Water quality detection device and method based on friction nanometer power generation
By combining multi-electrode triboelectric sensors with deep learning, the problem of triboelectric signals being affected by various physicochemical factors has been solved, enabling rapid and accurate detection of heavy metal ions, which is suitable for field applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-01
AI Technical Summary
Existing liquid detection methods based on triboelectric signals have limited information content in a single signal channel, and triboelectric signals are easily affected by the coupling of various physicochemical factors, making it difficult to simultaneously identify heavy metal ion species and quantify their concentrations.
By combining multi-electrode triboelectric sensors and deep learning, rich temporal and spatial location feature information is obtained through multi-channel triboelectric signals. A category and concentration identification model is constructed to achieve accurate identification and precise quantification of heavy metal ions.
It enables rapid and accurate detection of heavy metal ion types and concentrations. The device is simple, portable, and self-powered, making it suitable for field applications. It lowers the detection limit and improves the accuracy of identification and quantification.
Smart Images
Figure CN121955097A_ABST
Abstract
Description
Water quality detection device and method based on triboelectric nanogenerator Technical Field
[0001] This invention relates to the field of water quality testing, and in particular to a water quality testing device and method based on triboelectric nanogenerator. Background Technology
[0002] Water resources are essential for sustaining human life, and ensuring a safe and hygienic drinking water supply is of paramount importance. However, with the advancement of industrialization and urbanization, water pollution has become increasingly serious. According to relevant reports, if effective water quality monitoring measures are not implemented within five years, the survival and health of nearly half the world's population will be threatened. This situation highlights the severity of the water pollution problem and urgently requires attention. Among the many sources of water pollution, the indiscriminate discharge and improper treatment of heavy metal ions and pesticide residues in industrial wastewater are among the main causes. Therefore, detecting heavy metal ions in water is of great significance for water pollution control.
[0003] Traditional methods for heavy metal ion detection include atomic absorption spectrometry (AAS), atomic emission spectrometry (AES), inductively coupled plasma mass spectrometry (ICP-MS), and X-ray fluorescence spectrometry (XRF). However, these methods require expensive equipment, time-consuming sample preparation, and complex analytical procedures. Furthermore, the fixed equipment limits their application outside the laboratory. Therefore, various rapid detection techniques have been developed. One of the most commonly used methods is colorimetry, which offers advantages such as rapid visual identification, portability, and ease of operation. However, the accuracy of detection is largely influenced by the observer's subjective judgment. Notably, electrochemical techniques have seen rapid development in the field of heavy metal ion detection. In particular, the introduction of nanomaterials into the design of electrochemical biosensors has significantly improved their sensitivity, ushering in a new stage of development. Compared to traditional methods, these rapid methods offer significant advantages in practical applications; however, they are not easy to manufacture, and their sensitivity and accuracy are somewhat reduced. Moreover, in terms of power supply in sensor design, frequent charging or battery replacement is inconvenient. Therefore, there is an urgent need for a sensing strategy that can accurately and sensitively detect heavy metal ions, while being easy to manufacture, self-powered, and suitable for practical field applications.
[0004] Triboelectric nanogenerators (TENGs) can convert environmental mechanical energy into electrical signals and have been widely used in energy harvesting and self-powered sensing. With their advantages of self-powering, low cost, and environmental friendliness, TENGs are particularly suitable for portable and field sensing applications. Among various TENG structures, liquid-solid triboelectric nanogenerators (LS-TENGs) maintain stable performance even under complex conditions, exhibiting excellent environmental adaptability. Based on these advantages, LS-TENGs have been successfully applied in multiple fields, including electrochemical sensing and biomedicine. Existing research proposes a biomimetic electronic tongue system based on liquid-solid triboelectric nanogenerators (LS-TENGs), which can effectively distinguish different types of liquids through triboelectric response. However, the triboelectric response of LS-TENGs is highly sensitive to the physicochemical properties of the liquid, which directly affect the charge distribution and charge transfer behavior at the liquid-solid interface. Specifically, droplets containing different types and concentrations of heavy metal ions differ in conductivity and spreading behavior on the surface of the triboelectric layer, thus affecting the charge distribution and charge transfer behavior at the liquid-solid interface; simultaneously, changes in ion concentration further affect the interfacial charge transfer process through charge shielding effects. The above factors cause the type and concentration of heavy metal ions to have a coupled effect on the generation of triboelectric signals, making it difficult to simultaneously achieve accurate identification of ion types and precise quantification of concentrations based on signals obtained from a single triboelectric detection unit. Summary of the Invention
[0005] The purpose of this invention is to provide a water quality detection device and method based on triboelectric nanogenerators. This addresses the technical problems in existing liquid detection methods based on triboelectric signals, such as the limited information content of a single signal channel and the susceptibility of triboelectric signals to the coupling effects of various physicochemical factors, making it difficult to simultaneously identify heavy metal ion species and quantify their concentrations.
