Device life prediction method and system based on working state change curve

Abstract

The invention relates to the technical field of device service life prediction, in particular to a device service life prediction method and system based on a working state change curve, and provides the following scheme that the working state change curve of a device in the working process is obtained, resampling of a flow curve is achieved based on accumulative volume mapping, and the service life of the device is predicted. Hysteretic flyback recheck is carried out in the neighborhood of the candidate demarcation points to determine a final demarcation point and construct and update an equivalent window. Further, correction is carried out according to the volume difference among a plurality of updating equivalent windows, when the difference value is within a tolerance range, the difference value is directly adopted, otherwise, a correction factor is introduced to correct the updating window, and compensation is carried out in combination with the self-cleaning recovery characteristic of the device during periodic switching, so that the updating equivalent window which is more practical is obtained. Accurate characterization of the attenuation and recovery process of the device is realized, and a reliable basis is provided for service life prediction and operation and maintenance strategies of the device.

Description

Technical Field

[0001] This invention relates to the field of device life prediction technology, and in particular to a device life prediction method and system based on operating state change curves. Background Technology

[0002] In existing technologies, device performance evaluation primarily relies on flow monitoring and empirical threshold judgments during operation. However, these methods generally suffer from insufficient accuracy, failing to accurately reflect performance changes at different operating stages, leading to unreliable judgments regarding cleaning and replacement timing. Furthermore, existing flow curve processing methods are susceptible to data fluctuations, resulting in significant errors in identifying the device's degradation process and affecting the calculation of the updated equivalent window. Moreover, existing technologies fail to adequately characterize the device's self-cleaning function during cycle switching, causing discrepancies between window update results and actual operating conditions. These deficiencies collectively contribute to inaccurate device lifetime prediction, thereby impacting system stability and maintenance efficiency.

[0003] To address the above issues, this application presents a method and system for predicting device lifespan based on operating state change curves. Summary of the Invention

[0004] The technical problem this invention aims to solve is to address the shortcomings of existing technologies by providing a method and system for predicting device lifetime based on operating state change curves. This method acquires the operating state change curves of the device during operation, resamples the flow rate curve based on cumulative volume mapping, and performs hysteresis scan verification within the neighborhood of candidate boundary points to determine the final boundary point and construct an updated equivalent window. Furthermore, corrections are made based on the volume differences between multiple updated equivalent windows. When the difference is within the tolerance range, it is directly adopted; otherwise, a correction factor is introduced to modify the updated window, and compensation is made by combining the self-cleaning recovery characteristics of the device during cycle switching, thereby obtaining an updated equivalent window that more closely reflects reality.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A device lifetime prediction method based on operating state change curves is applied to an MCU in an offline fine filtration system. The MCU maintains an updated equivalent window and a device lifetime table corresponding to the updated equivalent window. The updated equivalent window is assigned values ​​based on the device's operating state change curve. The method includes:

[0007] Obtain the status information of the current equivalent window and the working object of the current work cycle;

[0008] Based on the status information, an update equivalent window is determined. After the current work cycle ends, the update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the work status change curve within the current work cycle. The size of the first equivalent window is the same as that of the current equivalent window, and the size of the second equivalent window is the same as that of the update equivalent window.

[0009] The updated equivalent window is modified based on the first equivalent window and the second equivalent window and then replaces the current equivalent window.

[0010] The device lifespan is obtained by comparing the current equivalent window with the device lifespan table.

[0011] Based on the status information, determine the update equivalent window, including:

[0012] Based on the status information of the working object of the device, the impurity accumulation rate of the current working cycle is obtained, wherein the status information includes at least one of turbidity value, sand content, particle size distribution and dissolved solids content;

[0013] Based on the impurity accumulation rate and the current equivalent window, a predicted decay is calculated, and the predicted decay is applied to the current equivalent window to obtain an updated equivalent window.

[0014] Based on the impurity accumulation rate and the current equivalent window, the predicted decay is calculated, including:

[0015] The current equivalent window is divided into multiple continuous sub-windows, wherein the length of the continuous sub-windows is calculated based on the changing trend of the impurity accumulation rate;

[0016] For each consecutive sub-window, the attenuation contribution value is calculated based on the corresponding impurity accumulation rate, and the attenuation contribution values ​​are summed to obtain the first attenuation curve.

[0017] Align the first attenuation curve with the predicted working state change curve within the current working cycle to obtain the hysteresis compensation factor.

[0018] The first attenuation curve is compensated according to the hysteresis compensation factor to obtain the second attenuation curve;

[0019] The predicted attenuation amount is determined based on the descending slope of the second attenuation curve and the cumulative attenuation area.

[0020] The status information also includes the influent flow rate. The update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the operating status change curve of the monitoring device within the current working cycle, including:

[0021] The working status change curve of the device in the current working cycle is obtained, and the working status change curve is compared with the reference cleaning curve to obtain the candidate boundary point, wherein the reference cleaning curve is calculated based on the current equivalent window and the inlet water flow rate.

[0022] For each candidate boundary point, the working state change curve is resampled in the neighborhood of the candidate boundary point, and hysteresis backscan verification is performed based on the resampling results to obtain the final boundary point.

[0023] Based on the position of the final dividing point on the working state change curve, the updated equivalent window is divided to obtain a first equivalent window and a second equivalent window, and the first equivalent window is assigned a value based on the current equivalent window.

[0024] The working state change curve is compared with the baseline cleaning curve to obtain candidate boundary points, including:

[0025] A first pointer and a second pointer are configured in the time axis corresponding to the working state change curve and the reference cleaning curve, wherein the first pointer is used to traverse the working state change curve and the second pointer is used to traverse the reference cleaning curve;

[0026] The first and second pointers are advanced within a preset candidate time window, and the time scaling factor and stability index within the corresponding candidate time window are calculated according to the advancement mode. The stability index includes the variance and range of the candidate time window.

[0027] When the time scaling factor is greater than or equal to the preset time scaling threshold, and the offset value between the stability index and the tolerance band is greater than the preset deviation threshold, the average time of the corresponding candidate time window is used as the candidate boundary point, wherein the tolerance band is calculated based on the minimum cleaning volume of the benchmark cleaning curve.

[0028] The propulsion mode includes:

[0029] The first pointer advances synchronously with the second pointer;

[0030] The first pointer advances fast while the second pointer advances slowly.

[0031] The first pointer advances slowly while the second pointer advances quickly.

[0032] The advance frequency of the pointers corresponding to the fast forward and the slow forward is calculated based on the average slope of the working state change curve and the reference cleaning curve in the corresponding candidate time window.

[0033] The step of resampling the working state change curve within the neighborhood of the candidate boundary point and performing hysteresis scan verification based on the resampling results includes:

[0034] The cumulative volume is calculated based on the working state change curve. The working state change curves corresponding to the neighborhood of the candidate boundary point are resampled based on the cumulative volume to obtain candidate pairs corresponding to the uplink and downlink segments. The volume span of the candidate pairs is greater than or equal to the minimum cleaning volume.

[0035] Align and interpolate the uplink and return segments on the volume axis, calculate the flow difference between the uplink and return segments under a consistent volume grid, and obtain the corresponding pseudohysteresis area by integration.

[0036] The pseudohysteresis areas of each candidate boundary point are sorted in descending order, and the first candidate boundary point is output as the final boundary point. The descending sorting also includes weighting the pseudohysteresis areas based on the area of ​​the neighborhood of the candidate boundary point.

[0037] The updated equivalent window is corrected based on the first equivalent window and the second equivalent window, including:

[0038] Calculate the volume difference between the first equivalent window and the second equivalent window;

[0039] If the volume difference is less than or equal to a preset tolerance threshold, the second equivalent window is used as the corrected updated equivalent window.

[0040] If the volume difference is greater than a preset tolerance threshold, a correction factor is calculated based on the volume difference, and the updated equivalent window is weighted and assigned a value based on the correction factor to obtain the corrected updated equivalent window.

