Enterprise AI Analysis

The bKSHHO-KELM Prediction Model

This study introduces the bKSHHO-KELM model, a novel approach for robust microseismic and blasting signal recognition. It leverages two key innovations:

• Kernel Search Harris Hawks Optimization (KSHHO): An enhanced variant of the Harris Hawks Optimization (HHO) algorithm that incorporates kernel functions. This improves the algorithm's global search capability and prevents it from getting trapped in local optima, addressing a common limitation of original HHO.
• Binary KSHHO (bKSHHO) for Feature Selection: KSHHO is adapted into a binary version specifically for feature selection tasks. It efficiently identifies the most discriminative feature subsets from complex signal data.
• Kernel Extreme Learning Machine (KELM): Integrated as the core classifier, KELM is a powerful machine learning method that utilizes kernel functions to handle nonlinear relationships in the data, enhancing its expressive power and generalization ability.

Together, bKSHHO performs intelligent feature selection, feeding optimized features to KELM for highly accurate classification of microseismic and blasting events.

Validation & Superior Performance

The proposed bKSHHO-KELM model demonstrated outstanding performance across multiple evaluation stages:

• KSHHO Optimization: Evaluated against ten state-of-the-art algorithms on the IEEE CEC 2022 benchmark functions. KSHHO achieved the highest average rank (3.75) and showed superior convergence performance, effectively avoiding local optima.
• Microseismic Signal Prediction: Applied to real-world microseismic and blasting signal data from the Linglong Gold Mine. The model achieved a remarkable 95.625% accuracy, 93.964% recall, 92.632% precision, and an F1 score of 0.931.
• Comparative Advantage: bKSHHO-KELM significantly outperformed seven classic classifiers (LightGBM, XGBoost, CatBoost, RandomF, AdaBoost, FKNN, KELM) and seven other KELM-based optimized variants, demonstrating superior stability and prediction accuracy for microseismic hazard early warning.

Advancing Mine Safety: Discussion & Outlook

While bKSHHO-KELM demonstrates promising performance, the research acknowledges several avenues for future work:

• Enhanced Optimization Robustness: Exploring more robust initialization strategies for bKSHHO (e.g., opposition-based learning) to mitigate sensitivity to initial population randomness.
• Broader Hazard Scope: Expanding the model's capabilities to include multi-class classification or anomaly detection for rare/unknown seismic events beyond binary microseismic/blasting signals.
• Advanced Signal Processing: Integrating techniques like time-frequency analysis, wavelet transforms, or deep learning-based feature extraction to better distinguish genuine seismic events from operational noise.
• Comprehensive Datasets: Developing more diverse datasets encompassing a broader spectrum of seismic and non-seismic sources to improve model generalizability in complex mining environments.
• Hybrid Deep Learning Approaches: Exploring combinations of bKSHHO-KELM's interpretability and efficiency with deep learning's powerful feature extraction capabilities.

Strategic Imperative for Mine Operators

The bKSHHO-KELM prediction model offers a highly effective and accurate solution for the critical challenge of microseismic hazard early warning in deep mining. Its superior performance in identifying microseismic and blasting signals positions it as an indispensable tool for enhancing mine safety management.

By leveraging this technology, mining operations can significantly improve real-time risk assessment, prevent potential disasters, and ensure the safety of personnel and equipment. This represents a strategic investment in operational resilience and a proactive approach to deep mining challenges.