Coal mine compliance monitoring model optimization method and system based on data cycle feedback

Abstract

The invention discloses a coal mine compliance monitoring model optimization method and system based on data cycle feedback, and relates to the technical field of coal mine production monitoring, and the method comprises the steps: constructing an initial compliance monitoring model; collecting and preprocessing detection data in a coal mine production process in real time; according to the preprocessed detection data, optimizing the initial compliance monitoring model based on a data cycle feedback mechanism to obtain an optimization model; and judging whether the coal mine production is compliant or not based on the optimization model. According to the method, the detection data is preprocessed, the accuracy of the data is improved, meanwhile, the compliance monitoring model is continuously updated and optimized according to the real-time data, the accurate compliance monitoring model is obtained, and the accuracy of judging whether the coal mine production is compliant or not is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine production monitoring, and specifically to an optimization method and system for a coal mine compliance monitoring model based on data loop feedback. Background Art

[0002] Currently, during the coal mine production monitoring process, it is necessary to continuously update and optimize the corresponding compliance monitoring model according to real-time data. Otherwise, it will lead to errors in the results of determining whether the coal mine production is compliant. At the same time, as the production process of coal mine enterprises continues to advance, the data timeliness is relatively high, the data differences between enterprises are large, and the data quality is uneven, and there are various defects in the platform detection data itself.

[0003] In addition, during the process of data pushing, transmission, and reception, the data is easily affected by noise data, data value loss, data conflicts, etc. Therefore, it is necessary to preprocess the collected large data set to ensure the accuracy and value of the big data analysis and prediction results. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to provide an optimization method and system for a coal mine compliance monitoring model based on data loop feedback, preprocess the detection data, improve the accuracy of the data, and at the same time continuously update and optimize the compliance monitoring model according to real-time data to obtain an accurate compliance monitoring model and improve the accuracy of judging whether the coal mine production is compliant.

[0006] To solve the above technical problem, the present invention provides the following technical solution: An optimization method for a coal mine compliance monitoring model based on data loop feedback, including: constructing an initial compliance monitoring model; collecting and preprocessing the detection data during the coal mine production process in real time; optimizing the initial compliance monitoring model based on the data loop feedback mechanism according to the preprocessed detection data to obtain an optimized model; and judging whether the coal mine production is compliant based on the optimized model.

[0007] As a preferred solution of the method for optimizing the coal mine compliance monitoring model based on data loop feedback of the present invention, wherein: the process of establishing the initial compliance monitoring model includes the following steps: setting a residual module integrating a multi-head attention mechanism; setting a data segmentation rule; setting a convolutional layer, and performing a convolution operation on the initial data processed according to the data segmentation rule through the convolutional layer to obtain N data blocks, adding a position encoding to each data block to obtain a data block sequence; inputting the data block sequence into a pre-constructed encoder for data processing, and measuring the gap between the predicted result and the actual result corresponding to the initial data based on a predefined self-supervised loss function during the data processing, and updating the parameters of the initial compliance monitoring model to obtain the initial compliance monitoring model.

[0008] As a preferred solution of the method for optimizing the coal mine compliance monitoring model based on data loop feedback of the present invention, wherein: the construction process of the encoder includes the following steps: determining the type and dimension of the input data; determining the reconstruction target; based on a preset data table, and according to the type and dimension of the input data and the reconstruction target, selecting encoder parameters to obtain an initial encoder;

[0009] Training the initial encoder, continuously adjusting the network parameters through forward propagation and backward propagation, monitoring the loss value and reconstruction error during the training process, and making the network converge. When the convergence value is less than a preset threshold, an encoder is obtained.

[0010] As a preferred solution of the method for optimizing the coal mine compliance monitoring model based on data loop feedback of the present invention, wherein: the detection data includes production volume, power consumption, and personnel activity conditions; the preprocessing includes removing duplicate values, processing missing values, processing outliers, data smoothing, and data merging processing.

[0011] As a preferred solution of the method for optimizing the coal mine compliance monitoring model based on data loop feedback of the present invention, wherein: the data merging processing includes the following steps: extracting features from each sub-data in the detection data, and numerically processing the feature indicators to obtain index values; the feature indicators include data keywords, data size, data recording time, data type, and data source address; arbitrarily selecting two sub-data and calculating the probability that the two sub-data can be merged;

[0012]

[0013] wherein, P i,j is the probability that the i-th sub-data and the j-th sub-data can be merged; T i,a is the a-th index value of the i-th sub-data; T j,a is the a-th index value of the j-th sub-data; max(T a) is the maximum value of the a-th index value among several sub-data in the detection data; two sub-data with probabilities greater than the preset probability threshold are merged to obtain the detection data after data merging processing.

