Rapid detection method for mercury concentration of surface sediment in floating rice planting area

Abstract

The invention discloses a floating rice planting area surface sediment mercury concentration rapid detection method, which comprises: S1, obtaining magnetic parameter data of a target floating rice planting area surface sediment, and obtaining physicochemical index data of the target floating rice planting area surface sediment; s2, inputting the obtained magnetic parameter data and physicochemical index data of the surface sediment of the target floating rice planting area into an MDN artificial intelligence model, and training and constructing the MDN artificial intelligence model; and S3, predicting the mercury concentration in the surface sediment of the floating rice planting area to be detected through the magnetic parameters, the physicochemical indexes and the constructed MDN artificial intelligence model. According to the method, the relationship among environmental magnetic parameters, physicochemical indexes and the floating rice planting area surface sediment mercury concentration is established through artificial intelligence, a floating rice planting area surface sediment mercury concentration rapid detection model is constructed, and a new method and a new thought are provided for floating rice planting area surface sediment mercury concentration detection.

Description

Technical Field

[0001] The present invention relates to the field of mercury pollution detection research, and particularly to a rapid detection method for mercury concentration in the surface sediment of a floating rice planting area based on environmental magnetics. Background Art

[0002] In the surface sediment of water bodies, heavy metal pollution is a common problem. Among them, mercury poses a potential threat to aquatic ecosystems and human health due to its significant biological toxicity, accumulation, and amplification effects.

[0003] In this context, floating rice emerged as an innovative planting mode. It uses technologies such as floating beds, floating islands, and floating boards to enable rice to grow suspended on the water surface. This mode has shown remarkable effects in solving the ecological restoration problems in coal mining subsidence water areas. However, the sediment in coal mining subsidence water areas may be rich in heavy metal pollutants including mercury, which poses a potential threat to the planting safety of floating rice.

[0004] To ensure the planting safety of floating rice in coal mining subsidence areas and avoid its being damaged by mercury pollution in the sediment, the rapid detection of mercury concentration in the surface sediment of floating rice planting areas is of crucial significance for the mercury pollution control in coal mining surface subsidence water areas and the protection of human health.

[0005] Traditional heavy metal mercury concentration detection methods are usually complex and time-consuming. The environmental magnetic measurement method, with its characteristics of simplicity, rapidity, economy, safety, and multi-functionality, shows great application potential in the field of environmental pollution monitoring and treatment. In recent years, numerous studies have shown that there is an inherent relationship between heavy metal elements and magnetic minerals. This relationship provides a new idea for the rapid detection of mercury concentration in sediments using environmental magnetic methods.

[0006] Based on this discovery, the present invention deeply explores the correlation between magnetic parameters and mercury concentration, and is committed to establishing a response relationship model between magnetic parameters and mercury concentration. Through this model, we can accurately and quickly evaluate the mercury concentration and its pollution status in the sediment, thereby providing strong technical support for the ecological environment protection and human health in floating rice planting areas. Summary of the Invention

[0007] The purpose of the present invention is to provide a rapid detection method for mercury concentration in the surface sediment of a floating rice planting area, so as to be able to rapidly evaluate the mercury concentration and its pollution status in the sediment.

[0008] To this end, the present invention provides a rapid detection method for mercury concentration in the surface sediment of a floating rice planting area, including: S1. Obtain the magnetic parameter data of the surface sediment in the target floating rice planting area, and obtain the physical and chemical index data of the surface sediment in the target floating rice planting area; S2. Input the obtained magnetic parameter data and physical and chemical index data of the surface sediment in the target floating rice planting area into the MDN artificial intelligence model, and train and construct the MDN artificial intelligence model; S3. Predict the mercury concentration in the surface sediment of the floating rice planting area to be measured through the magnetic parameters, physical and chemical indexes, and the constructed MDN artificial intelligence model.

[0009] The present invention uses a magnetic instrument to test the magnetic parameters of the surface sediment in the floating rice planting area to obtain magnetic parameter data; uses an AMA254 mercury analyzer, a PHSJ-4A pH meter, and a DDSJ-308A to analyze and detect the mercury concentration and physical and chemical indexes in the above surface sediment to obtain preliminary data; inputs the magnetic parameter data, physical and chemical indexes, and the mercury concentration data into the MDN artificial intelligence model, corrects the preliminary measurement data based on the mercury concentration to be measured, and predicts the mercury concentration to be measured through the magnetic parameters and the MDN artificial intelligence model, so as to obtain the output result. This technology can achieve rapid detection of mercury concentration in surface sediment with high precision and low detection limit, and provides a method for quickly detecting the mercury concentration in the surface sediment of the planting area.

