Intelligent commercial water drinking system and water purification method thereof

Abstract

The application relates to the technical field of water treatment, in particular to an intelligent commercial water drinking system and a water purification method thereof, which comprises the following steps: obtaining a drinking water machine initialization message triggered after replacement of a filter material, and obtaining drinking water machine water purification parameters, wherein the drinking water machine water purification parameters comprise water quality detection data, filter material use time and use interval time; after the water quality detection data, the filter material use time and the use interval time are pretreated, a to-be-analyzed data set is obtained; the to-be-analyzed data set is input into a preset filter core life prediction model, the filter core life prediction model dynamically calculates a filter core life prediction result according to the correlation between the data in the to-be-analyzed data set; the type of the filter material is obtained, and filter core maintenance suggestions corresponding to each filter material type are generated according to the filter core life prediction result. The application has the effect of realizing accurate and intelligent prediction and active management of the health state of the filter core.

Description

Technical Field

[0001] This application relates to the technical field of water treatment, and in particular to an intelligent commercial drinking water system and its water purification method. Background Technology

[0002] Commercial drinking water purifiers, also known as commercial integrated water purifiers, are efficient and continuous water supply devices designed for public environments such as offices, factories, schools, and hospitals. They provide stable, safe, and sufficient drinking water for groups of dozens to hundreds of people. By integrating purification, heating, and water storage functions, they can simultaneously provide hot boiling water and room temperature water that can be drunk directly, in order to meet the diverse drinking water needs of users. They are a key infrastructure for ensuring drinking water safety and convenience in modern commercial settings.

[0003] Currently, mainstream commercial direct drinking water equipment generally adopts multi-stage physical filtration combined with point-of-use disinfection technology to purify water. First, a pre-filter such as PP cotton removes large particulate impurities such as sediment and rust from the water; then, an activated carbon filter adsorbs residual chlorine, discoloration, odor, and organic pollutants from the water; finally, a reverse osmosis membrane filter, driven by pressure, efficiently filters out dissolved pollutants such as heavy metal ions, inorganic salts, bacteria, and viruses, thus producing pure direct drinking water. Some equipment also adds an ultraviolet sterilization module to irradiate the water at the point of use to further eliminate the risk of microbial contamination and ensure drinking water safety.

[0004] However, as the core consumable of the purification system, the performance degradation and actual service life of the filter element are affected by a variety of dynamic factors such as the amount of water treated, fluctuations in the quality of the incoming water, and the frequency of use. The maintenance of existing equipment usually relies on fixed time cycles, which may result in premature replacement of the filter element before its performance is fully exhausted, causing unnecessary waste of resources and increased operation and maintenance costs; or, when the filter element has actually failed, there may be no timely warning and replacement, resulting in a decline in purification effect, and users may continue to drink substandard water, posing potential safety hazards. Summary of the Invention

[0005] In order to achieve accurate, intelligent prediction and proactive management of the health status of filter cartridges, this application provides an intelligent commercial drinking water system and its water purification method.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: A water purification method for an intelligent commercial drinking water system, the water purification method for the intelligent commercial drinking water system comprising: Get the water dispenser initialization message triggered after the filter material is replaced, and get the water dispenser purification parameters, wherein the water dispenser purification parameters include water quality test data, filter material usage time and usage interval time; After preprocessing the water quality test data, the filter material usage time, and the usage time interval, a dataset to be analyzed is obtained. The dataset to be analyzed is input into a preset filter life prediction model, and the filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed. Obtain the filter material type and generate filter maintenance recommendations for each filter material type based on the filter life prediction results.

[0007] Preferably, the step of preprocessing the water quality testing data, the filter material usage time, and the usage time interval to obtain the dataset to be analyzed specifically includes: Based on the water quality test data and the usage time of the filter material, the degree of single filter cartridge wear is generated; The dataset to be analyzed is obtained by aligning the timestamps based on the single filter cartridge wear level and the usage time interval.

[0008] Preferably, the filter cartridge life prediction model is trained using the following method: The correlation between filter cartridge interval time and filter cartridge life loss during continuous use under different water quality conditions is obtained from the historical usage data of filter cartridges, and a corresponding loss curve is constructed based on the correlation. The loss inflection point is obtained from the loss curve. According to different water quality conditions, the corresponding calculation weights are set for each loss inflection point, the filter cartridge interval time and the filter cartridge life loss. The calculated weights under different water quality conditions are used to train the initial model to obtain the filter cartridge life prediction model.

[0009] Preferably: the step of inputting the dataset to be analyzed into a preset filter life prediction model, wherein the filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed, specifically includes: The filter life prediction model is input into the single filter wear level and the usage time interval from the dataset to be analyzed. The calculation weights obtained by the filter life prediction model based on each usage time interval are obtained, and the filter life prediction result is calculated based on the single filter wear level and the calculation weights.

[0010] Preferably: the step of obtaining the filter material type and generating filter element maintenance recommendations for each filter material type based on the filter element lifespan prediction results specifically includes: Obtain the estimated lifespan of the filter cartridge for each of the aforementioned filter material types; If the estimated lifespan of the filter cartridge is lower than the lifespan threshold of the filter material, then a filter cartridge maintenance recommendation is generated based on the water quality test data and the remaining filter material type.