[0014] As a preferred solution of the coal mine compliance monitoring model optimization method based on data loop feedback according to the present invention, wherein: the optimized model is compared with the initial compliance monitoring model to obtain a comparison result and display it.

[0015] As a preferred solution of the coal mine compliance monitoring model optimization method based on data loop feedback according to the present invention, wherein: comparing the optimized model with the initial compliance monitoring model includes the following steps: obtaining the first model parameters constituting the optimized model and the second model parameters constituting the initial compliance monitoring model; comparing the model parameters belonging to the same category among the first model parameters and the second model parameters to obtain a comparison result.

[0016] To further solve the above technical problems, the present invention provides the following technical solution: a coal mine compliance monitoring model optimization system based on data loop feedback, including: a construction module for constructing an initial compliance monitoring model; a collection and processing module for real-time collecting and preprocessing the detection data in the coal mine production process; an optimization module for optimizing the initial compliance monitoring model based on the data loop feedback mechanism according to the preprocessed detection data to obtain an optimized model; a judgment module for judging whether the coal mine production is compliant based on the optimized model.

[0017] A computer device includes a memory and a processor, the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the coal mine compliance monitoring model optimization method based on data loop feedback as described above are implemented.

[0018] A computer-readable storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, the steps of the coal mine compliance monitoring model optimization method based on data loop feedback as described above are implemented.

[0019] Advantages of the present invention: The present invention provides an optimization method and system for a coal mine compliance monitoring model based on data loop feedback. By constructing an initial compliance monitoring model, detecting data in the coal mine production process is collected and preprocessed in real time, and then the initial model is optimized based on the data loop feedback mechanism. Finally, the optimized model is used to determine whether the coal mine production is compliant. This method preprocesses the detection data, improves the accuracy of the data, and continuously updates and optimizes the compliance monitoring model according to the real-time data, thereby improving the accuracy of determining whether the coal mine production is compliant. Through the data loop feedback mechanism, the system can continuously self-improve and optimize, quickly respond to changes, adapt to the new environment, and continuously improve the competitiveness and value of the monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0021] Figure 1 It is the overall flowchart of the optimization method for the coal mine compliance monitoring model based on data loop feedback provided by an embodiment of the present invention.

[0022] Figure 2 It is the flowchart of constructing the initial compliance monitoring model of the optimization method for the coal mine compliance monitoring model based on data loop feedback of the present invention.

[0023] Figure 3 It is the block diagram of the optimization system of the optimization method for the coal mine compliance monitoring model based on data loop feedback of the present invention.

[0024] Figure 4 It is the schematic diagram of the monitoring and early warning model for suspected production suspension and illegal mining behavior of coal mine enterprises of the optimization method for the coal mine compliance monitoring model based on data loop feedback of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0027] Example 1. Referring to Figures 1 to 4 , an embodiment of the present invention provides an optimization method for a coal mine compliance monitoring model based on data loop feedback, including:

[0028] S1: Construct an initial compliance monitoring model.

[0029] It should be noted that the initial compliance monitoring model, as a basic framework, is used to evaluate the compliance in the coal mine production process. The initial compliance monitoring model is generated based on the data obtained within a cycle and is relatively accurate in judging whether the coal mine is compliant in the current cycle. However, with changes in the production process, production scale, etc., the initial compliance monitoring model needs to be continuously optimized and updated.

[0030] The initial compliance monitoring model refers to a convolutional neural network model.

[0031] The initial compliance monitoring model includes a coal mine enterprise power consumption classification prediction calculation model. Through the joint analysis of real-time monitoring data and a large amount of historical data of the coal mine enterprise's power monitoring data, personnel positioning data, and safety monitoring data, and based on the analysis of data fluctuations and changes in the time dimension, it evaluates the coal mine's excavation activity time interval, operation range, etc., and calculates the normal fluctuation range and predicted average value of the power consumption of the coal mine enterprise's excavation operations, the power consumption of other production support systems, and the basic living power consumption, providing basic data support for the monitoring and early warning of coal mine production behavior.

[0032] An application scenario example is as follows:

[0033] Evaluation and calculation of the average power consumption of coal mine excavation operations: The number of excavation working faces and operation time of the coal mine can be jointly determined through the changes in the data of personnel going up and down the well, the changes in the number of personnel at the working face, and the real-time changes in the mine power consumption data. Through the clustering analysis and correlation analysis of a large amount of data, and circularly correcting the calculation results, the normal fluctuation range and evaluation mean value are determined.

