Practical approach for integrated fault diagnosis in rotating machinery using vibration signals

In manufacturing, researchers have conducted extensive studies to prevent malfunctions of rotating equipment during operation via various Prognostics and Health Management (PHM) techniques. However, most recent studies require sufficient fault data for training, and they usually focus on only a single failure mode or component. Although they may offer novel academic insights, they often have limited practical field applications, mainly due to a lack of ability to access and cope with data in various operating environments. Therefore, it is necessary to develop a method that can detect multiple failure modes early with minimum or at least reasonable data acquisition. To overcome these challenges, this study presents a practical approach for integrated fault diagnosis of rotating machinery by establishing a multi-step diagnostic framework. The framework adopted physically interpretable frequency-domain features extracted from vibration signals, specifically the shaft rotational frequency and bearing characteristic frequencies (BPFO and BPFI), each considered up to the third harmonic, to jointly diagnose shaft and bearing faults. Health indicators were then computed using Mahalanobis distance with respect to a baseline distribution, enabling reliable fault diagnosis using only normal-condition data. To confirm universal field applications of the suggested method, the performance was validated in various test cases with different scenarios. The results proved that the approach was useful for practical purposes by achieving high accuracy, improving from 88% to near-perfect performance. This approach allows for the detection and diagnosis of anomalies and their specific types in both the shaft and rolling element bearings, relying solely on normal operating data.