Elastic Collisions in Complementarity-based Time-stepping Methods T. Preclik? , U. Rüde? ? Friedrich-Alexander University of Erlangen-Nürnberg Chair of Computer Science 10 (System Simulation), Cauerstr. 6, 91058 Erlangen, Germany e-mails: tobias.preclik@informatik.uni-erlangen.de, ulrich.ruede@informatik.uni-erlangen.de web page: http://www10.informatik.uni-erlangen.de/ T. Preclik? , U. Rüde? (? LSS) 1 / 10 Introduction Rigid multibody systems with non-smooth contact dynamics. T. Preclik? , U. Rüde? (? LSS) 2 / 10 Introduction Rigid multibody systems with non-smooth contact dynamics. 2.5 position velocity 2 1.5 1 0.5 0 -0.5 -1 -1.5 -1 -0.5 0 0.5 1 time T. Preclik? , U. Rüde? (? LSS) 2 / 10 Introduction Rigid multibody systems with non-smooth contact dynamics. Switching points separate smooth system evolutions (e.g. stick-slip transitions and impacts). 2.5 position velocity 2 1.5 1 0.5 0 -0.5 -1 -1.5 -1 -0.5 0 0.5 1 time T. Preclik? , U. Rüde? (? LSS) 2 / 10 Introduction Rigid multibody systems with non-smooth contact dynamics. Switching points separate smooth system evolutions (e.g. stick-slip transitions and impacts). Time-stepping methods do not detect switching points in contrast to event-driven schemes. T. Preclik? , U. Rüde? (? LSS) 2 / 10 Introduction Rigid multibody systems with non-smooth contact dynamics. Switching points separate smooth system evolutions (e.g. stick-slip transitions and impacts). Time-stepping methods do not detect switching points in contrast to event-driven schemes. Talk is in the context of complementarity-based time-stepping schemes by Anitescu and Potra [1997]. [1997] M. Anitescu and F.A. Potra. Formulating Dynamic Multi-Rigid-Body Contact Problems with Friction as Solvable Linear Complementarity Problems. ASME Nonlinear Dynamics, 14(3):231–247, 1997. T. Preclik? , U. Rüde? (? LSS) 2 / 10 Elastic Collisions in Time-Stepping Methods Restitution hypotheses are only meaningful for non-simultaneous collisions. Application of restitution hypotheses can cause constraint violation for simultaneous collisions. Figure: Newton’s cradle integrated with Moreau’s midpoint rule at t. T. Preclik? , U. Rüde? (? LSS) 3 / 10 Elastic Collisions in Time-Stepping Methods Restitution hypotheses are only meaningful for non-simultaneous collisions. Application of restitution hypotheses can cause constraint violation for simultaneous collisions. Figure: Newton’s cradle integrated with Moreau’s midpoint rule at (t + 12 δt)− . T. Preclik? , U. Rüde? (? LSS) 3 / 10 Elastic Collisions in Time-Stepping Methods Restitution hypotheses are only meaningful for non-simultaneous collisions. Application of restitution hypotheses can cause constraint violation for simultaneous collisions. Figure: Newton’s cradle integrated with Moreau’s midpoint rule at (t + 21 δt)+ . T. Preclik? , U. Rüde? (? LSS) 3 / 10 Elastic Collisions in Time-Stepping Methods Restitution hypotheses are only meaningful for non-simultaneous collisions. Application of restitution hypotheses can cause constraint violation for simultaneous collisions. Figure: Newton’s cradle integrated with Moreau’s midpoint rule at t + δt. T. Preclik? , U. Rüde? (? LSS) 3 / 10 Solution Set of a Simultaneous Collision v0− = v , v1− = v2− = 0 E − = 21 mv 2 perfectly elastic collisions λ1 λ2 v0+ v1+ v2+ E+ 2 3 mv 1 3 mv 1 3v 1 3v 1 3v Newton’s collision hypothesis Collision resolution mv 0 0 v 0 1 − 3E E− 4 3 mv 2 3 mv − 31 v 2 3v 2 3v E− Poisson’s collision hypothesis mv mv 0 0 v E− “like Newton’s cradle” inelastic collision Table: Collision resolution options for three spheres in a row without gaps. T. Preclik? , U. Rüde? (? LSS) 4 / 10 Solution Set of a Simultaneous Collision (cont.) T. Preclik? , U. Rüde? (? LSS) 5 / 10 Non-dispersive Solutions An ideal Newton’s Cradle is free of dispersion as noted e.g. in Herrmann and Seitz [1982] and thus collisions can be treated as occurring successively. Recipe for computing a non-dispersive solution: Let fi,j : R6nb → R6nb calculate the two-body collision between bodies i, j and f : R6nb → R6nb be the composition of all two-body collision operators then the fixed point iteration xn+1 = f (xn ), where x0 = ġ(q(t − ), ϕ(t − )) (possibly) converges to the non-dispersive solution and we can assume ġ(q(t + ), ϕ(t + )) = limn→∞ xn . Does not necessarily converge for perfectly elastic collisions (billiard frame, sphere trapped between walls). T. Preclik? , U. Rüde? (? LSS) 6 / 10 Consequences for Time-Stepping Methods Option 1: Resolve time of impact for elastic collisions. Integrate non-smooth system until elastic impact occurs, resolve elastic impact and continue integration. Would be step backwards towards event-driven schemes. Fails on accumulative switching points. Option 2: Integrate impacts inelastically and correct velocities after time step according to restitution hypothesis. No constraint violation. Elastic response shifted in time by up to δt. Limits collision frequency to T. Preclik? , U. Rüde? (? LSS) 1 δt . 7 / 10 Summary Restitution hypothesis for Newton’s cradle. Flexible restitution hypothesis insertion. Suggestion for treatment of elastic collisions in time-stepping methods. T. Preclik? , U. Rüde? (? LSS) 8 / 10 Questions? Comments? Suggestions? . . . T. Preclik? , U. Rüde? (? LSS) 9 / 10 References M. Anitescu and F.A. Potra. Formulating Dynamic Multi-Rigid-Body Contact Problems with Friction as Solvable Linear Complementarity Problems. ASME Nonlinear Dynamics, 14(3):231–247, 1997. F. Herrmann and M. Seitz. How does the ball-chain work? American Journal of Physics, 50(11):977–981, 1982. T. Preclik. Elastic Collisions in Complementarity-based Time-stepping Methods. Technical report, Friedrich-Alexander University Erlangen-Nuremberg, December 2010. C. Studer. Augmented time-stepping integration of non-smooth dynamical systems. PhD thesis, ETH Zürich, 2008. T. Preclik? , U. Rüde? (? LSS) 10 / 10
© Copyright 2024 Paperzz