[0034] Evaluation and calculation of the basic living power consumption: The power consumption with the lowest power and stable fluctuations for a long time in a year is regarded as the power consumption during coal mine shutdown and production suspension. Among them, shutdown refers to coal mine enterprises that are shut down, and production suspension refers to coal mine enterprises with normal production.

[0035] Evaluation and calculation of the power consumption of other production support systems: The power consumption of the production support system is obtained by subtracting the basic living power consumption from the total power consumption of the coal mine.

[0036] The initial compliance monitoring model includes a monitoring and early warning model for coal mining enterprises' illegal production beyond capacity and intensity. Based on the result data generated by the classification and estimation calculation model of coal mining enterprises' power consumption, the enterprises' historical power monitoring data, the power monitoring data of coal mines in the same region, the personnel positioning monitoring data, and the water hazard monitoring data, the comparative analysis and year-on-year analysis methods are used to dynamically capture the abnormal power consumption of the enterprises. Cluster analysis is used to group coal mines from the spatial dimension and the enterprise similarity dimension of regional coal mines, and the power consumption data of coal mines is compared horizontally to judge whether there is any behavior of production beyond capacity in the coal mines.

[0037] The application scenarios are illustrated as follows:

[0038] When the power consumption data of a coal mining enterprise surges both year-on-year and month-on-month, the number of people going down the mine increases year-on-year, and the production capacity power consumption ratio (total power consumption / monthly approved production capacity) is significantly abnormal compared with that of coal mines in the same region, it is determined that there is a suspected illegal behavior of production beyond capacity.

[0039] Based on the personnel location information, personnel job types information, the number of working faces, etc., judge whether there are the following major hidden dangers of over-intense production in the coal mine:

[0040] ① The number of levels in simultaneous production underground in the coal mine exceeds 2, or the number of coal mining and coal (semi-measurement rock) roadway driving faces operating simultaneously in a mining (panel) area exceeds the provisions of the "Coal Mine Safety Regulations";

[0041] ② The coal mine fails to formulate or strictly implement the underground labor quota system, or the number of people working in a single shift at the excavation and mining operation sites exceeds 20% of the relevant national staffing limits.

[0042] Based on the personnel positioning data of a large number of coal mining enterprises, cluster analysis is carried out on the concentrated areas of personnel activities. Combining the names of personnel sub-stations and personnel job types information, judge the suspected excavation and mining operation sites. Through the correlation analysis of the changes in the number of people at the suspected excavation and mining operation sites and the changes in power consumption data, the excavation and mining operation sites of the coal mine can be further determined. When the excavation and mining operation sites exceed the requirements of the regulations and specifications, an early warning information of over-intense illegal production is generated. When there is an over-limit alarm at the excavation and mining operation sites and the number of overstaffed people is more than 20% of the relevant national staffing limits, an early warning information of over-intense illegal production is generated.

[0043] As Figure 4 shown, the initial compliance monitoring model includes a monitoring and early warning model for suspected illegal behaviors of production during suspension in coal mining enterprises. By monitoring the power consumption monitoring data of suspended coal mining enterprises, the normal fluctuation range of the basic guarantee power consumption of the coal mine is analyzed and calculated. When the power consumption of the coal mine exceeds a certain threshold of the normal range, it is determined that the coal mining enterprise is suspected of having illegal behaviors of production during suspension.

[0044] The application scenarios are illustrated as follows: Monitoring of illegal behaviors of production during suspension in coal mining enterprises.

[0045] Integrate existing data sources such as sensors, monitoring systems, and production management systems in coal mines to collect detection data, and use real-time data transmission technologies (such as MQTT, WebSocket) to ensure that the data can be quickly transmitted to the data processing center. The detection data includes enterprise power monitoring data, basic data, personnel positioning data, safety monitoring data, hydrogeological monitoring data, major equipment monitoring data, and spatial geographic information data. Preprocess the detection data to ensure data quality. Optimize the initial compliance monitoring model based on the preprocessed data according to the data loop feedback mechanism to obtain an optimized model; the core of the data loop feedback mechanism lies in its iterative nature. By continuously collecting data, analyzing data, generating feedback, and applying feedback, the system can continuously improve and optimize itself. This iterative process helps to quickly respond to changes, adapt to new environments, and continuously improve the competitiveness and value of the system. Among them, the collected data is the preprocessed data obtained. Various methods can be used for data analysis, such as statistical analysis, data mining, machine learning, etc., to discover patterns and trends in the data. Based on the results of data analysis, corresponding feedback information is generated. The feedback information is an evaluation and suggestion regarding system performance, user behavior, monitoring business processes, etc. These feedback information facilitate the identification of potential problems and determine the direction of improvement. Apply the feedback and optimize according to the improvement direction. The optimization process includes: Parameter adjustment: Adjust the model parameters to improve performance. Algorithm improvement: Introduce more advanced machine learning or deep learning algorithms. Feature engineering: Add or modify input features to improve the model effect. Iterative loop: Repeat the evaluation, feedback, and optimization steps until the model performance reaches a satisfactory level. Continuously improve the accuracy and adaptability of the model through continuous learning, deploy the optimized model to the production environment, and monitor the real-time data. Judge whether the production activities are compliant according to the model output. Once a violation is detected, immediately trigger the alarm mechanism and notify the relevant personnel to take corresponding measures. Record all compliance judgment results and regularly generate compliance reports for management reference.