[0010] In addition to the purposes, features, and advantages described above, the present invention has other purposes, features, and advantages. The following will refer to the drawings for a further detailed description of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The specification drawings constituting a part of this application are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0012] Figure 1 is a flowchart of the rapid detection method for mercury concentration in the surface sediment of a floating rice planting area based on environmental magnetism of the present invention.

[0013] Figure 2 is an execution flowchart of the rapid detection method for mercury concentration in the surface sediment of a floating rice planting area based on environmental magnetism in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] The following will refer to the drawings and combine with embodiments to detail the present invention.

[0015] With reference to Figure 1 and Figure 2 , the rapid detection method for mercury concentration in an embodiment of the present invention includes the following steps S100-S600.

[0016] S100, Acquisition of Magnetic Parameter Data

[0017] Use a magnetic instrument to measure the magnetic parameters of surface sediments to obtain preliminary magnetic parameter data. The magnetic properties of the sediments Wm are [χ, χ ARM , SIRM, HIRM, SOFT, χ ARM / χ, χ ARM / SIRM, SIRM / χ].

[0018] Specifically, wrap approximately 0.5 g of sediment samples with plastic wrap and place them in a non-magnetic cylindrical plastic box dedicated for magnetic parameter testing for testing. Measure the magnetic susceptibility at low frequency (976 Hz; χ), and then perform mass normalization at room temperature using the MFK1-FAK bridge system to obtain the χ value. Use a DTECH 2000AF demagnetizer (ASC Scientific, Carlsbad, CA, USA) to apply an anhysteretic remanent magnetization (ARM) with a peak alternating magnetic field of 100 mT and a DC bias magnetic field of 0.04 mT. The measured value is expressed as the ARM magnetic susceptibility (χ ARM ), obtained by dividing the remanent magnetization by the stable magnetic field value; the mass magnetic susceptibility χ = a × 10 6 / m, where a is the Cabacho reading at 976 Hz and m is the sample mass; perform an isothermal remanent magnetization (IRM) experiment using a MMPM10 pulsed magnetizer. The IRM is measured using a JR-6 two-speed rotating magnetometer (AGICO). The IRM measured in a 1 T magnetic field is called the saturation isothermal remanent magnetization (SIRM). The "hard" IRM (HIRM) and "soft" IRM (SOFT) are calculated using the following formulas respectively:

[0019] SIRM = IRM 1000mT reading value × 1115 × 10 / m;

[0020] SOFT = (SIRM - IRM -20mT ) / 2;

[0021] HIRM = (SIRM + IRM -300mT ) / 2.

[0022] S200, Acquisition of Physical and Chemical Index Data and Mercury Concentration Data

[0023] Use an AMA254 mercury analyzer, a PHSJ-4A pH meter, a DDSJ-308A conductivity meter, and an SX-5-12 muffle furnace to measure the samples to obtain preliminary physical and chemical index data and mercury concentration data. The physical and chemical indices Wp are [pH, conductivity, LOI, Eh].

[0024] Specifically, use an AMA254 mercury analyzer to accurately measure the mercury concentration in the sample. Accurately weigh 10 g of the sample and pour it into a 50 mL beaker. Inject 25 mL of freshly prepared deionized water, and use a glass rod to stir well for 1 min to mix the sample and water evenly. Use a PHSJ-4A pH meter and a DDSJ-308A conductivity meter to measure the supernatant in the beaker after standing for 30 minutes in sequence. After the instrument readings are stable, record the corresponding values. Repeat the measurement three times for a single sample, and take the average value as the final pH and conductivity.

[0025] Gradually dry 5 g of the sample in an oven at 105 ± 2 °C until it reaches a constant weight; take several clean empty crucibles and burn them in a muffle furnace until they reach a constant weight, and accurately weigh them. Use an electronic analytical balance to accurately weigh 3 g of the dried sample (accurate to 0.0001 g, the same below), put it into the burned crucible, and burn it in the muffle furnace at 550 ± 5 °C for 6 h. After cooling, use an analytical balance to accurately weigh the weight of the burned sample together with the crucible.

