Rapid analysis method and system for heavy metals in water

Abstract

The invention relates to a water quality heavy metal rapid analysis method and system.The analysis method comprises the steps that an actual water quality sample is tested through anodic stripping voltammetry to obtain an electric sensing signal corresponding to the content of heavy metal in the sample; and processing and analyzing the obtained electric sensing signal to obtain a corresponding result of the concentration of the heavy metal in the water quality sample and the electric sensing signal, namely performing qualitative analysis, analyzing the signal, determining the type of the heavy metal according to a dissolution peak position, and determining the concentration of the heavy metal according to the dissolution peak position. According to a pre-trained qualitative recognition model, the type of a working electrode of an electric sensing signal source is determined, then a corresponding pre-trained quantitative regression model is selected based on the type of the working electrode and the type of the heavy metal, and the concentration of the heavy metal in a sample is accurately and quantitatively analyzed. Compared with the prior art, the method has the advantages that chemical-free quantitative analysis is realized through a rapid quantitative analysis technology, the detection process is simplified, the generation of waste liquid is reduced, and the problem of secondary environmental pollution is effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and particularly to a rapid analysis method and system for heavy metals in water quality. Background Art

[0002] Water is the source of life. The heavy metal content in water quality is one of the important indicators for evaluating the water environment quality and an important parameter for determining the degree of water environmental pollution. Portable detection equipment based on anodic stripping voltammetry (ASV) has a low manufacturing cost and can achieve highly sensitive on-site detection of heavy metals in water quality, becoming one of the mainstream equipment for on-site water quality detection. However, this equipment needs to rely on the standard curve method or the standard addition method to achieve quantitative analysis, and it needs to be recalibrated after several analyses. At the same time, due to the difficult-to-control processes such as working electrode polishing and material modification, the detection signals between different electrodes cannot be reproduced, resulting in the loss of generality of the established standard curve or model. Conducting multiple standard sample detections or matrix spike detections on-site requires users to carry a large amount of relevant reagents, which is time-consuming, laborious, and cumbersome to operate. Moreover, the disposal of the used standard solution also poses a huge secondary environmental problem, and it still cannot well meet the requirements of on-site rapid detection and emergency detection.

[0003] Therefore, it is necessary to establish an intelligent algorithm that can perform quantitative calculations without relying on on-site standard sample analysis, so as to achieve rapid and low-reagent detection of heavy metals in on-site water quality. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned defects existing in the prior art and provide a rapid analysis method and system for heavy metals in water quality. Through a rapid quantitative analysis technology that does not require on-site standard sample analysis and establishment of a standard curve, the present invention realizes reagent-free quantitative analysis, which not only simplifies the detection process but also reduces the generation of waste liquid, effectively avoiding the problem of secondary environmental pollution.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] The first aspect of the present invention provides a rapid analysis method for heavy metals in water quality, including the following steps:

[0007] Using anodic stripping voltammetry to test an actual water quality sample to obtain an electrical sensing signal corresponding to the types and contents of heavy metals in the sample;

[0008] Process and analyze the obtained electrical sensing signals to obtain the corresponding results of the heavy metal concentration and the electrical sensing signals in the water quality sample, where: First, perform qualitative analysis, that is, use a pre-trained support vector machine or artificial neural network model to analyze the signals to determine the working electrode category for heavy metal sensing signal acquisition. Then, select the corresponding pre-trained quantitative regression model based on the electrode category, and then perform accurate quantitative analysis of the heavy metal concentration in the sample through the pre-trained quantitative regression model.

[0009] Furthermore, when analyzing the signals, select the optimal hyperparameters of the machine learning model through an optimization algorithm.

[0010] Furthermore, the optimization algorithm includes particle swarm optimization or genetic algorithm.

[0011] Furthermore, before performing qualitative analysis, perform preprocessing steps on the obtained electrical sensing signals. The preprocessing steps include at least one of the following processes:

[0012] Filtering processing to eliminate noise and interference;

[0013] Baseline correction to calibrate the baseline drift of the signal;

[0014] Signal enhancement to improve the signal-to-noise ratio of the signal;

[0015] Perform dimensionality reduction processing on the electrical sensing signals using principal component analysis to extract key features.

