基于时间序列位图的滚动轴承异常检测

Study on rolling bearings anomaly detection based on time series bitmaps

针对机械故障快速实时异常检测问题,提出一种基于时间序列位图的新方法。对时域振动信号运用静态等区间符号化方法进行"粗粒化"处理,得到时域信号的符号矩阵,对时域波形的采样点和小波段分别进行特征统计,并用单层和双层时间序列位图对滚动轴承正常状态和异常状态进行显示。针对实验数据得到的检测结果验证了该方法的准确性和有效性。该方法克服了公式法异常检测中计算过程复杂和检测结果抽象的不足,用可视化位图的方式实现了机械运行状态的直观快速显示,降低了对现场异常检测和故障诊断操作人员的要求,是对现场机械实时在线异常检测方法的有力补充。

Aiming at achieving fast and real-time online anomaly detection on mechanical failure, a kind of new anomaly detection method based on time series bitmap technology is proposed. Static equal interval symbolic method is carried on to obtain the coarse graining processing of the vibration signal in time domain. Then characteristic statistical analysis is taken on sampling points and short vibration waveform. Normal condition and abnormal condition of the rolling bearing are showed by one-level and two-level time series bitmaps. The simulations and detection result based on the laboratory data prove this method accurate and effective. Our method overcomes traditional detect methods＇ defects such as complex algorithms or abstract result. The condition of mechanical system could be showed intuitive and simple. Thus judgment could be easily made for the operator or they didn＇ t need so much professional knowledge. This method was an effective supplement to traditional anomaly detection methods.