Water quality monitoring method, device, storage medium and program product

Abstract

The present application discloses a water quality monitoring method, device, storage medium and program product, relating to the technical field of water quality treatment. The method includes: preprocessing the water quality data collected by multiple sensors to obtain standardized water quality data, where the multiple sensors are arranged at key nodes of a clean workshop; inputting the standardized water quality data into a water quality prediction model trained based on a convolutional neural network model to obtain water quality change trend data; when the water quality change trend data reaches a warning threshold, triggering a warning mechanism and inputting the water quality data into a Bayesian fault model to obtain a water quality optimization strategy. Since the present application can accurately predict the water quality change trend by using the water quality prediction model after collecting the water quality data, it can give an early warning when the water quality may be abnormal and make adjustments through the water quality optimization strategy. It avoids the lag situation existing in the existing water quality monitoring methods, so as to be able to respond to water quality problems in advance and ensure the quality and efficiency of the production water in the clean workshop.

Description

Technical Field

[0001] The present application relates to the technical field of water quality treatment, and particularly to a water quality monitoring method, device, storage medium and program product. Background Art

[0002] With the continuous improvement of the requirements for a clean environment in modern industrial production, especially in the fields of electronics, medicine, precision manufacturing, etc., the precision requirements for water quality have reached an unprecedented level. Therefore, in the operation of clean workshops, the water quality guarantee of the water supply system is crucial.

[0003] Traditional water quality monitoring methods generally have many limitations. For example, manual sampling and detection have problems such as long time intervals and difficulty in reflecting water quality changes in real time, which easily lead to water quality problems during the interval between two detections but cannot be discovered in time. And some conventional on-line monitoring devices can only monitor a small number of water quality parameters, and have insufficient comprehensive monitoring capabilities for the complex water use environment and various key water quality indicators (such as microorganisms, particles, ion concentration, etc.) in clean workshops. When water quality changes are detected, serious water quality problems have already occurred, showing a lag.

[0004] Therefore, the existing water quality monitoring methods are generally in a passive monitoring state when water quality changes, with a lag, difficult to actively prevent, and easily leading to serious consequences such as production losses and product quality degradation. Summary of the Invention

[0005] The main purpose of the present application is to provide a water quality monitoring method, device, storage medium and program product, aiming to solve the technical problems that the existing water quality monitoring methods are generally in a passive monitoring state when water quality changes, with a lag, difficult to actively prevent, and easily leading to production losses.

[0006] To achieve the above object, the present application proposes a water quality monitoring method, which includes:

[0007] Preprocess the water quality data of the clean workshop collected by multiple sensors to obtain standardized water quality data, and the multiple sensors are pre-set at key nodes of the clean workshop;

[0008] Input the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, and the water quality prediction model is obtained by training based on a convolutional neural network model;

[0009] When the water quality change trend data reaches the warning threshold, trigger a warning mechanism;

[0010] Based on the warning mechanism, input the water quality data into a preset Bayesian fault model to obtain a corresponding water quality optimization strategy.

[0011] In one embodiment, the step of preprocessing the water quality data of a clean workshop collected by multiple sensors to obtain standardized water quality data includes:

[0012] Collect the water quality data of the clean workshop by multiple sensors at a preset sampling frequency, where the multiple sensors include at least one of a microbial sensor, a particle count sensor, an ion concentration sensor, a pH value sensor, and a dissolved oxygen sensor;

[0013] Perform data preprocessing on the water quality data to obtain standardized water quality data, where the data preprocessing includes outlier processing, missing value filling, and standardization processing.

[0014] In one embodiment, the step of performing data preprocessing on the water quality data to obtain standardized water quality data includes:

[0015] When the data points of the water quality data exceed a preset fluctuation range, determine that the data points are outliers;

[0016] Remove the outliers to obtain the cleaned water quality data;

[0017] Fill in the missing values of the cleaned water quality data by a preset numerical filling method to obtain the filled water quality data;

[0018] Map the filled water quality data to a standard numerical interval to obtain standardized water quality data.

[0019] In one embodiment, the step of inputting the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data includes:

[0020] Extract the water quality indicators of the standardized water quality data, where the water quality indicators include at least one of pH value, conductivity, dissolved oxygen, turbidity, temperature, and ion concentration;

[0021] Calculate the statistical indicators of the water quality indicators by a statistical analysis method, where the statistical indicators include at least one of mean, standard deviation, and coefficient of variation;

[0022] Input the water quality indicators and the statistical indicators into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is trained by a convolutional neural network model with historical water quality data as training samples.

[0023] In one embodiment, the training process of the water quality prediction model includes:

[0024] Obtain the historical water quality indicators and historical statistical indicators of the historical water quality data;

[0025] Classify the historical statistical indicators according to the time dimension to obtain linear data and non-linear data;

[0026] Perform linear fitting on the linear data and regression on the non-linear data by the random forest regression method to obtain standardized data;

[0027] Input the standardized data and the historical water quality indicators into a convolutional neural network model for correlation training to obtain a water quality prediction model.

[0028] In one embodiment, the step of inputting the water quality data into a preset Bayesian fault model based on the warning mechanism to obtain a corresponding water quality optimization strategy includes:

[0029] Based on the warning mechanism, start a preset Bayesian fault model, and the Bayesian fault model analyzes through the causal relationship and probability between water quality data;

[0030] Diagnose the water quality data through the Bayesian fault model and a preset expert system to determine the water treatment fault links and fault causes of the water quality data in the clean workshop;

[0031] Generate a corresponding water quality optimization strategy according to the water treatment fault links and fault causes;

[0032] Warn the maintenance personnel according to the water quality optimization strategy.

[0033] In one embodiment, the step of diagnosing the water quality data through the Bayesian fault model and a preset expert system to determine the water treatment fault links and fault causes of the water quality data in the clean workshop includes:

[0034] Use the water quality indicators as observation nodes and sensor failures as hidden nodes through the Bayesian fault model;

[0035] Determine the causal relationship between water quality data according to the hidden nodes;

[0036] Based on the historical water quality data, define the conditional probability table between the causal relationship and the observation nodes;

[0037] Diagnose the water quality data through a preset expert system based on the conditional probability table to obtain the water treatment fault links and fault causes of the water quality data in the clean workshop.

[0038] In addition, to achieve the above object, the present application also proposes a water quality monitoring device, and the device includes:

[0039] A preprocessing module for preprocessing the water quality data of a clean workshop collected by multiple sensors to obtain standardized water quality data, where the multiple sensors are pre-set at key nodes of the clean workshop;

[0040] A water quality prediction module for inputting the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is obtained by training based on a convolutional neural network model;

[0041] An early warning mechanism module for triggering an early warning mechanism when the water quality change trend data reaches an early warning threshold;

[0042] An optimization strategy module for inputting the water quality data into a preset Bayesian fault model based on the early warning mechanism to obtain a corresponding water quality optimization strategy.

[0043] In addition, to achieve the above object, the present application also proposes a storage medium, the storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, the steps of the water quality monitoring method as described above are implemented.