[0016] Furthermore, the acquisition method of the pre-trained electrode category qualitative recognition support vector machine or artificial neural network model includes:

[0017] Collect a series of water quality samples with known heavy metal types and concentrations;

[0018] Based on anodic stripping voltammetry, use different working electrodes to test all known water quality samples and record the electrical sensing signals of each sample;

[0019] Assign corresponding electrode category labels to each electrical sensing signal to form a heavy metal electrode category label training data set;

[0020] Train a support vector machine or artificial neural network model through the electrode category label training data set.

[0021] Furthermore, the acquisition method of the pre-trained quantitative regression model includes:

[0022] Collect a series of water quality samples with known heavy metal concentrations of different types;

[0023] Use anodic stripping voltammetry to test these samples and record the electrical sensing signals of each sample;

[0024] Corresponding to each electrical sensing signal is a corresponding heavy metal concentration value. Based on each electrode category, an electrical sensing signal-concentration dataset for different heavy metal types is constructed, and the heavy metal type-electrical sensing signal-concentration datasets corresponding to different types of electrode types form an electrical sensing signal-concentration database;

[0025] By dividing the electrical sensing signal-concentration database into a training set, a validation set, and a test set, select a quantitative regression model and initialize the parameters according to the data characteristics and requirements, train the model with the training set, evaluate and optimize the hyperparameters with the help of the validation set during this period, and finally evaluate with the test set. If it meets the standard, it is determined as a pre-trained quantitative regression model.

[0026] Furthermore, the determination of the heavy metal type is determined by the stripping peak position. For example, the stripping peak position of lead is -0.5V. For lead, the signal near -0.5V is selected as the input variable.

[0027] Furthermore, it also includes an adaptive learning process, and the adaptive learning process includes:

[0028] Integrate the newly measured electrical sensing signal and its corresponding electrode category and / or heavy metal concentration information into the current electrode type label training dataset and the electrical sensing signal-concentration database;

[0029] Select data from the updated database for retraining the model to update the model parameters;

[0030] Test the performance of the model through cross-validation or on an independent validation set to ensure that the retrained model maintains or improves the prediction accuracy;

[0031] Replace the original model with the model that has passed the verification for actual rapid analysis of heavy metals in water quality;

[0032] Periodically repeat the above steps to form a continuous adaptive learning loop.

[0033] The second aspect of the present invention provides a rapid water quality heavy metal analysis system, including:

[0034] An anodic stripping voltammetry rapid water quality heavy metal detection module for testing actual water quality samples to obtain electrical sensing signals corresponding to the heavy metal content in the samples;

[0035] A signal processing and analysis module is used to process and analyze the electrical sensing signals to obtain the corresponding results of the heavy metal concentration in the water quality sample and the electrical sensing signals, where: first, qualitative analysis is performed, that is, a pre-trained support vector machine or artificial neural network model is used to analyze the signals to determine the working electrode category for obtaining the sensing signals. Then, a corresponding pre-trained quantitative regression model is selected based on the electrode category. Then, the pre-trained quantitative regression model is used to perform precise quantitative analysis on the heavy metal concentration in the sample.

[0036] Furthermore, the signal processing and analysis module further includes:

[0037] An optimization algorithm component, which is used to select the best hyperparameters of the machine learning model through an optimization algorithm when selecting the best hyperparameters of the machine learning model;

[0038] A model training component is used to train a support vector machine or an artificial neural network model. The model training process includes: collecting a series of water quality samples with known heavy metal types, testing the water quality samples using anodic stripping voltammetry with different working electrodes, recording the electrical sensing signals of each sample, assigning corresponding electrode category labels to each electrical sensing signal to form an electrode category label training data set, and training a support vector machine or an artificial neural network model through the electrode category label training data set.

[0039] An adaptive learning component is used to implement an adaptive learning process. The adaptive learning process includes: integrating the newly measured electrical sensing signals and their corresponding electrode categories and / or concentration information into the current electrode category label training data set and the electrical sensing signal-concentration database; selecting data from the updated database for retraining the model to update the model parameters; testing the performance of the model through cross-validation or on an independent validation set to ensure that the retrained model maintains or improves the prediction accuracy; replacing the original model with the verified model for actual rapid analysis of heavy metals in water quality; periodically repeating the above steps to form a continuous adaptive learning loop.