[0006] First, this application provides a water quality detection device based on triboelectric nanogenerator, including a multi-electrode triboelectric sensor, wherein the multi-electrode triboelectric sensor includes a substrate, metal electrodes and a triboelectric layer;
[0007] Two or more metal electrodes are disposed between the substrate and the triboelectric layer. The metal electrodes are arranged at equal intervals along the length of the triboelectric layer. During water quality detection, the droplets are guided to slide along the length of the triboelectric layer. During the sliding process, the droplets pass through the areas of several metal electrodes in sequence, and time-related triboelectric signals are generated at each metal electrode.
[0008] Optionally, it may also include a signal acquisition module and a data processing module;
[0009] The input end of the signal acquisition module is connected to several metal electrodes, and the output end of the signal acquisition module is connected to the data processing module. The signal acquisition module is used to acquire triboelectric signals at several metal electrodes, and the data processing module is used to extract and analyze the acquired triboelectric signals and output the type and concentration information of heavy metal ions in the droplet to be tested.
[0010] Optionally, the metal electrode is made of copper, the width of the metal electrode is 1 cm, and the spacing between adjacent metal electrodes is 2 cm.
[0011] Optionally, the triboelectric layer is made of FEP film, and the substrate is made of acrylic acid.
[0012] Optionally, the number of metal electrodes is four.
[0013] Secondly, this application provides a water quality detection method based on triboelectric nanogenerator, using the aforementioned water quality detection device based on triboelectric nanogenerator, with the following specific steps:
[0014] S1: Construct a category and concentration recognition model based on deep learning. The category and concentration recognition model takes multi-channel triboelectric signals as input and outputs the types and concentrations of heavy metal ions in the droplet.
[0015] S2: Obtain training samples and train the category and concentration recognition model;
[0016] S3: A droplet containing the heavy metal ions to be measured is introduced into the surface of the multi-electrode triboelectric sensor, and the droplet slides along its length on the surface of the triboelectric layer, and the time-correlated multi-channel triboelectric signals generated at each metal electrode are collected simultaneously.
[0017] S4: Input the multi-channel triboelectric signal collected in step S3 into the model trained in step S2 to identify the type and concentration of heavy metal ions.
[0018] Optionally, the category and concentration identification model includes a data preprocessing module, a feature extraction module, a category identification module, and a concentration identification module;
[0019] The data preprocessing module is used to standardize the original triboelectric signal and mix in Gaussian noise. The feature extraction module is used to extract the feature information of the preprocessed triboelectric signal. The category identification module takes the output of the feature extraction module as input and the types of heavy metal ions identified at several metal electrodes as output. The concentration identification module takes the outputs of the category identification module and the feature extraction module as input and the concentration of heavy metal ions identified at several metal electrodes as output.
[0020] Optionally, the feature extraction module extracts the following feature information: peak current, current amplitude, and current signal duration information; average slope of the rising segment of the center peak of the current signal, average slope of the falling segment of the center peak, and the ratio of the average slope of the rising segment of the center peak to the average slope of the falling segment of the center peak; as well as the full width at half maximum (FWHM) and root mean square (RMS) value of the current signal.
[0021] Optionally, the data preprocessing module includes a normalization layer, a noise injection layer, and a sample mixing layer, and the feature extraction module includes a fully connected layer, a residual layer, and an attention mechanism layer.