[0044] In addition, to achieve the above object, the present application also provides a computer program product, the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the water quality monitoring method as described above are implemented.

[0045] One or more technical solutions proposed by the present application have at least the following technical effects: The water quality monitoring method of the present application includes: preprocessing the water quality data of a clean workshop collected by multiple sensors to obtain standardized water quality data, where the multiple sensors are pre-set at key nodes of the clean workshop; inputting the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is obtained by training based on a convolutional neural network model; triggering an early warning mechanism when the water quality change trend data reaches an early warning threshold; based on the early warning mechanism, inputting the water quality data into a preset Bayesian fault model to obtain a corresponding water quality optimization strategy.

[0046] Since after the present application collects water quality data through multiple sensors, the water quality prediction model can accurately predict the water quality change trend, and give an early warning and generate a water quality optimization strategy for adjustment when the water quality may be abnormal. It avoids the lag of the existing water quality monitoring methods, so as to be able to respond to water quality problems in advance and ensure the quality and efficiency of the production water in the clean workshop. Description of the Drawings

[0047] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments in line with this application, and are used together with the specification to explain the principles of this application.

[0048] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a schematic flowchart provided for the first embodiment of the water quality monitoring method of this application;

[0050] Figure 2 It is a schematic flowchart provided for the second embodiment of the water quality monitoring method of this application;

[0051] Figure 3 It is the overall flowchart architecture diagram of the water quality monitoring process provided by this application;

[0052] Figure 4 It is a schematic diagram of the module structure of the water quality monitoring device in the embodiment of this application.

[0053] The realization of the purpose of this application, functional features and advantages will be further described in conjunction with the embodiments with reference to the accompanying drawings. Detailed implementation manners

[0054] It should be understood that the specific embodiments described here are only used to explain the technical solutions of this application and are not used to limit this application.

[0055] To better understand the technical solutions of this application, the following will be described in detail in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0056] Traditional manual sampling and detection usually collect water samples from water supply points or water-using equipment in the plant at fixed time intervals (such as once a day or several times a week), and then send them to the laboratory for analysis. The biggest drawback of this method is the time lag. During the interval between two samplings, if the water quality suddenly changes, such as the rapid growth of microorganisms or the accidental mixing of impurities in the water, it is difficult to detect in time. Moreover, the manual operation process is prone to introduce errors, including problems such as contamination during sampling and improper sample preservation, affecting the accuracy of the detection results.

[0057] While some on - line monitoring devices can obtain data in real - time, their functions are relatively single. Commonly, they are some simple sensors that can only measure a few basic parameters such as pH value and conductivity. For the monitoring of microbial content, which is crucial for modern clean workshop production, existing devices may be difficult to achieve high - precision detection and cannot distinguish different types of microorganisms and their potential hazards to production. For example, in terms of particle counting, the monitoring of tiny particles (especially those with a particle size smaller than a certain degree) is not accurate enough, and these tiny particles may have a serious impact on precision manufacturing. For the monitoring of specific ion concentrations in water, generally only a few common ions are targeted, and it is difficult to effectively measure trace ions that need to be strictly controlled in some special production processes.

[0058] Therefore, this application provides a solution. By using a variety of high - precision sensors to comprehensively monitor water quality parameters, it is more comprehensive than the single or small - number - parameter monitoring of the prior art. At the same time, by using cloud computing and big data analysis technologies, the problem that the prior art is difficult to process massive and complex data is overcome. And it can explore the data value, accurately predict the trend of water quality changes, be able to respond to water quality problems in advance, and improve production continuity and efficiency.

[0059] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, water quality prediction and water quality warning functions, such as a personal computer, a cloud server, etc., or an electronic device that can implement the above functions, a water quality monitoring device that executes the water quality monitoring method of this application, etc. This embodiment does not limit this. Hereinafter, taking the cloud server as an example, this embodiment and the following embodiments will be described.

[0060] Based on this, Embodiment 1 of this application is proposed. Embodiment 1 of this application provides a water quality monitoring method, referring to Figure 1 , Figure 1 which is the process schematic diagram provided for Embodiment 1 of the water quality monitoring method of this application.

[0061] In this embodiment, the water quality monitoring method includes steps S10 - S40:

[0062] Step S10: Pre - process the water quality data of the clean workshop collected by multiple sensors to obtain standardized water quality data, and the multiple sensors are pre - set at key nodes of the clean workshop.

[0063] It should be noted that the clean workshop can be a special workshop dedicated to water quality treatment - related operations. In the clean workshop, there are strict environmental control requirements, the purpose of which is to ensure that the treated water quality meets high standards. Therefore, in order to ensure the water quality of the water service system in the clean workshop, it is necessary to monitor its water quality.

[0064] Due to the extremely high requirements for water quality in the production of clean workshops, traditional monitoring methods can generally only detect a limited number of parameters, and the accuracy is limited. Before conducting water quality monitoring in this application, during the construction or renovation stage of the clean workshop, it is necessary to plan the installation locations of water quality sensors (i.e., multi-sensors) for collecting water quality data at various locations in the clean workshop according to the water usage process of the workshop, the water quality requirements of different production areas, and the pipeline network layout. The installation locations generally need to be typical locations (i.e., key nodes) in the clean workshop that can reflect water quality changes. For example, for a clean workshop for electronic chip manufacturing, sensors are installed at key locations such as each processing link of the ultrapure water preparation system (such as before and after the reverse osmosis device, before and after the ion exchange column, etc.), the main water supply pipeline leading to the production workshop, and the water inlets of each chip manufacturing device. For microbial sensors, they are mainly installed near water storage devices prone to microbial growth and the return water pipeline to monitor the reproduction of microorganisms in real time.

[0065] After installing each sensor in accordance with compliance requirements, it is also necessary to calibrate each sensor using standard calibration solutions first. For example, for microbial sensors, use standard samples with known microbial concentrations to adjust the sensitivity and detection range of the sensors so that their measurement results match the standard values. For pH sensors, calibrate them using standard buffer solutions with pH values of 4.0, 7.0, and 10.0 to ensure the measurement accuracy of the sensors in different acidity and alkalinity environments.

[0066] At this time, the cloud server can be connected to the corresponding sensors, and appropriate parameters such as the sampling frequency and signal amplification factor can be set according to the type of sensor and the characteristics of the output signal. At the same time, metadata settings such as the location information and type identifier of the sensors are input into the data acquisition module.

[0067] Furthermore, to ensure the stable transmission of data between the cloud server and the sensors, a data concentrator can also be set between the cloud server and the sensors to cache the data. For example, when a short-term network failure occurs, the data concentrator can cache the data within a certain period of time and continue to transmit it after the network resumes to avoid data loss.

[0068] After configuring the above settings for the clean workshop, real-time water quality monitoring can be started.

[0069] It can be understood that water quality data can be various quantitative information used to describe the quality characteristics of water bodies, such as data on water temperature, chromaticity, turbidity, transparency, acidity and alkalinity (pH value), dissolved oxygen content, and concentrations of heavy metal ions (such as mercury, cadmium, lead, etc.).