[0040] Furthermore, the rapid heavy metal analysis system for water quality further includes:

[0041] An electromagnetic heating and water cooling temperature control pretreatment module, which is used to control the temperature of the water quality sample and can perform digestion treatment on the water quality sample as needed to obtain a water quality sample at the target temperature;

[0042] An electrolyte addition module, which is used to quantitatively add electrolytes to the water quality sample to be measured. Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1) The beneficial effects of the present invention are first reflected in its revolutionary improvement in on-site water quality heavy metal detection. Through a rapid quantitative analysis technique that eliminates the need for on-site standard sample analysis and the establishment of standard curves, the present invention realizes reagent-free quantitative analysis. This not only simplifies the detection process but also reduces the generation of waste liquid, effectively avoiding the problem of secondary environmental pollution. It makes water quality detection more environmentally friendly and efficient, and is of great significance for on-site emergency detection and routine monitoring.

[0044] 2) Secondly, by introducing a machine learning-based on-site rapid detection method, the present invention greatly improves the intelligence level of heavy metal detection equipment. It reduces the operation steps of quantitative analysis and significantly shortens the detection time, enabling the detection results to be immediately displayed after the measurement of the electrical sensing signal of unknown samples. This rapid response ability has great practical value for occasions that require rapid decision-making, such as the rapid response to water source pollution accidents, and can provide a scientific basis for water quality safety in a timely manner.

[0045] 3) The electrode qualitative intelligent decision-making - quantitative precise analysis mode constructed by the present invention effectively solves the problem of electrode instability by innovatively integrating qualitative and quantitative algorithms. Compared with the traditional method of constructing quantitative models, the present invention improves the generalization ability of the model at the algorithm level, reduces the dependence on a large amount of data, and improves the stability and accuracy of the model. It not only improves the reliability of the detection results but also provides a new direction for the development of water quality detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of the rapid analysis method for heavy metals in water quality in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The present invention will be described in detail below with reference to the drawings and specific embodiments. Features such as component models, material names, connection structures, control methods, algorithms, etc. that are not clearly stated in the present technical solution are regarded as common technical features disclosed in the prior art.

[0048] Embodiment 1

[0049] By introducing a machine learning-based on-site rapid detection method, this embodiment greatly improves the intelligence level of heavy metal detection equipment. It reduces the operation steps of quantitative analysis and significantly shortens the detection time, enabling the detection results to be immediately displayed after the measurement of the electrical sensing signal of unknown samples. This rapid response ability has great practical value for occasions that require rapid decision-making, such as the rapid response to water source pollution accidents, and can provide a scientific basis for water quality safety in a timely manner.

[0050] The rapid analysis method for heavy metals in water quality in this embodiment is shown in Figure 1 , and includes the following steps:

[0051] The anodic stripping voltammetry is used to test actual water quality samples to obtain an electrical sensing signal corresponding to the types and contents of heavy metals in the samples;

[0052] The obtained electrical sensing signal is processed and analyzed to obtain the corresponding results of the heavy metal concentration and the electrical sensing signal in the water quality sample, where: first, qualitative analysis is performed, that is, a pre-trained support vector machine or artificial neural network model is used to analyze the signal to determine the working electrode category for obtaining the heavy metal sensing signal, and then a corresponding pre-trained quantitative regression model is selected based on the electrode category, and then the heavy metal concentration in the sample is accurately quantitatively analyzed through the pre-trained quantitative regression model.

[0053] When analyzing the signal, the optimal hyperparameters of the machine learning model are selected through an optimization algorithm. Specifically, the optimization algorithm includes particle swarm optimization or genetic algorithm.

[0054] Before performing qualitative analysis, a preprocessing step is performed on the obtained electrical sensing signal. The preprocessing step includes at least one of the following processes: filtering to eliminate noise and interference; baseline correction to calibrate the baseline drift of the signal; signal enhancement to improve the signal-to-noise ratio; and dimensionality reduction processing of the electrical sensing signal using principal component analysis to extract key features.