[0011] The second objective of this invention is achieved through the following technical solution: A water purification device for an intelligent commercial drinking water system, the water purification device for the intelligent commercial drinking water system comprising: The parameter acquisition module is used to acquire the water dispenser initialization message triggered after the filter material is replaced, and to acquire the water purification parameters of the water dispenser, wherein the water purification parameters of the water dispenser include water quality test data, filter material usage time and usage interval time; The data processing module is used to preprocess the water quality test data, the filter material usage time, and the usage time interval to obtain the dataset to be analyzed. The lifespan prediction module is used to input the dataset to be analyzed into a preset filter lifespan prediction model. The filter lifespan prediction model dynamically calculates the filter lifespan prediction result based on the correlation between the data in the dataset to be analyzed. The maintenance suggestion module is used to obtain the filter material type and generate filter element maintenance suggestions corresponding to each filter material type based on the filter element life estimation results.

[0012] The above-mentioned objective three of this application is achieved through the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the water purification method of the intelligent commercial drinking water system described above.

[0013] The fourth objective of this application is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the water purification method of the above-described intelligent commercial drinking water system.

[0014] In summary, this application includes at least one of the following beneficial technical effects: 1. By comprehensively collecting multi-dimensional parameters such as water quality, usage time and usage interval, and using a preset prediction model to analyze their correlation, it changes the traditional method of relying on fixed-cycle filter replacement and can make personalized and dynamic predictions of the remaining life of the filter. 2. By combining cumulative usage time with real-time water quality data, the degree of filter cartridge wear per use is calculated. Combined with time-series alignment analysis based on usage time intervals, the assessment of filter cartridge performance degradation moves from macro-statistics to the micro-event level, providing a high-quality data foundation for accurate prediction. 3. By comparing the prediction results with the thresholds of each filter material type and integrating water quality data with the status information of other filter cartridges in the system, targeted maintenance suggestions containing problem location and optimization strategies can be generated, realizing an upgrade from replacement alarms to maintenance prescriptions, and improving the initiative and systematic nature of operation and maintenance. Attached Figure Description

[0015] Figure 1 This is a flowchart of a water purification method for an intelligent commercial drinking water system according to an embodiment of this application.

[0016] Figure 2 This is a schematic diagram of the water purification device of an intelligent commercial drinking water system according to one embodiment of this application.

[0017] Figure 3 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0018] The present application will be further described in detail below with reference to the accompanying drawings.

[0019] In one embodiment, such as Figure 1 As shown, this application discloses a water purification method for an intelligent commercial drinking water system, which specifically includes the following steps: S10: Obtain the water dispenser initialization message triggered after the filter material is replaced, and obtain the water dispenser's water purification parameters, including water quality test data, filter material usage time, and usage interval.

[0020] In this embodiment, the water dispenser initialization message refers to the instruction to restart the prediction of the remaining lifespan of all or one type of filter material in the filter cartridge. Water dispenser parameters refer to the operating parameters generated during the use of the water dispenser. Water quality test data refers to the quality of the water source entering the water dispenser. Filter material usage time refers to the duration of each type of filter material when installed in the water dispenser. Usage interval time refers to the time between completing one water source filtration and purification cycle and the next water source purification cycle.

[0021] Specifically, after all filter media have been replaced, such as during the initial installation of the water dispenser, or after it is determined that one or more filter media need to be replaced and the replacement has been completed, an initialization message is triggered for the replaced filter media. During continuous use of the water dispenser, the water quality is monitored by sensors, and this data is recorded as water quality detection data. After replacing the filter media, the usage time of each filter media is continuously recorded, obtaining the usage time for each type of filter media. Furthermore, the time is recorded synchronously each time water filtration and purification begins, such as when someone uses the water dispenser to fill water, when the water dispenser filters water on a scheduled basis, and after the water purification and filtration is completed. The time between the completion of one water purification cycle and the next water purification cycle is also recorded as the usage interval.

[0022] S20: After preprocessing the water quality test data, filter material usage time, and usage interval, the dataset to be analyzed is obtained.

[0023] In this embodiment, the dataset to be analyzed refers to a dataset that records data for analyzing and predicting the remaining lifespan of filter materials.

[0024] Specifically, after obtaining water quality testing data, filter material usage time, and usage intervals, preprocessing is performed according to the subsequent filter material prediction plan. This includes preliminary data analysis, filtering out abnormal data, and timestamp sorting. After data preprocessing, the data is stored according to time periods to obtain the dataset to be analyzed.

[0025] S30: Input the dataset to be analyzed into the preset filter life prediction model. The filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed.

[0026] In this embodiment, the filter life prediction model refers to a pre-trained model used to predict the remaining life of each type of filter based on the operating parameters of the water dispenser.

[0027] Specifically, after obtaining the dataset to be analyzed, the dataset is input into the filter life prediction model for analysis. The model judges the degree of influence of each type of parameter on the filter life based on the data in the dataset to be analyzed according to the training rules, and then dynamically calculates the remaining life of each filter material based on the degree of influence, which is used as the filter life prediction result.

[0028] S40: Obtain the filter material type and generate filter maintenance recommendations for each filter material type based on the filter life prediction results.

