations in the influent water quality.

[0099] Furthermore, the duration output subunit obtains the relative pollution level ΔE and the water quality impact coefficient α, which are then transformed using a natural logarithm function and multiplied by the basic cleaning duration to obtain the optimal cleaning duration. The calculation formula is as follows:

[0100] ;

[0101] In the formula, T cl The calculated optimal cleaning time is the final output cleaning execution time, dynamically adjusted based on the filter cartridge fouling level and influent water quality to ensure the best cleaning effect with minimal water consumption and membrane damage risk. S base The baseline cleaning time is the reference time required to clean a brand new water purifier filter under standard test conditions. It is obtained by cleaning experiments and is the basic reference value for cleaning time. ln() is the natural logarithm function, and ΔE×α is the comprehensive influencing factor, which comprehensively reflects the combined influence of the degree of pollution of the water purifier filter and the quality of the incoming water. The larger the value, the longer the required cleaning time.

[0102] By using the natural logarithm function to achieve nonlinear adaptive adjustment of the cleaning time, it is possible to ensure that the cleaning time is not too long and wastes water when the pollution is light, while ensuring that the cleaning time is sufficient to achieve the cleaning effect when the pollution is heavy, thus achieving precise dynamic control of the cleaning time.

[0103] Furthermore, the pulse frequency calculation unit simultaneously acquires four types of parameters: the pressure difference data between the inlet and outlet water of the water purifier filter element collected by the inlet and outlet water pressure sensors, the pressure difference characteristic threshold, the preset minimum pulse frequency, and the maximum pulse frequency. The acquired parameters are then used to calculate the optimal pulse frequency. The specific calculation formula is as follows:

[0104] ;

[0105] In the formula, f pu The optimal cleaning pulse frequency is the final output. A higher value indicates a more concentrated water flow pulse and stronger impact force during rinsing, suitable for severely clogged scenarios. A lower value indicates a gentler water flow during rinsing, suitable for mildly clogged scenarios or scenarios with a high risk of damage. min f max These are the minimum and maximum pulse frequencies, respectively, both calibrated through membrane fatigue experiments. P in P out These are the inlet and outlet water pressures of the water purifier filter element, respectively, which are collected by inlet and outlet water pressure sensors. △P char The pressure difference characteristic threshold is calibrated through a pollution gradient experiment to reflect the sensitivity of pressure difference to pulse frequency regulation.

[0106] Furthermore, P in —P out The pressure difference between the inlet and outlet of the water purifier filter element is the working pressure difference of the filter element, which directly reflects the degree of clogging of the water purifier filter element: the greater the pressure difference, the more serious the clogging inside the water purifier filter element, and the stronger the pulse flushing force is required.

[0107] The term is an exponential decay term with a value range of (0,1]. It decays exponentially as the pressure difference increases, achieving a non-linear increase in pulse frequency with the degree of blockage: the frequency increases slowly when the pressure difference is small, and the frequency rises rapidly when the pressure difference is close to the threshold, which meets the actual needs of water purifier filter flushing. That is, mild blockage only requires gentle pulses, while severe blockage requires dense and strong impact pulses.

[0108] The parameter safety verification unit verifies the safety range of the generated optimal cleaning time and pulse frequency, corrects parameters that exceed the allowable working range of the water purifier filter element to the safe range, and outputs the final execution parameters to the electric control water pump and flushing valve actuator.

[0109] Specifically, the parameter safety verification unit presets safe operating ranges for two types of parameters: cleaning time and pulse frequency, respectively, which are calibrated through durability and membrane fatigue impact tests. The cleaning time range includes the minimum allowable time to ensure the minimum rinsing effect and the maximum allowable time to avoid excessive rinsing and damage to the membrane. The pulse frequency range includes the minimum allowable frequency to ensure the impact force of water flow and the maximum allowable frequency to avoid membrane fiber breakage.

[0110] During verification, the optimal cleaning time T generated by the calculation is first... cl Compared with the interval threshold, if it is lower than the minimum value, it is directly corrected to the minimum allowable duration; if it is higher than the maximum value, it is corrected to the maximum allowable duration; if it is within the interval, the original calculated value is retained. Similarly, for the optimal pulse frequency f...pu Boundary checks are performed, and parameters that exceed the range are uniformly truncated and corrected to the corresponding range endpoint values ​​to ensure that the output parameters always comply with the water purifier filter design safety specifications.