[0070] Therefore, when the clean workshop is operating normally, various sensors can continuously collect the above water quality data according to the preset sampling frequency, and then transmit the digitized water quality data with metadata to the data concentrator in real time. The data concentrator receives data from each sensor, and preliminarily sorts and caches the data according to classification rules such as sensor type and water supply area, facilitating subsequent unified processing and transmission. Then the data concentrator transmits the cached data to the cloud server through the fiber optic network. During the transmission process, the SSL / TLS encryption protocol can be used to encrypt the data to ensure the security of the data during network transmission. After receiving the encrypted data, the cloud server restores the data using the corresponding decryption key and stores it in a pre-configured database.

[0071] It should be noted that a distributed computing environment based on Hadoop and Spark can also be built on the cloud server, and database systems such as MySQL, HBase, and MongoDB can be installed and configured. Create the corresponding database table structure for storing different types of data such as sensor information, water quality data, and user permissions. At the same time, set the resource allocation strategy of the cloud server and configure the trigger conditions for automatic resource scaling. For example, when the length of the data processing task queue exceeds a certain threshold or the storage utilization rate reaches a certain percentage, the server nodes or storage capacity are automatically increased.

[0072] It should be understood that the standardized water quality data can refer to the data obtained after standardizing or normalizing the water quality data.

[0073] In a specific implementation, when the clean workshop is operating normally, various sensors can continuously collect the above water quality data according to the preset sampling frequency. Then the cloud server performs standardization or normalization processing on these water quality data to obtain the standardized water quality data.

[0074] In a feasible implementation manner, step S10 of this embodiment may include the steps of: collecting the water quality data of the clean workshop by multiple sensors according to the preset sampling frequency, where the multiple sensors include at least one of a microbial sensor, a particle counting sensor, an ion concentration sensor, a pH value sensor, and a dissolved oxygen sensor; performing data preprocessing on the water quality data to obtain the standardized water quality data, and the data preprocessing includes outlier processing, missing value filling, and standardization processing.

[0075] It should be noted that the multi-sensors may include various water quality sensors. For example, the microbial sensor can accurately detect the microbial content in water, and based on the fluorescence detection principle, it can sensitively capture the microbial metabolism signal; the particle counting sensor can continuously count the particles in the water sample passing through the detection window and analyze the particle size, without missing tiny particle impurities; the ion concentration sensor measures the concentration change of specific ions through an ion-selective electrode; the pH sensor can measure the acidity and alkalinity (pH value) of the solution in the water sample; the dissolved oxygen sensor can detect the dissolved oxygen content in the water body. At the same time, when these sensors collect water quality data, metadata such as timestamps and sensor location codes can also be added.

[0076] These sensors comprehensively cover key water quality parameters such as microorganisms, particles, ion concentration, pH value, and dissolved oxygen, achieving high-precision and all-round real-time monitoring of water quality, ensuring that any water quality changes that may affect production can be detected in a timely manner. Through the above-mentioned high-precision multi-parameter sensor fusion, all-round and real-time monitoring of water quality can be achieved, overcoming the limitations of single-parameter or few-parameter monitoring in traditional monitoring methods.

[0077] It can be understood that the preset sampling frequency is the frequency of collecting water quality data of the clean workshop set on various sensors. For example, for the pH sensor and ion concentration sensor with relatively slow changes, a relatively low sampling frequency can be set (collecting data every 10 minutes); while for the microbial content and particle counting sensors that may change rapidly, a higher sampling frequency is set (collecting data every minute).

[0078] In this embodiment, after receiving the water quality data, the cloud server can first perform data cleaning operations on the water quality data, such as identifying and removing outliers in the data, filling in missing values in the data, etc. Then, the cleaned water quality data is standardized to obtain standardized water quality data, which facilitates subsequent data analysis.

[0079] In another feasible embodiment, the step of preprocessing the water quality data to obtain standardized water quality data in this embodiment includes: when the data points of the water quality data exceed the preset fluctuation range, determining that the data points are outliers; removing the outliers to obtain the cleaned water quality data; filling in the missing values of the cleaned water quality data by a preset numerical filling method to obtain the filled water quality data; mapping the filled water quality data to the standard numerical interval to obtain the standardized water quality data.

[0080] It should be noted that the preset fluctuation range can be a pre-set upper and lower limit interval that allows normal fluctuations in water quality.

[0081] Exemplarily, for microbial sensor data, if the change in microbial concentration at a certain data point is several times (i.e., the preset fluctuation range) more than the normal fluctuation range compared to the data at adjacent time points, and this change does not conform to the physical laws of water quality changes (such as no accompanying changes in temperature, pH value, etc.), then this data point is determined to be an outlier and is removed.

[0082] It can be understood that the preset numerical filling method can be an algorithm preset for filling these missing parts when there are missing values in the water quality data set. For example, the linear interpolation method based on adjacent data points or the mean filling method based on the statistical law of historical data, etc. This embodiment does not limit this.

[0083] In this embodiment, after the cloud server receives the water quality data, it first determines whether the data points of the water quality data exceed the preset fluctuation range. If they exceed the preset fluctuation range, the data points are determined to be outliers; and these outliers are removed to obtain the cleaned water quality data. Then, according to the time series characteristics of the data and other relevant parameters, the linear interpolation method based on adjacent data points or the mean filling method based on the statistical law of historical data is used to fill the missing values in the cleaned water quality data to obtain the filled water quality data. Finally, the cleaned water quality data is standardized to map water quality parameters of different magnitudes and units to a standard numerical range (for example, using Min - Max normalization for different types of data: (X - Min) / (Max - Min), where X is the data point of the water quality data to make the units uniform), to obtain the standardized water quality data. Thus, through data processing, the water quality data becomes more reasonable and accurate, facilitating subsequent data analysis.

[0084] Step S20: Input the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is trained based on a convolutional neural network model.

[0085] It should be noted that the water quality prediction model can be a model used to estimate and simulate the changes in water body quality. Through the water quality prediction model, the changes in water quality when the water body in a clean workshop is polluted in the future can be predicted.

[0086] It can be understood that the water quality change trend data can be data predicting the change direction of water body quality within a certain period. For example, trends such as the increase in the concentration of certain harmful substances.

[0087] It should be understood that the Convolutional Neural Network (CNN) model can be a model for deep learning based on water quality data.

[0088] Considering the huge and complex amount of data generated in the above-mentioned clean workshop, traditional methods are difficult to process massive and complex data; moreover, traditional methods also have the problem of not deeply mining data. Therefore, this embodiment can perform efficient storage and processing of massive data based on cloud servers and hybrid database storage. By means of machine learning and deep learning (convolutional neural network) algorithms, a water quality prediction model is constructed. Through in-depth data mining by the model, the internal relationship between water quality parameters and the change trend in the short term in the future can be accurately analyzed, and the time points of the impact of the accelerated microbial reproduction rate and ion concentration fluctuations on production can be known in advance, providing solid data support for subsequent decision-making, which is difficult to achieve by traditional methods.

