KR102266838B1 - Monitoring technique for leaching water of burial sites using field sensors and machine learning - Google Patents

Abstract

The present invention relates to a method for monitoring leachate in burial sites using on-site sensors and machine learning and, more specifically, to a concentration measurement method that simultaneously uses various on-site sensor measurement values and analyzes the same using the deep neural network (DNN) technology, which is a kind of the artificial neural network (ANN) technology to precisely determine NH4-N and Cl- concentrations in groundwater, and to a method for monitoring leachate in burial sites using the same.

Description

Monitoring technique for leaching water of burial sites using field sensors and machine learning

The present invention relates to a method for monitoring leachate in a buried site using in-situ sensors and machine learning. _{Specifically, NH 4} -N and Cl ^- concentrations in groundwater by using multiple field sensor measurements simultaneously and analyzing them using Deep Neural Network (DNN) technology, a type of artificial neural network (ANN) technology. It relates to a concentration measurement method and a method for monitoring leachate in the burial site using the same in order to accurately identify

In Korea, since the first outbreak of foot-and-mouth disease in 1934, a total of five cases of foot-and-mouth disease occurred from March 2000 to April 2010. In addition to foot-and-mouth disease, avian influenza has also been continuously occurring until recently.

From 2010 to 2018, 8 cases of foot-and-mouth disease killed 380,000 cattle and pigs, and 7 cases of avian influenza killed 69 million chickens and ducks. Including 470,000 pigs slaughtered due to the recent African swine fever, 70 million livestock have been culled each year over the past decade.

Most of the slaughtered livestock were disposed of by land filling, and through this, nearly 5,000 burial sites were created nationwide.

Leakage of leachate from burial sites for livestock causes contamination of groundwater, which can be an obstacle to the use of groundwater, as well as a problem in terms of disease control, so continuous monitoring is required.

Continuous groundwater measurement sensors, which are field sensors, can be installed in aquifers to monitor groundwater contamination. When leachate leaks from the burial site, the most sensitive changes in groundwater are ammonia nitrogen (NH ₄ -N) and chloride ion (Cl ^- ) concentrations. _{However, the accuracy of the NH 4} -N and Cl ^- concentrations obtained through the on-site sensor as described above is very low, so it is difficult to apply it to the determination of leachate leakage.

In order to solve the above problems, groundwater can be collected regularly and analyzed in a laboratory, which is a separate place. The above method can obtain high-accuracy results, but since continuous measurement is not possible, there is a gap in groundwater monitoring, a large cost is required for continuous observation, and errors or objectivity problems may occur due to professionalism due to human intervention.

Patent Document 1 relates to an unmanned automatic alkalinity measurement system and method. Patent Document 1 discloses an unmanned automatic alkalinity measurement system, comprising: a sample container for containing a sample to be measured; a water intake unit for receiving the sample to be measured; a test solution input unit for inputting a test solution for measuring the measurement target sample; a controller for storing the data measured by the test solution and controlling the unmanned automatic alkalinity measurement system; a drain unit for draining the sample to be measured from the sample container after the measurement of the sample to be measured is completed; and a dilution unit installed inside the sample container to evenly dilute the sample to be measured and the test solution, wherein the test solution is an acid solution capable of changing the pH of the sample. A measurement system is provided.

Patent Document 1 measures leachate using an unmanned automatic sensor, but does not recognize the accuracy of _{NH 4} -N and Cl ^{- concentrations recognized in the present invention, and simply measures alkalinity.} It does not meet the criteria for determining leakage of acupuncture water.

Patent Document 2 relates to a method for determining the outflow of leachate from a livestock burial site. Patent Document 2 includes the steps of 1) measuring the concentration of a material derived from a livestock carcass from the groundwater of the livestock burial site; 2) measuring the concentration of total organic carbon from the groundwater of step 1); 3) calculating the ratio of the concentration of the animal carcass-derived material to the concentration of total organic carbon; and 4) comparing the ratio of the concentration of the carcass derived material to the concentration of the total organic carbon to the ratio of the concentration of the carcass derived material to the total organic carbon concentration of the leachate inside the burial site. provides a method for determining whether leachate has leaked from livestock burial sites.

Patent Document 2 provides a method for determining a leak, but does not recognize a method for measuring the exact concentration of a material derived from a livestock carcass, which is the basis of the determination, and problems associated therewith.

