A Method and System for Predicting the Operating Conditions of Downhole Waterproof and Airtight Walls Based on GRNN Neural Network

Abstract

This application relates to the field of safety monitoring in coal mine goaf areas with water accumulation, and particularly to a method and system for predicting the operating conditions of underground waterproof and airtight walls based on a GRNN neural network. The method includes the following steps: sample data acquisition; outlier elimination processing; data standardization processing; data classification; model construction; and prediction result output. The system for predicting the operating conditions of underground waterproof and airtight walls based on a GRNN neural network includes: a data acquisition module; a data anomaly processing module; a data standardization processing module; a model judgment module; and an operating condition confirmation module. The beneficial effects of this invention are that by eliminating and standardizing outlier data, the misleading nature of monitoring data outliers is eliminated, thereby improving the accuracy of waterproof and airtight wall monitoring data; and by improving the accuracy of predicting the operating conditions of various monitoring parameters of waterproof and airtight walls.

Description

Technical Field

[0001] This application relates to the field of safety monitoring of coal mine goaf areas with water accumulation, and in particular to a method and system for predicting the working conditions of underground waterproof and airtight walls based on GRNN neural networks. Background Technology

[0002] With my country's new stage of coal development requirements, the research and application of underground water storage technology in coal mines have been widely promoted. The rational and standardized use of groundwater resources avoids the waste of mine water inflow. At the same time, the scientific storage of groundwater using local resources has effectively improved the production conditions in coal mine production areas in western China that are severely short of water, generating enormous social value.

[0003] To strengthen safety monitoring of waterlogged goaf areas in coal mines, artificial waterproof sealing walls, as crucial waterproofing facilities, are a priority monitoring area. However, most monitoring efforts focus on waterproof sealing construction methods, water source distribution monitoring, and coal pillar damage monitoring. While this has improved the level of digital monitoring of waterproof sealing walls and alleviated the pressure of manual inspections, it cannot effectively predict the operational status of the waterproof sealing walls. In the field of intelligent monitoring of waterproof sealing walls, some research institutions analyze the degree of water erosion damage to the coal pillars on both sides of the waterproof sealing wall by monitoring vibration and current signals, using the principle of differences in vibration signal propagation speed and current magnitude. Other researchers monitor the distribution of groundwater sources using nuclear magnetic resonance (NMR) miniature sensors and drilling cores of surrounding rock, based on NMR porosity characteristics and water signals. It is evident that all of these methods employ non-contact monitoring techniques, have incomplete monitoring parameters, and cannot rule out the possibility of outliers in the monitoring data. Such methods result in poor accuracy of the detection data, leading to a decrease in the accuracy of waterproof sealing condition predictions. Summary of the Invention

[0004] The technical problem to be solved by this invention is: how to improve the accuracy of predicting the working conditions of waterproof and airtight walls.

[0005] Therefore, this invention provides a method and system for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] A method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks includes the following steps:

[0008] Sample data collection: Historical data on gas leakage from waterlogged goaf areas underground, historical data on deformation of sealed walls, historical data on the internal stress of water-retaining coal pillars, and historical data on the seepage pressure of water flow inside coal pillars.

[0009] Outlier removal: Outliers in the sample data are removed to obtain the sample dataset;

[0010] Data standardization: The sample dataset is standardized using the mean and variance method;

[0011] Data classification: The standardized sample dataset is divided into a training set and a test set;

[0012] Model building: The training set is used as the input variable of the GRNN neural network for training. The trained GRNN neural network is then tested using the test set to obtain a prediction model for the working condition of the underground waterproof and airtight wall.

[0013] Prediction result output: The prediction model of the downhole waterproof sealing wall outputs the current working condition of the downhole waterproof sealing wall.