[0026] The calculation formula for the loss on ignition is as follows:

[0027]

[0028] In the formula: m b is the mass of the dried sediment sample + empty crucible before burning (g); m a is the mass of the sediment sample + empty crucible after burning (g).

[0029] Use a portable ORP composite electrode measuring instrument to measure the redox potential Eh.

[0030] S300. Initial data preprocessing

[0031] Specifically, eliminate abnormal data and missing data, and screen the data.

[0032] Furthermore, step S300 further includes step S310.

[0033] S310. Importance analysis of magnetic parameters and physicochemical indexes

[0034] Specifically, use the random forest method to conduct an importance analysis of the magnetic parameter Wm and the physicochemical index Wp. Obtain the weights of different magnetic parameters and physicochemical indexes on the mercury concentration in the sediment through the data dimensionality reduction algorithm, and select the magnetic parameters with larger weights as the input factors of the model. The specific method is as follows:

[0035] Use VIM to represent the variable importance score, and use GI to represent the Gini index. Assume there are J features, I decision trees, and C categories. The calculation formula for the Gini index of the q node of the i-th tree is:

[0036]

[0037] Among them, C represents C categories, and P qc represents the proportion of category c in node q.

[0038] Feature X j The importance of feature X at node q of the i-th tree, that is, the change in the Gini index before and after the split of node q is:

[0039]

[0040] Among them, and respectively represent the Gini indices of the two new nodes after the split;

[0041] If the nodes where feature Xj appears in decision tree i are the set Q, then the importance of Xj in the i-th tree is:

[0042]

[0043] Assume there are I trees in RF, then

[0044]

[0045] Finally, normalize all the obtained importance scores

[0046]

[0047] Through importance analysis, relatively important magnetic parameters and physical and chemical indexes can be effectively screened out, improving the prediction accuracy of the model.

[0048] Construction of S400 and MDN artificial intelligence models

[0049] MDN (Mixture Density Network) is composed of a standard neural network and a mixture density model. The parameters of the mixture density model, the mean, variance, and mixture coefficients are output by the neural network.

[0050] Specifically, the screened magnetic parameter data, physical and chemical index data, and mercury concentration data are input into the MDN artificial intelligence model. Based on gradient descent and backpropagation, the initial measurement data is corrected, and the mercury concentration to be measured is predicted through the magnetic parameter data, physical and chemical index data, and the MDN artificial intelligence model.

[0051] Furthermore, step S400 further includes the following steps S410 - S420.

[0052] S410. Take the detection result of the mercury concentration in the surface sediment of the floating rice planting area as the output target, and take the screened magnetic parameter data Wm and physical and chemical index data Wp as input parameters to construct the MDN artificial intelligence model;

[0053] S420. Based on the screened magnetic parameter data and physical and chemical index data, determine the model training data set and the model validation data set, and use the model training data set and the model validation data set to train and validate the MDN artificial intelligence model to obtain the final MDN artificial intelligence model.

[0054] S420 also includes the following steps S421 - S422.

[0055] S421. Division of the data set

[0056] Specifically, randomly divide the screened magnetic parameter data and physical and chemical index data into two groups, with 70% for the training group and 30% for the test group. The magnetic parameters Wm included in the input factors are [χ, χ ARM , SIRM, HIRM, SOFT, χ ARM / χ, χ ARM / SIRM, SIRM / χ], the physical and chemical indexes Wp are [pH, conductivity, LOI], and the output factor is the mercury concentration.

[0057] S422. Training of the loss function

[0058] Specifically, take the physical and chemical index data and magnetic parameter data of the training data set as the input variables of the MDN artificial intelligence model, and the mercury concentration as the output variable. When training the MDN, the selected loss function is the maximum likelihood estimation. The basic idea of maximum likelihood is to make the probability of the observed data appear the largest by adjusting the weights and biases of the model. For a set of independent and identically distributed observed data (D = x 1 , x 2 , …, x n ), the likelihood function of the GMM is the product of the probability density function values (pdf) of each data point in the model. Mathematically expressed as:

[0059]

[0060] Among them, (L) is the likelihood function, (Θ) represents the model parameters. In the case of the MDN, it is the parameters (π_i, μ_i, σ_i) of each Gaussian component, and p(x n |Θ) is the probability density value of (x n ) calculated according to the current parameters.

