Confined Water Rising Zone Height Prediction Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the height of a confined water rising zone in coal mines, such as the BP neural network method, face issues like local minimization, low convergence speed, and limited consideration of influencing factors, leading to inaccurate predictions and poor generalization.
Innovation Solution
A method combining grey relational analysis (GRA), entropy weight method (EWM), fuzzy comprehensive evaluation, principal component analysis (PCA), particle swarm optimization (PSO), and support vector regression (SVR) to screen and weight correlation factors, establishing a comprehensive height prediction model for the confined water rising zone.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If BP neural network method is used to predict confined water rising zone height, then prediction can be performed, but prediction accuracy is poor due to local minimization and convergence to local minima
Solution Approach 1:
The patent transforms the prediction problem from direct height prediction to predicting the development coefficient of the confined water rising zone, which ranges from 0 to 1. This parameter transformation resolves the local minimization problem by converting the prediction target into a normalized coefficient that reflects the proportion of the water rising zone height relative to the aquifer thickness, thereby improving both prediction accuracy and convergence reliability
Solution Approach 2:
The patent introduces the development coefficient as an intermediary parameter between the input correlation factors and the final height prediction. This intermediary transforms the complex direct prediction into a two-step process: first predicting the development coefficient, then calculating the height. This intermediary approach avoids the local minimization trap by working with a normalized, bounded parameter space
2Adaptability or versatility
If BP neural network with many influencing factors is used, then prediction coverage is improved, but model complexity increases and training difficulty increases
Solution Approach 1:
The patent extracts and selects only the most significant correlation factors from the numerous potential influencing factors. By using grey relational analysis to identify and extract the key factors (mining depth, aquifer thickness, permeability coefficient, fault strength index, fault fractal dimension, water pressure, surface tension coefficient, fracture coefficient, and density), the model achieves high prediction coverage while maintaining manageable complexity
Solution Approach 2:
The patent segments the prediction model into two independent parts: a grey relational analysis module for factor selection and an SVR prediction module for actual prediction. This segmentation allows the model to handle multiple influencing factors systematically, first identifying the relevant ones through GRA, then predicting using SVR, thereby reducing overall model complexity while maintaining comprehensive prediction coverage
3Ease of manufacture
If traditional empirical formula is used, then calculation is simple, but prediction accuracy is insufficient and generalization is poor
Solution Approach 1:
The patent replaces traditional empirical formulas with a support vector regression (SVR) model that uses kernel functions to capture nonlinear relationships. This substitution maintains calculation simplicity while dramatically improving prediction accuracy and generalization capability, as the SVR model can adapt to complex geological conditions without requiring complicated manual calculations
Data Source
AI summary
Provided is a method and system for predicting a height of a confined water rising zone. The method includes: obtaining sample data; dividing the sample data into a training sample and a test sample; calculating a degree of correlation between a height and a correlation factor value sequence; screening correlation factors according to the degree of correlation to obtain screened correlation factors; calculating weights of the screened correlation factors using an entropy weight method (EWM); obtaining standardized screened correlation factor value sequences according to correlation factor value sequences corresponding to the screened correlation factors; calculating a value of each indicator according to the standardized screened correlation factor value sequences and the weights; and obtaining a height prediction model of a confined water rising zone based on principal component analysis (PCA)-particle swarm optimization (PSO)-support vector regression (SVR), the value of each indicator, and the test sample.