[0029] Specifically, the type of filter material installed in the water dispenser is obtained as the filter material type. After obtaining the estimated lifespan of each filter material type, the remaining lifespan of each type is used to assess the overall purification capacity of the water dispenser for the current water quality, determining whether replacement is necessary. The assessment result is then output as a filter material maintenance suggestion. Optionally, to reduce information interference caused by frequent output, a message indicating the need for filter material replacement can be sent to relevant personnel or displayed on the display panel only when replacement of one or more filter materials is required.

[0030] In this embodiment, the initialization message of the water dispenser triggered after the filter material replacement is used as the starting point, ensuring the accuracy and traceability of the monitoring cycle. This provides a foundation for subsequent calculations based on the actual usage cycle, improving the reliability of the starting point for lifespan prediction. Furthermore, by comprehensively acquiring water quality testing data, filter material usage time, and usage interval time, a multi-parameter fusion acquisition method enhances the perception of the filter cartridge's working environment, enabling multi-dimensional and multi-faceted prediction of the filter cartridge's remaining lifespan. Water quality data reflects the pollution load of the water source on the filter cartridge, usage time reflects mechanical fatigue, and usage interval may affect the growth of microorganisms inside the filter cartridge or the recovery of material performance. Data from at least three dimensions collectively constitute dynamic variables affecting the filter cartridge's lifespan, allowing for a more comprehensive assessment of the filter cartridge's wear status. Furthermore, by preprocessing this data to form a dataset for analysis, a standardized and structured input is provided for subsequent model analysis, reducing the possibility of model calculation errors due to data format chaos or noise interference. By inputting the dataset into the filter cartridge lifespan prediction model, the model can perform dynamic calculations based on the correlations between the data. This model can learn and quantify the different impacts of use per unit time on filter lifespan under different water qualities, thus achieving dynamic and accurate prediction of the remaining lifespan of the filter cartridge. Its accuracy is significantly improved compared to the fixed-cycle method. Finally, it generates targeted maintenance recommendations based on the type of filter material. Since different filter materials, such as PP cotton, activated carbon, and RO membranes, have different functions and failure modes, it can provide clear guidance such as: if the RO membrane filter cartridge has an estimated remaining lifespan of 30%, it is recommended to focus on the influent water quality; if the pre-filter PP cotton cartridge is 85% used, it is recommended to prepare for replacement. In summary, it can dynamically predict the lifespan of the filter cartridges based on the actual usage of the water dispenser, thereby reducing the drinking water safety risks caused by premature filter cartridge failure, and also reducing resource waste and maintenance costs caused by premature replacement due to underutilization of the filter cartridges, achieving a balanced optimization of safety and economy.

[0031] In one embodiment, step S20 involves preprocessing the water quality testing data, filter material usage time, and usage interval to obtain the dataset to be analyzed, specifically including: S21: Based on water quality test data and filter material usage time, generate the degree of single filter cartridge wear.

[0032] In this embodiment, the degree of single filter cartridge wear refers to the measurable amount of degradation in the performance or expected lifespan of the filter cartridge after completing a specific water filtration and purification operation, which is related to the actual workload of this operation.

[0033] Specifically, each triggering of the water purification function, such as when a user draws water or when the device produces water on a timed schedule, constitutes an independent usage event. During each event, the corresponding filter material usage time is recorded synchronously, i.e., the precise duration of the event from start to finish, for example, 2.5 minutes. Water quality testing data is also collected during the event, such as the average and maximum TDS values ​​of the influent or readings of specific contaminants. Generating the degree of filter cartridge wear per use involves establishing a mapping relationship between water quality, time, and wear, allowing the construction of a baseline wear coefficient table for different filter material types. This table defines the baseline wear value generated per unit of usage time under standard water quality conditions. In actual calculations, a basic wear is first calculated based on the filter material usage time of the current event. Further, the actual water quality testing data for the current event is compared with standard water quality conditions to calculate the water quality impact factor. For example, if the influent TDS value is higher than the standard value, a coefficient greater than 1 is used to weight and amplify the basic wear; conversely, a coefficient equal to or less than 1 may be used. Finally, by multiplying the base loss by the water quality impact factor, the degree of filter cartridge wear in a single event is obtained. This value reflects the calibrated loss of the filter cartridge caused by a specific usage time under specific water quality conditions. In this way, the cumulative usage time is broken down and transformed into a series of micro-loss events labeled with water quality load, providing high-precision data units for subsequent refined and differentiated cumulative lifespan calculations.

[0034] S22: After aligning the timestamps according to the degree of single filter cartridge wear and the time interval of use, the dataset to be analyzed is obtained.

[0035] In this embodiment, timestamp alignment refers to the operation of associating, sorting, and combining each generated single filter wear level data point with its corresponding usage time interval in a strict chronological order to construct a structured data set with complete temporal context information.

