Time-series data exception detection method and system thereof

Abstract

The invention discloses a time-series data exception detection method and a system thereof. The method comprises the steps of receiving time-series data collected by one or more than one sensor arranged in a machine, wherein latest observation data in the time-series data is taken as to-be-monitored data; calculating the relative outlier distance between parameters in the to-be-monitored data and proper-length time-series corresponding parameters which are cached in the machine, so that outlier data with outlier identifications is further obtained; screening exception observation data from the outlier data by utilizing the correlation of parameter values in the proper-length time-series data, and positioning exception parameters in the exception observation data. According to the time-series data exception detection method, the time-series data exception detection accuracy is increased, and the exception detection correctness is guaranteed.

Description

technical field [0001] The invention belongs to the field of data processing, and in particular relates to a time series data anomaly detection method and system thereof. Background technique [0002] With the continuous development of sensing technology, more and more devices have realized the intelligence of devices by installing sensors. As time goes by, the data detected by the sensor forms a time series, that is, sequential data. Anomaly detection of time series data is an important basis and basis for early warning of equipment failure, anomaly location, and failure analysis. [0003] During the operation of the equipment, it usually produces two kinds of outlier data: 1) mutation data: a sudden change in the operation mode of the equipment will cause a sudden change in the sensing data, which is generated by the normal operation of the equipment; 2) abnormal data: During the operation of the equipment, one or several components fail, or the acquired data is abnormal...

AI Technical Summary

Problems solved by technology

[0004] The current anomaly detection of time series data mainly has the following shortcomings: 1) It cannot distinguish between mutation data and abnormal data

Both of these outlier data are considered abnormal data in most anomaly detection methods, leading to abnormal false positives

2) Unable to locate abnormal parameters

Some anomaly detection methods can only detect anomalies, but cannot locate the abnormal parameters

Images