As such, the prior art _{does not recognize a problem with respect to the accuracy of NH 4} -N and Cl ^- concentration measurement, and thus does not provide a solution thereto. For continuous burial leachate leak monitoring, continuous measurement using on-site sensors is the only solution. First of all, it is necessary to prepare a technology that can accurately determine the concentration _{of NH 4} -N and Cl ^- , which are factors that indicate pollution, from the on-site sensor measurement values that indicate the physical and chemical state of groundwater.

Republic of Korea Patent Publication No. 10-1334307 (2013.11.22) Republic of Korea Patent Publication No. 2012-0111220 (2012.10.10)

An object of the present invention is to solve the above problems, and an object of the present invention is to provide a method for continuously monitoring the leakage of leachate from a livestock burial site.

_{Specifically, 1) accurately calibrate the NH 4} -N and Cl ^- concentrations measured through an on-site sensor in groundwater, which had very low accuracy, 2) perform this in real time, 3) perform the above process unattended, and 4) The purpose of this is to provide a burial site leachate monitoring technique that can determine whether the leachate leaks in real-time and unattended.

In order to solve the above problems, the present invention uses several field sensor measurement values at the same time and analyzes them using Deep Neural Network (DNN) technology, which is a kind of artificial neural network (ANN) technology. In order to accurately determine the concentration of _{NH 4} -N and Cl ^{- in the} The present invention also provides a sensor simplification based on the training, prediction and sensitivity analysis of the model for estimating _{ammonia nitrogen (NH 4} -N) and chlorine ion (Cl ^- ) concentrations using in-situ sensor measurements and deep neural networks.

Specifically

The present invention relates to a livestock burial site and its surrounding groundwater level; water temperature; pH; oxidation-reduction potential (ORP); electric conductivity (EC); total dissolved solids (TDS); a concentration of at least one of cations comprising Ca, Na, K, and Mg; Cl, CO ₃ , SO ₄ Concentration of at least one or more of anions including; a measurement unit including one or more sensors capable of measuring at least one or more from the group comprising; a signal processing unit for processing the result measured by the measuring unit through an artificial neural network; It provides a device for measuring leachate in livestock burial sites including a control unit.

_{The measuring unit may also measure ammonia nitrogen (NH 4} -N) and chlorine ion (Cl ⁻ ) concentrations in the livestock burial site and its surrounding groundwater, and the measuring unit may measure unmanned and real-time.

The artificial neural network is at least one of CNN (Convolution Neural Network), DNN (Deep Neural Network), RNN, DBN, LSTM, GAN, and Softmax models, wherein the response variables are NH ₄ -N and Cl ^- concentrations, and the explanatory variables are It is a value measured by the on-site sensor in the measurement unit. Preferably, the artificial neural network is a CNN or a DNN, preferably a NARX-DNN, and in the case of a CNN, it is composed of a convolutional layer, a fully connected layer, and a softmax in the order, and a max pooling layer exists after the convolutional layer.

The measuring part is As, cyanide, mercury, organophosphorus, phenol, lead, hexavalent chromium Cr ⁺⁶ , trichloroethylene, tetrachloroethyl, 1.1.1-trichloroethane, benzene, toluene, ethylbenzene, xylene, dissolved oxygen , sodium ion, potassium ion, calcium ion, magnesium ion, sulfate ion, bicarbonate ion, carbonate ion, nitrate ion, NO _3- N, at least one or more of the total number of E. coli can also be measured.

In the present invention, the measuring device includes: a data transmitting/receiving unit capable of transmitting and receiving the result measured by the measuring unit by wire or wirelessly according to the request of the control unit; A display unit for displaying data processing information may be added, and the signal processing unit may be located separately from the measuring unit.

The present invention provides a method for determining whether livestock burial site leachate leaks using the burial site leachate measuring device,

a) disposing the measuring unit in and around the livestock burial site to determine the state;

b) receiving at least one measurement signal from the measurement unit;

c) determining whether the livestock burial site and its surroundings are leaking leachate through the artificial neural network in the signal processing unit based on the measured signal;

It also provides a method for determining whether leachate leaks at livestock burial sites, including

The measurement signal of step b) may be transmitted to a signal processing unit located separately through the data transmission/reception unit, processed through the artificial neural network in the signal processing unit, and only the processed result may be notified to the user, and the measurement signal of step b) It may include a step of pre-processing by the control unit before transmission to the signal processing unit, wherein the pre-processing includes noise removal, Fourier, Laplace, Octave Band Levels, sharpness, roughness, envelope, basis size, It is at least one selected from the group including frequency component analysis or time-dependent rate analysis signal processing to extract characteristic values through tonality, fluctuation strength, damping, and natural frequency analysis .

