Gear_fault_feature_image_data.zip

This dataset provides the original feature image used in the study “Development of Feature Images Considering Fault Characteristics for Gearbox Condition Classification.” The image was constructed based on vibration signals collected from a gearbox system under multiple fault conditions.Using a Genetic Algorithm (GA), 20 diagnostic features were selected from three domains — time, cepstrum, and frequency — due to their high discriminative power. These features were then arranged into a 4×5 matrix to form a feature image, which visually encodes domain-specific signal characteristics.- Features 1–5 (top two rows) represent time-domain features such as RMS, standard deviation, kurtosis factor, and smoothness.- Features 6–15 (middle two rows) are cepstrum-domain features that highlight periodicity and modulation, particularly useful for gear and bearing faults.- Features 16–20 (bottom row) are frequency-domain features such as frequency center and RMS frequency, which reflect energy distribution.This image serves as an input to a CNN model before RGB synthesis, preserving the raw structural pattern of selected features. It enables classification of six gearbox conditions including normal, single, and compound faults. The data corresponds to Figure 5 in the manuscript and represents Sensor #1.<br>Please refer to the related article for detailed methodology and fault scenarios.<br>