[0022] Optionally, the category recognition module includes a fully connected layer, a linear layer, and a normalization layer connected in sequence;
[0023] The concentration recognition module includes several recognition sub-modules for recognizing the concentration of metal ions in the triboelectric signals of several metal electrodes, and each recognition sub-module includes a splicing layer, a fully connected layer and a linear layer.
[0024] Because of the adoption of the above technical solution, the present invention has the following advantages:
[0025] 1. This application enables rapid and accurate detection of the types and concentrations of heavy metal ions. The device is simple, portable, low-cost, and self-powered, and can achieve precise concentration quantification over a wide concentration range with an extremely low detection limit.
[0026] 2. This application employs a multi-electrode triboelectric detection structure to generate multi-channel triboelectric signals with temporal sequence and spatial position correlation during the droplet sliding process, which significantly enriches the dimensions of detection information.
[0027] 3. This application effectively reduces the coupling interference of multiple physicochemical factors on a single triboelectric signal by extracting features from multi-channel triboelectric signals and combining them with deep learning analysis, thereby improving the accuracy and reliability of heavy metal ion identification and concentration quantification.
[0028] 4. The system of the present invention has a simple structure, requires no external power supply, has good portability and environmental adaptability, and is suitable for rapid on-site detection of heavy metal ions in aquatic environments.
[0029] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0030] The accompanying drawings of this invention are described below.
[0031] Figure 1 shows the positions of the four metal electrodes in the water quality testing device of the present invention.
[0032] Figure 2 shows the current response of the four metal electrodes in the water quality testing device of the present invention.
[0033] Figure 3 is a quantification diagram of the features extracted by the water quality detection method of the present invention.
[0034] Figure 4 is a quantification diagram of the complex features extracted by the water quality detection method of the present invention.
[0035] Figure 5 shows the category and concentration recognition model based on deep learning of this invention.
[0036] Figure 6 is a structural diagram of multiple modules in the category and concentration identification model of the present invention.
[0037] Figure 7 is a confusion diagram of the six different heavy metal ions of the present invention.
[0038] Figure 8 is a comparison chart of the normalized concentrations of the predicted concentration and the actual concentration of the present invention.
[0039] Figure 9 is a confusion matrix diagram of ion type classification in real water samples according to the present invention.
[0040] Figure 10 shows the prediction error of actual water sample concentration according to the present invention. Detailed Implementation
[0041] The present invention will be further described below with reference to the accompanying drawings and embodiments. The terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the embodiments of the present invention. Furthermore, unless otherwise expressly specified and limited, the terms "connected" and "linked" should be interpreted broadly, for example, they can refer to a fixed connection, a detachable connection, an integral connection, an electrical connection, or a signal connection; they can refer to a direct connection or an indirect connection through an intermediate medium.
[0042] Example 1:
[0043] As shown in Figure 1, a water quality detection device based on triboelectric nanogenerator includes a multi-electrode triboelectric sensor, a signal acquisition module, and a data processing module. The multi-electrode triboelectric sensor includes a substrate, metal electrodes, and a triboelectric layer.
[0044] Two or more metal electrodes are disposed between the substrate and the triboelectric layer. The metal electrodes are arranged at equal intervals along the length of the triboelectric layer. During water quality detection, the droplets are guided to slide along the length of the triboelectric layer. During the sliding process, the droplets pass through the areas of several metal electrodes in sequence, and time-related triboelectric signals are generated at each metal electrode.
[0045] The input end of the signal acquisition module is connected to several metal electrodes, and the output end of the signal acquisition module is connected to the data processing module. The signal acquisition module is used to acquire triboelectric signals at several metal electrodes, and the data processing module is used to extract and analyze the acquired triboelectric signals and output the type and concentration information of heavy metal ions in the droplet to be tested.
[0046] In this embodiment, there are four metal electrodes, the metal electrodes are made of copper, the width of the metal electrodes is 1 cm, and the spacing between adjacent metal electrodes is 2 cm; the triboelectric layer is made of FEP film, and the substrate is made of acrylic.