[0014] By adopting the above technical solutions, the misleading nature of outlier monitoring data is eliminated through abnormal data elimination processing. Standardization processes prevent changes in large and small values ​​between different data types due to the addition of new data, eliminating the need for data redefinition and thus improving the accuracy of waterproof and airtight wall monitoring data. Furthermore, by predicting the operating conditions of various monitoring parameters for waterproof and airtight walls, setting different maximum water level limits for waterlogged mining areas, and establishing an autonomous analysis and judgment system, the accuracy of predicting the operating conditions of waterproof and airtight walls is improved.

[0015] Furthermore, in the outlier removal process, the outlier removal method adopts the 3σ principle. In the formula, σ is the standard deviation; n is the total number of samples; X i Let represent one of the sample values; μ is the sample mean. If the probability of the data falling within the interval (μ-3σ, μ+3σ) is 0.9974, then the residual error X of the monitored data values ​​is calculated. b Determine X b =|X i The magnitude between -μ| and 3σ, if X b >3σ means eliminating the outlier.

[0016] Furthermore, in the data standardization process, the formula used to standardize the sample data is as follows: The standardized sample data is denoted as X. k .

[0017] Furthermore, in the data standardization process, the standardized sample X... k Perform cluster analysis.

[0018] Furthermore, the types of cluster analysis are determined according to national regulations on risk levels. After determining the types of clusters, the cluster types are analyzed based on an expert knowledge base to determine the working condition type.

[0019] Furthermore, the dataset after clustering analysis is split into a training set and a test set. The training set is used for sample training, and the test set is used for performance testing of the trained GRNN neural network.

[0020] Furthermore, the ratio of the training set to the test set is 8:2.

[0021] Furthermore, the parameters required for the GRNN neural network include the input variables, the transfer function of the pattern layer, the transfer function of the summation layer, the expansion rate of the radial basis function, and the maximum allowed number of neurons.

[0022] Furthermore, the training process of the GRNN neural network is as follows: the input of the GRNN neural network is X. k X k =[q 1k ....q 4k b 1k ....b 4k , z 1k ....z 4k s 1k ...s 4k ], where q 1k ...q 4k This represents the standardized monitoring values ​​for four types of gases; b 1k ......b 4k This represents the standardized surface strain monitoring value; z 1k ......z 4k The values ​​are the standardized internal stress monitoring values ​​of the coal pillar; s 1k ......s 4k To determine the standardized internal water flow permeability pressure value of the coal pillar, the trained GRNN neural network can accurately identify which cluster the input data belongs to.

[0023] A downhole waterproof sealing wall condition prediction system based on GRNN neural network includes:

[0024] The data acquisition module is used to collect data from the installed multi-gas, strain, stress and seepage pressure sensors to obtain real-time data and historical data related to the operating status of the waterproof and airtight wall.

[0025] The data anomaly handling module is used to eliminate outliers in the acquired big data.

[0026] The data standardization module is used to standardize data that has already been processed for outliers.

[0027] The model determination module is used to set the parameters required for the GRNN neural network, including input variables, pattern layer transfer function, summation layer transfer function, radial basis expansion rate, and maximum allowed number of neurons.

[0028] The operating condition confirmation module is used to identify which cluster the input data belongs to the trained GRNN neural network and issue an early warning based on the identification results.

[0029] The beneficial effects of this invention are as follows: It uses data from multiple sensors as the basis for determining the operating conditions of waterproof and airtight walls, resulting in more comprehensive and stable data signal monitoring. By eliminating abnormal data, it removes misleading values ​​from the monitoring data and standardizes the data, preventing changes in large and small values ​​due to the addition of new data, thus eliminating the need for data redefinition and improving the accuracy of waterproof and airtight wall monitoring. Furthermore, by predicting the operating conditions of various monitoring parameters of the waterproof and airtight wall, it sets different maximum water level limits for waterlogged goaf areas and establishes an autonomous analysis and judgment system. This system adaptively provides water level safety warnings under different operating conditions, prompting for drainage requirements in goaf areas and achieving a safe water level in waterlogged goaf areas. Attached Figure Description

[0030] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0031] Figure 1 This is a schematic diagram of the working condition prediction system for the underground waterproof and airtight wall of the present invention.