[0036] Specifically, after calculating the wear level of a single filter cartridge, a timestamp is added to that data point. This timestamp typically records the end time of the current usage event. Simultaneously, the usage time interval recorded in the preceding steps also carries a time attribute, representing the idle time between the end of the previous event and the start of the current event. These two dimensions of information are then integrated. For example, a record in the following format is generated and stored: Serial Number N, Timestamp Tn, Single Filter Cartridge Wear Level Ln, Interval with Previous Event In-1. Here, Tn is the end time of the Nth event, Ln is the wear value of that event, and In-1 is the waiting time before the start of the current event, i.e., the duration from T(n-1) to the start point of event Tn. In this way, independent, discrete wear events and intervals are linked together into a time series reflecting the device's usage pattern. All historical records, such as all records since the last filter cartridge replacement, are arranged in ascending order by timestamp Tn, forming the original time-series queue of the dataset to be analyzed. In this process, features can also be extracted from the interval time In-1, for example, it can be divided into categories such as short intervals (e.g., <1 hour), medium-long intervals (e.g., 1-12 hours), and long intervals (e.g., >12 hours), and this classification can be used as a new feature field for that record. The final dataset to be analyzed is a multi-column data table, with records containing at least key fields such as timestamp, interval duration / category, and degree of single-event damage. This dataset not only records the amount of damage, but more importantly, it encodes the rhythm and pattern of damage occurrence, providing crucial time-series dependency information for the predictive model to learn the dynamic laws of filter performance degradation.

[0037] In this embodiment, the granularity of loss assessment is refined, and the accuracy of time-series data analysis is enhanced, thus providing a better data foundation for subsequent high-precision predictions. By generating the degree of single-use filter cartridge wear based on water quality testing data and filter material usage time, continuous, macroscopic monitoring data is transformed into discrete, event-based loss evaluation units. For example, it can analyze the specific loss values ​​generated by using the filter cartridge for 2 hours when the water TDS is 300 ppm and by using it for 3 hours when the water TDS is 100 ppm. This quantification of single-use loss degree improves the system's understanding of the instantaneous changes in filter cartridge workload, enabling the model to capture the key impact of high-intensity usage periods on filter cartridge lifespan, thereby enhancing the predictive model's sensitivity to sudden or periodic water quality deterioration. By aligning the timestamps based on the single-use filter cartridge wear degree and usage time interval, the temporal relationship between loss events and idle intervals can be accurately established. For example, in commercial environments, water dispensers exhibit intermittent usage characteristics, such as high load during working hours and zero load at night or on holidays. This allows the model to further analyze patterns such as whether filter performance partially recovers after a period of high-wear use followed by long periods of inactivity, or whether frequent short-interval start-stop cycles exacerbate filter wear. By constructing this timestamped dataset for analysis, the quality of the data input to the prediction model is improved, enabling it to reflect not only the amount of wear but also when and at what pace the wear occurs, thereby enhancing the reliability and stability of the final lifespan prediction results.

[0038] In one embodiment, in step S30, the filter life prediction model is trained using the following method: S301: Obtain the correlation between filter cartridge interval time and filter cartridge life loss during continuous use under different water quality conditions from the historical usage data of the filter cartridge, and construct the corresponding loss curve based on the correlation.

[0039] In this embodiment, the filter cartridge historical usage refers to the operation and replacement records collected from a large number of identical devices in real commercial scenarios, covering the complete lifecycle of the filter cartridge from brand new installation to complete failure. The correlation refers to a statistical or mathematical model, derived through data analysis, showing how the device's usage interval pattern affects the filter cartridge's performance degradation rate under different water quality conditions. The loss curve plots the performance degradation trajectory with cumulative usage time or cumulative treated water volume on the horizontal axis and the normalized performance index or remaining lifespan percentage of the filter cartridge on the vertical axis.

[0040] Specifically, the historical dataset required for model training comprises detailed logs from multiple water dispensers recorded during long-term operation. Each log entry includes a timestamp, incoming water quality data, water production duration, and the time interval between preceding and following events. This historical data is first grouped according to water quality, for example, into low-hardness, medium-hardness, and high-hardness groups based on average incoming water TDS. Within each water quality group, the algorithm further analyzes the distribution patterns of filter cartridge interval times, such as frequent short intervals, regular daytime use with long intervals, and random intervals, and their correlation with the final filter cartridge lifespan, i.e., the actual lifespan. For example, through data mining, it was found that in the high-hardness group, devices that frequently experienced long periods of inactivity, such as intervals exceeding 24 hours, had significantly shorter actual total filter cartridge lifespans than devices with more uniform interval times. Based on this analysis, the algorithm fits a representative performance degradation curve, i.e., a loss curve, for each typical water quality-interval pattern combination. These curves can be represented by mathematical functions, such as exponential decay functions and piecewise linear functions. The process of constructing this curve involves extracting a large amount of raw operating data into several models that can characterize the typical paths of filter element wear under different operating conditions, providing an intuitive and quantitative basis for subsequent identification of key attenuation stages and setting differentiated calculation weights.

[0041] S302: Obtain the inflection point of loss from the loss curve. Based on different water quality conditions, set corresponding calculation weights for each loss inflection point, filter cartridge interval time, and filter cartridge life loss.

[0042] In this embodiment, the loss inflection point refers to the critical point on the loss curve where the performance degradation rate changes significantly, such as the turning point from a slow linear degradation phase to an accelerated degradation phase. The calculated weights are important parameters in the model used to quantify the influence of different factors, such as specific inflection point characteristics and specific time interval patterns, on the final lifetime prediction result.

