Anomaly Detection for Equipment Condition via Frequency Spectrum Entropy

Some of the critical and practical issues regarding the problem of condition monitoring of mobile equipment have been discussed, and an anomaly detection method without priori knowledge has been proposed. The method involves setting amplitude benchmark via spectrum amplitude in normal condition and obtaining the maximum entropy value in abnormal condition. The condition identification is achieved through estimating the amount of anomaly information in spectrum, and a measure of anomaly condition is given by the anomaly degree derived from entropy value dividing the maximum value. The sensitivity, stability and computation load of the method have been also discussed, and the method is validated on an experimental test-bed that the test bearings with different fault diameter support the motor shaft.