[0055] Specifically, the acquisition method of the pre-trained support vector machine or artificial neural network model includes: collecting a series of water quality samples with known heavy metal types and concentrations; based on the anodic stripping voltammetry, using different working electrodes to test all known water quality samples and recording the electrical sensing signal of each sample; assigning corresponding electrode category labels to each electrical sensing signal to form a heavy metal electrode category label training data set; and training a support vector machine or artificial neural network model through the heavy metal electrode category label training data set.

[0056] In terms of the specific principle, the principle of obtaining a pre-trained support vector machine or artificial neural network model is based on the data-driven idea of machine learning and the electrochemical detection characteristics of anodic stripping voltammetry. Anodic stripping voltammetry can cause different heavy metals to produce stripping peaks at specific potentials, thereby outputting electrical sensing signals related to the types and contents of heavy metals. By collecting water quality samples with known heavy metal types and concentrations and testing them using this method, a series of electrical sensing signals reflecting heavy metal information can be obtained. Assigning corresponding electrode category labels to these electrical sensing signals is because different working electrodes have differences in the sensitivity and selectivity of heavy metal detection, and the electrode category is closely related to the heavy metal detection characteristics. In this way, a training dataset for heavy metal electrode category labels is constructed. Support vector machine and artificial neural network models have powerful pattern recognition and data fitting capabilities. Using this training dataset to train them enables the models to learn the internal association patterns between electrical sensing signals, electrode categories, heavy metal types, and concentrations, and finally obtain a pre-trained model that can be used for subsequent qualitative analysis of heavy metals in water quality, that is, to determine the working electrode category and heavy metal type.

[0057] During specific training, first, water quality samples with known heavy metal types and concentrations are collected, and electrical sensing signals are obtained through testing with different working electrodes using anodic stripping voltammetry, and electrode category labels are assigned to them to form a training dataset for heavy metal electrode category labels. Then, the dataset is divided, with a part used as the training set and a part used as the validation set. For the support vector machine model, initialize model parameters such as the kernel function type (such as linear kernel, radial basis kernel, etc.), penalty factor, etc. Use the electrical sensing signals in the training set as input features and the corresponding electrode category labels as output targets. Continuously adjust the model parameters through optimization algorithms (such as genetic algorithm, sequential minimal optimization algorithm) so that the model can classify and predict the electrode category as accurately as possible on the training set; during the training process, use the validation set to evaluate the model performance and adjust the model hyperparameters according to the evaluation results to avoid overfitting.

[0058] During specific training, for the artificial neural network model, first design the network structure, determine the number of hidden layers, the number of neurons in each layer, etc., select a suitable activation function (such as ReLU function), and initialize the network weights and biases. Input the training set data into the network, calculate the output result through forward propagation, calculate the loss function (such as cross-entropy loss function) according to the difference between the output result and the true label, then calculate the gradient using the backpropagation algorithm, and update the network parameters through an optimizer (such as stochastic gradient descent, Adam optimizer, etc.); similarly, use the validation set for evaluation and tuning during the training process. Finally, when the model achieves good performance on the validation set, the training is completed.

[0059] In specific implementation, the method for obtaining the pre-trained quantitative regression model includes: collecting a series of water quality samples of different types with known heavy metal concentrations; testing these samples using anodic stripping voltammetry and recording the electrical sensing signals of each sample; corresponding each electrical sensing signal to a corresponding heavy metal concentration value, constructing different heavy metal type electrical sensing signal-concentration datasets based on each electrode category, and forming an electrical sensing signal-concentration database with the heavy metal type-electrical sensing signal-concentration datasets corresponding to different types of electrode types; by dividing the electrical sensing signal-concentration database into a training set, a validation set, and a test set, selecting a quantitative regression model and initializing parameters according to the data characteristics and requirements, training the model with the training set, evaluating and tuning the hyperparameters with the help of the validation set during this period, and finally evaluating with the test set. If the standard is met, it is determined as the pre-trained quantitative regression model.

[0060] In specific implementation, the determination of the heavy metal type is determined by the stripping peak position. For example, the stripping peak position of lead is -0.5V, and for lead, the signal near -0.5V is selected as the input variable.