[0046] Beneficial effects of the above technical solution: Preprocess the detection data to improve the accuracy of the data. At the same time, continuously update and optimize the compliance monitoring model according to the real-time data to obtain an accurate compliance monitoring model, and improve the accuracy of judging whether coal mine production is compliant.

[0047] Furthermore, the process of establishing the initial compliance monitoring model includes the following steps:

[0048] S101: Set a residual module with a fused multi-head attention mechanism;

[0049] S102: Set the data segmentation rule;

[0050] S103: Set up a convolutional layer. Perform a convolution operation on the initial data processed based on the data segmentation rule through the convolutional layer to obtain N data blocks, and add position encoding to each data block to obtain a data block sequence;

[0051] S104: Input the data block sequence into a pre-constructed encoder for data processing. During the data processing, measure the gap between the predicted result and the actual result corresponding to the initial data based on a pre-defined self-supervised loss function, and update the parameters of the initial compliance monitoring model to obtain the initial compliance monitoring model.

[0052] The working principle and beneficial effects of the above technical solution: Multi-Head Attention: Used to learn information in multiple different representation subspaces of the sequence. Residual Block: Used to prevent the problem of gradient disappearance in deep networks and is achieved by adding skip connections. The residual block is expressed as: F(x) = H(x) + x, where F(x) represents the output of the residual block, H(x) represents the transformation of the input x inside the residual block (including multiple convolutional layers and non-linear activation functions), and x is the input directly passed through the skip connection. Based on the residual block, the gradient propagation is smoother, the network is deeper, and the number of parameters is smaller. Combining method: In each Transformer encoder layer, first apply the multi-head attention mechanism. Then, add the output of the multi-head attention to the input of this layer through a residual connection. Finally, stabilize the training process through Layer Normalization.

[0053] The data segmentation rule refers to segmenting according to specific attributes of the data (such as type, size, access frequency, etc.) to meet different data processing requirements. For example, data of the same type is used as a segmentation set, data with the same access frequency is used as a segmentation set, or segmentation is performed according to a data table generated based on size, and data of different sizes are used as different segmentation sets.

[0054] Purpose of setting data segmentation rules: To segment the input data into smaller chunks (or sequences) for processing long sequences or high-dimensional data. This includes: Segmenting data according to the dimensions of the data (such as time, space) or a fixed-size window. Ensure that each data chunk contains sufficient information for meaningful processing. Set up convolutional layers to extract local features from the data chunks. Use one-dimensional, two-dimensional, or three-dimensional convolutional layers, depending on the dimension of the data. Apply appropriate padding and stride to ensure that the output size meets expectations. Activation functions (such as ReLU) are used to increase non-linearity. Add positional encoding to each data chunk. Since the Transformer model itself does not process the sequential information of the sequence, positional encoding is used to provide the model with the position information of the data chunk in the original data. Use sine and cosine functions to generate positional encoding, which can generate unique encodings for each position. Add the positional encoding to the embedded representation of the data chunk. Input the data chunk sequence into a pre-built encoder for data processing. The encoder based on Transformer contains multiple residual modules that integrate multi-head attention mechanisms. The data chunk sequence passes through the encoder layers, and each layer applies multi-head attention, residual connections, and layer normalization. The encoder can handle the dependencies between data chunks and extract global information. The encoder includes an incremental encoder. In a motor control system, to accurately control the speed and position of the motor, it is necessary to obtain the rotation information of the motor in real time. The controller adjusts the control strategy of the motor by comparing the actual position signal output by the encoder with the set target position signal to achieve high-precision position control. The encoder includes an absolute encoder. Inside the absolute encoder, there is a unique encoding sequence corresponding to each possible position of the device. When the position of the device changes, the encoder outputs a corresponding digital signal representing the new position information, which is used to monitor the position information of the device. Measure the gap between the predicted result and the actual result based on a predefined self-supervised loss function. Self-supervised learning: Train the model by designing tasks without explicit labels. Loss function: It can be mean squared error (MSE) or cross-entropy loss, etc., depending on the nature of the task. In some cases, use contrastive loss (such as Triplet Loss) to encourage the model to learn the similarities and differences between data chunks. Update model parameters: Use the backpropagation algorithm to optimize the loss function and update the weights of the encoder and related layers. Use an optimizer (such as Adam) to accelerate the training process. Obtain an initial compliance monitoring model. After sufficient training iterations, the model will be able to generate compliance predictions based on the input data. This system integrating multi-head attention, residual modules, convolutional layers, positional encoding, and self-supervised learning will be able to process complex coal mine production data and evaluate its compliance in real time.