[0032] Figure 2 This is a technical roadmap for the method of predicting the working conditions of a waterproof and airtight wall in wells, as invented in this invention.

[0033] Figure 3 This is a flowchart of the GRNN neural network training process in this invention.

[0034] Figure 4 This is a flowchart for confirming the maintenance results of the waterproof and airtight wall in this invention. Detailed Implementation

[0035] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0036] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the 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 of the invention. Furthermore, features defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0037] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0038] Reference Figure 1 A downhole waterproof and airtight wall condition prediction system based on GRNN neural network includes a data acquisition module, a data anomaly processing module, a data standardization processing module, a model judgment module, and a condition confirmation module.

[0039] Reference Figure 1 The data acquisition module is used to collect data from the installed multi-gas, strain, stress and seepage pressure sensors to obtain real-time data on the operating status of the waterproof and airtight wall.

[0040] Reference Figure 1 The data acquisition module is equipped with a waterproof and airtight wall monitoring system, which includes a multi-sensor data acquisition module, including multi-gas sensors (CO, CO2, CH4, O2), surface strain sensors, borehole stress sensors, and seepage pressure sensors.

[0041] Reference Figure 1Multiple gas sensors (CO, CO2, CH4, O2) are installed at the top of the sealed wall to monitor gas leakage from the waterlogged goaf. Surface strain sensors are installed at the four corners of the sealed wall to monitor its deformation. Borehole stress sensors are installed on both sides of the sealed wall at coal pillars 2m from the floor to monitor the internal stress of the water-retaining coal pillars. Pressure sensors are installed on both sides of the sealed wall at coal pillars 0.5m and 1m from the floor to monitor the pressure of water seepage inside the coal pillars. After the sensors are deployed underground, their power lines are connected. The signal lines of the multiple gas sensors are connected to the KX18(D) intrinsically safe mine control box; the signal lines of the surface strain, borehole stress, and pressure sensors are connected to the YGSJ12(C) fiber optic demodulator; and the signals from the KX18(D) intrinsically safe mine control box and the YGSJ12(C) fiber optic demodulator are connected to the KJJ32(A) switch. After all the collected data signals reach the switch, the switch will upload these signals to the monitoring system. Finally, the monitoring system will complete the prediction of the data analysis results.

[0042] Reference Figure 1 In this embodiment, to simulate the working state of the waterproof sealing wall as closely as possible, the entire sealing wall was monitored for a period of one month, ultimately collecting a dataset of 1440 samples. Each sample contains information such as the current monitoring gas data, sealing wall deformation data, and internal stress and seepage pressure data of the coal pillar. The collected monitoring data is sufficient to fully characterize various operating conditions of the underground artificial sealing wall.

[0043] The data anomaly handling module is used to eliminate outliers in the acquired large dataset. During the acquisition of the various monitoring parameters, outliers may appear in the monitoring data due to equipment malfunctions and external factors specific to the underground environment. Outliers refer to data that deviates from the normal monitoring range. In actual data analysis, ignoring outliers can lead to incorrect analysis results and reduce the accuracy of the trained neural network. Therefore, it is necessary to filter the original monitoring data to eliminate outliers.

[0044] Outlier data elimination methods employ the 3σ principle. In the formula, σ is the standard deviation; n is the total number of samples; X i This represents one of the sample values; μ is the sample mean. According to the mathematical interpretation of the 3σ principle, if the data is distributed in the interval (μ-3σ, μ+3σ), the probability is 0.9974. This means that the probability of data exceeding this interval is less than 0.3%, which is the residual error X for calculating the monitored data values. b Determine X b =|X i The magnitude between -μ| and 3σ, if X b>3σ means eliminating the outlier.

