Source-linked AI summary
Stiffness Analysis of Overconstrained Parallel Manipulators
Anatoly Pashkevich, Damien Chablat, Philippe Wenger
TL;DR
The paper tackles stiffness analysis for flexible-link, overconstrained parallel manipulators, including singular postures. It combines FEA-evaluated multidimensional virtual springs with a kinetostatic solution strategy, and demonstrates the approach through architecture comparisons and Orthoglide validation. The proposed model achieves about 5% accuracy across almost all workspace points, with errors up to 20% near a flat singularity.
Problem
Standard stiffness methods cannot be applied directly to overconstrained architectures with redundant constraints, while accurate FEA entails high computational effort.
Method
The method combines FEA-evaluated 6-dof virtual springs for flexible links with a kinetostatic solution strategy that computes stiffness for overconstrained and singular configurations.
Results
The method compares 3-PUU and 3-PRPaR manipulators, confirms advantages of the parallelogram-based architecture, and validates the Orthoglide design against FEA.
Takeaways & Limitations
The approach provides a systematic stiffness-analysis method for overconstrained parallel manipulators while avoiding workspace-wide model remeshing.
Takeaways & Limitations
The current scope assumes similar kinematic chains and base-to-foot actuator placement, and neglects gravity effects important for heavy manipulators.
Abstract
from arXiv · showhide
The paper presents a new stiffness modeling method for overconstrained parallel manipulators with flexible links and compliant actuating joints. It is based on a multidimensional lumped-parameter model that replaces the link flexibility by localized 6-dof virtual springs that describe both translational/rotational compliance and the coupling between them. In contrast to other works, the method involves a FEA-based link stiffness evaluation and employs a new solution strategy of the kinetostatic equations for the unloaded manipulator configuration, which allows computing the stiffness matrix for the overconstrained architectures, including singular manipulator postures. The advantages of the developed technique are confirmed by application examples, which deal with comparative stiffness analysis of two translational parallel manipulators of 3-PUU and 3-PRPaR architectures. Accuracy of the proposed approach was evaluated for a case study, which focuses on stiffness analysis of Orthoglide parallel manipulator.
1. Introduction
The paper addresses stiffness modeling for overconstrained parallel manipulators, where redundant constraints complicate standard analysis. It proposes a FEA-informed lumped model and solution strategy supporting flexible links, coupling, and singular postures.
- Parallel manipulators offer high accuracy, low mass/inertia, and high structural rigidity for positioning and machining applications.
- FEA is accurate but computationally expensive because repeated workspace analyses require remeshing, whereas MSA reduces cost using flexible beam elements.
- VJM represents component flexibility with localized virtual springs added to rigid manipulator models.
- The proposed method uses FEA-evaluated 6-dof virtual springs with translational, rotational, and coupled compliance, plus a kinetostatic strategy for overconstrained and singular configurations.
- Orthoglide and related mechanisms are overconstrained because redundant chains impose repeated platform-rotation constraints.
2.2. Basic Assumptions
Each kinematic chain is modeled as a serial structure containing rigid transformations, actuated and passive joints, and localized virtual springs. This representation standardizes different architectures while retaining actuator, foot, and leg compliance.
- The modified VJM replaces manipulator legs with serial-chain models containing rigid links, passive joints, and virtual springs.
- The leg spring uses six coordinates: three translational and three rotational virtual-joint deflections.
- The chain transformation combines base, actuator, foot, universal-joint, leg, spring, and tool transformations to determine end-effector location.
- Although the kinematic model contains 24 variables, some virtual springs are redundant and are retained initially for computational convenience.
2.3. Differential kinematic model
The differential and kinetostatic models relate virtual-joint and passive-joint variations to six-dimensional end-effector motion and external loads. Eliminating internal variables yields Cartesian stiffness, with special treatment required near singular chain postures.
- 2.3. Differential kinematic model: The differential kinematic equation maps virtual-joint and passive-joint variations through Jacobians to six-dimensional end-effector translation and rotation.
- 2.3. Differential kinematic model: A semianalytical product-factorization method computes Jacobian derivatives without the awkward expressions produced by straightforward differentiation.
- 2.4. Kinetostatic and Stiffness Models: The 6×6 component stiffness matrices are non-diagonal, allowing translational–rotational compliance coupling that ordinary lumped models omit.
- 2.4. Kinetostatic and Stiffness Models: The virtual springs use an aggregated 19×19 stiffness matrix, while passive-joint reactions are set to zero.