[0055] According to some embodiments of the present invention, the residual module includes:

[0056] A direct connection layer, which is used to directly connect and allow the input of the module to be directly added to the output of the module;

[0057] A normalization layer, which is used to adjust the data distribution of the output of the intermediate layer to make it have stable mean and variance;

[0058] An activation function, which is used to introduce non-linearity so that the network can learn complex mapping relationships;

[0059] A dropout layer, which is used to prevent overfitting.

[0060] Working principle and beneficial effects of the above technical solution: Direct connection layer (residual connection). The direct connection layer allows the input of the module to be directly added to the output of the module, which helps to alleviate the problem of vanishing gradients or exploding gradients in deep networks, enabling the network to be deeper and easier to train. Application: Residual Network (ResNet) is a typical application of this connection method. In each residual block, the input is directly connected to the output through a "shortcut" and added to the result after being processed by multiple convolutional layers or fully connected layers. Normalization layer (such as Batch Normalization, BN) is used to adjust the data distribution of the intermediate layer output to have stable mean and variance. This helps to accelerate the training process, improve the convergence speed of the model, and can reduce the sensitivity of the model to the initial parameters. Application: Batch normalization is usually added after the convolutional layer or fully connected layer and before the activation function. It can significantly improve the training stability and performance of deep learning models. Activation function is used to introduce non-linearity so that the neural network can learn complex mapping relationships. A neural network without an activation function can only represent linear relationships, which limits its ability to handle complex problems. Commonly used activation functions: ReLU (Rectified Linear Unit): Directly outputs for positive inputs and outputs 0 for negative inputs. Sigmoid: Maps the input to the interval (0, 1) and is commonly used in the output layer of binary classification problems. Tanh: Maps the input to the interval (-1, 1), similar to Sigmoid but with a wider output range. Leaky ReLU, PReLU, etc.: Variants of ReLU used to solve the problem of vanishing gradients when the input is negative. Application: Activation functions are usually located after each convolutional layer or fully connected layer. Dropout layer is used to prevent overfitting by randomly discarding (i.e., setting to 0) a part of the neurons (and their connections) in the network. This can be regarded as training a collection of multiple different sub-networks, each sub-network being a simplification of the original network. During testing, all neurons are retained, but their outputs are multiplied by a scaling factor related to the dropout rate. Application: The Dropout layer is usually added after the fully connected layer, but can also be used after the convolutional layer. The dropout rate is a hyperparameter that needs to be adjusted according to the specific task. By combining and optimizing these components, a powerful and efficient deep learning model can be constructed to process complex coal mine production data.

[0061] Further, the construction process of the encoder includes the following steps:

[0062] Determine the type and dimension of the input data;

[0063] Determine the reconstruction target;

[0064] Based on a preset data table, select encoder parameters according to the type and dimension of the input data and the reconstruction target to obtain an initial encoder. The encoder parameters include the network structure, activation function and number of layers, loss function, and optimizer of the encoder part;

[0065] Use the training data to train the initial encoder, continuously adjust the network parameters through forward propagation and backward propagation, monitor the loss value and reconstruction error during the training process, and make the network converge. When the convergence value is less than the preset threshold, obtain the encoder.

