Application of CBM+ to Shot-Blast Rotating Components Based on Current Signal

Key Points

• The study aims to evaluate fault diagnosis and prediction techniques for shot-blast rotating components based on current signal analysis.
• Monitored current and voltage data from induced motors at one-second intervals.
• Classified current data into categories: no load, normal, warning, risk, and startup.
• Utilized feature extraction techniques like mean, standard deviation, RMS, max, and EMA difference.
• Applied machine learning models including Random Forest, SVM, KNN, and LightGBM.
• Evaluated model performance using t-SNE visualization and cost function based on F1-score.
• LightGBM showed the best performance among the machine learning models applied.
• Defined overload conditions occurring within 2 minutes as indicators of potential failure.
• Confirmed that early detection based on equipment condition is achievable.
• The proposed technique aids in determining preventive maintenance timings, minimizing unexpected failures.

Abstract