[0111] In this embodiment, by jointly regulating the degree of pollution and water quality parameters, the cleaning time and pulse frequency are adaptively and dynamically adjusted. The optimal cleaning parameters can be matched according to the actual clogging state of the water purifier filter and the inlet water quality conditions. While ensuring the cleaning effect, unnecessary water consumption and membrane damage risk are reduced, further improving the adaptability and rationality of the cleaning strategy.

[0112] Example 5, refer to Figure 1-4 As shown, the cleaning effect closed-loop verification module includes a pollution improvement calculation unit, a threshold comparison unit, a verification result output unit, a parameter iteration optimization unit, and a strategy feedback adjustment unit.

[0113] During system operation, the pollution improvement calculation unit obtains the pollution state index E before and after cleaning. By calculating the ratio of the change in the pollution index to the pollution index before cleaning, the pollution improvement rate index, which reflects the proportion of pollution elimination after cleaning, is obtained. The specific calculation formula is as follows:

[0114] ;

[0115] In the formula, γ is the pollution improvement rate, with a value range of (−∞, 1], reflecting the improvement ratio of the filter element's pollution state after cleaning, and E pre The filter element contamination status index before cleaning is calculated and output by the filter element contamination status reconstruction module, E. post The filter element contamination status index is calculated and output by the filter element contamination status reconstruction module after cleaning.

[0116] The threshold comparison unit obtains the pollution improvement rate index and the cleaning effect judgment threshold calibrated in the laboratory. Abnormal negative values ​​are removed through linear rectification operation to obtain the verification value used to judge the cleaning effect. The specific calculation formula is as follows:

[0117] ;

[0118] In the formula, S is the cleaning effect verification value, which ranges from [0, +∞), used to determine whether the cleaning meets the requirements; γ is the pollution improvement rate; and δ threshold The threshold for judging the cleaning effect is a fixed parameter calibrated in the laboratory. It represents the minimum improvement rate required to determine whether the cleaning is qualified. ReLU() is a linear rectified function with one-sided suppression characteristics: when the input value γ—δ threshold When the value is greater than 0, the original value is output; when the input value is γ - δ, the original value is output. thresholdWhen the value is ≤0, the output is 0, which can clearly distinguish whether the cleaning meets the standard and eliminate the interference of negative improvement rate caused by abnormal increase in the pollution index, thus improving the robustness of the verification logic.

[0119] Specifically, when S > 0, it means that the pollution improvement rate exceeds the judgment threshold, the cleaning effect meets the standard, and no secondary cleaning is required. When S ≤ 0, it means that the pollution improvement rate does not reach the threshold, the cleaning effect does not meet the standard, and a secondary cleaning process needs to be triggered. At the same time, abnormal negative improvement rate cases with γ < 0 are filtered out to avoid misjudgment.

[0120] The verification result output unit classifies the cleaning effect level based on the comparison between the cleaning effect verification value and the judgment threshold. For verification results that meet the threshold requirements, the cleaning is marked as completed, and the current cleaning process ends. For verification results that do not meet the threshold requirements, a secondary optimization cleaning process is directly triggered. At the same time, the verification results are synchronized to the parameter iteration optimization unit and the strategy feedback adjustment unit.

[0121] The parameter iteration optimization unit receives the cleaning effect verification results and the parameter data used in this cleaning. Through gradient descent iteration of the verification data, it updates the function coefficients of the cleaning duration calculation and pulse frequency calculation, and continuously optimizes the calculation logic of cleaning duration and pulse frequency.

[0122] The strategy feedback adjustment unit receives the cleaning effect verification results and the comprehensive reward value Z of this cleaning strategy. It iteratively updates the index weights of the multi-objective fusion calculation through verification data, and continuously adjusts the balance logic of the three objectives of flux recovery, water consumption control and membrane damage protection to improve the adaptability and rationality of subsequent cleaning strategies.

[0123] It should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should also be within the scope of protection of this invention.