[0061] In the case of the GMM, p(x n |Θ) is given by the following formula:

[0062]

[0063] Wherein, N(x|μ,σ 2 ) represents the pdf of the Gaussian distribution, and M is the number of Gaussian distributions.

[0064] Since directly maximizing the likelihood function is mathematically intractable, especially when products are involved, in practice, maximizing the logarithm of the likelihood is usually chosen:

[0065]

[0066] Maximizing the log-likelihood is carried out through gradient descent and other optimization algorithms, where the gradients of the parameters are calculated by the backpropagation algorithm. During training, it is necessary to ensure that the mixing coefficients are non-negative and sum to 1, and the variances are positive.

[0067] Step S500: Model performance evaluation

[0068] Specifically, the test set data is used to evaluate the prediction performance of the mercury concentration in the surface sediment of the floating rice planting area of the MDN artificial intelligence model, and further verify the generalization ability of the prediction model of the mercury concentration in the surface sediment of the floating rice planting area, so as to verify the accuracy of the model. The evaluation method used is the Nash efficiency coefficient (NSE), and the specific calculation formula is as follows:

[0069]

[0070] Where n is the number of samples, and y i is the measured value of the mercury concentration in the surface sediment of the floating rice planting area for the i-th sample, is the predicted value of the mercury concentration in the surface sediment of the floating rice planting area for the i-th sample, is the average value of the predicted values of the mercury concentration in the surface sediment of the floating rice planting area for the i-th sample.

[0071] Table 1 is a list of the prediction performance evaluation of the mercury concentration in the surface sediment of the floating rice planting area by the MDN artificial intelligence model of the present invention on the dataset of water quality indicators and magnetic parameters.

[0072] Table 1

[0073]

[0074] The results show that among the first-class factors, the verification R of model 2 with the input factor of Wp is higher, reaching 0.633. However, comparing the training and verification R of the second-class factors, the simulation effects of the two models of the first-class factors are relatively poor.

[0075] S600, Mercury concentration prediction

[0076] The magnetic parameter data and physicochemical index data of the surface sediment samples can be input into the established MDN artificial intelligence model to predict their mercury concentration.

[0077] Table 2

[0078] Model 1 Model 2 Model 3 Input factor Wm Wp Wm, Wp Verify R value 0.272 0.633 0.706

[0079] Table 2 is a list of the prediction accuracy evaluation of the MDN artificial intelligence model of the present invention on the dataset of water quality indicators and magnetic parameters for the mercury concentration in the surface sediment of the floating rice planting area.

[0080] In summary, the rapid detection method for mercury concentration in the surface sediment of the floating rice planting area based on environmental magnetism provided by the present invention has the following technical effects:

[0081] After the model is trained, for the subsequent mercury concentration monitoring of the surface sediment in the floating rice planting area, only the easily obtainable magnetic parameter data and physicochemical index data of the surface sediment need to be detected, and then the data is input into the pre-trained model to obtain the undetected mercury concentration data, so as to achieve the purpose of reducing the detection cost and realizing the rapid detection of mercury concentration.

[0082] The above are only the embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modifications, substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for rapid detection of mercury concentration in surface sediments in floating rice planting areas, characterized in that: include: S1. Obtaining magnetic parameter data of surface sediments in the target floating rice planting area, and obtaining physical and chemical index data of surface sediments in the target floating rice planting area; S2, inputting the magnetic parameter data and physical and chemical index data of the surface sediments of the target floating rice planting area into the MDN artificial intelligence model, and training and constructing the MDN artificial intelligence model; S3. The mercury concentration in the surface sediments of the floating rice planting area to be tested is predicted through magnetic parameters, physical and chemical indicators and the constructed MDN artificial intelligence model.

2. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 1, characterized in that: The magnetic parameter data of the surface sediments in the target floating rice planting area include various magnetic parameters Wm measured by magnetic instruments and calculated, where Wm is [χ, χ ARM 、SIRM、HIRM、SOFT、χ ARM / χ、χ ARM / SIRM, SIRM / χ], Mass magnetic susceptibility χ = a × 10 6 / m, where a is the Carbachol reading at 976 Hz and m is the sample mass; χ ARM It is the mass magnetic susceptibility measured under the conditions of non-hysteresis remanent magnetization ARM applied with a peak alternating magnetic field of 100 mT and a DC bias magnetic field of 0.04 mT; Saturation isothermal remanent magnetism SIRM is the isothermal remanent magnetism IRM measured in a 1T magnetic field: HIRM and SOFT are calculated using the following formula: SOFT = (SIRM - IRM -20mT ) / 2; HIRM=(SIRM+IRM -300mT ) / 2, where IRM -20mT IRM is the isothermal remanent magnetization measured in a -20 mT magnetic field. -300mT is the isothermal remanent magnetization measured in a -300 mT magnetic field.

3. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 1, characterized in that: The mercury concentration data and physical and chemical index data Wp of the surface sediments in the target floating rice planting area are obtained as [pH, EC, LOI, Eh], including using AMA254 mercury analyzer to measure mercury concentration, using PHSJ-4A acidity meter and DDSJ-308A conductivity meter to measure pH value and conductivity EC in turn, using SX-5-12 muffle furnace to measure loss on ignition LOI, and using portable ORP composite electrode meter to measure redox potential Eh.

4. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 1, characterized in that: The obtained initial magnetic parameter data and physical and chemical index data are processed, outliers and missing values ​​are eliminated, feature importance analysis is performed on the obtained magnetic parameters, and magnetic parameters with a large proportion of importance are screened out.

5. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 4, characterized in that: The calculation method for selecting the magnetic parameters with the largest importance is as follows: VIM is used to represent the variable importance score, and GI is used to represent the Gini index. Assuming there are J features, I decision trees, and C categories, the calculation formula for the Gini index of the i-th tree node q is: Among them, C means there are C categories, P qc Indicates the proportion of category c in node q; Feature X j The importance of node q in the i-th tree, that is, the change in the Gini index before and after the node q branches is: in, and Respectively represent the Gini index of the two new nodes after branching; If the node where feature Xj appears in decision tree i is set Q, then the importance of Xj in the i-th tree is: Assuming there are I trees in RF, then: Finally, all the obtained importance scores are normalized.

6. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 4, characterized in that: The construction of the MDN artificial intelligence model includes: S21, using the mercury concentration detection result in the surface sediment of the floating rice planting area as the output target, and using the physical and chemical index data and magnetic parameter data of the surface sediment of the floating rice planting area after screening as input parameters; S22, based on the mercury concentration data in the surface sediments of the floating rice planting area, the screened physical and chemical index data, and the magnetic parameter data, determine the model training data set and the model verification data set; randomly divide the mercury concentration data in the surface sediments of the floating rice planting area, the screened physical and chemical index data, and the magnetic parameter data into a training group and a test group; S23. By presetting the MDN function, the model structure is constructed, and the above-screened physical and chemical index data and magnetic parameter data are used as input variables of the MDN artificial intelligence model. The Adam optimization algorithm is used to automatically adjust the learning rate of each parameter according to the estimated second-order moment of the parameter, and the optimal loss function of the model is selected. The ReLU function is selected as the optimal activation function of the MDN model to construct the final MDN artificial intelligence model; S24. Use the test set data to evaluate the prediction performance of the MDN artificial intelligence model for mercury concentration in surface sediments in floating rice planting areas, further verify the generalization ability of the model for predicting mercury concentration in surface sediments in floating rice planting areas, and verify the accuracy of the model.

7. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 6, characterized in that: In step S23, the optimal loss function of the model is selected as follows: Where (L) is the likelihood function, (Θ) represents the model parameters. In the case of MDN, it is the parameter of each Gaussian component (π i , μ i ,σ i ), where (N(x|μ,σ 2 )) represents the pdf of the Gaussian distribution, M is the number of Gaussian distributions, and during the training process, the mixing coefficients are guaranteed to be non-negative and sum to 1, and the variance is positive, D = x1, x2, ..., x n is a set of independent and identically distributed observations.

8. The method for rapid detection of mercury concentration in surface sediments in floating rice planting areas according to claim 6, characterized in that: In step S24, the evaluation method of the MDN artificial intelligence model for predicting the mercury concentration in the surface sediment of the floating rice planting area is the Nash efficiency coefficient method NSE, and the calculation formula is as follows: Where n is the number of samples, y i is the measured value of mercury concentration in the surface sediment of the floating rice planting area of ​​the i-th sample, is the predicted value of mercury concentration in surface sediments of the floating rice planting area of ​​the i-th sample, is the average of the predicted values ​​of mercury concentration in surface sediments of the floating rice planting area for the ith sample.

Images