[0041] The device lifespan is obtained by comparing the current equivalent window with the device lifespan table, including:

[0042] Input the current equivalent window into the device life table and retrieve the corresponding life reference value;

[0043] The remaining lifespan of the device is determined based on the lifespan reference value, and the device lifespan is output.

[0044] A device life prediction system based on operating state change curves, the system being applied to an MCU in an offline fine filtration system, includes a data acquisition module, a processing module, and a life assessment module, wherein:

[0045] The data acquisition module is used to acquire the status information of the water sample to be filtered and the working status change curve of the device in the current working cycle, and transmit the data to the calculation and processing module;

[0046] The computational processing module, located in the MCU, is used to maintain and update the equivalent window and the device lifetime table, and to update, correct, divide, and predict the decay amount of the equivalent window.

[0047] The lifespan assessment module is used to compare the updated equivalent window with the device lifespan table to obtain the remaining lifespan of the device.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] This invention, by introducing a volume axis-based resampling method and a hysteresis retracement verification mechanism, effectively reduces interference caused by instantaneous fluctuations and uneven sampling during the identification of candidate boundary points, thereby improving the stability and accuracy of boundary point positioning. By comparing the differences between the uplink and downlink segments under a consistent volume grid and using the pseudo-hysteresis area as the criterion, it not only enhances the identifiability of the device's attenuation and recovery characteristics but also ensures that the boundary point selection results are consistent with the actual operation process. Attached Figure Description

[0050] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0051] Figure 1 This is a schematic diagram illustrating the lifespan degradation of the device according to an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the updated equivalent window segmentation in an embodiment of the present invention;

[0053] Figure 3 This is an exemplary application scenario diagram of an embodiment of the present invention;

[0054] Figure 4 This is a schematic flowchart illustrating a device lifetime prediction method based on a working state change curve, according to an embodiment of this application.

[0055] Figure 5 This is a schematic diagram illustrating the segmentation process of updating the equivalent window in an embodiment of this application;

[0056] Figure 6 This is a schematic diagram illustrating the principle of candidate boundary point screening in an embodiment of this application;

[0057] Figure 7 This is a schematic diagram illustrating the correction principle of updating the equivalent window in the embodiments of this application. Detailed Implementation

[0058] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0059] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0060] This application applies to equipment products with periodic effluent characteristics and multi-factor water quality disturbance conditions. Their life prediction process is limited by the differences in water quality in the operating environment and the dynamic fluctuations in the effluent performance of the equipment. In the process of equipment life assessment, there are prediction bias problems caused by the nonlinear cumulative effect of water quality and random disturbances in effluent flow.

[0061] Application scenarios include, but are not limited to:

[0062] A household or commercial water purification system with multi-stage series connection and periodic backwashing function;

[0063] Offline water treatment devices that use high-precision equipment for pretreatment or fine treatment in industrial pipelines;

[0064] The performance of reverse osmosis pre-protection devices operating in complex water quality environments is affected by the dual coupling of dissolved solids content and particulate matter concentration.

[0065] The selection of application scenarios is based on the common challenges in device lifetime prediction, and each application scenario must possess at least one of the following characteristics:

[0066] The device decays at a non-uniform rate, exhibiting multi-stage decline or hysteresis compensation behavior.

[0067] The outflow curve is affected by external inlet pressure and instantaneous water usage behavior, resulting in short-term fluctuations or false anomalies.

[0068] Traditional lifetime estimation methods based on cumulative flow thresholds suffer from problems such as being overly conservative or delaying alarms.

[0069] The device degradation is a non-structural geometrical abrupt change, manifesting only as a minimal perturbation-induced degradation of the equivalent cleaning capability.

[0070] It should be noted that the device life prediction method proposed in this application does not rely on the fixed geometry of the device or a single water quality index as a prerequisite, but is designed for working scenarios with complex water quality disturbances and constraints on nonlinear changes in effluent performance.

[0071] It is worth noting that the update equivalent window update and hysteresis scan verification strategy of this application is not designed for a specific model of device or a specific water treatment system. It can also be applied to scenarios where dynamic fluctuations in water quality parameters cause cumulative bias in the prediction model; and scenarios where the device performance exhibits a time-dependent decay pattern under cyclic operating conditions.

[0072] In some alternative embodiments, the device in this application can be understood as a filter element.

[0073] Taking filter elements as an example, it can be understood that this application mainly targets filter elements used for oil filtration, that is, filter elements that filter oil. Oil filtration filter elements are widely used in hydraulic systems, lubrication systems, and various industrial equipment. Their main function is to remove particulate impurities and colloidal deposits from the oil to ensure the normal operation of system components and extend the service life of equipment. Unlike water filtration, oil usually has higher viscosity and a more complex composition of contaminants, including metal abrasive particles, carbon soot particles, oxide deposits, and additive degradation products. These contaminants can cause a more significant clogging effect inside the filter element.

[0074] In the following content, the water sample to be filtered and water body information can both be understood as the oil to be filtered and oil information.

[0075] refer to Figure 1 , Figure 1 This is a schematic diagram illustrating the device lifespan decay provided in the embodiments of this application.

[0076] Figure 1 The diagram illustrates a device comprising three filtration cycles: t1, t2, and t3. As the operating time progresses, the device's effluent performance gradually declines, with varying rates of flow rate decline across different cycles: In the initial cycle t1, the internal filter media is relatively clean, resulting in a low rate of decline; after entering the intermediate cycle t2, impurities gradually accumulate, accelerating the rate of flow rate decline; and in the final cycle t3, the device approaches saturation, causing a rapid decrease in flow rate and entering a critical stage of its lifespan.

[0077] It should be noted that, Figure 1 The decay curves shown are for illustrative purposes only, mainly used to demonstrate the basic principles and stage characteristics of device lifespan decay, and do not represent the precise operating state of the actual device. In practical applications, the device lifespan curve will be affected by many factors such as influent water quality, operating conditions, pressure conditions, and device materials, and its decay pattern may exhibit complex characteristics such as multi-segmentation, nonlinearity, or local fluctuations.

[0078] Understandably, in this application, changes in effluent flow rate can be considered a direct representation of the device's operating state. By performing equivalent processing on the effluent flow rate, it can be mapped to a refresh equivalent window. In other words, the refresh equivalent window, as a unified metric, reflects the filtration performance and attenuation level of the device in different cycles. When the device is in its initial cycle, the refresh equivalent window value is relatively large, corresponding to a stable effluent flow rate close to the ideal baseline. As the device gradually accumulates contaminants, the refresh equivalent window shrinks accordingly, until it significantly decreases in the critical cycle, corresponding to a sharp drop in effluent flow rate.

[0079] It should be noted that the aforementioned updated equivalent window is not a real physical quantity, but a characterization index obtained through modeling and transformation based on the effluent flow rate curve, used to simplify the description and calculation of the device's lifespan degradation process. The method in this application establishes a mapping relationship between the current equivalent window and the device's lifespan, thereby enabling unified lifespan prediction across different devices and operating environments. It can be understood that in this application, the MCU always maintains a current equivalent window, which can be understood as the updated equivalent window of the previous operating cycle. In each operating cycle, this application updates the current equivalent window based on the updated equivalent window of the current operating cycle to achieve the lifespan prediction objective of this application.

[0080] It is easy to understand that during the water filtration operation of the device, the outflow rate of a certain cycle is not a linear change. Therefore, the corresponding renewal equivalent window is not a constant value in the corresponding cycle. Under normal circumstances, it can be assumed that the renewal equivalent window is maintained at a relatively high level at the beginning of the cycle. As the device gradually accumulates pollutants, the outflow rate slowly decreases, and the renewal equivalent window also gradually decreases accordingly.

[0081] Therefore, the updated equivalent window is not a static fixed quantity, but a dynamic quantity that evolves continuously with time and cumulative usage. This application accurately extracts features reflecting the device's degradation state by characterizing this dynamic change process and combining it with processing steps such as candidate boundary point identification, local resampling, and hysteresis retracement verification, thereby achieving reliable prediction of the device's lifespan.