[0066] The working principle and beneficial effects of the above technical solution: First, it is necessary to clarify the type of input data, such as whether it is image, time series, text, or other types of data. Different types of data may require different preprocessing methods and network structures. Dimension: Determine the dimension of the input data, including its shape (such as the height, width, and number of color channels of an image), sequence length, or number of features, etc. Determine the reconstruction target, that is, determine the judgment of whether the coal mine production is compliant after extracting the data features. The preset data table is a corresponding data table for the type and dimension of the input data, the reconstruction target, and the network structure, activation function and number of layers, loss function, and optimizer of the encoder part. According to the type and dimension of the input data, select an appropriate network layer type (such as convolutional layer, fully connected layer, recurrent layer, etc.). Design the stacking method of the network layers and determine the number of neurons or filter size of each layer. Add an activation function after the output of each layer to introduce non-linearity. Commonly used activation functions include ReLU, Sigmoid, Tanh, etc. Select an appropriate activation function according to the network structure and task requirements. Determine the number of layers of the encoder. Too many layers may lead to overfitting, while too few layers may not be able to learn enough features. Preset data table: A preset data table or configuration file can be created to list the recommended network structures, activation functions, and number of layers, etc. under different data types and dimensions for quick selection and adjustment. Select an appropriate loss function according to the reconstruction target. For example, for pixel-level reconstruction tasks, commonly used loss functions include mean squared error (MSE) or cross-entropy loss (for binary or probabilistic data). Select an optimization algorithm to update the network parameters, such as SGD, Adam, RMSprop, etc. Configure the parameters of the optimizer, such as learning rate, momentum, etc. Use the training data to train the initial encoder, input the training data into the encoder, and calculate the reconstruction error through forward propagation. Use the backpropagation algorithm to calculate the gradient and update the network parameters. Monitor the loss value and reconstruction error during the training process, and a validation set can be used to evaluate the generalization ability of the model. When the loss value is less than the preset threshold or the performance on the validation set no longer improves significantly, stop training. After training is completed, obtain the trained encoder model. It is convenient to obtain an accurate encoder.

[0067] S2: Real-time collect and preprocess the detection data during the coal mine production process.

[0068] Specifically, the detection data includes production volume, power consumption, and personnel activity conditions; the preprocessing includes removing duplicate values, handling missing values, handling outliers, data smoothing, and data merging processing.

[0069] The working principle and beneficial effects of the above technical solution: By preprocessing the detection data, duplicate data can be removed, missing data can be supplemented, outlier data can be deleted, the data can be smoothed to reduce the random noise of the data, and the data can be merged to facilitate the integration of the data and improve the data processing efficiency. When removing duplicate values, data analysis and summarization are performed based on a pivot table. By dragging relevant fields into the row area and setting the count summarization method, which records are duplicates can be identified. When handling missing values, it is achieved based on linear interpolation or polynomial interpolation. Outliers are identified and processed based on descriptive analysis methods or the Z-score method. Data smoothing is achieved based on at least one of the moving average method, exponential smoothing, and filters. Data merging processing is used to determine the probability of mergeable data, and merging is performed if it is greater than the preset probability threshold.

[0070] Further, the data merging processing includes the following steps:

[0071] Feature extraction is performed on each sub-data in the detection data, and the feature indicators are numerically processed to obtain index values;

[0072] The feature indicators include data keywords, data size, data recording time, data type, and data source address;

[0073] Arbitrarily select two sub-data and calculate the probability that the two sub-data can be merged;

[0074]

[0075] where P i,j is the probability that the i-th sub-data and the j-th sub-data can be merged; T i,a is the a-th index value of the i-th sub-data; T j,a is the a-th index value of the j-th sub-data; max(T a ) is the maximum value of the a-th index value among several sub-data in the detection data; the first index value is the numerical processing result of the data keyword; the second index value is the numerical processing result of the data size; the third index value is the numerical processing result of the data recording time; the fourth index value is the numerical processing result of the data type; the fifth index value is the numerical processing result of the data source address;

[0076] Merge the two sub-data with a probability greater than the preset probability threshold to obtain the detection data after data merging processing.

[0077] Working principle and beneficial effects of the above technical solution: Feature extraction is performed on each sub-data in the detection data, and the feature indicators are numerically processed to obtain index values; the feature indicators include data keywords, data size, recording time of the data, type of the data, and source address of the data; any two sub-data are randomly selected, the probability that the two sub-data can be merged is calculated, and the two sub-data with a probability greater than the preset probability threshold are merged to obtain the detection data after data merging processing. The efficient integration of data is realized, which is convenient for improving the data processing efficiency. Among them, the preset probability threshold is 0.85, which is preset according to the data merging requirements, and the preset range is (0.5, 1). The higher the data merging requirements, the larger the preset probability threshold.

[0078] S3: Based on the preprocessed detection data, the initial compliance monitoring model is optimized based on the data loop feedback mechanism to obtain an optimized model.

[0079] S4: Based on the optimized model, it is judged whether the coal mine production is compliant.

[0080] The optimized model is compared with the initial compliance monitoring model, and the comparison result is obtained and displayed:

[0081] Comparing the optimized model with the initial compliance monitoring model includes the following steps:

[0082] Obtain the first model parameters that make up the optimized model and the second model parameters that make up the initial compliance monitoring model;

[0083] Compare the model parameters of the same category in the first model parameters and the second model parameters to obtain a comparison result.

