Abstract: Approximate Entropy (ApEn) provides a new dimensionless factor to denote the irregularity of time series from the view of complexity measuring. Only short time-series are needed to calculate the ApEn of random signal or ascertain signals with high anti-interference ability. The conception, nature and a practical fast algorithm are introduced. Successful application has been achieved to analyze the looseness fault in bearing bushing of turbo-generator set. The results show that ApEn has high ability to quantify the complexity of signals, thereby providing an effective technology for condition monitoring and fault diagnosis of mechanical equipment.