[0089] In a specific implementation, the cloud server first performs deep learning on water quality data through a convolutional neural network model to obtain a water quality prediction model that can predict the change direction of water body quality within a certain period. After processing to obtain standardized water quality data, the standardized water quality data can be input into the water quality prediction model to obtain data on the change trend of the water quality of the water body in the clean workshop in the future for a period of time.

[0090] In a feasible implementation manner, step S20 of this embodiment may include the steps of: extracting water quality indicators of the standardized water quality data, where the water quality indicators include at least one of pH value, conductivity, dissolved oxygen, turbidity, temperature, and ion concentration; calculating statistical indicators of the water quality indicators through statistical analysis methods, where the statistical indicators include at least one of mean, standard deviation, and coefficient of variation; and inputting the water quality indicators and the statistical indicators into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is obtained by training through a convolutional neural network model with historical water quality data as training samples.

[0091] It should be noted that water quality indicators can be a series of parameters used to measure the water body quality status. For example, the pH value is used to reflect the acidity and alkalinity of water, the conductivity represents the ability of the solution to conduct current, the temperature is the degree of hotness and coldness of the water body, and the ion concentration can refer to the content of specific ions in the solution. Through the above water quality indicators, the water body quality can be comprehensively evaluated.

[0092] It can be understood that the statistical analysis method can be a method for statistically analyzing representative and accurate data of each water quality indicator in the standardized water quality data. For example, descriptive statistical analysis includes calculating the mean, median, standard deviation, etc. to summarize the central tendency and dispersion degree of the data; for example, inferential statistical analysis can estimate the overall characteristics and conduct hypothesis testing through sample data to determine whether there is a relationship between two or more variables, etc.

[0093] Among them, statistical indicators can be parameters used to describe the characteristics of statistical analysis data. The mean is the average of standardized water quality data, reflecting the average level or center position of standardized water quality data; the standard deviation measures the degree of dispersion of standardized water quality data relative to the mean, indicating the dispersion of standardized water quality data; the coefficient of variation is the ratio of the standard deviation to the mean, which is used to compare the degree of dispersion of two or more groups of data with different means or different units, and can more intuitively reflect the relative degree of variation of the data.

[0094] These statistical indicators can reflect the average level, fluctuation and stability of water quality within a certain time range. For example, by calculating the standard deviation of the ion concentration of water supplied to a production area, we can understand the degree of change in ion concentration and determine whether there is a potential risk of water pollution.

[0095] At the same time, machine learning algorithms (such as support vector machines, random forests, etc.) and deep learning algorithms (such as convolutional neural networks) can also be used to establish water quality prediction models.

[0096] Among them, the training process of the water quality prediction model described in this embodiment includes: obtaining historical water quality indicators and historical statistical indicators of historical water quality data; classifying the historical statistical indicators according to the time dimension to obtain linear data and nonlinear data; performing linear fitting on the linear data and regressing the nonlinear data through the random forest regression method to obtain standardized data; inputting the standardized data and the historical water quality indicators into the convolutional neural network model for correlation training to obtain a water quality prediction model.

[0097] It should be noted that historical water quality data can be various quantitative information used to describe the historical water quality characteristics of the clean room; historical water quality indicators can be a series of parameters used to measure the historical water quality conditions of the clean room; historical statistical indicators can be parameters used to describe and statistically analyze the characteristics of the historical data of the clean room.

[0098] Historical water quality data can be used as training samples to learn the intrinsic relationship between water quality parameters and the patterns of change over time. For example, by analyzing the correlation between water temperature, microbial content, particle counts and other data in the past few months and water quality deterioration events, a support vector machine model is established to predict whether water quality will exceed the normal range in the short term in the future. For image data of particle counts, a convolutional neural network model is used to extract and analyze features to improve the recognition ability and prediction accuracy of tiny particle changes.

[0099] Specifically, the training process of the water quality prediction model includes the following steps:

[0100] 1. Obtain historical water quality data (for example, use data from the past two years for basic training), and process it to obtain historical water quality indicators and historical statistical indicators. Then, classify the above historical statistical indicators according to the time dimension, and overall divide them into linear data and non-linear data;

[0101] 2. For linear data, directly perform linear fitting; for non-linear data, select the random forest regression method for regression to obtain standardized data. Among them, the water quality prediction model is applicable to non-linear and multi-feature data types.

[0102] 3. Then, use the standardized data, temperature, pH value, ion concentration and other historical water quality indicators as data inputs, and set the goal to predict the health status of water quality at different time periods.

[0103] 4. Then, use the time of the most recent two months as the test data, and start correlation training by selecting the optimal learning rate, number of trees, hidden layer size and other parameters by comparing the results of different parameters.

[0104] 5. Finally, use the mean square error, mean absolute error, coefficient of determination, etc. to verify the model results, determine the rationality, and obtain the water quality prediction model.

[0105] In this embodiment, by extracting the water quality indicators of the standardized water quality data and calculating the statistical indicators of the water quality indicators through statistical analysis methods; then inputting the water quality indicators and statistical indicators into the preset water quality prediction model to obtain the water quality change trend data. Thus, the water quality change trend can be accurately predicted, the water quality problem can be dealt with in advance, and the production continuity and efficiency can be improved.

[0106] Step S30: When the water quality change trend data reaches the warning threshold, trigger the warning mechanism.

[0107] It should be noted that the warning threshold can be a value preset in the cloud server for warning of excessive water quality changes. When the cloud server monitors that the water quality change trend data reaches or exceeds this warning threshold, an alarm signal will be sent.

[0108] Among them, the warning threshold can be set according to the requirements of different production processes for water quality. For example, for the ultrapure water in an electronic chip manufacturing workshop, the microbial content threshold is set to not exceed 1 colony forming unit per milliliter, and the number of particles with a particle size greater than 0.1 micrometers per liter does not exceed 10, etc. This embodiment does not limit this.

[0109] It is understandable that the early warning mechanism can be a program set for warning managers. For example, warning messages can be sent to managers in various ways such as text messages, emails, and mobile app push notifications. The warning messages can detail abnormal water quality parameters, the location where the problem occurs (e.g., precisely shown on the network diagram of the plant's water supply pipeline through sensor location coding), the possible scope of influence (analyzed based on the relevance between the water supply pipeline network and the water usage area), and the assessment of the severity of the problem (based on the degree of deviation of the water quality parameters from the threshold and the potential impact on production), etc.

[0110] Furthermore, the above data can be visualized, and the real-time water quality data can be presented to managers in intuitive forms such as charts and dashboards on the cloud server. For example, the changing trend of the microbial content in a certain water usage area over time can be shown through a line chart, and the pH value differences between different areas can be compared through a bar chart. Managers can view these real-time data visualization interfaces anytime and anywhere through the web terminal or mobile app to timely understand the water quality status in the plant.

[0111] Step S40: Based on the early warning mechanism, input the water quality data into a preset Bayesian fault model to obtain the corresponding water quality optimization strategy.