[0043] Specifically, after obtaining the wear curves representing different operating conditions, the training program uses mathematical methods to automatically identify the wear inflection points on each curve. For example, the inflection point where the wear gradually becomes rapid can be obtained by calculating the first derivative of the curve, or a threshold for the decay rate can be set, and the starting point exceeding this threshold can be defined as the inflection point. The significance of identifying the inflection point is that it marks the critical state where the filter cartridge enters a period of rapid performance degradation from a stable operating period. Furthermore, the goal of training is to enable the model to predict when the inflection point will occur and the rate of degradation after the inflection point. To this end, the model needs to configure different computational weights for different input features. The algorithm analyzes historical data and summarizes patterns: for example, in the mode of high hardness water quality and extremely irregular usage intervals, the inflection point occurs earlier, and the slope of the curve after the inflection point (decay weight) is very large; while in the mode of low hardness water quality and stable daily use, the inflection point occurs later, and the decay weight is also smaller. The training process is to adjust the parameters inside the model so that when the model receives a set of features describing the current water quality and recent interval patterns, the predicted inflection point position and subsequent decay rate can match these patterns summarized from the historical wear curves. These finalized internal parameters constitute the set of computational weights that the model dynamically invokes when facing different situations.

[0044] S303: Train the initial model with the calculation weights under different water quality conditions to obtain the filter cartridge life prediction model.

[0045] Specifically, after completing the aforementioned data analysis and feature extraction, such as constructing curves, identifying inflection points, and deriving weight patterns, the data is used to guide model training. Training data consists of features X and labels (Y). Features X include water quality sequences, usage time series, and interval time series from historical data. Labels Y can include the final actual remaining lifespan of the filter cartridge, as well as higher-level labels derived from the aforementioned analysis, such as whether the inflection point has passed and which decay stage it is currently in. A suitable initial model is selected, such as an LSTM model capable of handling time-series data. In training iterations, a batch of (X, Y) data is input into the model each time, and the model calculates the predicted value based on the current parameters. The difference between the predicted value and the true Y is calculated using a loss function, such as mean squared error loss, and optionally an additional penalty term for inflection point prediction errors. The optimization algorithm backpropagates based on this difference, adjusting millions of parameters in the model. Simultaneously, the correspondence between inflection points and decay weights obtained in the aforementioned steps under different water quality and interval modes can be incorporated into the training process as prior knowledge or constraints. For example, a regularization term can be added to the loss function to penalize situations where the inflection point features predicted by the model are significantly inconsistent with prior knowledge. After a sufficient number of rounds of data training and parameter optimization, the model's predictive ability tends to stabilize, and the loss function is reduced to a low level, thus obtaining the filter life prediction model.

[0046] In this embodiment, historical data is actively extracted to reveal the true trajectory of filter cartridge performance changes over time or usage under different water quality conditions. The resulting loss curves visually demonstrate that filter cartridge loss is often non-linear and may accelerate after a certain critical point. The curve construction based on historical real data ensures that the model's learning objective aligns with the actual physicochemical changes, enhancing the physical interpretability of the model's predictions. By identifying loss inflection points from the loss curves—which typically correspond to the critical state of accelerated filter cartridge performance degradation, possibly indicating severe pore blockage, near-saturation of adsorption capacity, or the formation of a fouling layer on the membrane surface—the model can provide early warnings of precipitous performance drops, rather than just the average rate of degradation. This is crucial for preventative maintenance. Based on different water quality conditions, corresponding calculation weights are assigned to each loss inflection point, filter cartridge interval time, and filter cartridge lifespan loss. This allows the filter cartridge to reach its loss inflection point later under high-quality water conditions, with idle time potentially having a positive weight on performance recovery; while under poor water quality conditions, the inflection point occurs earlier, and idle time may have a negative weight due to contaminant deposition. By assigning differentiated weights to this key feature, the trained model acquires context-aware capabilities. Furthermore, by training the initial model using these calculated weights, the resulting filter life prediction model improves its robustness in the face of varying water quality and usage patterns, reducing the likelihood of inaccurate predictions in new environments due to over-reliance on data from a single operating condition.

[0047] In one embodiment, in step S30, the dataset to be analyzed is input into a preset filter life prediction model. The filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed, specifically including: S31: Input the single filter element wear level and usage time interval from the dataset to be analyzed into the filter element life prediction model.

[0048] Specifically, when a prediction needs to be generated, such as when a scheduled task is triggered or a user query is requested, the latest dataset to be analyzed is read from storage. This data is then input into the filter life prediction model. After receiving the input data, the model loads its pre-trained parameters and begins to perform forward propagation calculations.

[0049] S32: Obtain the calculation weights calculated by the filter life prediction model based on each usage time interval, and calculate the filter life prediction result based on the degree of filter wear per use and the calculation weights.

[0050] Specifically, after the model receives the input time-series data, it first analyzes the usage time interval patterns in the input sequence. Based on what it has learned during training, the model understands the meaning of different interval patterns. For example, a long interval may mean that contaminants have a longer time to accumulate and concentrate inside the filter cartridge, so the degree of single-use filter cartridge wear that follows may indicate the onset of more severe clogging, and this event should be given a higher computational weight. Conversely, events in a continuous and stable operating state may have a lower weight. The model calculates a dynamic, context-dependent weight Wi for each single-use filter cartridge wear level Li in the sequence through its internal mechanism. After obtaining all weights Wi, the model performs a weighted cumulative calculation. One approach is to calculate the filter cartridge lifespan prediction result, such as the remaining lifespan percentage R, as: R = 1 - Σ(Li×Wi) / Z, where Z is a lifespan normalization base preset according to the filter cartridge type. Another approach is for the model to directly output the predicted value through its final fully connected layer or regression layer. The total wear of the filter element is a weighted sum of wear from all historical events, with the weights dynamically determined by the specific circumstances of each event. Through this mechanism, the model can distinguish the different impacts of uniform wear and intermittent impact wear on the long-term lifespan of the filter element. Ultimately, the model outputs a filter element lifespan prediction, a quantitative and intelligent judgment that incorporates understanding of time-series patterns, rather than simply a summation.