[0045] The data standardization module is used to standardize data that has already been processed for outliers. The formula used for standardizing sample data is as follows: In the formula, σ is the standard deviation; X i Represents one of the sample values; μ is the sample mean, and the standardized sample is denoted as X. k .

[0046] The model evaluation module is used to set the parameters required for the GRNN neural network, including input variables, pattern layer transfer function, summation layer transfer function, radial basis function (RBF) propagation rate, and the maximum allowed number of neurons. Input variables include: gas leaking from the goaf at the joint between the waterproof sealing wall and the coal pillar; deformation of the waterproof sealing wall under multiple combined forces; and the internal stress and seepage pressure of the coal pillars on both sides of the waterproof sealing wall. Output variables represent values ​​for different waterproof sealing wall operating conditions. Based on the actual underground environment requirements, the target error, RBF propagation rate, and maximum allowed number of neurons are set for training.

[0047] The working condition confirmation module is used to identify which cluster the input data belongs to by the trained GRNN neural network, which means it can identify which type of waterproof and airtight wall working condition it belongs to. The working condition confirmation module sets different maximum water level limits for each type of waterproof and airtight wall working condition, so that it can adaptively provide water level safety prompts when encountering different waterproof and airtight strength levels, prompting the drainage requirements of the goaf area, and achieving the effect of safe water level in the water-filled goaf area.

[0048] Reference Figure 2 A method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks includes the following steps:

[0049] Sample Data Acquisition: Outliers in the sample data are eliminated to obtain a sample dataset. Through a multi-sensor data acquisition module, historical data on gas samples leaking from the underground water-filled goaf are collected from multiple gas sensors (CO, CO2, CH4, O2), surface strain sensors, borehole stress sensors, and seepage pressure sensors. Historical data on the deformation of the sealed wall, the internal stress of the water-retaining coal pillar, and the seepage pressure of water flow within the coal pillar are also collected. In this embodiment, the entire sealed wall was monitored for a period of one month. Each sample includes the current monitoring gas data, sealed wall deformation data, and internal stress and seepage pressure data of the coal pillar.

[0050] Outlier removal: Let X i For a sample at a certain time, 0≤i≤1440, X k =q 1k……q 4k b 1k ...b 4k , z 1k ...z 4k s 1k ...s 4k ], where q 1k ……q 4k This represents the standardized monitoring values ​​for four types of gases; b 1k ...b 4k This represents the standardized surface strain monitoring value; z 1k ...z 4k The values ​​are the standardized internal stress monitoring values ​​of the coal pillar; s 1k ...s 4k This represents the standardized internal water seepage pressure value of the coal pillar. Outlier elimination was performed on the monitoring values ​​for all the above time periods. If the residual error of any monitoring data value in a given time period is greater than 3σ, the monitoring data is considered an outlier, eliminated, and the entire time period and its corresponding value (X) are deleted. i sample.

[0051] Outlier data elimination methods, using the 3σ principle, are as follows:

[0052] According to the mathematical interpretation of the 3σ principle, the probability of data falling within the interval (μ-3σ, μ+3σ) is 0.9974. This means that the probability of data falling outside this interval is less than 0.3%, which is the residual error X in calculating the monitored data values. b Determine X b =|X i The magnitude between -μ| and 3σ, if X b >3σ means eliminating the outlier.

[0053] Data standardization: using the mean and variance method The sample dataset is standardized, and the standardized sample is denoted as X. k .

[0054] Data classification: The specific process of dividing the standardized sample dataset into training and test sets is as follows: First, cluster analysis is performed on the standardized dataset. The types of clusters are determined according to the national risk level regulations. Based on the national risk level regulations, the strength levels of waterproof and airtight walls are divided into C, B, and A levels from low to high.