[0112] It should be noted that the Bayesian fault model can be a model based on Bayesian theory for analyzing the fault-related situations in a clean plant. Through the Bayesian fault model, data such as the occurrence of faults, the types of faults, the scope of influence of faults, and which fault is most likely to occur in the clean plant after the water quality changes can be determined, which helps to make maintenance decisions in advance and reduce maintenance costs.

[0113] It is understandable that the water quality optimization strategy can be a set of measures for improving and enhancing the water body quality and for emergency handling of faults.

[0114] Traditional fault diagnosis generally relies more on the experience judgment of on-site personnel, and the warning information is also incomplete. It can only prompt simple parameter over-limit, and cannot accurately evaluate the possible location, scope of influence, and severity of the fault, resulting in untimely or inaccurate countermeasures. However, through the Bayesian fault model in this application, the causal relationship between water quality data can be quickly analyzed to determine the root cause of the fault, such as filter blockage, disinfection equipment failure, pipeline leakage, etc., so as to take repair measures in time to avoid production interruption or product quality decline caused by water quality problems.

[0115] Exemplarily, for the faults in a clean workshop, the corresponding countermeasures may include: according to the fault diagnosis results, the cloud server provides emergency treatment suggestions to the management personnel. If it is a problem of excessive microorganisms, it is recommended to immediately increase the intensity or frequency of disinfection treatment, such as increasing the ultraviolet disinfection time or increasing the dosage of disinfectant. For the problem of particle pollution, it is recommended to check whether the filters in the corresponding area are blocked, and if necessary, perform temporary bypass or replace the filter element operation. At the same time, for serious water quality problems that may affect production, it is recommended to suspend the operation of relevant production equipment to avoid producing unqualified products.

[0116] At the same time, the cloud server can assist in formulating a maintenance plan according to the fault type and severity. For some simple faults, such as sensor faults, it can provide operation guides for replacing sensors and lists of required tools and accessories. For complex faults, such as pipeline leaks or water treatment equipment failures, it can arrange maintenance personnel and maintenance time according to the fault location and possible maintenance steps, and estimate the resources and costs required for maintenance. During the maintenance process, the water quality data can be continuously monitored to evaluate the maintenance effect and ensure that the water quality returns to normal.

[0117] After the fault repair is completed, the cloud server can further provide adjustments to the water use strategy. It can dynamically adjust the water use strategy of the clean workshop according to the real-time water quality monitoring data and analysis results. For example, if the water quality in a certain area has been maintained at a high standard and the water consumption is stable, it is recommended to appropriately reduce the water quality monitoring frequency in this area, reduce the working load of the sensors, and at the same time adjust the intensity of water treatment. Reduce the dosage of disinfectant or extend the filtration cycle to reduce the operating cost. On the contrary, if the water quality in a certain water use link fluctuates, it is recommended to add additional treatment steps, such as adding an activated carbon filtration device in the water supply pipeline or adjusting the working pressure and flow rate of the reverse osmosis membrane to ensure the water use safety of key production equipment.

[0118] At the same time, the water supply distribution can be optimized. If the production task in a certain production area decreases and the water consumption decreases, the opening of the water supply valve can be adjusted to reduce the water supply in this area, and the excess water resources can be reasonably allocated to other areas with greater water use requirements. During the process of adjusting the water supply distribution, it is necessary to fully consider the different water quality requirements of different areas to avoid water quality problems caused by changes in water supply.

[0119] In specific implementation, when the early warning mechanism is triggered, it indicates that the water quality in the clean workshop is abnormal. At this time, the cloud server can input the water quality data into a preset Bayesian fault model, output the fault diagnosis results and arrange for maintenance personnel to repair, and at the same time generate corresponding water quality optimization strategies to adjust the water supply distribution.

[0120] In the technical solution provided in this embodiment, when the clean workshop is operating normally, various sensors can continuously collect the above water quality data at a preset sampling frequency. Then, after the cloud server standardizes or normalizes these water quality data, standardized water quality data is obtained. Since the cloud server has previously performed deep learning based on the water quality data through a convolutional neural network model to obtain a water quality prediction model that can predict the change direction of the water body quality over a certain period of time, the standardized water quality data can be input into the water quality prediction model to obtain data on the water quality change trend of the water body in the clean workshop in the next period of time. If the water quality change trend data reaches the warning threshold, the warning mechanism is triggered. At this time, it indicates that the water quality in the clean workshop is abnormal. The cloud server can input the water quality data into a preset Bayesian fault model, output the fault diagnosis result, arrange for maintenance personnel to repair, and at the same time generate a corresponding water quality optimization strategy to adjust the water supply distribution. Since in this embodiment, after collecting water quality data through multiple sensors, the water quality prediction model can accurately predict the water quality change trend, give an early warning when the water quality may be abnormal, and generate a water quality optimization strategy for adjustment. It avoids the lag of the existing water quality monitoring methods, so as to be able to respond to water quality problems in advance and ensure the quality and efficiency of the production water in the clean workshop.

[0121] Based on the above-mentioned first embodiment of the present application, the second embodiment of the present application is proposed. In the second embodiment of the present application, the same or similar content as the above-mentioned first embodiment can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 2 , Figure 2 which is the schematic flow chart provided for the second embodiment of the water quality monitoring method of the present application.

[0122] Step S40 of this example includes steps S41 to S44:

[0123] Step S41: Based on the warning mechanism, start a preset Bayesian fault model, and the Bayesian fault model analyzes through the causal relationship and probability between water quality data.

[0124] It should be noted that the causal relationship between water quality data can be the relationship corresponding to that a change in one water quality index will directly cause a change in another or multiple water quality indexes. And the probability is a measure of the likelihood of these causal relationships occurring.

[0125] Exemplarily, an increase in the organic matter content in water may cause a decrease in dissolved oxygen because the decomposition of organic matter consumes oxygen. By learning the causal relationship and probability between water quality data, the Bayesian fault model can better understand the water quality change mechanism, predict the water quality change trend, and thus more effectively formulate water quality protection and management strategies.

[0126] Step S42: Diagnose the water quality data through the Bayesian fault model and a preset expert system to determine the water treatment fault links and fault causes in the clean workshop for the water quality data.

[0127] It should be noted that the expert system can be a rule base storing a large amount of knowledge and experience of domain experts regarding water quality problems. The rules of the expert system can be used to assist the Bayesian fault model in fault diagnosis, further helping to confirm the fault causes and providing more accurate guidance for maintenance personnel.

[0128] Exemplarily, if the concentration of a certain specific ion in the water is detected to increase abnormally, and at the same time, the abnormal operation of water-using equipment in a specific area is accompanied, it can be judged whether the water supply pipeline in this area has corroded and leaked, resulting in the mixing of external ions into the water supply system according to the expert experience rules of the expert system.

[0129] It can be understood that the water treatment fault link can be the resolution of problems occurring in the entire water treatment system in the clean workshop. The fault cause can be the factor leading to the fault in this link, such as equipment aging, microbial growth blocking pipelines or filter materials, etc.