[0082] refer to Figure 2 , Figure 2 This is a schematic diagram of the updated equivalent window segmentation provided in an embodiment of this application.

[0083] Figure 2 exist Figure 1 Based on this, the correlation between the performance degradation law of the device in different working cycles and the update equivalent window is further shown.

[0084] It is understandable that in the actual water filtration process of the device, the effluent flow rate does not decrease linearly as ideally, but rather exhibits a nonlinear change due to multiple factors such as water quality fluctuations, pump pressure changes, and valve opening and closing. Based on the modeling method proposed in this application, this complex flow rate change can be understood as a process of reducing the update equivalent window:

[0085] At the beginning of each cycle, the pore structure is not yet significantly blocked, and the renewal equivalent window is similar to that of the previous cycle, exhibiting a high and stable effluent flow rate.

[0086] As the usage time increases, impurities gradually accumulate inside the device, reducing the effective filtration area and causing the replacement equivalent window to gradually shrink, resulting in a decrease in the outflow rate.

[0087] It should be noted that, Figure 2 The diagram shown is merely an illustrative example of the aforementioned principle, intended to visually demonstrate the trend of the update equivalent window gradually shrinking over time, and is not limited to the specific change curve in a real application.

[0088] refer to Figure 3 , Figure 3 This is an exemplary application scenario diagram provided for an embodiment of this application.

[0089] Figure 3 The application scenario shown includes a water filtration device 100, a device 101, an inlet 102, an outlet 103, and a processor 104. Solid lines represent water flow channels, and dashed lines represent data transmission. Specifically:

[0090] The inlet 102 is used to receive the raw water to be filtered;

[0091] Device 101 is installed inside water filtration device 100 and is used to filter raw water;

[0092] The outlet 103 is connected to the device 101 and is used to output the purified water filtered by the device 101.

[0093] The processor 104 collects operating parameters related to the state of the device 101 through sensors, such as water flow rate and operating time, and predicts the device lifespan based on the operating parameters.

[0094] Understandable Figure 3 The application scenarios shown are merely illustrative examples of embodiments of this application, used to help understand the application of the device lifetime prediction method and system proposed in this application.

[0095] In practical applications, the specific structure and functional modules of the water filtration device 100 can be adjusted according to different product forms or usage requirements, for example:

[0096] The processor 104 can be built into the water filtration device 100 or set independently in an external control unit and communicate with the water filtration device 100 via wired or wireless means.

[0097] The operating parameters related to the status of device 101 are not limited to effluent flow rate and operating time, but may also include water quality indicators, pressure changes and valve start-stop times.

[0098] The outlet 103 can be further connected to a water storage tank, water supply pipeline or end-use water appliance to meet different application scenarios.

[0099] therefore, Figure 3 The application scenarios shown do not constitute a limitation on the scope of protection of this application. The methods and systems of this application are also applicable to other types of water filtration equipment and device structures.

[0100] The following section, with reference to the accompanying drawings, describes a device lifetime prediction method based on an operating state change curve provided in an embodiment of this application. Figure 4 The method shown is applied to the MCU of an offline fine filtration system. The MCU maintains a current equivalent window and a device lifetime table corresponding to the current equivalent window. The current equivalent window is assigned a value based on the device's operating state change curve. The specific steps are as follows:

[0101] S1: Obtain the status information of the current equivalent window and the device working object in the current work cycle;

[0102] In this embodiment, the working object of the device can be understood as the water sample to be filtered. The state information of the water sample to be filtered can include indicators such as turbidity, hardness, and TDS (total dissolved solids), or it can be working parameters obtained by flow sensor and time accumulation. The specific types can be flexibly configured according to the application scenario, as long as they can support the evaluation of the working state of the device to a minimum. This application does not impose any further limitations.

[0103] By collecting the above information at the beginning of the work cycle, boundary conditions can be provided for subsequent updates to the equivalent window, thereby avoiding lifetime prediction errors caused by water quality differences.

[0104] S2: Based on the status information, determine the update equivalent window. After the current work cycle ends, the update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the work status change curve of the monitoring device during the current work cycle.

[0105] In this embodiment, the working state change curve can be understood as the change curve of the outflow rate, which is recorded in real time by the sensor and processed after the cycle ends. By calculating the cumulative outflow volume, the working state change curve can be resampled at the candidate boundary point, and the validity of the boundary point can be verified by using hysteresis scan. If the curve shows different attenuation characteristics before and after the boundary point, the updated equivalent window is divided into a first equivalent window and a second equivalent window.

[0106] S3: Modify the updated equivalent window according to the first equivalent window and the second equivalent window, and replace the current equivalent window;

[0107] In this embodiment, the first and second equivalent windows represent the flow performance at different stages within the cycle. After superimposing and comparing them, the MCU obtains the correction result through interpolation, weighted averaging, or difference correction. The corrected updated equivalent window replaces the current equivalent window and is stored as the starting state for the next working cycle. In this way, errors caused by short-term abnormal fluctuations can be dynamically eliminated, thereby ensuring the continuity and stability of lifetime prediction.

[0108] S4: Compare the current equivalent window with the device life table to obtain the device life;

[0109] In this embodiment, the device lifetime table can be obtained from accelerated aging tests in the laboratory, statistical analysis of big data from actual work, or training of empirical models, and stored in the non-volatile memory of the MCU. When the current equivalent window is obtained, the MCU can quickly obtain the quantitative value of the remaining lifetime of the device by looking up the table or interpolation, avoiding complex calculation burdens and ensuring that the prediction results can be output in real time in the embedded environment, so that users can keep track of the device status in a timely manner.

[0110] As is understood, MCU refers to Microcontroller Unit, which is essentially an embedded control chip that integrates a processor core, memory, and peripheral interfaces. In this embodiment, the MCU is used to execute the various steps of the device lifetime prediction method and maintain the correspondence between the current equivalent window and the device lifetime table. The MCU can be any 8-bit, 16-bit, or 32-bit model, or it can be a dedicated chip with low-power characteristics; the specific type can be selected according to the application scenario.

[0111] Before delving into the specific technical details of the steps, this application embodiment needs to reiterate that the core problem faced by the device lifetime prediction in this application is not simply lifetime curve fitting, but rather how to obtain an accurate judgment on the device's degradation trend based solely on limited real-time working data in an offline fine filtration environment where massive historical data is lacking.

[0112] Understandably, in typical household or distributed water purification devices, MCUs are often limited by storage and computing capabilities, lacking the conditions for long-term accumulation of large-scale samples. If relying on traditional lifespan prediction methods based on big data regression or long-term cumulative data, problems such as model convergence difficulties and insufficient environmental adaptability will occur, resulting in significant deviations in prediction results.

[0113] The lifetime prediction logic proposed in this embodiment is not based on all historical data. Instead, it normalizes the complex flow decay process within a single cycle by updating the definition of the equivalent window. In this way, the performance of the device in each cycle is mapped to an equivalent value corresponding to the device's lifetime table, thus enabling cross-cycle comparison and updates even with limited data.

[0114] It is easy to understand that even if the MCU only stores limited recent period information, it can still maintain the continuity of predictions through the inheritance and correction of window values, without requiring complete historical curves. In other words, the method in this application, through structured modeling, circumvents the hardware constraints of insufficient data storage and improves the usability of the prediction method in offline working scenarios.

[0115] Furthermore, this application emphasizes not static assumptions about water quality or the environment, but rather projects the combined effects of flow disturbances, water quality fluctuations, and pressure changes onto the shrinking trend of the updated equivalent window through the segmentation and updating of periodic windows. This mapping method not only reduces the impact of external disturbances on lifetime prediction but also provides quantifiable and iterative indicators for constructing device lifetime tables. In engineering practice, the device lifetime table does not need to be an exact global function, but rather gradually approximates the actual decay law through continuous comparison of different periodic windows, maintaining high adaptability and robustness even under different water quality and usage conditions.

[0116] Next, we will further elaborate on the principle of updating the equivalent window in this application.