Also, the measurement signal of step b) may include white noise, pink noise, brown noise, and impulse.

In addition, the present invention can transmit the livestock burial site location information to determine the state of step a) together with the measurement signal of step b), and a separate burial site synthesis when notifying the user of the processed result It can also be transmitted to the management system at the same time.

The processed result includes information on the location of the livestock burial site to determine the state, the leachate concentration according to the surrounding location of the burial site, and whether the leachate leaks.

The artificial neural network has already learned about the location information of the livestock burial site to determine the state, the measurement signal, and the state of the burial site and its surroundings, and transfers the measurement signal and the result of the determination of leakage to a separate server computer. The step of building a database for learning by storing it in is added. The measurement result of ammonia nitrogen (NH ₄ -N) and chlorine ion (Cl ^- ) measured separately from the measurement signal is fed back to the signal processing unit so that the learning database is continuously updated, and the updated database is the re-creation of the artificial neural network. used as learning materials.

The present invention solves the problems of the prior art, and can provide continuous and accurate leakage of leachate from livestock burial sites.

_{Specifically, 1) accurately measure ammonium nitrogen (NH 4} -N) and chlorine ion (Cl ^- ) concentrations in groundwater, which had very low accuracy, 2) measure them in real time, 3) perform the above process unattended, 4) Based on this, it is possible to provide a burial site leachate monitoring technique that can determine whether the leachate is leaking in real time and unattended.

The present invention has an additional advantage in that it is possible to select a measurement value having an important influence among the sensors to be measured, and to simplify the sensor to be measured later based on the result by using the result of the artificial neural network while maintaining the accuracy of the system. The method of determining the state of the sewer pipe according to the present invention has the advantage of collecting signals about the flow rate and sediment without affecting the current sewage pipe facility without the need to directly install it in the sewage pipe, such as a flow meter. The acoustic analysis terminal of the present invention has the advantage of very easy operation because it is sufficient to be disposed at both ends of the sewage pipe to determine the state.

The method for determining whether a livestock burial site leachate leaks according to the present invention has the advantage of allowing a signal processing unit or a server to collect signals by transmitting a signal in real time or substantially equivalent to real time in a wired/wireless manner. Data can be collected and transmitted all at once or, if necessary, can be transmitted in real time, and can be transmitted wirelessly through a controller including a sensor, which has the advantage of convenient data management.

The method for determining whether a livestock burial site leachate leaks according to the present invention has the advantage of solving the conventional inaccurate measurement method by using a noise-resistant artificial neural network method rather than a simple pattern analysis method for the transmitted signal.

The method for determining the state of the sewer pipe according to the present invention is at least one of CNN (Convolution Neural Network), DNN (Deep Neural Network), RNN, DBN, LSTM, GAN, and Softmax models, and in the case of CNN, fully connected with the convolutional layer It is possible to actually process the collected big data by using a model composed of layers and softmax in the order of which a max pooling layer exists after the convolutional layer. A result that could not be processed by the regression technique used in the prior art can be provided within a time that can be actually utilized.

In addition, the method for determining whether livestock burial site leachate leaks according to the present invention has the advantage of continuously increasing the reliability of diagnosis through continuous re-learning as well as initial learning. In addition, the accuracy can be further improved by feeding back the analyzed concentration value separately from the current measurement result.

1 is a conceptual diagram of an artificial neural network-based burial groundwater NH ₄ -N and Cl ^{-concentration estimation method according to the present invention.}
2 is a schematic diagram of an artificial intelligence technique-based leachate leak determination model according to the present invention.
3 is an ANN model training network structure.
4 is a NARX-DNN model training network structure.
5 is a detailed exemplary diagram of the NARX-DNN model learning network structure applied to the present invention.
6 is a prediction result of nitrate nitrogen concentration using the deep neural network and in-situ sensor data (water level, EC, and water temperature) according to the first embodiment.
7 is a chlorine ion concentration prediction result using the deep neural network and in-situ sensor data (water level, EC, and water temperature) according to the second embodiment.
8 is a prediction result of nitrate nitrogen concentration using a deep neural network and in-situ sensor data (water level, EC, and water temperature) according to the third embodiment.