[0130] Exemplarily, if the microbial content suddenly increases and at the same time the pH value decreases, the Bayesian fault model determines the possibility that a certain water treatment link fails, resulting in microbial growth and changing the acidity and alkalinity of the water. Combining with the expert system, various possible fault causes are considered, such as filter blockage, disinfection equipment failure, pipeline leakage, etc. At the same time, combining the changes in different sensor data, the probability of each fault cause is calculated. Through the above method, the possible fault root cause can be quickly located.

[0131] In a feasible implementation manner, step S42 of this example includes the steps of: taking the water quality index as an observation node through the Bayesian fault model, and taking the sensor fault as a hidden node; determining the causal relationship between water quality data according to the hidden node; defining a conditional probability table between the causal relationship and the observation node based on the historical water quality data; diagnosing the water quality data based on the conditional probability table through a preset expert system to obtain the water treatment fault links and fault causes in the clean workshop for the water quality data.

[0132] It should be noted that the observation node can be that the Bayesian fault model takes the water quality index that can be directly measured as a visible and obtainable reference; while the sensor fault is taken as a hidden node because the sensor fault is difficult to be directly observed, but it will affect the observation of the water quality index.

[0133] By establishing this model relationship with water quality indicators as observed nodes and sensor failures as hidden nodes, the known observed data of water quality indicators can be used to infer whether the sensor has failed and information such as the probability of failure, so as to improve the accuracy and reliability of water quality monitoring.

[0134] It can be understood that the conditional probability table can be the probability result of the causal relationship of water quality, the relationship between the above water quality indicators and whether the sensor fails.

[0135] Specifically, for the sake of understanding, the analysis steps of the Bayesian fault model are taken as an example for illustration:

[0136] 1. Determine the nodes. Define the observed nodes as pH value, dissolved oxygen, and water quality turbidity; define the hidden nodes as sensor failure and data transmission anomaly.

[0137] 2. Determine the edges. Confirm the causal relationship of water quality data. For example, the abnormal water quality parameters caused by sensor failure; the relationship such as the missing water quality data caused by data transmission anomaly.

[0138] 3. Define the conditional probability table. Based on historical water quality data, define the probability that the sensor (S1) is normal for parameters such as pH value to be normal as 95%; the probability that the sensor fails for abnormal pH value is 80%, and the conditional probability table corresponding to the pH value can be established, such as:

[0139] P(pH = normal|S1 = normal) = 0.95;

[0140] P(pH = abnormal|S1 = normal) = 0.05;

[0141] P(pH = normal|S1 = failure) = 0.20;

[0142] P(pH = abnormal|S1 = failure) = 0.80.

[0143] 4. Observe the current data performance for fault diagnosis and decision-making. If it is judged that the PH is abnormal, the probability of sensor failure is 80%; at the same time, multi-dimensional judgment by combining other parameters can improve the judgment accuracy.

[0144] In this embodiment, first, the water quality indicators are used as observation nodes through the Bayesian fault model, and the sensor faults are used as hidden nodes; and the causal relationships between water quality data are determined according to the hidden nodes. Then, based on the historical water quality data, the conditional probability table between the causal relationships and the observation nodes is defined. Finally, through a preset expert system, the water quality data is diagnosed based on the conditional probability table, and the water treatment fault links and fault causes of the water quality data in the clean workshop are obtained. By establishing such a model relationship with water quality indicators as observation nodes and sensor faults as hidden nodes, the known water quality indicator observation data can be used to infer whether the sensor fails and information such as the probability of failure, so as to improve the accuracy and reliability of water quality monitoring.

[0145] Step S43: Generate corresponding water quality optimization strategies according to the water treatment fault links and fault causes.

[0146] It should be noted that the water quality optimization strategy can also be further combined with the long-term production plan of the clean workshop to plan water quality guarantee measures in advance. For example, in the off-season of production, when the number of operating production equipment decreases and the water consumption decreases, the entire water service system is maintained and optimized. The water treatment equipment can be comprehensively inspected and maintained, the aging components can be replaced, and at the same time, the operation mode of the water supply system can be adjusted to reduce unnecessary water resource waste and water treatment costs. Before the peak season of production arrives, according to the expected production scale and the increase in water demand, the water quality guarantee measures are adjusted in advance, such as increasing the water quality monitoring frequency, strengthening the water treatment intensity, and stocking key water treatment consumables, etc., to ensure the stability of water quality during high-load production.

[0147] At the same time, the cloud server can also regularly conduct a comprehensive analysis of the water quality monitoring data and system operation data, summarize the water quality change rules and the problems existing in the system operation. According to the analysis results, continuous improvements are made to the sensor layout, data acquisition parameters, data analysis model, warning threshold, water use strategy, etc. For example, if it is found that the water quality change trend in a certain area deviates greatly from the existing water quality prediction model, the water quality data in this area is re-collected and analyzed, and the parameters of the water quality prediction model are optimized to improve the prediction accuracy. Through the above continuous improvement method, the ability of the cloud server to monitor and manage the water quality of the clean workshop can be continuously improved, ensuring the smooth progress of production and the stability of product quality.

[0148] Step S44: Give a warning to the maintenance personnel according to the water quality optimization strategy.

[0149] In this embodiment, when a water quality anomaly warning occurs, the Bayesian fault model and the rule-based expert system can be activated to analyze the causal relationships and probabilities between water quality data, and determine the fault links and causes. This can not only quickly locate faults, but also optimize the water use strategy based on real-time data and long-term planning, realizing an integrated innovation from warning to decision support, effectively ensuring production and optimizing operating costs.

[0150] Exemplarily, to facilitate understanding of the implementation process of the water quality monitoring method obtained by combining the above-mentioned Embodiment 1 and Embodiment 2, please refer to Figure 3 , Figure 3 , which is the overall process architecture diagram of the water quality monitoring process provided by this application. Specifically:

[0151] 1. Data collection. In terms of the system architecture, first, at key nodes of the water service system in the clean workshop (such as positions like the water supply system, the connection of water-using equipment, the return water pipeline, the inlet and outlet of water treatment equipment, etc.), install a variety of high-precision sensors such as fluorescence detection-based microbial sensors, laser scattering particle counting sensors, ion-selective electrode sensors, pH sensors, and dissolved oxygen sensors. These sensors can collect multi-dimensional data of water quality in real time (such as temperature, flow rate, pH value, microbial concentration, dissolved oxygen concentration, dissolved sulfur concentration, particulate matter concentration, metal ion concentration, etc.), and then convert the analog signals obtained by the sensors into digital signals, and add metadata such as time stamps and sensor location information.

[0152] 2. Data storage. Set up a data concentrator to receive data from the sensors and perform preliminary sorting and caching (such as intermittent type caching and continuous type caching). Then, through an optical fiber network encrypted with the SSL / TLS protocol, securely transmit the data to the cloud server to ensure the confidentiality and integrity of the data transmission process.