[0051] In this embodiment, the accurate application of the prediction model is achieved, transforming the complex patterns of the model training phase into executable and interpretable real-time computational logic, ensuring the real-time nature and reliability of the prediction results. By acquiring the computational weights calculated by the filter life prediction model based on each usage time interval, the model dynamically matches or calculates the corresponding computational weights based on the currently input, specific usage time interval data. For example, for a long idle interval, the model may use a negative weight representing the risk of microbial growth or a positive weight representing the solidification of contaminant deposits; for frequent short intervals, it may use a weight representing the mechanical stress from frequent start-stop cycles. Through this dynamic weight matching mechanism, the model can respond in real-time to the actual working conditions of the equipment, thereby enhancing the individual adaptability and contextual accuracy of the prediction. The final comprehensive calculation yields the filter life prediction result based on the single filter wear level and the computational weights. The dynamic weights determined by the time intervals representing the working mode and recovery effect are used for comprehensive calculation. This calculation method effectively simulates the cumulative process of filter life loss: Total loss = Σ (loss amount per event × influence coefficient of the specific time environment before and after the event). This makes the filter life prediction result output by the model based on the accumulation after a detailed evaluation of each usage event, which greatly improves the consistency between the prediction result and the actual physical degradation of the filter, making the maintenance recommendations more scientific and timely.

[0052] In one embodiment, step S40, namely obtaining the filter material type and generating filter element maintenance recommendations for each filter material type based on the filter element lifespan prediction results, specifically includes: S41: Obtain the estimated lifespan of the filter cartridge for each filter material type.

[0053] Specifically, in a commercial water dispenser equipped with multi-stage filtration, such as PP cotton, activated carbon, RO membrane, and post-activated carbon filter, each filter element works collaboratively, but their failure rates and replacement cycles differ. The predictive model can perform a unified calculation for the entire machine during operation, but at the output or intermediate layers, it generates independent filter life estimates for each filter element and associates these results with specific physical filter element instances. The current device's filter element layout information is stored in memory or a configuration file, for example: Slot 1: PP cotton filter element, serial number SN001; Slot 2: Pre-activated carbon filter element, serial number SN002; ... After the predictive model completes its calculations, it queries this configuration table to retrieve the various output values ​​and assign them to the corresponding filter material types. For example, the system records: Material type: PP cotton, estimated remaining lifespan: 18%; Material type: RO membrane, estimated remaining lifespan: 65%. The ability to provide precise maintenance recommendations down to the specific type of filter element gives management actions a clear direction, avoiding vague maintenance prompts and laying a data foundation for refined and low-cost operation and maintenance.

[0054] S42: If the estimated lifespan of a filter cartridge is lower than the lifespan threshold of the filter material, filter cartridge maintenance recommendations will be generated based on water quality test data and the type of remaining filter material.

[0055] In this embodiment, the lifespan threshold is a preset critical value for different filter material types, used to trigger different levels of maintenance alarms. The filter cartridge maintenance recommendation is an actionable guidance scheme generated based on diagnostic results and integrated with system context information.

[0056] Specifically, at least two lifespan thresholds can be preset for each filter material type: a warning threshold and a replacement threshold. Each estimated result is compared with the threshold for its corresponding material type. For example, the replacement threshold for PP cotton is set at 20%, and the replacement threshold for RO membrane is set at 25%. When an estimated result for a certain filter type is detected, such as PP cotton remaining at 15% below its replacement threshold, an emergency replacement recommendation is triggered. If the result, such as RO membrane remaining at 35%, falls between the warning threshold and the replacement threshold, a warning is generated. When generating recommendations, the content is dynamically adjusted based on water quality testing data and the remaining filter material type. For example, when the RO membrane lifespan declines abnormally rapidly, while recommending RO membrane replacement, the recent influent TDS history can be analyzed. If it remains consistently high, the recommendation can include the explanation: "Long-term high influent hardness has been detected; it is recommended to simultaneously check the status of the upstream PP cotton and activated carbon filter cartridges." The remaining filter material type refers to other filter cartridges in the system that have not triggered a replacement alarm. When generating recommendations for a specific filter cartridge, the status of its upstream and downstream filter cartridges is evaluated. For example, when recommending the replacement of pre-activated carbon filters, if it is found that the lifespan of the downstream RO membrane is also nearing its threshold, the recommendation can be added. For instance, if the lifespan of the downstream RO membrane has also been reduced to 30%, it is recommended to evaluate and replace the RO membrane simultaneously to achieve better overall water purification results and a longer system stability period. Through this recommendation generation logic that combines root cause analysis and system cascading considerations, the output filter maintenance recommendations can provide comprehensive maintenance decision support information that includes problem identification, impact assessment, and optimization strategies, significantly improving the professionalism and proactive prevention capabilities of operation and maintenance work.