[0117] In one example, determining the update equivalent window based on the state information includes:

[0118] S2.1: Based on the state information of the water sample to be tested, obtain the impurity accumulation rate of the current working cycle;

[0119] Specifically, in the process of predicting the lifespan of a device, relying solely on the downward trend of the outflow rate is often insufficient to reflect the true rate of device blockage, because the properties of impurities carried by different water bodies vary greatly, and even with the same rate of flow rate decrease, the actual load on the device may be completely different.

[0120] For example, when the particulate matter in the water is small in size and has a high concentration, the pores of the filter media are more prone to micro-clogging in a short period of time. If the main impurities in the water are large suspended particles, these impurities may be quickly trapped on the surface of the device, forming a blockage layer in the short term, but can be partially released under certain flushing action. Therefore, when updating the equivalent window, it is necessary to first quantify the impurity accumulation rate of the current working cycle based on the status information, and use this as the basis for calculating the attenuation amount.

[0121] In this embodiment, the status information includes at least one of the following: turbidity value, sand content, particulate matter size distribution, and dissolved solids content. These parameters can comprehensively reflect the characteristics of impurities in the raw water. Turbidity value reflects the overall impact of suspended particles on optical properties and can be considered as an indicator of the overall particle concentration in the water. Sand content directly determines the total amount of inorganic particles that the device needs to retain per unit time. Particulate matter size distribution reveals the correlation between the fineness of particles in the water and the risk of clogging; smaller particles are easier to penetrate the pores of the filter media. Dissolved solids content reflects crystallizable salts or chemical impurities in the water. These substances will affect the device's permeability in the form of deposition or scaling during long-term operation. By combining and evaluating the above information, the comprehensive load level that the device needs to cope with in the current working cycle, i.e., the impurity accumulation rate, can be obtained. This allows subsequent lifespan prediction to no longer rely solely on flow signals but incorporates the quality characteristics of the water itself, improving the accuracy and adaptability of lifespan estimation.

[0122] Furthermore, after collecting the aforementioned status information, the MCU performs structured processing on the data. For example, turbidity values ​​can be collected by an optical turbidity sensor and mapped to a standardized numerical range; sand content can be estimated by combining pressure difference changes with a particle monitor; particle size distribution can be obtained through a laser particle size analyzer or statistical derivation based on flow velocity disturbances; and dissolved solids content can be indirectly calculated using a conductivity sensor. Due to cost and volume constraints in offline fine filtration environments, the MCU does not retain all raw data for extended periods. Instead, it uses a preset algorithm to map the aforementioned raw data indicators into a periodic impurity accumulation rate and updates it at the end of the current working cycle in conjunction with an equivalent update window.

[0123] S2.2: Calculate the predicted decay amount based on the impurity accumulation rate and the current equivalent window, and apply the predicted decay amount to the current equivalent window to obtain an updated equivalent window;

[0124] Specifically, the updated equivalent window can be understood as the mapping value of the device's effective water purification capacity in the current working cycle, while the impurity accumulation rate represents the actual load on the device caused by the water body in that cycle. If the original flow curve is used directly for calculation, it will often be affected by instantaneous fluctuations or differences in usage habits, resulting in unstable prediction results.

[0125] In this embodiment, by establishing a correspondence between the impurity accumulation rate and the updated equivalent window, a theoretical predicted attenuation is first calculated. This predicted attenuation is then applied to the current equivalent window to obtain the updated equivalent window. By combining water characteristics with device performance, prediction biases that may be caused by single-dimensional data are avoided. This ensures that the update process considers both the actual filtration performance of the device and the impact of water quality differences on the attenuation rate.

[0126] Furthermore, the calculation process for predicting the attenuation amount is not a fixed value replacement, but rather a periodic correction method. Specifically, after each cycle, the MCU analyzes the device's operating status change curve for that cycle, and, combined with the impurity accumulation rate, calculates an attenuation ratio corresponding to the device's lifetime schedule.

[0127] For example, when the impurity accumulation rate is detected to be significantly higher than that under standard water conditions, the MCU will amplify the attenuation ratio accordingly, even if the decrease in effluent flow rate is limited, to reflect potential clogging of the device pores. Conversely, when the impurity accumulation rate is low, the predicted attenuation is reduced to avoid over-predicting a shortened device lifespan. The resulting updated equivalent window will replace the original window value and be used for the next lifespan prediction iteration.

[0128] In one example, the predicted decay is calculated based on the impurity accumulation rate and the current equivalent window, including:

[0129] S2.2.1: Divide the current equivalent window into multiple continuous sub-windows, wherein the length of the continuous sub-windows is calculated based on the changing trend of the impurity accumulation rate;

[0130] Specifically, the current equivalent window is divided into multiple continuous sub-windows because the impurity accumulation process is not linear but exhibits phased differences depending on the water quality and working conditions. If the overall window is still used as the analysis object, it is easy to mask situations where the local pollution rate is too fast, or to fail to reflect buffer periods that occur during operation. By subdividing the updated equivalent window into continuous sub-windows, each sub-window can correspond to the impurity accumulation state within a specific time period, making subsequent predictions more refined and dynamically adaptable.

[0131] In this embodiment, the length of the sub-window is not fixed, but is adaptively calculated based on the changing trend of the impurity accumulation rate. When an increase in the content of particulate matter or dissolved solids in the water is detected, the sub-window will be divided more densely to capture sudden device performance degradation; conversely, when the water quality is relatively stable, the sub-window will be appropriately lengthened to reduce the unnecessary computational burden caused by data dispersion.

[0132] S2.2.2: For each continuous sub-window, calculate the attenuation contribution value according to the corresponding impurity accumulation rate, and sum the attenuation contribution values ​​to obtain the first attenuation curve;

[0133] Specifically, the impurity accumulation rate of each sub-window can be regarded as the load intensity of the device at that stage. After the load intensity accumulates to a certain level, it will be directly reflected in the downward trend of the outflow rate or the increase of the pressure difference.

[0134] In this embodiment, the attenuation contribution value within a given time period is obtained by combining the impurity accumulation rate with the sub-window duration, and then the first attenuation curve is obtained by summing these values ​​one by one. This allows for a complete depiction of the device's performance degradation trajectory over time, rather than relying solely on single-point water quality parameters to infer the overall attenuation. Furthermore, since each sub-window retains its independent impurity characteristics, this summation method avoids diluting extreme water quality in the global prediction, thus more accurately reflecting the device's attenuation pattern during actual operation.

[0135] S2.2.3: Align the first attenuation curve with the predicted working state change curve within the current working cycle to obtain the hysteresis compensation factor;

[0136] Specifically, the first attenuation curve is aligned with the predicted working state change curve within the current working cycle because the performance degradation of the device does not immediately manifest when contaminants enter the filter material, but rather exhibits a certain lag, such as a significant increase in resistance only appearing after the pores are gradually filled.

[0137] S2.2.4: Compensate the first attenuation curve according to the hysteresis compensation factor to obtain the second attenuation curve;

[0138] In this embodiment, the compensation process is not just a simple time offset, but also includes adjustments to the attenuation amplitude. That is, considering the energy consumption difference caused by lag, the curve's downward slope and intercept are adjusted to make the overall shape of the curve consistent with the actual flow curve. In this way, the processed second attenuation curve not only matches the working conditions in the time dimension, but also truly maps the attenuation trend of the device in terms of attenuation amplitude, ensuring the accuracy and reliability of the prediction model.

[0139] S2.2.5: Determine the predicted attenuation amount based on the descending slope of the second attenuation curve and the cumulative attenuation area;

[0140] In this embodiment, by calculating the downward slope of the second attenuation curve, the rate of device performance decline per unit time can be obtained, thereby assessing the attenuation intensity under the current water quality conditions; by accumulating the attenuation area, the overall purification capacity lost by the device in one cycle can be quantified.

[0141] Next, we will further elaborate on the principle of updating the equivalent window segmentation in this application.