[0153] 3. Data processing. Then use the cloud server to obtain data and perform data preprocessing (such as cleaning, standardization, etc.) on the obtained water quality data. Next, classify the historical water quality data, and use statistical analysis, machine learning (support vector machine, random forest, etc.) and deep learning (such as the above-mentioned convolutional neural network) algorithms to conduct correlation tests on the data, and establish a big data model (i.e., a water quality prediction model) to predict the water quality change trend data.

[0154] Among them, the cloud server can be built based on platforms with distributed computing architectures such as Hadoop and Spark, combined with a hybrid database storage system composed of MySQL, HBase, MongoDB, etc. It can process a large amount of water quality data, and has good scalability, and can flexibly allocate resources according to the data volume and processing requirements.

[0155] 4. Data warning. When the water quality change trend data exceeds the preset threshold or shows an abnormal trend, threshold alarms are sent to the management personnel through various channels such as text messages, emails, and mobile application push, and the data is displayed, detailing the abnormal situation, problem location, scope of influence, and severity.

[0156] 5. Water quality diagnosis. Finally, the Bayesian fault model and the rule-based expert experience library are used for data diagnosis and model diagnosis to quickly determine the cause of water quality anomalies. At the same time, the water use strategy is optimized based on water quality data and production plans to achieve comprehensive management from daily monitoring, fault response to long-term water use planning, ensuring the quality and efficiency of the production water in the clean workshop.

[0157] Through the above water quality monitoring method, water quality detection can be carried out in fields with extremely high water quality requirements. For example, in the electronic industry, such as in the clean workshops for electronic chip manufacturing in the electronic industry, the water quality requirements are extremely high. This method can accurately monitor parameters such as the concentration of microorganisms, particles, and ions in water, prevent chips from being contaminated due to water quality problems, ensure the production quality and yield of chips, and meet the strict requirements for ultrapure water in the electronic production process. For example, in the pharmaceutical manufacturing field, such as in the pharmaceutical production process, the water quality directly affects the quality and safety of drugs. This method can real-time monitor various indicators of pharmaceutical water, ensure that impurities, microorganisms, etc. in the water are at an extremely low level, ensure that the drug production meets strict quality standards, which is particularly crucial for the production of drugs sensitive to water quality such as injections and eye drops. For example, in the field of precision optical instrument manufacturing, such as the production of precision optical components such as optical lenses requires high-purity water. By continuously monitoring parameters such as particle counting in water, scratches and contamination of the optical surface by tiny particles are avoided, ensuring the accuracy and imaging quality of optical instruments, and providing strong support for the production of high-precision optical products. For example, in the chemical industry, such as in chemical clean workshops, accurate water quality monitoring helps to control chemical reaction conditions. This method can ensure that parameters such as the pH value and ion concentration of water are appropriate, prevent equipment corrosion and adverse chemical reactions, and ensure the safety and stability of the chemical production process.

[0158] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the water quality monitoring method of this application. Based on this technical concept, more forms of simple transformations are within the protection scope of this application.

[0159] This application also provides a water quality monitoring device. Please refer to Figure 4 , Figure 4 which is the module structure schematic diagram of the water quality monitoring device in the embodiment of this application; the water quality monitoring device includes:

[0160] A pretreatment module 401, which is used to preprocess the water quality data of the clean workshop collected by multiple sensors to obtain standardized water quality data, and the multiple sensors are pre-set at key nodes of the clean workshop;

[0161] A water quality prediction module 402, configured to input the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is obtained by training based on a convolutional neural network model;

[0162] An early warning mechanism module 403, configured to trigger an early warning mechanism when the water quality change trend data reaches an early warning threshold;

[0163] An optimization strategy module 404, configured to input the water quality data into a preset Bayesian fault model based on the early warning mechanism to obtain a corresponding water quality optimization strategy.

[0164] As an implementation manner, the preprocessing module 401 is further configured to collect water quality data of a clean workshop by multiple sensors at a preset sampling frequency, where the multiple sensors include at least one of a microbial sensor, a particle counting sensor, an ion concentration sensor, a pH value sensor, and a dissolved oxygen sensor; perform data preprocessing on the water quality data to obtain standardized water quality data, where the data preprocessing includes outlier processing, missing value filling, and standardization processing.

[0165] As an implementation manner, the preprocessing module 401 is further configured to determine that a data point is an outlier when the data point of the water quality data exceeds a preset fluctuation range; remove the outlier to obtain cleaned water quality data; perform missing value filling on the cleaned water quality data by a preset numerical filling method to obtain filled water quality data; map the filled water quality data to a standard numerical interval to obtain standardized water quality data.

[0166] As an implementation manner, the water quality prediction module 402 is further configured to extract water quality indicators of the standardized water quality data, where the water quality indicators include at least one of pH value, conductivity, dissolved oxygen, turbidity, temperature, and ion concentration; calculate statistical indicators of the water quality indicators by a statistical analysis method, where the statistical indicators include at least one of mean, standard deviation, and coefficient of variation; input the water quality indicators and the statistical indicators into a preset water quality prediction model to obtain water quality change trend data, where the water quality prediction model is obtained by training based on a convolutional neural network model with historical water quality data as training samples.

[0167] As an implementation manner, the water quality prediction module 402 is further configured to obtain historical water quality indicators and historical statistical indicators of historical water quality data; classify the historical statistical indicators according to the time dimension to obtain linear data and non-linear data; perform linear fitting on the linear data and perform regression on the non-linear data by the random forest regression method to obtain standardized data; input the standardized data and the historical water quality indicators into a convolutional neural network model for correlation training to obtain a water quality prediction model.

[0168] As an implementation manner, the early warning mechanism module 403 is further configured to start a preset Bayesian fault model based on the early warning mechanism, and the Bayesian fault model analyzes through the causal relationship and probability between water quality data; diagnose the water quality data through the Bayesian fault model and a preset expert system to determine the water treatment fault link and fault cause of the water quality data in the clean workshop; generate a corresponding water quality optimization strategy according to the water treatment fault link and fault cause; issue an early warning to maintenance personnel according to the water quality optimization strategy.

[0169] As an implementation manner, the optimization strategy module 404 is further configured to use the water quality indicator as an observation node and the sensor fault as a hidden node through the Bayesian fault model; determine the causal relationship between water quality data according to the hidden node; define a conditional probability table between the causal relationship and the observation node based on the historical water quality data; diagnose the water quality data based on the conditional probability table through a preset expert system to obtain the water treatment fault link and fault cause of the water quality data in the clean workshop.

[0170] Other embodiments or specific implementation manners of the water quality monitoring device of the present application can refer to the above method embodiments, and will not be elaborated here.

[0171] The water quality monitoring device provided by the present application adopts the water quality monitoring method in the above embodiment, and can solve the technical problem that the existing water quality monitoring method is generally in a passive monitoring state when the water quality changes, has hysteresis, is difficult to actively prevent, and is likely to cause production losses. Compared with the prior art, the beneficial effects of the water quality monitoring device provided by the present application are the same as those of the water quality monitoring method provided by the above embodiment, and other technical features in the water quality monitoring device are the same as the features disclosed in the above embodiment method, and will not be elaborated here.