[0057] In this embodiment, the maintenance strategy is differentiated, precise, and proactive, transforming prediction results into specific and actionable operation and maintenance instructions, and enabling the overall system status to be linked, thereby improving the efficiency and effectiveness of the entire maintenance operation. Since a water purification system contains various filter materials, such as PP cotton, activated carbon, and RO membranes, their functions, costs, replacement difficulties, and failure consequences differ. By obtaining the lifespan prediction results for each filter material type, and obtaining their lifespan predictions separately, each stage of filter material can be managed independently. If the lifespan prediction result of a filter material is lower than its lifespan threshold, a filter material maintenance suggestion is generated based on water quality testing data and the remaining filter material type. This suggestion, generated based on water quality testing data and the remaining filter material type, can prompt the replacement of a specific filter material. It may also combine the water quality data triggered by the warning to provide root cause analysis suggestions, such as rapid RO membrane lifespan decay, high current influent water hardness, suggesting checking the pre-treatment filter effect and considering the installation of a water softener. When determining to replace a specific filter material, the remaining filter material type can be considered. For example, when it's determined that the pre-filter PP cotton needs replacement, it might be recommended to check the activated carbon filter at the same time, as it may already be loaded with a significant amount of penetrating colloidal substances. Alternatively, when replacing the RO membrane, it might be suggested to replace the post-activated carbon filter simultaneously to ensure optimal taste. This approach, by suggesting the synergistic working relationship between filter elements, enhances the systematic and thoroughness of maintenance, reducing the risk of a decline in overall water purification efficiency or rapid contamination of newly replaced filters due to poor condition of other filters caused by replacing a single stage.

[0058] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0059] In one embodiment, a water purification device for an intelligent commercial drinking water system is provided, which corresponds one-to-one with the water purification method of the intelligent commercial drinking water system described in the above embodiments. For example... Figure 2 As shown, the water purification device of this intelligent commercial drinking water system includes a parameter acquisition module, a data processing module, a lifespan estimation module, and a maintenance suggestion module. Detailed descriptions of each functional module are as follows: The parameter acquisition module is used to acquire the water dispenser initialization message triggered after the filter material is replaced, and to acquire the water purification parameters of the water dispenser. The water purification parameters of the water dispenser include water quality test data, filter material usage time and usage interval time. The data processing module is used to preprocess water quality test data, filter material usage time, and usage interval to obtain the dataset to be analyzed. The lifespan prediction module is used to input the dataset to be analyzed into the preset filter lifespan prediction model. The filter lifespan prediction model dynamically calculates the filter lifespan prediction result based on the correlation between the data in the dataset to be analyzed. The maintenance suggestion module is used to obtain the filter material type and generate filter maintenance suggestions for each filter material type based on the filter life prediction results.

[0060] Optionally, the data processing module includes: The wear level acquisition submodule is used to generate the wear level of a single filter cartridge based on water quality test data and the usage time of the filter material; The data sorting submodule is used to align timestamps based on the degree of single filter cartridge wear and the time interval between uses to obtain the dataset to be analyzed.

[0061] Optionally, the water purification device for the intelligent commercial drinking water system may also include: The historical data preprocessing module is used to obtain the correlation between filter cartridge interval time and filter cartridge life loss during continuous use under different water quality conditions from the historical usage of filter cartridges, and to construct the corresponding loss curve based on the correlation. The weight calculation module is used to obtain the loss inflection point from the loss curve. According to different water quality conditions, the corresponding calculation weight is set for each loss inflection point, filter cartridge interval time and filter cartridge life loss. The model training module is used to train the initial model with the calculation weights under different water quality conditions to obtain the filter cartridge life prediction model.

[0062] Optional, the lifespan prediction module includes: The data input submodule is used to input the single filter element wear level and usage time interval from the dataset to be analyzed into the filter element life prediction model. The lifespan prediction submodule is used to obtain the calculation weights calculated by the filter lifespan prediction model based on each usage time interval, and to calculate the filter lifespan prediction result based on the degree of filter wear per use and the calculation weights.

[0063] Optional, the maintenance suggestion module includes: The prediction result acquisition submodule is used to obtain the predicted lifespan of the filter cartridge for each type of filter material. The maintenance suggestion generation submodule is used to generate filter cartridge maintenance suggestions based on water quality test data and the type of remaining filter material if the estimated lifespan of the filter cartridge is lower than the lifespan threshold of the filter material.

[0064] Specific limitations regarding the water purification device of the intelligent commercial drinking water system can be found in the limitations of the water purification method for the intelligent commercial drinking water system mentioned above, and will not be repeated here. Each module in the water purification device of the aforementioned intelligent commercial drinking water system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0065] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a water purification method for an intelligent commercial drinking water system.

[0066] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Get the water dispenser initialization message triggered after the filter material is replaced, and get the water dispenser's water purification parameters, including water quality test data, filter material usage time, and usage interval time. After preprocessing the water quality test data, filter material usage time, and usage interval, the dataset to be analyzed is obtained. Input the dataset to be analyzed into the preset filter life prediction model. The filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed. Obtain the filter material type and generate filter maintenance recommendations for each filter material type based on the filter life prediction results.