[0047] In this embodiment, compared with the single-electrode LS-TENG sensor, the multi-metal electrode structure in this application can sequentially collect triboelectric response signals at different spatial positions as the droplet slides down the surface of the triboelectric layer, thereby obtaining richer temporal and spatial positional feature information. This application can obtain more complete dynamic morphological change information during the droplet's movement, avoiding the problem of insufficient signal information caused by the single-electrode sensing unit, and providing a sufficient signal basis for subsequent accurate differentiation of heavy metal ion types and precise quantification of concentration.
[0048] In this embodiment, as shown in Figure 1, four metal electrodes are defined as electrodes 1 to 4, arranged from top to bottom. As the droplet slides down the array, each electrode sequentially generates a triboelectric signal, providing crucial information for detecting the type and concentration of heavy metal ions. Upon initial contact with the polymer surface, contact electrification occurs, forming an electric double layer at the interface. Electrons transfer from the droplet to the polymer, resulting in a negatively charged polymer surface and a positively charged droplet. As the droplet spreads and slides on the polymer film surface, positive charges gradually accumulate inside the droplet. When the droplet contacts the polymer surface above electrode 1, the excess positive charge generated by electrostatic induction induces electrons to flow from the copper electrode to ground until equilibrium is reached. This process generates a positive current signal. As the droplet continues to slide and leaves the copper electrode, the local electric field near the electrode changes, inducing charge recirculation and generating a reverse current signal. Finally, the droplet slides along the polymer surface, continuously accumulating positive charges until it detaches from the surface. A similar mechanism exists when the droplet slides past electrodes 2, 3, or 4. Upon contact with each electrode, interfacial charge transfer drives electrons to flow through the external circuit, generating current pulses. Partial detachment of the droplet causes electron backflow, resulting in currents of opposite polarity. As the droplet slides across all four electrodes, four time-resolved current pulses are generated, each corresponding to the interaction between the droplet and each electrode. These multi-electrode signals provide a more complete description of droplet dynamics, thus aiding in the differentiation of heavy metal ion types and their concentrations.
[0049] Example 2:
[0050] Figure 5 shows a water quality detection method based on triboelectric nanogenerators. Using the aforementioned water quality detection device based on triboelectric nanogenerators, the specific steps are as follows:
[0051] S1: Construct a category and concentration recognition model based on deep learning. The category and concentration recognition model takes a multi-channel triboelectric signal related to time and spatial location as input and the type and concentration of heavy metal ions in the droplet as output. The category and concentration recognition model includes a data preprocessing module, a feature extraction module, a category recognition module, and a concentration recognition module.
[0052] 1. Data Preprocessing Module: As shown in Figure 6(a), the data preprocessing module includes a normalization layer, a noise injection layer, and a sample mixing layer. The normalization layer is used to normalize the acquired multi-channel triboelectric signals to reduce the influence of factors such as small differences in droplet velocity on the signal amplitude. The noise injection layer is used to add a small amount of Gaussian noise to the original signal. The sample mixing layer uses the Mixup data augmentation method to improve the generalization ability and robustness of the model. The preprocessed four-channel triboelectric signals are spliced in the time dimension and expanded into a one-dimensional vector as the model input.
[0053] 2. Feature extraction module: As shown in Figure 5, the feature extraction module is used to extract the feature information of the preprocessed triboelectric signal. The feature extraction module includes a fully connected layer, a residual layer and a multi-head attention mechanism layer.
[0054] In this embodiment, as shown in Figure 2, the feature information extracted by the feature extraction module includes the peak current I. max and I min The current amplitudes h1, h2, h3, and h4, and the duration of the current signal are used to describe the overall amplitude and time-scale characteristics of the signal.
[0055] As shown in Figure 3, the feature information also includes the average slope R of the rising segment of the center peak of the current signal. u The average slope R of the descending segment of the central peak d The average slope R of the rising segment of the central peak u The average slope R of the descending segment of the central peak d The ratio S r , is used to describe the characteristics of waveform slope change.
[0056] As shown in Figure 4, the feature information also includes the full width at half maximum (FWHM), root mean square (RMS) value, skewness, or other parameters in Figure 4 of the current signal, which are used to describe the amplitude distribution and energy characteristics of the triboelectric signal.
