Observer-based fault detection and diagnosis for mechanical transmission systems with sensorless variable speed drives

Observer based approaches are commonly embedded in sensorless variable speed drives for the purpose of speed control. It estimates system variables to produce errors or residual signals in conjunction with corresponding measurements. The residual signals then are relied to tune control parameters to maintain operational performance even if there are considerable disturbances such as noises and component faults. Obviously, this control strategy outcomes robust control performances. However, it may produce adverse consequences to the system when faults progress to high severity. To prevent the occurrences of such consequences, this research proposes the utilisation of residual signals as detection features to raise alerts for incipient faults. Based on a gear transmission system with a sensorless variable speed drive (VSD), observers for speed, flux and torque are developed for examining their residuals under two mechanical faults: tooth breakage with different degrees of severities and shortage of lubricant at different levels are investigated. It shows that power residual signals can be based on to indicate different faults, showing that the observer based approaches are effective for monitoring VSD based mechanical systems. Moreover, it also shows that these two types fault can be separated based on the dynamic components in the voltage signals. COMADEM 2015 + X CORENDE different levels are investigated. It shows that residual signals can be generated to indicate different faults; showing that the observer based approaches are effective for monitoring VSD based mechanical systems. 2. Observer design 2.1 Induction motor model The squirrel cage IM model is obtained by describing the stator and rotor electrical circuits’ equations based on stator voltage vectors and stator and rotor currents victors , as follows : dds dd = us − Rsis ................................................................................... (1) ddr dd = −Rrir ........................................................................................ (2) ψs = Lsis + Lmir = Lm Lr ψr + σLsis .................................................... (3) Tem = J dωr dd + Tllld ............................................................................. (4) Where, us and is are stator voltage and current respectively, Rs and Rr represent stator and rotor resistances respectively, Lm, Ls, and Lr are the mutual, stator and rotor inductances respectively, σ is the motor leakage coefficient, Tem and Tllld depicts the electromagnetic and load torques respectively, j is the motor inertia, and ψs and ψr are stator and rotor fluxes respectively. 2.2 Speed and flux observer The sensorless VSD regulates the induction motor speed without speed feedback devices. The motor speed is mathematically estimated using terminal supply inputs and induction motor equations . Many approaches have been developed for induction motor speed estimation (1, . However, among those, Model Reference Adaptive System (MRAS) estimator is one of the most common schemes, due to its simplicity, stability, less computationally intensive, and low implementation cost . Therefore it is selected here to be used to estimate speed and flux for control and fault detection purposes simultaneously. Speed and flux MRAS observers, are generally applied by represented the rotor flux ψr components using two different set of equations. The first is independent of the speed and depicted as a reference model, while the other, denoted as an adjustable model, uses the speed value as an adjusting parameter that makes the difference between the two models vanishes. The general structure of an MRAS observer is represented in Figure 1. Figure 1. The general structure of MRAS observer The reference and adjustable models equations are described in the stationery reference frame and respectively represented as follows :