Enhanced network learning model with intelligent operator for the motion reliability evaluation of flexible mechanism

The evaluation of flexible mechanism involving multi-body dynamics with high nonlinearity and transients urgently requires an efficient evaluation method to enhance its reliability and safety. In this work, an enhanced network learning method (ENLM) is proposed to improve the modeling precision and simulation efficiency in flexible mechanism reliability evaluation, by introducing generalized regression neural network (GRNN) and multi-population genetic algorithm (MPGA) into extremum response surface method (ERSM). In the ENLM modeling, the ERSM is adopted to reasonably handle transients (time varying) problem in motion reliability analysis by considering one extreme value in whole response process; the GRNN is applied to address high-nonlinearity in surrogate modeling; the MPGA is utilized to find the optimal model parameters in ENLM modeling. In respect of the developed ENLM, the motion reliability of two-link flexible robot manipulator (TFRM) was evaluated, with regard to the related input random parameters to material density, elastic modulus, section sizes, and deformations of components. In term of this study, it is illustrated that (i) the comprehensive reliability of flexible robot manipulator is 0.951 when the allowable deformation is 1.8x10(-2) m; (ii) the maximum deformations of member-1 and member-2 obey normal distributions with the means of 1.45x10(-2) m and 1.69x10(-2) m as well as the standard variances of 6.77x10(-4) m and 4.08x10(-4) m, respectively. The comparison of methods demonstrates that the ENLM improves the modeling precision by 3.29% and reduces the simulation efficiency by 1.19 s under 10000 simulations, and the strengths of the ENLM with high modeling precision and high simulation efficiency become more obvious with the increase of simulations. The efforts of this study provide a learning-based reliability analysis way (i.e., ENLM) for the motion reliability design optimization of flexible mechanism and enrich mechanical reliability theory. (c) 2020 Elsevier Masson SAS. All rights reserved.