- 2.4. Kinetostatic and Stiffness Models: For a prescribed end-effector displacement, the kinetostatic equations determine external forces together with virtual-spring reactions and passive-joint variations.
- 2.4. Kinetostatic and Stiffness Models: The resulting stiffness matrix generally contains symmetric intrinsic stiffness and skew-symmetric loading-related components.
- 2.4. Kinetostatic and Stiffness Models: For translational manipulators, individual chain stiffness matrices can be rank-deficient while their aggregate rank reaches six, resisting all end-effector displacements.
2.5. Comparison with other results
The proposed methodology addresses overconstrained manipulators by solving kinetostatic equations per chain and jointly considering kinematics and static equilibrium. It also supports general 6-dof spring models and broader manipulator configurations.
- Comparison with other results: The method targets overconstrained mechanisms, for which standard stiffness analysis cannot be directly applied.The alternative system is non-square for architectures such as Orthoglide and Delta, preventing unique solution without architectural modification.
- Comparison with other results: A chain-wise solution strategy combines kinematic and static-equilibrium equations, enabling stiffness computation at singular postures without least-square pseudoinversions.The method detects motion subspaces that produce no force or torque reactions in the virtual springs.
- Comparison with other results: Redundant springs compensated by passive joints need not be manually eliminated from the model.This permits direct inclusion of general 6-dof virtual springs derived from FEA models.
- Comparison with other results: The technique can be generalized to different actuator locations and to non-similar kinematic chains.After computing the chain Jacobians, stiffness matrices are evaluated consistently and aggregated.
- Comparison with other results: Each kinematic chain uses one actuator-control spring and three 6-dof springs for transmission and link compliance.This is the adopted compliant-component structure for the kinematic-chain model.
3.1. Actuator compliance
Actuator compliance is modeled through the control-loop compliance and actuator-transmission compliance, reflecting both servomechanism mechanics and control effects.
- Actuator compliance: Actuator compliance is represented by one scalar control-loop parameter and one 6×6 actuator compliance matrix.The transmission’s mechanical flexibility is emphasized because screws, gears, shafts, and belts usually lie outside the feedback-control loop.
3.2. Link Compliance
Link compliance is represented by coupled 6-dof virtual springs, with parameters obtained analytically for beam approximations or through FEA for complex geometries.
- Link Compliance: Link compliance uses symmetric positive-definite 6×6 matrices that couple translational and rotational deformations.This differs from lumped models that neglect coupling and retain only selected one-dimensional deformations.
- Link Compliance: The simplest link model approximates the link as a beam and expresses nonzero compliance elements analytically from its geometry and material properties.The listed parameters include length, area, bending and torsional inertia, and Young’s and Coulomb’s modules.
- Link Compliance: For certain link shapes, a single-beam approximation can be insufficient and may require a serial chain of beams.The paper identifies this as an accuracy limitation of the simplified model.
- Link Compliance: The Orthoglide foot is evaluated by comparing alternative link stiffness models through compliance-matrix elements.The comparison is presented in Table 1.
3.3. FEA-based evaluation of stiffness parameters
For complex link geometries, the method evaluates stiffness parameters with a dedicated FEA model and derives matrix columns from separately applied loads and measured displacements.
- FEA-based evaluation of stiffness parameters: FEA evaluates link stiffness by applying three forces and three torques to a reference object and using the resulting linear and angular displacements.These displacement responses provide the columns of the stiffness matrix.
- FEA-based evaluation of stiffness parameters: For the Orthoglide foot, single-beam and four-beam approximations provide about 50% and 30% accuracy, respectively.The study also highlights mesh discretization and redundant reference-body motion data as accuracy considerations.
- FEA-based evaluation of stiffness parameters: FEA computational expense is limited because the proposed technique evaluates link stiffness only once rather than recomputing the entire manipulator for every posture.This contrasts with straightforward full-manipulator FEA modeling.
4. Application examples
Application examples compare Orthoglide-based 3-PUU and 3-PRPaR translational manipulators using stiffness models that account for their differing compliant architectures. The proposed approach captures parallelogram effects, handles singular postures, and achieves substantially improved agreement with FEA results.
- Simplified Orthoglide model: The flexible 3-PUU model computes Jacobians over four passive-joint variables while using FEA-derived link stiffness parameters.Unlike the nominal rigid model, the flexible model permits variations in all passive coordinates around their nominal values.