[0061] In specific implementation, it also includes an adaptive learning process, and the adaptive learning process includes:

[0062] Integrating the newly measured electrical sensing signals and their corresponding electrode category and / or heavy metal concentration information into the current electrode type label training dataset and the electrical sensing signal-concentration database;

[0063] Selecting data from the updated database for retraining the model to update the model parameters;

[0064] Testing the performance of the model through cross-validation or on an independent validation set to ensure that the retrained model maintains or improves the prediction accuracy;

[0065] Replacing the original model with the verified model for actual rapid analysis of heavy metals in water quality;

[0066] Periodically repeating the above steps to form a continuous adaptive learning cycle.

[0067] Specifically, the acquisition of the pre-trained quantitative regression model is based on the combination of anodic stripping voltammetry and machine learning. Anodic stripping voltammetry enables different heavy metals to undergo redox reactions at specific potentials, generating characteristic electrical sensing signals that are closely related to the types and concentrations of heavy metals. By collecting water quality samples with known heavy metal concentrations, testing them, and recording the electrical sensing signals, an electrical sensing signal-concentration dataset and database are constructed, providing a rich and accurate data basis for model training. At the machine learning level, the database is divided into a training set, a validation set, and a test set. According to the characteristics of the data (such as data distribution, number of features, etc.) and analysis requirements, a suitable quantitative regression model (such as linear regression, neural network regression, etc.) is selected, and the model parameters are initialized. The training set is used to train the model so that the model learns the mapping relationship between the electrical sensing signal and the heavy metal concentration; the validation set is used to evaluate the performance of the model on unseen data and adjust the hyperparameters to prevent overfitting; finally, the test set is used to comprehensively evaluate the generalization ability and accuracy of the model. If the expected standard is met, the model becomes a pre-trained quantitative regression model and can be used for subsequent prediction of heavy metal concentrations in water quality.

[0068] The determination of heavy metal types utilizes the characteristic that different heavy metals have specific stripping peak positions in anodic stripping voltammetry. Each heavy metal will exhibit an obvious stripping peak at a specific potential. For example, lead has a stripping peak near -0.5V. By selecting the signals near this peak position as input variables, it is possible to identify whether specific heavy metals such as lead are present in the sample, realizing the qualitative analysis of heavy metals.

[0069] The adaptive learning process is to enable the model to continuously adapt to new data and environmental changes. As newly measured electrical sensing signals and their related information are continuously integrated into the original dataset and database, the training data of the model is updated and expanded. Data is selected from the updated database to retrain the model, updating the model parameters so that it can learn new patterns and rules. The performance of the model is tested through cross-validation or an independent validation set to ensure that the retrained model does not degrade or even improves in terms of accuracy and generalization ability. The model that passes the validation replaces the original model, and this process is repeated periodically to form a continuous adaptive learning loop, ensuring that the model always maintains good performance during long-term use and achieving rapid and accurate analysis of heavy metals in water quality.

[0070] The electrode qualitative intelligent decision-making - quantitative precise analysis mode constructed in the present invention effectively solves the problem of electrode instability by innovatively integrating qualitative and quantitative algorithms. Compared with traditional methods for constructing quantitative models, the present invention improves the generalization ability of the model at the algorithm level, reduces the dependence on a large amount of data, and improves the stability and accuracy of the model. It not only improves the reliability of the detection results but also provides a new direction for the development of water quality detection technology.

[0071] The adaptive learning process adopted by the present invention realizes the rapid, accurate and continuously optimized analysis ability of heavy metals in water quality by continuously integrating new data and updating the analysis model, enhances the adaptability, robustness, resource allocation efficiency and decision-making support ability of the system, while reducing the operation cost and complexity, and provides an efficient and intelligent technical solution for water quality monitoring and management.

[0072] Example 2

[0073] A rapid water quality heavy metal analysis system in this embodiment includes:

[0074] A rapid anodic stripping voltammetry water quality heavy metal detection module for testing actual water quality samples to obtain an electrical sensing signal corresponding to the heavy metal content in the samples;

[0075] A signal processing and analysis module for processing and analyzing the electrical sensing signal to obtain the corresponding results of the heavy metal concentration and the electrical sensing signal in the water quality sample. Among them: first, qualitative analysis is carried out, that is, a pre-trained support vector machine or artificial neural network model is used to analyze the signal to determine the working electrode category for obtaining the heavy metal sensing signal. The heavy metal type is determined by the stripping peak position (for example, the stripping peak position of lead is -0.5V, and for lead, the signal near -0.5V is selected as the input variable). Then, a corresponding pre-trained quantitative regression model is selected based on the electrode category, and then the heavy metal concentration in the sample is accurately quantitatively analyzed through the pre-trained quantitative regression model.