[0172] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the water quality monitoring method in the above embodiment.

[0173] The computer-readable storage medium provided by the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system or device. The program code contained on the computer-readable storage medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0174] The above computer-readable storage medium may be included in the water quality monitoring device; it may also exist alone and not be assembled into the water quality monitoring device.

[0175] The above computer-readable storage medium carries one or more programs. When the above one or more programs are executed by the water quality monitoring device, the water quality monitoring device is caused to: preprocess the water quality data of the clean workshop collected by multiple sensors to obtain standardized water quality data, where the multiple sensors are pre-set at key nodes of the clean workshop; input the standardized water quality data into a pre-set water quality prediction model to obtain water quality change trend data, and the water quality prediction model is obtained by training based on a convolutional neural network model; trigger an early warning mechanism when the water quality change trend data reaches the early warning threshold; based on the early warning mechanism, input the water quality data into a pre-set Bayesian fault model to obtain a corresponding water quality optimization strategy.

[0176] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The above-mentioned programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, by connecting through the Internet using an Internet service provider).

[0177] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and this module, program segment, or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the boxes may occur in a different order than that marked in the accompanying drawings. For example, two consecutively represented boxes can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0178] The modules described in the embodiments of this application can be implemented in software or in hardware. Among them, the name of the module does not constitute a limitation to the unit itself in some cases.

[0179] The readable storage medium provided by this application is a computer-readable storage medium. The computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for performing the above-mentioned water quality monitoring method, and can solve the technical problems that the existing water quality monitoring methods are generally in a passive monitoring state when the water quality changes, there is a lag, it is difficult to actively prevent, and it is easy to cause production losses. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by this application are the same as those of the water quality monitoring method provided by the above-mentioned embodiments, and will not be elaborated here.

[0180] The present application also provides a computer program product, including a computer program which, when executed by a processor, implements the steps of the water quality monitoring method as described above.

[0181] The computer program product provided by the present application can solve the technical problems that the existing water quality monitoring methods are generally in a passive monitoring state when the water quality changes, there is a lag, it is difficult to actively prevent, and it is easy to cause production losses. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the water quality monitoring method provided by the above embodiments, and will not be elaborated here.

[0182] The above are only partial embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural transformation made by using the content of the specification and drawings of the present application under the technical concept of the present application, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present application.

Claims

1. A water quality monitoring method, characterized in that: The method comprises: Preprocessing the water quality data of the clean room collected by multiple sensors to obtain standardized water quality data, wherein the multiple sensors are pre-installed before and after the reverse osmosis device of the ultrapure water preparation system in the clean room, before and after the ion exchange column, the main water supply pipeline leading to the production workshop, and the water inlet of each chip manufacturing equipment; Inputting the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, wherein the water quality prediction model is obtained by training based on a convolutional neural network model; When the water quality change trend data reaches the warning threshold, triggering the warning mechanism; Based on the early warning mechanism, the water quality data is input into a preset Bayesian fault model to obtain a corresponding water quality optimization strategy; Among them, the step of inputting the water quality data into a preset Bayesian fault model based on the early warning mechanism to obtain a corresponding water quality optimization strategy includes: based on the early warning mechanism, starting a preset Bayesian fault model, and the Bayesian fault model analyzes the cause-effect relationship and probability between water quality data; defining the observation nodes as pH value, dissolved oxygen and water turbidity through the Bayesian fault model, defining the hidden nodes as sensor failure and data transmission anomaly, defining the cause-effect relationship of water quality data by the water quality parameter abnormality caused by sensor failure or the water quality data missing caused by data transmission anomaly, and defining a conditional probability table between normal sensor for normal observation node and sensor failure for abnormal observation node; diagnosing the water quality data through a preset expert system based on the conditional probability table to obtain the water treatment failure link and failure cause of the water quality data in the clean workshop; generating a corresponding water quality optimization strategy based on the water treatment failure link and failure cause; and issuing an early warning to maintenance personnel based on the water quality optimization strategy.

2. The method according to claim 1, characterized in that The step of preprocessing the water quality data of the clean room collected by multiple sensors to obtain standardized water quality data includes: Collecting water quality data of clean workshops by using multiple sensors at a preset sampling frequency, wherein the multiple sensors include at least one of a microbial sensor, a particle counting sensor, an ion concentration sensor, a pH sensor, and a dissolved oxygen sensor; The water quality data is preprocessed to obtain standardized water quality data, wherein the data preprocessing includes outlier processing, missing value filling and standardization processing.

3. The method according to claim 2, characterized in that The step of preprocessing the water quality data to obtain standardized water quality data includes: When a data point of the water quality data exceeds a preset fluctuation range, determining the data point as an abnormal value; Removing the abnormal values ​​to obtain cleaned water quality data; Filling missing values ​​in the cleaned water quality data using a preset value filling method to obtain filled water quality data; The filled water quality data is mapped to a standard numerical range to obtain standardized water quality data.

4. The method according to any one of claims 1 to 3, characterized in that The step of inputting the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data includes: Extracting water quality indicators of the standardized water quality data, wherein the water quality indicators include at least one of pH value, conductivity, dissolved oxygen, turbidity, temperature and ion concentration; Calculating the statistical index of the water quality index by a statistical analysis method, wherein the statistical index includes at least one of a mean, a standard deviation and a coefficient of variation; The water quality index and the statistical index are input into a preset water quality prediction model to obtain water quality change trend data. The water quality prediction model is obtained by training a convolutional neural network model using historical water quality data as training samples.

5. The method according to claim 4, characterized in that The training process of the water quality prediction model includes: Obtain historical water quality indicators and historical statistical indicators of historical water quality data; Classify the historical statistical indicators according to the time dimension to obtain linear data and nonlinear data; Performing linear fitting on the linear data and regressing the nonlinear data by random forest regression method to obtain standardized data; The standardized data and the historical water quality indicators are input into a convolutional neural network model for correlation training to obtain a water quality prediction model.

6. A water quality monitoring device, characterized in that: The water quality monitoring device implements the water quality monitoring method according to claim 1, and the device comprises: A preprocessing module, used to preprocess the water quality data of the clean room collected by multiple sensors to obtain standardized water quality data, wherein the multiple sensors are pre-set at key nodes of the clean room; A water quality prediction module, used to input the standardized water quality data into a preset water quality prediction model to obtain water quality change trend data, wherein the water quality prediction model is obtained by training based on a convolutional neural network model; An early warning mechanism module, used to trigger an early warning mechanism when the water quality change trend data reaches an early warning threshold; The optimization strategy module is used to input the water quality data into a preset Bayesian fault model based on the early warning mechanism to obtain a corresponding water quality optimization strategy.

7. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the water quality monitoring method according to any one of claims 1 to 5 are implemented.

8. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the water quality monitoring method according to any one of claims 1 to 5 are implemented.

Images