[0055] Due to variations in construction materials and strength grades of underground waterproof sealing walls in coal mines, the internal stress changes caused by gas leakage and water seepage also differ across mining areas. Therefore, even with the same monitoring parameters, the classification of monitoring data and the determination of operating conditions for waterproof sealing walls of different strength grades should be tailored to their specific characteristics. The same analytical logic cannot be applied to waterproof sealing walls of different strength grades. Therefore, after determining the risk level, cluster analysis is performed based on the expert knowledge base to identify the specific operating condition. These operating conditions are categorized as follows: hazardous gas concentration environment, waterproof sealing wall deformation, internal stress of coal pillars, and internal seepage pressure of coal pillars. The expert knowledge base is a core component of the expert system, primarily storing and managing knowledge acquired from books and the experience gained by experts in various fields through long-term practical work.

[0056] Based on the specific conditions of the underground waterproof sealing wall in different mining areas, danger thresholds are set for multiple gas sensors (CO, CO2, CH4, O2), surface strain sensors, borehole stress sensors, and seepage pressure sensors. When judging the working conditions, for example, if the gas sensors (CO, CO2, CH4, O2), surface strain sensors, and borehole stress sensors do not exceed their corresponding danger thresholds, but the seepage pressure sensor readings exceed the danger thresholds, then the artificial waterproof sealing wall in the underground environment is determined to be in a coal pillar seepage pressure condition.

[0057] The dataset from the clustering analysis is split into a training set and a test set, with a ratio of 8:2. The training set is used for sample training, while the test set is used to test the performance of the GRNN neural network trained on the training set.

[0058] Reference Figure 3 and Figure 4 Model construction: Setting the parameters required for the GRNN neural network, including input variables, pattern layer transfer function, summation layer transfer function, radial basis expansion rate, and maximum allowed number of neurons. Input variables include the following: gas leaking from the goaf at the joint between the waterproof sealing wall and the coal pillar, deformation of the waterproof sealing wall under multiple combined forces, and the internal stress and seepage pressure of the coal pillars on both sides of the waterproof sealing wall.

[0059] The mode layer transfer function is Summation layer transfer function The output variables represent values ​​for different waterproof and airtight wall conditions. Based on the actual downhole environment requirements, the training target error, radial basis function propagation rate, and maximum allowable number of neurons are set.

[0060] The training process of the GRNN neural network is as follows: The input of the GRNN neural network is X. kX k =q 1k ……q 4k b 1k ...b 4k , z 1k ...z 4k s 1k ...s 4k ], where q 1k ……q 4k This represents the standardized monitoring values ​​for four types of gases; b 1k ...b 4k This represents the standardized surface strain monitoring value; z 1k ...z 4k The values ​​are the standardized internal stress monitoring values ​​of the coal pillar; s 1k ...s 4k The standardized internal water flow permeation pressure value of the coal pillar is used for training according to the parameters of the designed model layer transfer function, summation layer transfer function, etc.

[0061] Prediction output: The trained GRNN neural network is a prediction model for the underground waterproof and airtight wall working condition. The prediction model for the underground waterproof and airtight wall working condition can accurately identify which cluster the input data belongs to, which means it can identify which type of waterproof and airtight wall working condition it belongs to.

[0062] The system inputs real-time parameters, including leaked gas from the goaf at the joint between the waterproof sealing wall and the coal pillar, deformation of the waterproof sealing wall under multiple combined forces, and internal stress and seepage pressure of the coal pillars on both sides of the waterproof sealing wall, after outlier and standardization processing. This data is used to predict the current operating conditions: hazardous gas concentration environment, waterproof sealing wall deformation, internal stress of the coal pillar, and internal seepage pressure of the coal pillar. Based on one of the concrete dam safety monitoring technical standards, a GRNN neural network trained with sufficient samples can predict these waterproof sealing wall strength parameters. Different maximum water level limits are set for each waterproof sealing wall operating condition, allowing the system to adaptively provide water level safety warnings under different waterproof sealing strength levels, prompting for goaf drainage and achieving a safe water level in the goaf. This patent only predicts the waterproof sealing wall operating conditions and does not currently cover the subsequent drainage pump station control.

