Training Parameters for Fuzzy Stochastic RBF Neural Network

Western China weakly cemented sandstones were tested for Cerchar abrasivity under dry, mud-saturated, and water-saturated states. The dataset contains network training parameter settings for an improved fuzzy stochastic RBF model that predicts abrasivity indices. It was authored by Yafeng Yao and last updated in April 2026.

Use Cases

• Train a neural network to predict the Cerchar Abrasivity Index (CAI) using inputs like hardness, wave velocity, porosity, and equivalent quartz content.
• Train a neural network to predict the Cerchar Abrasivity Ratio (CAR) from the same physical property inputs.
• Compare model performance metrics (MSE, MAE, R²) of the improved RBF against MLR, SVR, and traditional RBF models.
• Analyze the relationship between rock physical state (dry, mud-saturated, water-saturated) and its impact on CAI values.

Strengths

• Model achieved a prediction error of less than 5% and a Coefficient of Determination (R²) of 0.967 in engineering examples.
• Improved RBF model reduced MSE by up to 65.7% and MAE by up to 47.4% compared to benchmark models.

Limitations

• Dataset is very small at 5.5 KB, indicating limited sample size for training data.
• Scope is restricted to weakly cemented sandstones from western China, limiting geographic generalizability.

Provenance

Source

figshare, authored by Yafeng Yao.

Collection Method

Derived from Cerchar abrasivity tests and neural network training for prediction modeling.

Time Range

null

Freshness

Last updated in April 2026.

Geography

Weakly cemented sandstones from western China.