Probabilistic-based approach using Kernel Density Estimation for gap modeling in a statistical tolerance analysis

Edoh Goka (1, 2) , Pierre Beaurepaire (1) , Lazhar Homri (2) , Jean-Yves Dantan (2)

1 IP - Institut Pascal
2 LCFC - Laboratoire de Conception Fabrication Commande

The statistical tolerance analysis has become a key element used in the design stage to reduce the manufacturing cost, the rejection rate and to have high quality products. One of the frequently used methods is the Monte Carlo simulation, employed to compute the non-conformity rate due to its efficiency in handling the tolerance analysis of over-constrained mechanical systems. However, this simulation technique requires excessive numerical efforts. The goal of this paper is to improve this method by proposing a probabilistic model of gaps in fixed and sliding contacts and involved in the tolerance analysis of an assembly. The probabilistic model is carried out on the clearance components of the sliding and fixed contacts for their assembly feasibility considering all the imperfections on the surfaces. The kernel density estimation method is used to deal with the probabilistic model. The proposed method is applied to an over-constrained mechanical system and compared to the classical method regarding their computation time.

Keywords: Tolerance analysis; Optimization; Form defects; Contact types; Monte Carlo simulation; Kernel Density Estimation

Domains: Mathematics [math]

Edoh Goka, Pierre Beaurepaire, Lazhar Homri, Jean-Yves Dantan. Probabilistic-based approach using Kernel Density Estimation for gap modeling in a statistical tolerance analysis. Mechanism and Machine Theory, 2019, 139, pp.294-309. ⟨10.1016/j.mechmachtheory.2019.04.020⟩. ⟨hal-02321275⟩