[0142] Understandably, the segmentation of the updated equivalent window is based on the non-uniformity of performance degradation during the operating cycle of the device. The pollutant load experienced by the device during operation is often not a stable input, but rather fluctuates dynamically with changes in water turbidity, sand content, particulate matter size distribution, and dissolved solids content. These fluctuations result in phased performance degradation. If the entire operating cycle is used directly as a single calculation window, the impurity accumulation rate at different stages will be averaged, causing a significant deviation between the predicted results and the actual degradation performance.

[0143] In this embodiment, the aforementioned calculated update equivalent window can be understood as: the theoretical remaining filtering capacity of the device after the current filtering cycle.

[0144] It is easy to understand that residual filtration capacity is a quantitative indicator obtained by combining the impurity load that the device bears in the current cycle, the changes in the pore structure of the filter medium, and the pressure drop generated when water flows through the device. Since the device does not directly experience pore blockage, reduction in effective filtration area, or increase in internal flow resistance during the filtration process, if this situation is not dynamically corrected, it will often lead to distortion of lifespan prediction, and may result in premature device failure or overly conservative replacement. Therefore, the updated replacement equivalent window is essentially a mathematical characterization of the device's true residual filtration potential under the current water quality environment, and it can more accurately reflect the device's working status.

[0145] refer to Figure 5 , Figure 5 This is a schematic diagram of the segmentation process for updating the equivalent window provided in an embodiment of this application.

[0146] In one example, the status information also includes influent flow rate, and the update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the operating status change curve of the monitoring device within the current working cycle, including:

[0147] S2.3: Obtain the working status change curve of the device in the current working cycle, and compare the working status change curve with the reference cleaning curve to obtain the candidate boundary point, wherein the reference cleaning curve is calculated based on the current equivalent window and the inlet flow rate;

[0148] Specifically, the working state of the device varies at different stages. Especially between the middle and later stages of use, the pores of the filter medium gradually shrink, causing the working state change curve to deviate from the ideal baseline cleaning curve. Without introducing the baseline cleaning curve as a reference, it is impossible to accurately determine when the device performance reaches an inflection point, making it difficult to reasonably divide the update equivalent window.

[0149] In this embodiment, the outflow rate data of the current working cycle is collected in real time and plotted as a change curve. Then, based on the set inflow rate and the theoretical outflow performance of the device under the initial update equivalent window, a baseline cleaning curve is generated. The baseline cleaning curve reflects the flow rate attenuation trend when the device remains clean under ideal conditions. When the actual working state change curve deviates significantly from the baseline cleaning curve, these deviation points are marked as candidate boundary points. The time or working interval corresponding to these candidate boundary points often indicates a jump in the device's filtration resistance, thus becoming a reference for subsequently dividing the first and second equivalent windows.

[0150] Furthermore, in the process of comparing the working state change curve and the baseline cleaning curve, this embodiment introduces a dual-pointer screening method to determine candidate boundary points. Single-curve comparisons are often affected by local fluctuations; directly selecting the point with the greatest difference between adjacent curves as the candidate boundary point can easily lead to false boundary points. However, the dual-pointer approach can more robustly screen regions with persistent offset characteristics across the entire system.

[0151] refer to Figure 6 , Figure 6 A schematic diagram illustrating the principle of candidate boundary point selection provided in the embodiments of this application.

[0152] Figure 6 The diagram shows the deviation of the working status change curve (dashed line) from the baseline cleaning curve (solid line) within a working cycle, where the horizontal axis represents time and the vertical axis represents the instantaneous outflow rate at the corresponding moment.

[0153] It is understandable that, due to fluctuations in the contamination rate during actual use, the operating status change curve shows a step-like or sudden downward trend compared to the ideal baseline clean curve.

[0154] Figure 6 Seven candidate time windows are further illustrated, with a first pointer and a second pointer used to advance along the two curves. Through this differentiated advancement method using two pointers, the local residuals between the two curves can be calculated segment by segment, and candidate boundary points are marked at locations where the residual change rate reaches a preset threshold. Figure 6 The candidate boundary points include two candidate boundary points: candidate boundary point one and candidate boundary point two.

[0155] In one example, the working state change curve is compared with a baseline cleaning curve to obtain candidate boundary points, including:

[0156] S2.3.1: Configure a first pointer and a second pointer in the time axis corresponding to the working state change curve and the reference cleaning curve, wherein the first pointer is used to traverse the working state change curve and the second pointer is used to traverse the reference cleaning curve;

[0157] Specifically, when establishing a time axis correspondence between the working state change curve and the baseline cleaning curve, the introduction of a first pointer and a second pointer is to enable dynamic comparison of the two curves under the same time reference system. If only relying on overall statistical characteristics or comparison of discrete points, it is easy to lose the capture of curve details due to local fluctuations in the device's operation. However, by using a dual-pointer configuration, it is possible to ensure point-by-point traversal and synchronous / asynchronous advancement of the two curves, so that the flow difference within any time period can be detected in a timely manner.

[0158] S2.3.2: Advance the first pointer and the second pointer within a preset candidate time window, and calculate the time scaling factor and stability index within the corresponding candidate time window according to the advancement mode, wherein the stability index includes the variance and range of the candidate time window;

[0159] In this embodiment, by setting candidate time windows on the time axis, the risk of misjudgment caused by oversensitivity to single-point data can be avoided.

[0160] Understandably, the purpose of the candidate time window is to limit the advancement of the pointer within a certain range, allowing for a comprehensive evaluation of curve differences within that range. The time scaling factor is determined by comparing the rate and magnitude of change of the actual effluent flow curve and the baseline cleanliness curve within the candidate time window. Its significance lies in characterizing whether there is a significant time misalignment or delay between the actual and ideal curves. A large time scaling factor indicates that the deviation between the actual and ideal curves within the window is persistent rather than instantaneous; this persistent deviation often corresponds to the actual attenuation of the device. The stability index measures the degree of data fluctuation by calculating the variance and range within the candidate time window, thereby determining whether the curve's performance within that range is a smooth transition or subject to sharp fluctuations. By combining the time scaling factor and the stability index, both the persistence and fluctuation characteristics of the offset can be considered within the candidate time window, ensuring that the selected candidate boundary points have stable physical meaning, rather than merely being accidental results at the data level.

[0161] In some optional implementations, the advancement mode includes: the first pointer and the second pointer advancing synchronously; the first pointer advancing rapidly while the second pointer advancing slowly; the first pointer advancing slowly while the second pointer advancing rapidly; wherein the rapid advance and the slow advance correspond to the advancement frequency of the pointer, and the advancement frequency is calculated based on the average slope of the working state change curve and the reference cleaning curve in the corresponding candidate time window.

[0162] It is understood that the fast forward or slow forward does not change the physical time axis, but rather refers to using different sample steps and traversal densities on the working state change curve and the baseline cleaning curve respectively on the real time series of the same candidate time window to establish local registration.

[0163] Specifically, the candidate time window is fixed to the real time interval, and both the first and second pointers move monotonically forward within this interval without backtracking. When the working state change curve exhibits a steeper average downward trend within this time window, to avoid misjudging faster state progress as time misalignment, the first pointer uses a smaller sampling interval, and the second pointer uses a larger sampling interval. This allows the two curves to achieve a flexible one-to-many or many-to-one correspondence in morphological evolution without changing the real time coordinates. Conversely, when the average slope of the working state change curve within this time window is less than the average slope of the baseline cleaning curve, the second pointer increases its traversal frequency, and the first pointer decreases its traversal frequency. If the robust average slopes of the two curves within this time window are approximately the same, the two pointers advance synchronously, that is, they advance with the same sample step size at the same real time step. The aforementioned advancement frequency is adaptively calculated from the robust average slopes of the two curves within the candidate time window and is constrained by monotonous advancement and non-backtracking constraints to avoid spurious differences caused by jump point alignment.