[0067] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Get the water dispenser initialization message triggered after the filter material is replaced, and get the water dispenser's water purification parameters, including water quality test data, filter material usage time, and usage interval time. After preprocessing the water quality test data, filter material usage time, and usage interval, the dataset to be analyzed is obtained. Input the dataset to be analyzed into the preset filter life prediction model. The filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed. Obtain the filter material type and generate filter maintenance recommendations for each filter material type based on the filter life prediction results.

[0068] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0069] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0070] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A water purification method for an intelligent commercial drinking water system, characterized in that, The water purification method of the intelligent commercial drinking water system includes: Get the water dispenser initialization message triggered after the filter material is replaced, and get the water dispenser purification parameters, wherein the water dispenser purification parameters include water quality test data, filter material usage time and usage interval time; After preprocessing the water quality test data, the filter material usage time, and the usage time interval, a dataset to be analyzed is obtained. The dataset to be analyzed is input into a preset filter life prediction model, and the filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed. Obtain the filter material type and generate filter maintenance recommendations for each filter material type based on the filter life prediction results.

2. The water purification method of the intelligent commercial drinking water system according to claim 1, characterized in that, The process of preprocessing the water quality testing data, the filter material usage time, and the usage time interval to obtain the dataset to be analyzed specifically includes: Based on the water quality test data and the usage time of the filter material, the degree of single filter cartridge wear is generated; The dataset to be analyzed is obtained by aligning the timestamps based on the single filter cartridge wear level and the usage time interval.

3. The water purification method of the intelligent commercial drinking water system according to claim 2, characterized in that, The filter cartridge life prediction model was trained using the following method: The correlation between filter cartridge interval time and filter cartridge life loss during continuous use under different water quality conditions is obtained from the historical usage data of filter cartridges, and a corresponding loss curve is constructed based on the correlation. The loss inflection point is obtained from the loss curve. According to different water quality conditions, the corresponding calculation weights are set for each loss inflection point, the filter cartridge interval time and the filter cartridge life loss. The calculated weights under different water quality conditions are used to train the initial model to obtain the filter cartridge life prediction model.

4. The water purification method of the intelligent commercial drinking water system according to claim 3, characterized in that, The step of inputting the dataset to be analyzed into a preset filter life prediction model, wherein the filter life prediction model dynamically calculates the filter life prediction result based on the correlation between the data in the dataset to be analyzed, specifically includes: The filter life prediction model is input into the single filter wear level and the usage time interval from the dataset to be analyzed. The calculation weights obtained by the filter life prediction model based on each usage time interval are obtained, and the filter life prediction result is calculated based on the single filter wear level and the calculation weights.

5. The water purification method for the intelligent commercial drinking water system according to claim 1, characterized in that, The process of obtaining the filter material type and generating filter maintenance recommendations for each filter material type based on the filter lifespan prediction results specifically includes: Obtain the estimated lifespan of the filter cartridge for each of the aforementioned filter material types; If the estimated lifespan of the filter cartridge is lower than the lifespan threshold of the filter material, then a filter cartridge maintenance recommendation is generated based on the water quality test data and the remaining filter material type.

6. A water purification device for an intelligent commercial drinking water system, characterized in that, The water purification device of the intelligent commercial drinking water system includes: The parameter acquisition module is used to acquire the water dispenser initialization message triggered after the filter material is replaced, and to acquire the water purification parameters of the water dispenser, wherein the water purification parameters of the water dispenser include water quality test data, filter material usage time and usage interval time; The data processing module is used to preprocess the water quality test data, the filter material usage time, and the usage time interval to obtain the dataset to be analyzed. The lifespan prediction module is used to input the dataset to be analyzed into a preset filter lifespan prediction model. The filter lifespan prediction model dynamically calculates the filter lifespan prediction result based on the correlation between the data in the dataset to be analyzed. The maintenance suggestion module is used to obtain the filter material type and generate filter element maintenance suggestions corresponding to each filter material type based on the filter element life estimation results.

7. The water purification device for the intelligent commercial drinking water system according to claim 6, characterized in that, The data processing module includes: The wear level acquisition submodule is used to generate the wear level of a single filter cartridge based on the water quality test data and the usage time of the filter material; The data sorting submodule is used to align the timestamps according to the single filter element wear level and the usage time interval to obtain the dataset to be analyzed.

8. The water purification device for the intelligent commercial drinking water system according to claim 7, characterized in that, The water purification device of the intelligent commercial drinking water system also includes: The historical data preprocessing module is used to obtain the correlation between filter cartridge interval time and filter cartridge life loss during continuous use under different water quality conditions from the historical usage of filter cartridges, and to construct the corresponding loss curve based on the correlation. The weight calculation module is used to obtain the loss inflection point from the loss curve and set corresponding calculation weights for each loss inflection point, the filter cartridge interval time and the filter cartridge life loss according to different water quality conditions. The model training module is used to train the initial model with the calculated weights under different water quality conditions to obtain the filter cartridge life prediction model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the water purification method of the intelligent commercial drinking water system as described in any one of claims 1 to 5.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the water purification method of the intelligent commercial drinking water system as described in any one of claims 1 to 5.

Images