Hereinafter, embodiments in which those of ordinary skill in the art to which the present invention pertains can easily practice the present invention will be described in detail with reference to the accompanying drawings. However, when it is determined that a detailed description of a related well-known function or configuration may unnecessarily obscure the gist of the present invention in describing the operating principle of the preferred embodiment of the present invention in detail, the detailed description thereof will be omitted.

In addition, the same reference numerals are used throughout the drawings for parts having similar functions and functions. Throughout the specification, when it is said that a part is connected to another part, it includes not only a case in which it is directly connected, but also a case in which it is indirectly connected with another element interposed therebetween. In addition, the inclusion of a certain component does not exclude other components unless otherwise stated, but means that other components may be further included.

Hereinafter, the present invention will be described in more detail.

1 is a conceptual diagram of an artificial neural network-based burial groundwater NH ₄ -N and Cl ^- concentration estimation method according to the present invention, and FIG. 2 is a schematic diagram of a leachate leak determination model based on an artificial intelligence technique according to the present invention.

1 schematically shows a method for training a _{neural network to estimate NH 4} -N and Cl ^- concentrations only with sensor values by inputting field sensor values and setting laboratory measurements as outputs.

The present invention is a burial site leachate monitoring technique using on-site sensors and machine learning. As observation items of the sensor, ammonia nitrogen (NH ₄ -N), which changes most sensitively in groundwater when leachate leaks from the burial site, occurs; Chloride ions (Cl ⁻ ), and nitrate nitrogen (NO ₃ ⁻ N) were adopted. Ammonia nitrogen (NH ₄ -N) and chlorine ion (Cl ^- ) are preferred.

In addition, in order to improve and correct the reliability of the above three items measured from the sensor during the observation period, periodic groundwater sampling and water analysis are conducted and continuous concentration analysis is performed through ex-situ equipment construction. improved reliability.

In the present invention, the NARX-DNN model was specifically applied as an artificial neural network. 3 and 4 , the NARX-DNN model according to the present invention includes an input layer, a hidden layer, and an output layer, similar to a general artificial neural network (ANN) model. It has a learning network structure composed of Through this, it is possible to model the non-linear relationship between the response variable (water quality concentration at the time of runoff) and the explanatory variable (data on the indirect indication of runoff).

Both the ANN and NARX-DNN models have learning networks of input (x), hidden (h), and output layers (y), whereas the ANN model has one hidden layer, whereas in the NARX-DNN model, the learning network consists of multiple hidden layers. do.

In other words, it is called a deep neural network (DNN) model because the training network of the NARX-DNN model can be expanded compared to the ANN model to account for complex nonlinearities. As the NARX-DNN model according to the present invention learns the pattern of monitoring data based on the extended network, it is advantageous for learning the pattern of an actual natural phenomenon having a non-linear relationship between various external factors.

While the ANN model uses only the external environmental factor (x) as an input factor, the NARX-DNN model has better predictive performance than the ANN model as the past autoregressive data (y) is additionally used as an input factor.

5 is a detailed exemplary diagram of the NARX-DNN model learning network structure applied to the present invention. Referring to Figure 5, the deep neural network of the present invention is a general feed-forward deep neural network model, an input layer, an output layer, and learning consisting of two or more hidden layers. It has a network structure.

By using a deep neural network for NH ₄ -N and Cl ^- concentration estimation, it has the effect of interpreting non-linear relationships that are not possible with existing linear regression techniques. The training of the deep neural network used to calibrate the water quality sensor includes NH ₄ -N and Cl ^- concentrations through regular groundwater sampling and laboratory analysis as output values (y), and groundwater level, water temperature, EC, NO ₃ -N, NH ₄ -N, Cl ^-, etc. are used as input values (x).

Field measurement sensors installed in polluted sites may have various errors due to technical limitations, and when values including large errors are used as input data for deep neural networks, the NH ₄ -N and Cl ^- concentration estimation results have great uncertainty. Therefore, in order to accurately _{estimate the NH 4} -N and Cl ^- concentrations, it is necessary to figure out what input data the uncertainty of the estimation result is from, and to remove the input data with a large contribution to uncertainty.

Variance-based global sensitivity analysis can be used to determine the contribution of uncertainty in input data, and it is desirable to do this before applying artificial neural networks.

In addition, for the spread of on-site sensor-based burial site monitoring methods, sensors with high sensor costs and a relatively large contribution to uncertainty can be excluded.