Claims

1. A water purifier AI intelligent control system, characterized in that: The water purifier includes an inlet and outlet water pressure sensor, an inlet and outlet water total dissolved solids content sensor, a flow sensor, an electric water pump, and a flushing valve actuator. The water purifier is also equipped with a filter cartridge contamination status reconstruction module, a cleaning strategy reward evaluation module, a dynamic cleaning parameter generation module, and a cleaning effect closed-loop verification module. The filter cartridge fouling state reconstruction module is configured to collect data on influent viscosity, flow rate, membrane area, influent-outfluent pressure difference, and total dissolved solids content, and to perform physical constraint fusion processing on the multi-source data to obtain the fouling state index; The cleaning strategy reward evaluation module is configured to collect data on flux, water consumption, rated flushing water volume, peak pressure, and burst threshold before and after cleaning, obtain a comprehensive reward value through collaborative calculation, and output it to the dynamic cleaning parameter generation module and the cleaning effect closed-loop verification module. The dynamic cleaning parameter generation module receives the pollution index and reward value, and combines them with the calibrated benchmark value, water quality reference value, and pressure difference threshold to generate the optimal cleaning duration and pulse frequency, which are then output to the electric control water pump and flushing valve actuator. The cleaning effect closed-loop verification module collects the pollution index and effect judgment threshold before and after cleaning. The verification result is obtained by comparing the degree of pollution improvement with the threshold, which triggers the secondary cleaning process.

2. The AI ​​intelligent control system for a water purifier according to claim 1, characterized in that: The filter cartridge contamination state reconstruction module includes a fluid parameter acquisition unit, a water quality parameter acquisition unit, and a contamination state fusion calculation unit. The fluid parameter acquisition unit is configured to collect data on inlet water viscosity, inlet water flow rate, and inlet-outlet water pressure difference of the filter element according to the filter element operation monitoring requirements, and extract fluid resistance characteristics. The water quality parameter acquisition unit is configured to collect total dissolved solids content data of the filter cartridge inlet and outlet water, calculate the ratio of total dissolved solids content of inlet and outlet water, and extract membrane retention efficiency characteristics. The pollution state fusion calculation unit is configured to acquire fluid resistance characteristics, membrane retention efficiency characteristics, and effective membrane area parameters of the filter element. Through fusion calculation, the filter element pollution state index is obtained, which quantifies the degree of internal blockage of the filter element.

3. The AI ​​intelligent control system for a water purifier according to claim 2, characterized in that: The pollution state fusion calculation unit includes a fluid resistance characteristic normalization subunit, a retention efficiency characteristic conversion subunit, and a pollution index output subunit: The fluid resistance characteristic normalization subunit is configured to obtain parameters such as influent viscosity, influent flow rate, effective membrane area of ​​filter element, and influent-outfluent pressure difference, and to obtain the normalized fluid resistance value by calculating the ratio of fluid transport flux to filtration resistance. The retention efficiency characteristic conversion subunit is configured to obtain the ratio of total dissolved solids content in influent and effluent, and to obtain a standardized retention efficiency value through a square root nonlinear conversion operation. The pollution index output sub-unit is configured to obtain the normalized fluid resistance value and the standardized retention efficiency value, and obtain the pollution state index by multiplying the two types of features, thereby realizing the coupled quantification of multi-dimensional pollution features.

4. The AI ​​intelligent control system for a water purifier according to claim 1, characterized in that: The cleaning strategy reward evaluation module includes a flux recovery rate calculation unit, a water consumption cost calculation unit, a membrane damage risk calculation unit, and a multi-objective fusion output unit. The flux recovery rate calculation unit is configured to obtain the flux parameters of the filter element before and after cleaning and the rated flux parameters of the new filter element, and to obtain the flux recovery degree index by calculating the ratio of the change in flux before and after cleaning to the rated flux. The water consumption cost calculation unit is configured to obtain the parameters of cleaning water consumption and rated flushing water volume, and obtain the cleaning water consumption cost index by calculating the ratio of water consumption to the rated value and then performing a natural exponential nonlinear transformation. The membrane damage risk calculation unit is configured to obtain the peak cleaning pressure and the filter element burst pressure threshold parameters, and to obtain the membrane damage risk index by calculating the ratio of the peak pressure to the burst threshold. The multi-objective fusion output unit is configured to acquire flux recovery level indicators, cleaning water consumption cost indicators, and membrane damage risk indicators. It obtains a comprehensive reward value through multiplication of multiple indicators, thereby achieving automatic balance of multiple objectives.