[0076] An electromagnetic heating and water-cooling temperature control pretreatment module, which is used to control the temperature of the water quality sample and can digest the water quality sample as needed to obtain the water quality sample at the target temperature;

[0077] An electrolyte addition module for quantitatively adding electrolytes to the water quality sample to be measured.

[0078] The signal processing and analysis module further includes:

[0079] An optimization algorithm component, which is used to select the best hyperparameters of the machine learning model through optimization algorithms such as particle swarm optimization or genetic algorithm when selecting the best hyperparameters of the machine learning model;

[0080] A model training component for training a support vector machine or an artificial neural network model. The model training process includes: collecting a series of water quality samples with known heavy metal types and concentrations, testing all the known water quality samples using different working electrodes based on anodic stripping voltammetry, and recording the electrical sensing signals of each sample; assigning corresponding electrode category labels to each electrical sensing signal to form a heavy metal electrode category label training dataset; training a support vector machine or an artificial neural network model through the electrode category label training dataset; at the same time, collecting a series of water quality samples with known heavy metal concentrations of different types, testing these samples using anodic stripping voltammetry, recording the electrical sensing signals of each sample, corresponding each electrical sensing signal to a corresponding heavy metal concentration value, constructing different heavy metal type electrical sensing signal-concentration datasets based on each electrode category, and forming an electrical sensing signal-concentration database with the heavy metal type-electrical sensing signal-concentration datasets corresponding to different types of electrode types, thereby training a pre-trained quantitative regression model.

[0081] An adaptive learning component for implementing an adaptive learning process. The adaptive learning process includes: integrating newly measured electrical sensing signals and their corresponding electrode categories and / or heavy metal concentration information into the current electrode type label training dataset and the electrical sensing signal-concentration database; selecting data from the updated database for retraining the model to update the model parameters; testing the performance of the model through cross-validation or on an independent validation set to ensure that the retrained model maintains or improves the prediction accuracy; replacing the original model with the verified model for actual rapid analysis of heavy metals in water quality; periodically repeating the above steps to form a continuous adaptive learning loop.

[0082] In specific implementation, the rapid heavy metal analysis system for water quality proposed in this embodiment is an integrated hardware and software solution. The specific hardware description is as follows:

[0083] Anodic stripping voltammetry rapid heavy metal detection module for water quality:

[0084] This module includes a three-electrode system device, including a working electrode, an auxiliary electrode, and a reference electrode, for performing anodic stripping voltammetry tests. Different working electrodes can be selected according to actual needs to adapt to the detection of different heavy metals.

[0085] Equipped with a dedicated electrochemical workstation, which can precisely control the application of voltage and current and simultaneously precisely measure the electrical sensing signals.

[0086] Connected to a temperature control unit, which can ensure that the temperature of the electrode is stable within a suitable range during the test to improve the accuracy and stability of the test.

[0087] This module also includes a sample clamping and transfer device for automatically and precisely delivering the water quality sample near the electrode to ensure efficient testing.

[0088] Signal processing and analysis module:

[0089] It includes a high-performance data processing unit, such as an embedded computer or an industrial PC, for running complex data processing algorithms and machine learning models to process the electrical sensing signals obtained from the rapid detection module of heavy metals in water quality by anodic stripping voltammetry.

[0090] Integrates a dedicated digital signal processor (DSP) or a field-programmable gate array (FPGA) to achieve fast and efficient signal processing and meet the requirements of real-time analysis.

[0091] Equipped with a high-resolution analog-to-digital converter (ADC) for accurately converting electrochemical signals into digital signals for subsequent processing and analysis.

[0092] This module also contains a user interface, such as a touch screen or an LCD display, for displaying analysis results (including the type of working electrode, the type of heavy metal determined by qualitative analysis, and the heavy metal concentration obtained by quantitative analysis, etc.) and operation instructions to facilitate user use.

