A new synergy of singular spectrum analysis with a conscious algorithm to detect faults in industrial robotics

Riyadh Nazar Ali Algburi, Hongli Gao, Zaid Al-Huda

Abstract: Investigating vibration signals is an effective technique for assessing system malfunction. However, extraction of weak flaw characteristics of shaking signals with large noise is difficult. Therefore, a new approach using singular spectrum analysis (SSA) integrated with generalized structured shrinkage algorithm (GSSA) is proposed in this study for flaw diagnosis in an industrial robot. First, SSA allows the separation of complicated encoding signals to several interpretable elements involving a trend signal, set of cyclic oscillations, and residual (mixed) signal. Second, the concept of algorithm-conscious sparsity-assisted techniques is presented to improve flaw features, expand the model-conscious sparsity-assisted flaw detection, and enable an easy and scalable algorithm layout. Third, GSSA is built in algorithm-conscious techniques to address drawbacks of $$L_{1}$$ L 1 -norm penalty on the basis of flaw characteristic optimization techniques and describe generalized structured shrinkage regulators. GSSA is proposed to extract noise interference, discrete frequency interference, and cyclic impulse of a robot manipulator focused on the signal of rotary encoder. Fourth, a sequence of numerical simulations and four experimental cases are carried out. Finally, comparisons are made with model-conscious strategies, including window-group-lasso and basis pursuit denoising, to prove the GSSA advantages of weak fault enhancement features further.