[0084] Working principle and beneficial effects of the above technical solution: The optimized model is compared with the initial compliance monitoring model, and the comparison result is obtained and displayed. The comparison result includes the comparison result of the model parameters of the coal mine enterprise power consumption classification prediction calculation model included in the initial compliance monitoring model and the optimized coal mine enterprise power consumption classification prediction calculation model included in the optimized model. The comparison result of the over-capacity and over-intensity illegal production monitoring and early warning model of the coal mine enterprise included in the initial compliance monitoring model and the optimized over-capacity and over-intensity illegal production monitoring and early warning model of the coal mine enterprise included in the optimized model. The comparison result of the suspected production suspension and illegal mining behavior monitoring and early warning model of the coal mine enterprise included in the initial compliance monitoring model and the optimized suspected production suspension and illegal mining behavior monitoring and early warning model of the coal mine enterprise included in the optimized model. It is convenient to show which parameters of the model have changed, and when it is found that the changes are too large, it can also timely detect the errors in model updates and improve the accuracy of model update monitoring.

[0085] Embodiment 2, an embodiment of the present invention, provides an optimization system for a coal mine compliance monitoring model based on data loop feedback, including: a construction module for constructing an initial compliance monitoring model; a collection and processing module for collecting and preprocessing detection data during the coal mine production process in real time; an optimization module for optimizing the initial compliance monitoring model based on the data loop feedback mechanism according to the preprocessed detection data to obtain an optimized model; and a judgment module for judging whether the coal mine production is compliant based on the optimized model.

[0086] The beneficial effects of the above technical solution: The core of the data loop feedback mechanism lies in its iterative nature. By continuously collecting data, analyzing data, generating feedback, and applying the feedback, the system can continuously improve and optimize itself. This iterative process helps to quickly respond to changes, adapt to new environments, and continuously enhance the competitiveness and value of the system. Among them, the data collection is the preprocessed data obtained. Multiple methods are used for data analysis, such as statistical analysis, data mining, machine learning, etc., to discover patterns and trends in the data. Based on the results of data analysis, corresponding feedback information is generated. The feedback information is the evaluation and suggestions regarding aspects such as system performance, user behavior, and monitoring business processes. These feedback information facilitate the identification of potential problems and determine the direction of improvement. Apply the feedback and optimize according to the improvement direction. Improve the accuracy and adaptability of the model through continuous learning. The optimization process includes: parameter adjustment: adjusting the model parameters to improve performance. Algorithm improvement: introducing more advanced machine learning or deep learning algorithms. Feature engineering: adding or modifying input features to improve the model effect. Iterative loop: repeating the evaluation, feedback, and optimization steps until the model performance reaches a satisfactory level. Improve the accuracy and adaptability of the model through continuous learning, deploy the optimized model to the production environment, and monitor the real-time data. Judge whether the production activities are compliant according to the model output. Preprocess the detection data to improve the data accuracy, and at the same time continuously update and optimize the compliance monitoring model according to the real-time data to obtain an accurate compliance monitoring model, and improve the accuracy of judging whether the coal mine production is compliant.

[0087] Embodiment 3 is an embodiment of the present invention. What is different from the previous embodiment is that when a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs, etc., all kinds of media that can store program codes.

[0088] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in combination with an instruction execution system, apparatus, or device.

[0089] More specific examples (nonexhaustive list) of computer-readable media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which a program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0090] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0091] Embodiment 4, an embodiment of the present invention, provides an optimization method for a coal mine compliance monitoring model based on data loop feedback. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0092] To verify the effectiveness of the present invention, a set of comparative experiments was designed. The experiments selected 10 coal mine enterprises of different scales and used the existing technology and the present invention respectively for compliance monitoring for 6 months. During the experiment, this embodiment collected data on various key indicators, including the accuracy rate of compliance judgment, the detection time of abnormal behaviors, the false alarm rate, the missed alarm rate, etc. At the same time, this embodiment also simulated various production environment change situations to test the adaptability and update efficiency of the two methods.

[0093] During the experiment, the data was summarized and analyzed weekly, and a final statistical comparison was carried out after the experiment ended. To ensure the reliability of the data, this embodiment adopted the method of cross-validation, that is, the two methods were alternately used at different time periods to eliminate the influence of other factors. In addition, industry experts were invited to review the determination of abnormal situations to ensure the objectivity of the evaluation. The experimental results showed that the present invention had made significant improvements in various indicators, especially in terms of real-time performance and accuracy.

