Robust fault detection, location, and recovery of damaged data using linear regression and mathematical models

摘要

A new method using linear regression and mathematical models is proposed for fault detection, location and the recovery (FDIR) of damaged data. Linear regression was used to identify the mathematical model of the system from source data. For this reason the paper focuses on the class of fault whose impact on the system causes that the measured data are differ from the expected values. Based on the mathematical model of the process developed during identification from source data, the damage area can be pre-located, using proposed recurrence algorithm of fault recovery (FR) we can exactly estimated expected value on the damage area. Based on estimated expected value the damage area can be precisely located and the class of faults can be estimated. This paper proposes an innovative approach to recurrence self-tuning of fault recovery (FR) from parts incomplete or damaged data of the pre-allocated area.