5. The AI ​​intelligent control system for a water purifier according to claim 4, characterized in that: The multi-objective fusion output unit includes an indicator standardization subunit, a collaborative computation subunit, and a reward level division subunit: The index standardization subunit is configured to perform linear normalization on the flux recovery degree index, the cleaning water consumption cost index, and the membrane damage risk index, so that the three types of indexes are in the same numerical range. The collaborative computing subunit is configured to multiply the three standardized indicators to obtain the original reward value, thereby achieving collaborative optimization of multiple objectives. The reward level division sub-unit is configured to divide reward levels based on the original reward value, mark high reward value strategies as executable strategies, and mark low reward value strategies as strategies to be optimized.

6. The AI ​​intelligent control system for a water purifier according to claim 1, characterized in that: The dynamic cleaning parameter generation module includes a cleaning duration calculation unit, a pulse frequency calculation unit, and a parameter security verification unit. The cleaning time calculation unit is configured to acquire the filter cartridge contamination status index, the pre-calibrated baseline value of the contamination status of a brand new filter cartridge, the total dissolved solids content of the influent, the reference value of the total dissolved solids content of the local municipal influent, and the basic cleaning time parameters calibrated in the laboratory. The acquired parameters are then used to obtain the optimal cleaning time through a nonlinear adaptive calculation of a logarithmic function. The pulse frequency calculation unit is configured to acquire the pressure difference between the filter cartridge inlet and outlet, the pressure difference characteristic threshold calibrated in the laboratory, the pre-calibrated minimum pulse frequency, and the pre-calibrated maximum pulse frequency parameters, and to obtain the optimal pulse frequency by performing an adaptive interval mapping operation using an exponential function on the acquired parameters. The parameter safety verification unit is configured to verify the optimal cleaning time and optimal pulse frequency within a safe range, correct parameters that exceed the safe range to within the allowable range, and output the final execution parameters.

7. The AI ​​intelligent control system for a water purifier according to claim 6, characterized in that: The cleaning time calculation unit includes a pollution degree difference calculation subunit, a water quality impact factor calculation subunit, and a time output subunit: The pollution level difference calculation subunit is configured to obtain the pollution status baseline value of a brand new filter cartridge and the current pollution status index, and calculate the relative pollution level by the ratio of the difference to the baseline value. The water quality impact factor calculation subunit is configured to obtain the total dissolved solids content of the influent and the reference value, and then calculate the water quality impact coefficient by ratio calculation. The duration output subunit is configured to obtain the relative pollution level and water quality influence coefficient, and after transformation by the natural logarithm function, multiply it by the basic cleaning duration to obtain the optimal cleaning duration, thereby realizing the joint control of the cleaning duration by the pollution level and water quality parameters.

8. The AI ​​intelligent control system for a water purifier according to claim 1, characterized in that: The closed-loop verification module for cleaning effect includes a pollution improvement calculation unit, a threshold comparison unit, and a verification result output unit. The pollution improvement calculation unit is configured to obtain the filter element pollution status index before and after cleaning, and to obtain the pollution improvement rate index by calculating the ratio of the change in pollution index to the pollution index before cleaning. The threshold comparison unit is configured to obtain the pollution improvement rate index and the cleaning effect judgment threshold, and remove negative values ​​through linear rectification operation to obtain the cleaning effect verification value. The verification result output unit is configured to determine the cleaning effect based on the cleaning effect verification value, mark the verification result that meets the requirements as cleaning completed, and trigger a secondary cleaning process for the verification result that does not meet the requirements.

9. The AI ​​intelligent control system for a water purifier according to claim 8, characterized in that: The closed-loop verification module for cleaning effect also includes a parameter iteration optimization unit and a strategy feedback adjustment unit. The parameter iteration optimization unit is configured to feed back the cleaning effect verification results to the dynamic cleaning parameter generation module, and iteratively update the logarithmic function coefficients for calculating the cleaning duration and the exponential function coefficients for calculating the pulse frequency through the verification data, thereby optimizing the calculation logic of the cleaning duration and pulse frequency. The strategy feedback adjustment unit is configured to feed back the cleaning effect verification results to the cleaning strategy reward evaluation module. By iteratively optimizing the indicator weights of multi-objective fusion calculation through verification data, the rationality of subsequent cleaning strategies is improved.

Images