[0164] In this embodiment, the so-called fast forward and slow forward are achieved by adjusting the pointer advance frequency. This advance frequency is not a fixed value, but is dynamically calculated based on the average slope of the curves within the candidate time window. Specifically, the average slope of the curves reflects the rate of change of the effluent flow rate over time. When the average slope of the actual working state change curve within the candidate time window is greater than the average slope of the baseline cleaning curve, it indicates that the actual flow rate decreases faster, and the advance frequency of the first pointer needs to be reduced accordingly. When the average slope of the actual working state change curve is less than the average slope of the baseline cleaning curve, the advance frequency of the second pointer needs to be reduced. This slope-based adaptive adjustment method allows for a more reasonable morphological match between the two curves within the candidate time window, thereby avoiding error accumulation caused by curve misalignment in subsequent candidate boundary point extraction.

[0165] Those skilled in the art will understand that the average slope of the aforementioned curve can be obtained through methods such as outlier removal statistics within a window, piecewise linear fitting, or median trend estimation; the upper and lower limits of the propulsion frequency and the step quantization can be set according to the sampling period, MCU computing power, and noise level, as long as it satisfies the requirement of establishing a monotonic, backtrack-free local registration within a fixed real time window, and this application does not impose any further limitations.

[0166] S2.3.3: When the time scaling factor is greater than or equal to the preset time scaling threshold, and the offset value between the stability index and the tolerance band is greater than the preset deviation threshold, the average time of the corresponding candidate time window is taken as the candidate boundary point, wherein the tolerance band is calculated based on the minimum cleaning volume of the benchmark cleaning curve.

[0167] Furthermore, in the final determination process of candidate boundary points, this embodiment uses two aspects to ensure the validity of the boundary points:

[0168] On the one hand, the time scaling factor is required to exceed the time scaling threshold in order to verify that the difference between the actual working state change curve and the baseline cleaning curve is continuous and sufficiently significant;

[0169] On the other hand, it is required that the deviation of the stability index relative to the tolerance band exceeds a preset deviation threshold to ensure that the deviation is not a normal fluctuation within the benchmark range.

[0170] The tolerance band is calculated based on the minimum cleaning volume of the baseline cleaning curve. Its purpose is to provide a reasonable tolerance range for small deviations between curves, avoiding misjudging sensor noise or minor fluctuations in inlet water pressure as device performance degradation. In this way, the average time of the candidate time window is selected as the candidate dividing point, which can smooth out local jitter within the window and ensure that the final dividing point is closer to the true turning point of the device's performance state.

[0171] S2.4: For each candidate boundary point, the working state change curve is resampled in the neighborhood of the candidate boundary point, and hysteresis backscan verification is performed based on the resampling results to obtain the final boundary point.

[0172] Specifically, simply comparing with a baseline cleaning curve is prone to errors, especially when user water usage fluctuates or sensor measurements are unstable. A single deviation point may not accurately represent the true change in device performance. Therefore, this step resamples the data in the neighborhood of the candidate boundary point, recalculating the local trend of the operating state change curve with higher sampling accuracy within a certain range before and after the candidate boundary point. This method filters out occasional noise and yields a smoother, more representative local curve.

[0173] In this embodiment, after resampling, a hysteresis retracement verification method is used. This method examines not only the curve changes before the candidate boundary point but also the curve trends after the candidate boundary point. If the candidate boundary point exhibits a consistent trend change, such as a continuously increasing flow rate decrease rate or a sudden change in the attenuation slope, then the candidate boundary point can be confirmed as a genuine performance inflection point, rather than a false signal caused by noise. The finally selected candidate boundary points serve as the basis for dividing the first and second equivalent windows. This approach significantly improves the accuracy of the boundary points, reduces erroneous divisions caused by single-point anomalies, and thus ensures the stability and reliability of the subsequent lifetime prediction model.

[0174] In one example, the step of resampling the working state change curve within the neighborhood of the candidate boundary point and performing hysteresis scan verification based on the resampling results includes:

[0175] S2.4.1: Calculate the cumulative volume based on the working state change curve, and resample the working state change curve corresponding to the neighborhood of the candidate boundary point based on the cumulative volume to obtain candidate pairs corresponding to the uplink segment and the backlink segment, wherein the volume span of the candidate pair is greater than or equal to the minimum cleaning volume.

[0176] Specifically, direct comparison of the original time series within the neighborhood of candidate boundary points is easily affected by sampling rhythm, short-term water usage surges, and local noise, leading to incomparability between different neighborhoods. To obtain a scale directly related to the filtration process, the cumulative volume V is first obtained by integrating the effluent flow rate over time within this neighborhood. The curve is then resampled with a fixed volume step size to form Q(V), representing the relationship between the effluent flow rate Q and the cumulative volume V. Using volume as a resampling reference can transform the time scaling differences caused by different water usage rhythms and sampling intervals into a unified processing scale, making the neighborhoods of each candidate boundary point horizontally comparable and directly reflecting the instantaneous response within segments of equal effluent volume.

[0177] In this embodiment, in order to reliably extract the two natural segments corresponding to the uplink and downlink segments from Q(V), local trend stripping is first performed on the resampling results within the volume neighborhood of the candidate boundary point:

[0178] The neighborhood baseline is obtained by using moving median, low-pass smoothing, or piecewise linear fitting, and the fluctuation is calculated to highlight the positive and negative oscillations of natural perturbations relative to the baseline.

[0179] Subsequently, the discrete first-order symbol sequence of the fluctuation is scanned to identify continuous non-negative directional segments as candidates for the upward segment, and the immediately following continuous non-positive directional segments as candidates for the return segment; or when the directional is opposite, they are paired according to the mirror rule to ensure that the two segments follow each other on the volume axis and are in opposite directions.

[0180] Furthermore, to ensure both physical meaning and numerical stability, only candidate pairs satisfying the following pairwise constraints are retained:

[0181] The first constraint is that the volume span of each of the two segments is not less than the minimum clean volume threshold, and the ratio of their volume spans is within a preset range, in order to avoid distortion caused by extremely short segments or excessive imbalance.

[0182] The second constraint is that the magnitude difference between the two endpoints has sufficient distinguishability compared with the local noise level, and the threshold can be set by neighborhood robust variance or quantile range.

[0183] The third constraint is that the monotonicity within the fragment is checked by a small amount of fault tolerance (allowing slight inversions at individual sampling points), and morphological opening / closing operations are performed on the symbol sequence to remove glitches when necessary.

[0184] Uplink and downlink segments that meet the above conditions constitute candidate pairs. If multiple feasible pairs exist within the same neighborhood, the pair with more sufficient volume coverage, higher amplitude energy index, or closer location to the candidate boundary point is selected first, and the rest are used as backups for subsequent weighting.

[0185] The specific handling method is as follows:

[0186] Within the neighborhood interval before and after the candidate boundary point, the flow curve is resampled according to a fixed volume step size to obtain an upward segment corresponding to the decline in device performance and a return segment corresponding to the recovery after the cleaning process.

[0187] The two curves generated in this way not only eliminate the influence of uneven temporal distribution in the original data, but also reflect the symmetry of device decay and recovery at the same volume scale. Furthermore, the candidate pairs must be set such that their volume span is greater than or equal to the minimum clean volume to ensure that the resulting uplink and downlink curves cover a sufficient range, avoiding unrepresentative results due to excessively small sample intervals, thus ensuring the robustness and accuracy of subsequent calculations.

[0188] S2.4.2: Align and interpolate the uplink and return segments on the volume axis, calculate the flow difference between the uplink and return segments under a consistent volume grid, and obtain the corresponding pseudohysteresis area by integration;

[0189] Specifically, after resampling, the up-segment and down-segment need to be projected onto a unified volume grid. Since the two curves come from different time windows, direct comparison will result in inconsistent sampling point numbers and misaligned positions. Therefore, interpolation and alignment processing is required on the volume axis to make the two curves comparable under the same volume coordinates.