[0057] In this embodiment, as shown in Figure 6(b), the fully connected layer includes a linear layer, a normalization layer, a GELU activation function layer, and a recognition layer connected in sequence. As shown in Figure 5, the feature extraction module has 6 residual layers. As shown in Figure 6(c), the residual layer includes a fully connected layer, a linear layer, a normalization layer, and a recognition layer connected in sequence.
[0058] 3. Category recognition module: As shown in Figure 5, the category recognition module includes a fully connected layer, a linear layer and a normalization layer connected in sequence.
[0059] 4. Concentration recognition module: As shown in Figure 5, the concentration recognition module includes several recognition sub-modules for recognizing the concentration of metal ions in the triboelectric signals of several metal electrodes, as shown in Figure 6(d). Each recognition sub-module includes a splicing layer, a fully connected layer and a linear layer.
[0060] In this embodiment, there are four identification sub-modules, which are used to identify the triboelectric signals of four metal electrodes with different concentration ranges of metal ions. The identification sub-module includes a connection layer, four fully connected layers and a linear layer connected in sequence.
[0061] S2: Obtain model training samples and train the category and concentration recognition model;
[0062] In this embodiment, a heavy metal ion sensing dataset containing six different heavy metal ion solutions was obtained, totaling approximately 15,000 samples. 80% of the data was used for model training, and the remaining 20% was used for model testing.
[0063] S3: A droplet containing the heavy metal ions to be measured is introduced into the surface of the multi-electrode triboelectric sensor, and the droplet slides along its length on the surface of the triboelectric layer, and the time-correlated multi-channel triboelectric signals generated at each metal electrode are collected simultaneously.
[0064] In this embodiment, a micro-injection pump controls the syringe needle to form a droplet with a volume of approximately 50µL, which is then allowed to fall freely from a height of approximately 10cm onto the surface of the triboelectric layer and slide. The droplet passes through four metal electrode regions in sequence, generating corresponding triboelectric response signals, thereby forming a multi-channel triboelectric signal output with temporal sequence and spatial position information.
[0065] S4: Input the multi-channel triboelectric signal collected in step S3 into the model trained in step S2 to identify the type and concentration of heavy metal ions.
[0066] S5: Experiments and Verification:
[0067] S5.1: Model Recognition and Concentration Prediction Performance Verification: To verify the performance of the droplet-driven intelligent multi-metal electrode triboelectric detection system described in this application in terms of heavy metal ion species identification and concentration quantification, a two-stage analysis strategy was adopted based on the trained model. First, heavy metal ion types were identified, and then the corresponding ion concentrations were predicted based on the identification results. Experimental results show that this method can effectively distinguish multiple heavy metal ion species in water. As shown in Figure 7, the confusion matrix analysis shows that, under the experimental conditions, the accuracy rate of heavy metal ion species identification in water exceeds 98%. After completing ion type identification, the concentration of heavy metal ions in water was further predicted. To facilitate the evaluation of prediction performance, both the predicted and actual concentrations were normalized to a standard scale. As shown in Figure 8, the predicted and actual concentrations have a high degree of consistency, with an average absolute percentage error of less than 4%. This method effectively decouples heavy metal ion species and concentration information, and further completes the accurate quantification of concentration based on ion species identification, thereby achieving synchronous detection of heavy metal ions in water.
[0068] S5.2: Real Water Sample Detection Verification: To evaluate the detection performance of this invention under practical application conditions, real water samples collected from rivers, lakes, and tap water were used for testing, with lead ions as a representative example for analysis. Experimental results show that this method can first accurately determine the presence of specific heavy metal ions in the water sample and further predict the concentration of these ions. As shown in Figure 9, even with interference from other ions, the overall accuracy of heavy metal ion identification can still reach 91.2%. As shown in Figure 10, further analysis of the concentration prediction error in real water samples shows that the prediction error is controlled within 20% of the actual concentration. Compared with traditional analytical methods, the detection method described in this application significantly shortens the single detection time, with each measurement responding rapidly within 80 ms, and requiring no complex sample pretreatment or dedicated analytical instruments, indicating that the system is suitable for rapid on-site detection of heavy metal ions in water. Compared with traditional analytical methods, this method requires significantly less time. Each measurement is completed within a few seconds, without the need for complex pretreatment or dedicated instruments. These results demonstrate that devices combined with deep learning can be used for rapid on-site analysis of heavy metal ions in water.