As a result of applying the present invention specifically to the Anseong area, it was determined that the groundwater level, water temperature, and electrical conductivity (EC) were most suitable considering the uncertainty contribution and sensor cost, and using these as input data, NH ₄ -N and Cl ^- concentration was able to estimate successfully.

Based on these NH ₄ -N and Cl ^- , it is possible to monitor whether the leachate is leaking or not, and use the real-time monitoring technique for the whole country by combining this information and geographic information.

<Example 1 Nitric acid nitrogen (NO ₃ -N) prediction performance>

_{Nitric acid nitrogen (NO 3} -N) prediction was performed using all in-situ observation data. For the structure of the deep neural network for nitrate prediction, 4 hidden layers and 8 neurons (nodes) were used, and a total of 2000 training iterations were used.

In the prediction structure, the Adam optimizer was used to optimize the objective function, and the learning rate was 0.001. It was confirmed that the prediction performance was not high when the entire in-situ observation data was used as in the case of conventional ammonia nitrogen and chlorine ion concentration prediction.

Therefore, in order to block uncertainty and inaccuracy from input factors (in-situ data), a wide-area sensitivity analysis was performed.

As a result of the wide-area sensitivity analysis, the relative sensitivities were 5.186e-06, 2.807e-03, 1.334e-04, 1.247e-02, 2.027e-01 for water level, EC, water temperature, nitrate nitrogen, chlorine ion, and ammonia nitrogen, respectively. , and 7.818e-01.

This means that the uncertainty arising from the water level is the lowest and the uncertainty arising from the concentration of nitrate nitrogen, chlorine ion, and ammonia nitrogen is the greatest, and it is a conclusion consistent with the observation of inaccuracies in the existing in-situ chlorine ion and ammonia nitrogen concentration. .

Therefore, in order to predict the concentration of nitrate nitrogen, a final prediction was performed using a total of three in-situ sensor data, water level, EC, and water temperature as input data.

6 shows a training result and an ensemble prediction result. 6 is a prediction result of nitrate nitrogen concentration using a deep neural network and in-situ sensor data (water level, EC, and water temperature).

As can be seen in FIG. 6 , it can be seen that relatively accurate predictions were made for 10 data not used for nitrate nitric training.

<Example 2 Chlorine ion (Cl ^- ) and ammonia nitrogen (NH ₄ -N) prediction improvement and May 2020 additional data application result>

The work to confirm the predictive performance of the development model for the additionally secured data at the Anseong 1 burial site and the improvement of the existing development model reflecting the additionally secured data were made. The structure of the improved deep neural network for chlorine ion concentration prediction used 4 hidden layers and 6 neurons (nodes), and a total of 2000 training iterations were used.

In the prediction structure, the Adam optimizer was used to optimize the objective function, and the learning rate was 0.001. Unlike the existing model, the final prediction was made using only three in-situ sensor data, water level, EC, and water temperature as input data.

7 is a prediction result according to this. 7 is a result of prediction of chloride ion concentration using a deep neural network and in-situ sensor data (water level, EC, and water temperature).

As shown in FIG. 7, the concentration of chlorine ions is predicted relatively accurately, but the concentration of one chlorine ion in the latter period is located outside the 95% prediction interval. This prediction error is caused by the fact that the concentration is higher than the laboratory data used for training (laboratory measured chlorine ion concentration used for training is 100 mg/l or less). It suggests that the inclusion should make the forecast more sophisticated.

<Example 3>

The structure of the improved deep neural network for predicting the ammonia nitrogen concentration used 4 hidden layers and 6 neurons (nodes), and a total of 2000 training iterations were used. In the above prediction structure, the Adam optimizer is used to optimize the objective function, and the learning rate is 0.001.

As with the previous model, the final prediction was performed using water level, EC, and water temperature, which are only three in-situ sensor data as input data.

The prediction result is as shown in FIG. 8, and it can be confirmed that a stable prediction is made similarly to the existing result. In contrast, the additionally obtained in-situ sensor ammonia nitrogen concentration data shows a very large difference from the laboratory data, so it is judged that there is a problem in the reliability of the sensor data.