[0190] In this embodiment, the two curves are first reconstructed by interpolation at the same volume resolution, and then the instantaneous flow difference between them is calculated point by point under a consistent volume grid. By integrating the difference curve, a numerical index can be obtained, which represents the area enclosed by the actual flow curve and the ideal return curve in the neighborhood of the candidate boundary point, i.e., the pseudo-hysteresis area. The larger the pseudo-hysteresis area, the more significant the difference between the decrease and increase in flow within the neighborhood, and the more likely it corresponds to the actual cleanliness boundary point of the device. Enhancing the robustness of the judgment by energy quantification of the entire range can avoid the problem of single-point difference being affected by noise or local anomalies. The principle is to use the hysteresis phenomenon of flow with volume as a significant feature of the device performance state transition, thereby providing a quantitative basis for the final selection of candidate boundary points.

[0191] S2.4.3: Sort the pseudo-hysteresis areas of each candidate boundary point in descending order, and output the first candidate boundary point as the final boundary point. The descending order sorting also includes weighting the pseudo-hysteresis areas according to the area of ​​the neighborhood of the candidate boundary point.

[0192] Specifically, after obtaining the pseudo-hysteresis area of ​​all candidate boundary points, this embodiment sorts these candidate boundary points in descending order and selects the candidate boundary point with the largest area as the final output boundary point.

[0193] It is understandable that the hysteresis characteristics of the neighborhoods of different candidate boundary points vary. The larger the area value, the stronger the difference between the up-travel and down-travel curves in that neighborhood, and the more it represents a significant change in the working state of the device before and after that position, thus possessing sufficiency and reliability as the final boundary point.

[0194] Furthermore, to avoid amplifying local noise due to relying solely on area values, the ranking process also uses the overall area of ​​the neighborhood of each candidate boundary point as a weighting factor to assign a weighted value to the hysteresis area. The weighted ranking result not only considers the strength of local differences but also balances the size of the neighborhood coverage, ensuring that the finally selected boundary points possess both local abrupt change characteristics and sufficient representativeness on the global curve. This strategy effectively improves the robustness of boundary point identification, avoiding misjudgments caused by local anomalies or short-term fluctuations. Through this verification mechanism, the most reasonable cleaning time can be automatically identified among multiple candidate points, improving the system's intelligence level and user experience.

[0195] S2.5: Based on the position of the final dividing point on the working state change curve, the updated equivalent window is divided to obtain a first equivalent window and a second equivalent window, and the first equivalent window is assigned a value based on the current equivalent window;

[0196] In this embodiment, the final determined dividing point is used as the anchor point for dividing the updated equivalent window, that is, dividing the entire window into two parts: the first equivalent window and the second equivalent window.

[0197] Understandably, under the limitations of this application, the device often exhibits two distinct performance phases within a single operating cycle. In the initial phase, the filtration efficiency is relatively stable, and the effluent flow rate decreases slowly. However, in the later phase, due to rapid impurity accumulation, the device's degradation rate accelerates significantly. Without window segmentation, the lifetime prediction model can only extrapolate based on the overall curve, easily underestimating the early-stage lifetime or overestimating the later-stage lifetime, ultimately leading to prediction errors. Therefore, by segmenting the equivalent renewal window, the remaining filtration capacity of the device at different stages can be characterized more precisely.

[0198] In practice, based on the position of the final dividing point on the operating state change curve, the first half is defined as the first equivalent window, and the second half as the second equivalent window. The first equivalent window is then assigned a value based on the current equivalent window, serving as the baseline segment for subsequent lifetime prediction. Furthermore, the second equivalent window serves as a crucial input for subsequent decay curve prediction, reflecting the device's performance characteristics near failure. This segmentation and assignment method not only achieves segmented modeling of device performance but also allows the lifetime prediction process to incorporate actual performance at different stages, resulting in more realistic predictions.

[0199] Next, we will further elaborate on the principle of updating the equivalent window correction in this application.

[0200] refer to Figure 7 , Figure 7 A schematic diagram illustrating the correction principle of the updated equivalent window provided in the embodiments of this application.

[0201] In one example, the updated equivalent window is corrected based on the first equivalent window and the second equivalent window, including:

[0202] S3.1: Calculate the volume difference between the first equivalent window and the second equivalent window;

[0203] In this embodiment, the start and end points of the cumulative volume of the two updated equivalent windows are compared to calculate the corresponding volume span difference, thereby obtaining an index that can characterize the degree of consistency between the two. This processing method not only allows for direct numerical comparison of the results of different algorithms but also provides an objective basis for determining whether further corrections are needed.

[0204] S3.2: If the volume difference value is less than or equal to the preset tolerance threshold, the second equivalent window is used as the corrected updated equivalent window;

[0205] In this embodiment, when the calculated volume difference is within the allowable tolerance range, it means that the difference between the first equivalent window and the second equivalent window is small enough, and the two can be regarded as being consistent in a statistical sense. At this time, no additional complex correction operation is required, and the second equivalent window can be used directly.

[0206] The second equivalent window was chosen because it more comprehensively reflects the characteristics of device performance degradation and recovery, and its physical interpretation is closer to the actual cleaning process. This approach effectively reduces additional computational overhead and avoids distortion introduced by over-correcting the curve. Simultaneously, this decision logic improves the algorithm's robustness, enabling the system to maintain relatively stable output under different operating conditions. It ensures that a simplified path is used within the error range, thereby improving computational efficiency and practicality while reducing the complexity of subsequent processing.

[0207] S3.3: If the volume difference value is greater than the preset tolerance threshold, calculate the correction factor based on the volume difference value and assign a weighted value to the updated equivalent window based on the correction factor to obtain the corrected updated equivalent window;

[0208] In another case, if the volume difference exceeds the tolerance threshold, it indicates that there is a large deviation between the first and second equivalent windows. If either one is used directly, the final updated equivalent window will deviate from the actual working pattern.

[0209] To address this issue, this embodiment introduces the concept of a correction factor. The correction factor is calculated based on the volume difference, and its value reflects the relative reliability between two candidate windows. When the volume difference is large, the correction factor tends to increase the weight of the more stable window; conversely, when the volume difference is relatively small but still exceeds the tolerance range, the two windows are weighted in a relatively balanced manner. Specifically, the correction factor is combined with the updated equivalent window and weighted accordingly to obtain a corrected window that comprehensively considers the characteristics of both the first and second equivalent windows. This approach avoids the bias risk of relying entirely on a single result while fully utilizing the complementary characteristics of different algorithms to form an updated equivalent window that more closely approximates real-world conditions.

[0210] In one optional implementation, the correction of the updated equivalent window also includes modeling and compensating for the self-cleaning process of the device at the end of each working cycle. Specifically, when the device stops working or enters standby mode, it will experience a certain degree of self-recovery due to mechanisms such as fluid pressure drop, residual impurity sedimentation, or reverse osmosis flushing. This process will cause a slight rebound in the working state change curve before the start of the next working cycle. If this characteristic is ignored in the correction of the updated equivalent window, it will lead to a systematic deviation between the cumulative volume and the actual decay trajectory. Especially in multi-cycle continuous operation scenarios, this deviation will accumulate and eventually affect the accuracy of the updated equivalent window.

[0211] In this embodiment, firstly, based on the flow rate differences between the end point of different cycles and the start point of the next cycle in the device's historical operating data, typical self-cleaning recovery amplitudes are extracted, and a self-cleaning recovery factor related to cumulative volume or operating time is established. Then, when updating the equivalent window, this self-cleaning recovery factor is introduced into the correction process to dynamically adjust the upper and lower boundaries of the candidate time window. For example, when a candidate time window span is detected to cover a cycle switching point, a self-cleaning recovery amount is introduced within the cumulative volume range corresponding to that interval, thereby correcting the window boundary offset. In this way, the problem of underestimating or overestimating the volume due to the device's self-cleaning process can be avoided, allowing the updated equivalent window to more realistically reflect the attenuation and recovery patterns of the device under actual operating conditions.