[0093] Electromagnetic heating and water-cooling temperature control pretreatment module:

[0094] This module consists of an electromagnetic heater and a water-cooling system and can accurately control the temperature of the sample according to different test requirements.

[0095] Equipped with a high-precision temperature sensor and a control system, such as a PID controller, to achieve precise temperature regulation and ensure that the sample is pretreated at the target temperature.

[0096] Designed with a sample container and a reaction chamber to allow pretreatment of the sample at a specific temperature, such as digestion treatment, to improve the accuracy of subsequent detection.

[0097] Electrolyte addition module:

[0098] Contains a precise metering pump or titration device for quantitatively adding electrolytes to the water quality sample to ensure the accuracy of the addition amount.

[0099] Equipped with a flow sensor and control software to monitor and control the addition amount of electrolytes in real time and ensure the accuracy of the added electrolytes.

[0100] Optimization algorithm component:

[0101] This component is part of the signal processing and analysis module and contains software implementing optimization algorithms such as particle swarm optimization (PSO) and genetic algorithm (GA).

[0102] Through these algorithms, the system can automatically adjust and optimize the hyperparameters of machine learning models, such as the kernel function parameters and penalty factors of support vector machines, and the learning rate and number of hidden layer nodes of artificial neural networks, to improve the accuracy of analysis and the performance of the models.

[0103] Model training component:

[0104] It includes software tools for training support vector machine (SVM) and artificial neural network (ANN) models.

[0105] This component can process and analyze the electrical sensing signals obtained from anodic stripping voltammetry, and combine the known electrode category labels and heavy metal concentration information to train a support vector machine or artificial neural network model for qualitative analysis and a pre-trained quantitative regression model for quantitative analysis respectively.

[0106] Adaptive learning component:

[0107] This component includes software algorithms for updating and optimizing the analysis model.

[0108] It can integrate new electrical sensing signal data, electrode category information, and heavy metal concentration information, automatically update the electrode type label training dataset and the electrical sensing signal-concentration database, and perform model retraining.

[0109] It includes cross-validation and performance testing tools to ensure that the model maintains or improves prediction accuracy after update, and can periodically repeat the adaptive learning process to enable the system to continuously adapt to new water quality detection requirements.

[0110] The above description of the embodiments is for the convenience of those of ordinary skill in the art to understand and use the invention. Obviously, those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative efforts. Therefore, the present invention is not limited to the above embodiments, and the improvements and modifications made by those skilled in the art without departing from the scope of the present invention should be within the protection scope of the present invention.

Claims

1. A rapid analysis method for heavy metals in water, characterized in that: The following steps are involved: The actual water quality samples are tested by anodic stripping voltammetry to obtain the electrical sensing signals corresponding to the types and contents of heavy metals in the samples; The acquired electrical sensor signal is processed and analyzed to obtain the concentration of heavy metals in the water quality sample and the corresponding results of the electrical sensor signal, wherein: first, a qualitative analysis is performed, that is, a pre-trained support vector machine or artificial neural network model is used to analyze the signal to determine the type of working electrode used to obtain the heavy metal sensor signal, and then the corresponding pre-trained quantitative regression model is selected based on the electrode type, and then the concentration of heavy metals in the sample is accurately quantitatively analyzed through the pre-trained quantitative regression model.

2. A method for rapid analysis of heavy metals in water according to claim 1, characterized in that: When analyzing signals, an optimization algorithm is used to select the best machine learning model hyperparameters.

3. A method for rapid analysis of heavy metals in water according to claim 2, characterized in that: The optimization algorithm includes particle swarm optimization or genetic algorithm.

4. A method for rapid analysis of heavy metals in water according to claim 1, characterized in that: Before performing qualitative analysis, the acquired electrical sensor signal is subjected to a preprocessing step, and the preprocessing step includes at least one of the following processes: Filtering to eliminate noise and interference; Baseline correction to calibrate the baseline drift of the signal; Signal enhancement to improve the signal-to-noise ratio; Principal component analysis is used to reduce the dimension of the electrical sensor signal to extract key features.

5. A method for rapid analysis of heavy metals in water according to claim 1, characterized in that: The methods for obtaining pre-trained support vector machine or artificial neural network models include: Collect a series of water samples with known heavy metal types and concentrations; Based on anodic stripping voltammetry, all known water quality samples were tested using different working electrodes, and the electrical sensing signal of each sample was recorded; Assign a corresponding electrode category label to each electrical sensing signal to form a heavy metal electrode category label training data set; A support vector machine or artificial neural network model is obtained by training the electrode category label training data set.