[0094] Table 1 Comparison of evaluation indicators with the existing technology

[0095]

[0096]

[0097] Table 1 fully demonstrates the significant advantages of the present invention compared with the prior art solutions. In terms of the accuracy rate of compliance judgment for core indicators, the present invention reaches 93%, which is 8 percentage points higher than the prior art, reflecting higher reliability. The model update frequency has been increased from once a month to real-time update, greatly enhancing the real-time performance and flexibility of the system. The abnormal behavior detection time has been shortened from 48 hours to 6 hours, an increase of 87.5%, greatly improving the response speed and safety management ability of the system. The false alarm rate and missed alarm rate have decreased by 53.3% and 58.3% respectively, significantly reducing false positive and false negative situations and improving the monitoring accuracy. The data processing efficiency has increased by 4 times, from 1000 records per second to 5000 records per second, greatly improving the system's ability to process large-scale real-time data. The time for the model to adapt to the new environment has been shortened from 2 weeks to 3 days, a decrease of 78.6%, indicating that it has stronger environmental adaptability and learning ability. These data comprehensively reflect that the present invention has made remarkable progress in terms of accuracy, real-time performance, efficiency and adaptability, providing more powerful technical support for coal mine safety production and compliance management.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An optimization method for a coal mine compliance monitoring model based on data loop feedback, characterized in that Including: Construct an initial compliance monitoring model; Collect and preprocess the detection data in the coal mine production process in real time; Based on the preprocessed detection data, optimize the initial compliance monitoring model based on the data loop feedback mechanism to obtain an optimized model; Judge whether the coal mine production is compliant based on the optimized model.

2. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 1, wherein: The process of establishing the initial compliance monitoring model includes the following steps: Set a residual module with a fused multi-head attention mechanism; Set data segmentation rules; Set a convolutional layer, and perform a convolution operation on the initial data processed based on the data segmentation rules through the convolutional layer to obtain N data blocks, and add position encoding to each data block to obtain a data block sequence; Input the data block sequence into a pre-constructed encoder for data processing, and measure the gap between the predicted result and the actual result corresponding to the initial data based on a predefined self-supervised loss function during the data processing, and update the parameters of the initial compliance monitoring model to obtain the initial compliance monitoring model.

3. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 2, wherein: The construction process of the encoder includes the following steps: Determine the type and dimension of the input data; Determine the reconstruction target; Based on a preset data table, and according to the type and dimension of the input data and the reconstruction target, select encoder parameters to obtain an initial encoder; Train the initial encoder, continuously adjust the network parameters through forward propagation and backward propagation, monitor the loss value and reconstruction error during the training process, and make the network converge. When the convergence value is less than a preset threshold, obtain the encoder.

4. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 3, wherein: The detection data includes production volume, power consumption, and personnel activity conditions; the preprocessing includes removing duplicate values, handling missing values, handling outliers, data smoothing, and data merging processing.

5. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 4, characterized in that: The data merging processing includes the following steps: Extract features from each sub-data in the detection data, and numerically process the feature indicators to obtain index values; The feature indicators include data keywords, data size, data recording time, data type, and data source address; Arbitrarily select two sub-data and calculate the probability that the two sub-data can be merged; where P i,j is the probability that the i-th sub-data and the j-th sub-data can be merged; T i,a is the a-th index value of the i-th sub-data; T j,a is the a-th index value of the j-th sub-data; max(T a ) is the maximum value of the a-th index value among several sub-data in the detected data; Merge the two sub-data with a probability greater than the preset probability threshold to obtain the detection data after data merging processing.

6. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 5, characterized in that: Judging whether the coal mine production is compliant based on the optimized model specifically means comparing the optimized model with the initial compliance monitoring model, obtaining a comparison result and displaying it.

7. The method for optimizing the coal mine compliance monitoring model based on data loop feedback according to claim 6, characterized in that: Comparing the optimized model with the initial compliance monitoring model includes the following steps: Obtain the first model parameters that make up the optimized model and the second model parameters that make up the initial compliance monitoring model; Compare the model parameters of the same category in the first model parameters and the second model parameters to obtain a comparison result.

8. A system adopting the optimization method of the coal mine compliance monitoring model based on data loop feedback as described in any one of claims 1 to 7, characterized in that, Including: A construction module that constructs an initial compliance monitoring model; A collection and processing module that collects and preprocesses the detection data in the coal mine production process in real time; An optimization module that optimizes the initial compliance monitoring model based on the data loop feedback mechanism according to the preprocessed detection data to obtain an optimized model; A judgment module that judges whether the coal mine production is compliant based on the optimized model.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for optimizing the coal mine compliance monitoring model based on data loop feedback according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for optimizing the coal mine compliance monitoring model based on data loop feedback according to any one of claims 1 to 7.

Images