[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A water quality detection device based on triboelectric nanogenerator, characterized in that, The invention includes a multi-electrode triboelectric sensor, which comprises a substrate, metal electrodes, and a triboelectric layer. Two or more metal electrodes are disposed between the substrate and the triboelectric layer. The metal electrodes are arranged at equal intervals along the length of the triboelectric layer. During water quality detection, a droplet is guided to slide along the length of the triboelectric layer. During the sliding process, the droplet passes through several metal electrode regions in sequence, and a time-related triboelectric signal is generated at each metal electrode.
2. The water quality detection device based on triboelectric nanogenerator according to claim 1, characterized in that, It also includes a signal acquisition module and a data processing module; the input end of the signal acquisition module is connected to several metal electrodes, and the output end of the signal acquisition module is connected to the data processing module. The signal acquisition module is used to acquire triboelectric signals at several metal electrodes, and the data processing module is used to extract and analyze the features of the acquired triboelectric signals and output the type and concentration information of heavy metal ions in the droplet to be tested.
3. The water quality detection device based on triboelectric nanogenerator according to claim 1, characterized in that, The metal electrode is made of copper, the width of the metal electrode is 1 cm, and the spacing between adjacent metal electrodes is 2 cm.
4. The water quality detection device based on triboelectric nanogenerator according to claim 1, characterized in that, The triboelectric layer is made of FEP film, and the substrate is made of acrylic acid.
5. A water quality detection device based on triboelectric nanogenerator according to claim 1 or 3, characterized in that, The number of metal electrodes is four.
6. A water quality detection method based on triboelectric nanogenerators, characterized in that, The water quality detection device based on triboelectric nano-power generation as described in any one of claims 1-5 includes the following specific steps: S1: Constructing a category and concentration recognition model based on deep learning, wherein the category and concentration recognition model takes multi-channel triboelectric signals as input and the types and concentrations of heavy metal ions in the droplets as output; S2: Obtain training samples and train the category and concentration recognition model; S3: A droplet containing the heavy metal ions to be measured is introduced into the surface of the multi-electrode triboelectric sensor, and the droplet slides along its length on the surface of the triboelectric layer, and the time-correlated multi-channel triboelectric signals generated at each metal electrode are collected simultaneously. S4: Input the multi-channel triboelectric signal collected in step S3 into the model trained in step S2 to identify the type and concentration of heavy metal ions.
7. The water quality detection method based on triboelectric nanogenerator according to claim 6, characterized in that, The category and concentration identification model includes a data preprocessing module, a feature extraction module, a category identification module, and a concentration identification module. The data preprocessing module is used to standardize the original triboelectric signal and mix in Gaussian noise. The feature extraction module is used to extract the feature information of the preprocessed triboelectric signal. The category identification module takes the output of the feature extraction module as input and the types of heavy metal ions identified at several metal electrodes as output. The concentration identification module takes the outputs of the category identification module and the feature extraction module as input and the concentration of heavy metal ions identified at several metal electrodes as output.
8. The water quality detection method based on triboelectric nanogenerator according to claim 7, characterized in that, The feature extraction module extracts the following feature information: peak current, current amplitude, and current signal duration; average slope of the rising segment of the center peak, average slope of the falling segment of the center peak, and the ratio of the average slope of the rising segment of the center peak to the average slope of the falling segment of the center peak; as well as the full width at half maximum (FWHM) and root mean square (RMS) value of the current signal.
9. A water quality detection method based on triboelectric nanogenerator according to claim 7, characterized in that, The data preprocessing module includes a normalization layer, a noise injection layer, and a sample mixing layer, while the feature extraction module includes a fully connected layer, a residual layer, and an attention mechanism layer.
10. A water quality detection method based on triboelectric nanogenerator according to claim 7, characterized in that, The category identification module includes a fully connected layer, a linear layer, and a normalization layer connected in sequence; the concentration identification module includes several identification sub-modules for identifying the concentration of metal ions in several metal electrode triboelectric signals, and each identification sub-module includes a splicing layer, a fully connected layer, and a linear layer.