Research on missing value prediction of measured ERT data for coal mine based on a GRNN algorithm.

Area of Science:

• Geophysics
• Mining Engineering
• Data Science

Background:

• Long-term monitoring of coal seam floors using Electrical Resistivity Tomography (ERT) is crucial for detecting potential water inrush.
• Electrode disconnections in ERT systems lead to data loss, compromising early warning capabilities and hindering the identification of hidden dangers in coal mining operations.
• The challenging underground environment often prevents timely maintenance of disconnected electrodes, necessitating methods to handle missing data.

Purpose of the Study:

• To analyze the impact of electrode disconnection on ERT measured data in coal mining environments.
• To introduce and apply the General Regression Neural Network (GRNN) algorithm for predicting missing ERT data caused by electrode disconnections.
• To validate the effectiveness of the GRNN algorithm in restoring data integrity and accuracy for improved safety monitoring.

Main Methods:

• Analysis of the influence of electrode disconnection on Electrical Resistivity Tomography (ERT) data.
• Implementation of the General Regression Neural Network (GRNN) algorithm to predict missing data points.
• Conducting verification experiments in a water tank and applying the method to actual coal mining face data.

Main Results:

• The GRNN algorithm demonstrated high accuracy in predicting missing data, achieving 91.46% accuracy with 82.96% original data integrity and 82.45% accuracy with only 55.56% integrity.
• In a practical coal mining application, the GRNN method predicted data with 85.18% accuracy from a dataset with 73.8% integrity.
• The GRNN approach showed a 14.99% improvement in prediction accuracy compared to the traditional mean value interpolation method.

Conclusions:

• The GRNN algorithm is a viable and effective method for addressing data loss issues in long-term ERT monitoring of coal seam floors.
• Accurate prediction of missing data using GRNN enhances the reliability of early warning systems for water inrush and improves the identification of underground hazards.
• The proposed method contributes to safer and more efficient coal mining operations by ensuring data continuity and integrity in ERT monitoring systems.