- Manipulator geometry: The study compares simplified 3-PUU and 3-PRPaR Orthoglide architectures, with the latter using kinematic parallelograms corresponding to the prototype design.The 3-PRPaR architecture adds kinematic constraints relative to 3-PUU, motivating a quantitative stiffness comparison.
- Parallelogram-based manipulator: The parallelogram model removes one dependent passive joint, reduces the passive-joint Jacobians to 6x3, and represents the element with a reduced 5-dof virtual spring.The reduced spring contains two translational and three rotational components with coupling between them.
- Stiffness comparison: Approximately 10-fold higher rotational stiffness is obtained with parallelograms than with U-joints within the dexterous workspace.The comparison covers three typical non-singular postures and two singular postures for the parallelogram architecture.
- Accuracy validation: About 5% accuracy is achieved for almost all workspace points, versus about 25–30% for previous methods.Near the flat singularity at Q2, rotational-stiffness error rises to 20%.
- Accuracy validation: The results justify applying the developed technique at the pre-design stage, while further research is needed to include joint stiffness alongside link and actuator compliance.The extended parallelogram model improved translational-stiffness accuracy but retained limited rotational accuracy near Q2.
5. Conclusions
The paper proposes a systematic stiffness method for overconstrained parallel manipulators and demonstrates it through Orthoglide-family comparisons and FEA validation. The method also has a defined scope and several stated extensions.
- Conclusions: The method models flexible links with FEA-evaluated multidimensional lumped parameters covering translational, rotational, and coupled compliance.It also solves kinematic and static relations together for each chain before aggregating the results.
- Conclusions: It computes stiffness matrices for overconstrained mechanisms at arbitrary postures, including singular configurations and their neighborhoods.The numerical algorithm inherently handles redundant passive-joint springs.
- Conclusions: Application examples compare the stiffness of Orthoglide-family manipulators with U-joint-based and parallelogram-based links.The reported simulations confirmed advantages of the parallelogram-based architecture and supported the Orthoglide prototype design.
- Conclusions: The proposed model’s accuracy was evaluated by comparison with FEA modeling.
- Conclusions: The method was applied to mechanisms with similar kinematic chains and actuators between the base and foot, while broader architectures remain future work.Planned directions include experimental Orthoglide verification and modeling gravity effects in heavy manipulators.
Appendix A. SVD-based computing of stiffness matrix for a kinematic chain with passive joints
The appendix derives a stiffness matrix for a kinematic chain with passive joints by applying SVD to the passive-joint Jacobian and resolving the resulting rank-deficient system. The procedure accounts for passive-joint motions when mapping end-effector displacements to external forces.
- Model setup: The chain model relates external forces, passive-joint variations, and end-effector displacement through the passive-joint Jacobian and virtual-spring compliance.The virtual-spring compliance is represented by a 6x6 positive-definite symmetric matrix.
- SVD transformation: SVD factorizes the passive-joint Jacobian into orthogonal matrices and a quasi-diagonal matrix of non-negative singular values.The factorization separates full-rank and rank-deficient components.
- Rank-deficient solution: The transformed equations set force components associated with nonzero singular values to zero and solve the remaining components from the first matrix relation.This treatment covers ordinary postures and singular configurations, where the Jacobian rank can decrease.
- Stiffness construction: The final chain stiffness is obtained after restoring the original variables and eliminating the passive-joint directions that accept motion without force or torque reactions.Its rank depends on the number of passive joints and the chain posture.
- FEA parameter extraction: FEA displacement data under six force and torque load cases are fitted to translation and rotation parameters, then scaled into a 6x6 compliance matrix.The fitted transformation uses a Procrustes least-squares solution and SVD; the resulting matrix is symmetrized by averaging corresponding off-diagonal elements.
Appendix C. Compliance parameters of the Orthoglide links
The Orthoglide link compliance parameters are obtained using actuator data and FEA-based link simulations. The results identify the most compliant components among the modeled manipulator elements.
- Compliance parameters: For the Orthoglide manipulator, actuator compliance was evaluated as 5 10^-5 mm/N, while link compliance matrices were computed through FEA-based simulation.The reported matrices use millimetres, radians, newtons, and newton-millimetres for length, angle, force, and torque.
- Compliance comparison: The feet and parallelogram bars are the most compliant manipulator components.The remaining elements have rigidity 5–10 times higher.