[0063] In summary, this invention uses data from multiple sensors as the basis for determining the operating conditions of waterproof and airtight walls, resulting in more comprehensive and stable monitoring of data signals. By eliminating outlier data, it avoids misleading information from abnormal values ​​and standardizes the data to prevent changes in large and small values ​​due to the addition of new data, eliminating the need for data redefinition. By predicting the operating conditions of various monitoring parameters of the waterproof and airtight wall, different maximum water level limits are set for different waterlogged goaf areas. An autonomous analysis and judgment system is established, which adaptively provides water level safety warnings under different operating conditions, prompting for drainage requirements in the goaf area to achieve a safe water level in the waterlogged goaf area.

[0064] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined by the scope of the claims.

Claims

1. A method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks, characterized in that, Includes the following steps, Sample data collection: Historical data on gas leakage from waterlogged goaf areas underground, historical data on deformation of sealed walls, historical data on the internal stress of water-retaining coal pillars, and historical data on the seepage pressure of water flow inside coal pillars. Outlier removal: Outliers in the sample data are removed to obtain the sample dataset; Data standardization: The sample dataset is standardized using the mean and variance method; Data classification: The standardized sample dataset is divided into a training set and a test set; Model building: The training set is used as the input variable of the GRNN neural network for training. The trained GRNN neural network is then tested using the test set to obtain a prediction model for the working condition of the underground waterproof and airtight wall. Prediction result output: The prediction model of the downhole waterproof sealing wall outputs the current working condition of the downhole waterproof sealing wall; In the data standardization process, the standardized samples... X k Perform cluster analysis; The types of cluster analysis are determined based on national regulations on risk levels. After determining the types of clusters, the cluster types are analyzed based on an expert knowledge base to determine the working condition type.

2. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 1, characterized in that, In the outlier removal process, the outlier removal method adopts the 3σ principle. In the formula, σ is the standard deviation; n is the total number of samples; X i This represents one of the sample values; The sample mean is given if the data is distributed in... The interval probability is 0.9974, which is the residual error in calculating the monitoring data values. X b ,determination The magnitude between and 3σ, if X b >3σ means eliminating the outlier.

3. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 2, characterized in that, In the data standardization process, the formula used to standardize the sample data is: The standardized sample data is denoted as X k .

4. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 1, characterized in that, The dataset after clustering analysis is split into a training set and a test set. The training set is used for sample training, and the test set is used for performance testing of the trained GRNN neural network.

5. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 4, characterized in that, The ratio of the training set to the test set is 8:

2.

6. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 1, characterized in that, The parameters required for the GRNN neural network include input variables, pattern layer transfer function, summation layer transfer function, radial basis expansion rate, and maximum allowed number of neurons.

7. The method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks according to claim 6, characterized in that, The training process of the GRNN neural network is as follows: The input of the GRNN neural network is... X k , ,in, These represent the standardized monitoring values ​​for four types of gases. This represents the standardized surface strain monitoring value. These are the standardized stress monitoring values ​​inside the coal pillar; To determine the standardized internal water flow permeability pressure value of the coal pillar, the trained GRNN neural network can accurately identify which cluster the input data belongs to.

8. A prediction system based on the method for predicting the working conditions of downhole waterproof and airtight walls based on GRNN neural networks as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect data from the installed multi-gas, strain, stress and seepage pressure sensors to obtain real-time data and historical data related to the operating status of the waterproof and airtight wall. The data anomaly handling module is used to eliminate outliers in the acquired big data. The data standardization module is used to standardize data that has already been processed for outliers. The model determination module is used to set the parameters required for the GRNN neural network, including input variables, pattern layer transfer function, summation layer transfer function, radial basis expansion rate, and maximum allowed number of neurons. The operating condition confirmation module is used to identify which cluster the input data belongs to the trained GRNN neural network and issue an early warning based on the identification results.

Images