Claims (18)

level of groundwater in and around livestock burial sites; water temperature; pH; oxidation-reduction potential (ORP); electric conductivity (EC); total dissolved solids (TDS); a concentration of at least one of cations comprising Ca, Na, K, and Mg; Cl, CO ₃ , SO ₄ Concentration of at least one or more of anions including; a measurement unit including one or more sensors capable of measuring at least one or more from the group comprising;
a signal processing unit for processing the result measured by the measuring unit through an artificial neural network; and
control unit;
A method for determining whether livestock burial site leachate leaks using a livestock burial site leachate measuring device comprising:
a) disposing the measuring unit in and around the livestock burial site to determine the state;
b) receiving at least one measurement signal from the measurement unit;
c) determining whether the livestock burial site and its surroundings are leaking leachate through the artificial neural network in the signal processing unit based on the measured signal;
includes,
Before step c), applying a variance-based global sensitivity analysis method to the measurement signal to remove data having a large contribution to uncertainty from the measurement signal,
The step of removing data having a large contribution to uncertainty is a method of determining whether leachate leaks at livestock burial sites are to remove data whose dispersion-based wide area sensitivity analysis result is 1.247e-02 or higher. According to claim 1,
The method of determining whether the livestock burial site leachate leakage is that the measuring unit also measures _{the NH 4} -N and Cl ^- concentrations of the livestock burial site and its surrounding groundwater. According to claim 1,
The method for determining whether the leachate leak in the livestock burial site is to measure the measurement unit in real time and unattended. According to claim 1,
The artificial neural network is at least one of CNN (Convolution Neural Network), DNN (Deep Neural Network), RNN, DBN, LSTM, and GAN models,
In this case, the response variables are NH ₄ -N and Cl ^- concentrations, and the explanatory variable is a value measured by the measurement unit, a method for determining whether leachate leaks from a livestock burial site. 5. The method of claim 4,
The artificial neural network is a NARX-DNN model, a method for determining whether leachate leaks in livestock burial sites. According to claim 1,
a data transmission/reception unit capable of transmitting and receiving the result measured by the measurement unit by wire or wirelessly according to the request of the control unit;
a display unit for displaying data processing information;
A method for determining whether there is a leak of leachate from a livestock burial site to which a standard has been added. According to claim 1,
The method of determining whether the signal processing unit is located separately from the measurement unit leachate leakage. delete According to claim 1,
A method of determining whether leachate leaks in livestock burial sites by transmitting the measurement signal of step b) to a signal processing unit located separately through the data transmitting and receiving unit, processing the signal through the artificial neural network in the signal processing unit, and then notifying the user of only the processed result. According to claim 1,
Method for determining whether leachate leaks in livestock burial sites, comprising the step of pre-processing in the control unit before transmitting the measurement signal of step b) to the signal processing unit. 11. The method of claim 10,
The pre-processing includes noise removal, Fourier, Laplace, Octave Band Levels, sharpness, roughness, envelope, basis size, tonality, Fluctuation Strength, Damping, Natural Frequency. Frequency) A method for determining whether leachate leaks at least one or more livestock burial sites selected from the group including frequency component analysis to extract characteristic values through analysis or signal processing for time-dependent change rate analysis. According to claim 1,
The measurement signal of step b) is a signal including white noise, pink noise, brown noise, and impulse. A method for determining whether leachate leaks in livestock burial sites. According to claim 1,
A method for determining whether leachate leaks in a livestock burial site by transmitting the location information of the livestock burial site to determine the state of step a) together with the measurement signal of step b). 10. The method of claim 9,
A method for determining whether livestock burial site leachate leaks, which is also transmitted to a separate burial site comprehensive management system when the processed result is notified to the user 10. The method of claim 9,
The method for determining whether leachate leaks at the livestock burial site includes information on the location of the livestock burial site for which the processed result is to be determined, the leachate concentration according to the surrounding location of the burial site, and whether the leachate leaks. According to claim 1,
The artificial neural network is a method for determining whether the livestock burial site leachate leakage, which has already been learned about the location information of the livestock burial site to determine the state, the measurement signal, and the state of the burial site and its surroundings. According to claim 1,
A method for determining whether leachate leaks at livestock burial sites is added by storing the measurement signal and the result of determining whether leakage is present in a separate server computer and building a database for learning. 18. The method of claim 17,
The measurement result of ammonia nitrogen (NH ₄ -N) and chlorine ion (Cl ^- ) measured separately from the measurement signal is fed back to the signal processing unit so that the learning database is continuously updated, and the updated database is the re-creation of the artificial neural network. A method for determining whether a livestock burial site leachate leak is being used as a learning material.

Images