6. A method for rapid analysis of heavy metals in water according to claim 5, characterized in that: The methods for obtaining pre-trained quantitative regression models include: Collect a series of different types of water samples with known heavy metal concentrations; The samples were tested using anodic stripping voltammetry, and the electrical sensing signal of each sample was recorded; For each electrical sensor signal, a corresponding heavy metal concentration value is corresponding to the electrical sensor signal. Based on each electrode type, an electrical sensor signal-concentration data set of different heavy metal types is constructed. The heavy metal type-electrical sensor signal-concentration data sets corresponding to different types of electrodes constitute an electrical sensor signal-concentration database. By dividing the electrical sensor signal-concentration database into training set, validation set, and test set, the quantitative regression model is selected and the parameters are initialized according to the data characteristics and requirements. The model is trained with the training set, and the hyperparameters are evaluated and tuned with the help of the validation set. Finally, the test set is used for evaluation. If it meets the requirements, it is determined as a pre-trained quantitative regression model.

7. A rapid analysis method for heavy metals in water according to claim 1, characterized in that: Also included is an adaptive learning process, the adaptive learning process comprising: Integrate the newly measured electrical sensor signal and its corresponding electrode type and / or heavy metal concentration information into the current electrode type label training data set and the electrical sensor signal-concentration database; Select data from the updated database for retraining the model to update the model parameters; Ensure that the retrained model maintains or improves its prediction accuracy by cross-validating or testing the model’s performance on an independent validation set; The verified model will replace the original model for actual rapid analysis of heavy metals in water; Repeat the above steps periodically to form a continuous adaptive learning cycle.

8. A rapid analysis system for heavy metals in water, characterized in that: include: Anodic stripping voltammetry water quality heavy metal rapid detection module is used to test actual water quality samples to obtain electrical sensing signals corresponding to the heavy metal content in the samples; The signal processing and analysis module is used to process and analyze the electrical sensor signal to obtain the concentration of heavy metals in the water quality sample and the corresponding results of the electrical sensor signal, wherein: first, a qualitative analysis is performed, that is, a pre-trained support vector machine or artificial neural network model is used to analyze the signal to determine the working electrode category for obtaining the sensor signal, and then the corresponding pre-trained quantitative regression model is selected based on the electrode category, and then the concentration of heavy metals in the sample is accurately quantitatively analyzed through the pre-trained quantitative regression model.

9. A rapid heavy metal analysis system for water quality according to claim 8, characterized in that: The signal processing and analysis module also includes: An optimization algorithm component, wherein the optimization algorithm component is used to select the best machine learning model hyperparameters through an optimization algorithm when selecting the best machine learning model hyperparameters; A model training component is used to train a support vector machine or an artificial neural network model. The model training process includes: collecting a series of water samples with known heavy metal types, testing the water samples using anodic stripping voltammetry with different working electrodes, recording the electrical sensing signal of each sample, assigning a corresponding electrode category label to each electrical sensing signal, forming an electrode category label training data set, and training the support vector machine or artificial neural network model through the electrode category label training data set; An adaptive learning component is used to implement an adaptive learning process, which includes: integrating the newly measured electrical sensor signals and their corresponding electrode categories and / or concentration information into the current electrode category label training data set and the electrical sensor signal-concentration database, selecting data from the updated database for model retraining to update model parameters, testing the performance of the model through cross-validation or on an independent validation set to ensure that the retrained model maintains or improves the accuracy of the prediction, replacing the original model with the validated model for actual rapid analysis of heavy metals in water quality, and periodically repeating the above steps to form a continuous adaptive learning cycle.

10. A rapid heavy metal analysis system for water quality according to claim 8, characterized in that: The water quality heavy metal rapid analysis system also includes: An electromagnetic heating and water-cooling temperature control pretreatment module, which is used to control the temperature of the water sample and can digest the water sample as needed to obtain a water sample at a target temperature; The electrolyte adding module is used to quantitatively add electrolyte to the water quality sample to be tested.

Images