[0212] Furthermore, the stability of the self-cleaning recovery factor can be incorporated into the correction process to differentiate the windows for different cycles. For device models or operating conditions with significant self-cleaning recovery characteristics, the weight of the cleaning correction factor is appropriately increased to fully compensate for volume differences caused by cycle switching; while for cases with weak recovery effects, a lower weight is used to avoid over-correction. This not only improves the adaptability of the correction method to different devices and operating conditions but also ensures that the updated equivalent window reflects the actual operating state of the device without deviating from the original data characteristics due to over-correction. Through this improvement, the final updated equivalent window exhibits better stability and universality in multi-cycle operating environments, thus providing more reliable input data for subsequent device life prediction and maintenance strategies.

[0213] In one example, the device lifetime is obtained by comparing the current equivalent window with the device lifetime table, including:

[0214] S4.1: Input the current equivalent window into the device life table and retrieve the corresponding life reference value;

[0215] S4.2: Determine the remaining lifespan of the device based on the lifespan reference value, and output the device lifespan.

[0216] In one example, this application provides a device life prediction system based on a working state change curve. The system is applied to an MCU in an offline fine filtration system and includes a data acquisition module, a processing module, and a life assessment module, wherein:

[0217] The data acquisition module is used to acquire the status information of the water sample to be filtered and the working status change curve of the device in the current working cycle, and transmit the data to the calculation and processing module;

[0218] The computational processing module, located in the MCU, is used to maintain and update the equivalent window and the device lifetime table, and to update, correct, divide, and predict the decay amount of the equivalent window.

[0219] The lifespan assessment module is used to compare the updated equivalent window with the device lifespan table to obtain the remaining lifespan of the device.

[0220] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A device lifespan prediction method based on operating state change curves, applied to an MCU in an offline fine filtration system, characterized in that, The MCU stores a device lifetime table and a current equivalent window, wherein the current equivalent window is assigned a value based on the device's operating state change curve, and the method includes: Obtain the status information of the working object of the device; Based on the status information, an update equivalent window is determined. After the current work cycle ends, the update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the work status change curve during the current work cycle. The size of the first equivalent window is the same as that of the current equivalent window, and the size of the second equivalent window is the same as that of the update equivalent window. The updated equivalent window is modified based on the first equivalent window and the second equivalent window and then replaces the current equivalent window. The device lifespan is obtained by comparing the current equivalent window with the device lifespan table.

2. The device life prediction method based on the operating state change curve according to claim 1, characterized in that, Based on the status information, determine the update equivalent window, including: Based on the status information of the working object of the device, the impurity accumulation rate of the current working cycle is obtained, wherein the status information includes at least one of turbidity value, sand content, particle size distribution and dissolved solids content; Based on the impurity accumulation rate and the current equivalent window, the predicted decay amount is calculated, and the predicted decay amount is applied to the current equivalent window to obtain an updated equivalent window.

3. The device life prediction method based on the operating state change curve according to claim 2, characterized in that, Based on the impurity accumulation rate and the current equivalent window, the predicted decay is calculated, including: The current equivalent window is divided into multiple continuous sub-windows, wherein the length of the continuous sub-windows is calculated based on the changing trend of the impurity accumulation rate; For each consecutive sub-window, the attenuation contribution value is calculated based on the corresponding impurity accumulation rate, and the attenuation contribution values ​​are summed to obtain the first attenuation curve. Align the first attenuation curve with the predicted working state change curve within the current working cycle to obtain the hysteresis compensation factor. The first attenuation curve is compensated according to the hysteresis compensation factor to obtain the second attenuation curve; The predicted attenuation amount is determined based on the descending slope of the second attenuation curve and the cumulative attenuation area.

4. The device life prediction method based on the operating state change curve according to claim 1, characterized in that, The status information also includes the influent flow rate. The update equivalent window is divided into a first equivalent window and a second equivalent window by monitoring the operating status change curve of the monitoring device within the current working cycle, including: The working status change curve of the device in the current working cycle is obtained, and the working status change curve is compared with the reference cleaning curve to obtain the candidate boundary point, wherein the reference cleaning curve is calculated based on the current equivalent window and the inlet water flow rate. For each candidate boundary point, the working state change curve is resampled in the neighborhood of the candidate boundary point, and hysteresis backscan verification is performed based on the resampling results to obtain the final boundary point. Based on the position of the final dividing point on the working state change curve, the updated equivalent window is divided to obtain a first equivalent window and a second equivalent window, and the first equivalent window is assigned a value based on the current equivalent window.

5. The device life prediction method based on the operating state change curve according to claim 4, characterized in that, The working state change curve is compared with the baseline cleaning curve to obtain candidate boundary points, including: A first pointer and a second pointer are configured in the time axis corresponding to the working state change curve and the reference cleaning curve, wherein the first pointer is used to traverse the working state change curve and the second pointer is used to traverse the reference cleaning curve; The first and second pointers are advanced within a preset candidate time window, and the time scaling factor and stability index within the corresponding candidate time window are calculated according to the advancement mode. The stability index includes the variance and range of the candidate time window. When the time scaling factor is greater than or equal to the preset time scaling threshold, and the offset value between the stability index and the tolerance band is greater than the preset deviation threshold, the average time of the corresponding candidate time window is used as the candidate boundary point, wherein the tolerance band is calculated based on the minimum cleaning volume of the benchmark cleaning curve.

6. The device life prediction method based on the operating state change curve according to claim 5, characterized in that, The propulsion mode includes: The first pointer advances synchronously with the second pointer; The first pointer advances fast while the second pointer advances slowly. The first pointer advances slowly while the second pointer advances quickly. The advance frequency of the pointers corresponding to the fast forward and the slow forward is calculated based on the average slope of the working state change curve and the reference cleaning curve in the corresponding candidate time window.

7. The device life prediction method based on the operating state change curve according to claim 4, characterized in that, The step of resampling the working state change curve within the neighborhood of the candidate boundary point and performing hysteresis scan verification based on the resampling results includes: The cumulative volume is calculated based on the working state change curve. The working state change curves corresponding to the neighborhood of the candidate boundary point are resampled based on the cumulative volume to obtain candidate pairs corresponding to the uplink and downlink segments. The volume span of the candidate pairs is greater than or equal to the minimum cleaning volume. Align and interpolate the uplink and return segments on the volume axis, calculate the flow difference between the uplink and return segments under a consistent volume grid, and obtain the corresponding pseudohysteresis area by integration. The pseudohysteresis areas of each candidate boundary point are sorted in descending order, and the first candidate boundary point is output as the final boundary point. The descending sorting also includes weighting the pseudohysteresis areas based on the area of ​​the neighborhood of the candidate boundary point.

8. The device life prediction method based on the operating state change curve according to claim 1, characterized in that, The updated equivalent window is corrected based on the first equivalent window and the second equivalent window, including: Calculate the volume difference between the first equivalent window and the second equivalent window; If the volume difference is less than or equal to a preset tolerance threshold, the second equivalent window is used as the corrected updated equivalent window. If the volume difference is greater than a preset tolerance threshold, a correction factor is calculated based on the volume difference, and the updated equivalent window is weighted and assigned a value based on the correction factor to obtain the corrected updated equivalent window.

9. The device life prediction method based on the operating state change curve according to claim 1, characterized in that, The device lifespan is obtained by comparing the current equivalent window with the device lifespan table, including: Input the current equivalent window into the device life table and retrieve the corresponding life reference value; The remaining lifespan of the device is determined based on the lifespan reference value, and the device lifespan is output.

10. A device life prediction system based on operating state change curves, used to implement the device life prediction method based on operating state change curves as described in any one of claims 1-9, characterized in that, The system is applied to the MCU of the offline fine filtration system, including a data acquisition module, a processing module, and a lifespan assessment module, wherein: The data acquisition module is used to acquire the status information of the water sample to be filtered and the working status change curve of the device in the current working cycle, and transmit the data to the calculation and processing module; The computational processing module, located in the MCU, is used to maintain and update the equivalent window and the device lifetime table, and to update, correct, divide, and predict the decay amount of the equivalent window. The lifespan assessment module is used to compare the updated equivalent window with the device lifespan table to obtain the remaining lifespan of the device.

Images