Combination of the finite element method and data mining techniques to design and optimize bearings

  1. Lostado-Lorza, R. 2
  2. Escribano-García, R. 2
  3. Fernández-Martínez, R. 1
  4. Illera-Cueva, M. 2
  5. Mac Donald, B.J. 3
  1. 1 Universidad del País Vasco/Euskal Herriko Unibertsitatea
    info

    Universidad del País Vasco/Euskal Herriko Unibertsitatea

    Lejona, España

    ROR https://ror.org/000xsnr85

  2. 2 Universidad de La Rioja
    info

    Universidad de La Rioja

    Logroño, España

    ROR https://ror.org/0553yr311

  3. 3 Dublin City University
    info

    Dublin City University

    Dublín, Irlanda

    ROR https://ror.org/04a1a1e81

Libro:
Advances in Intelligent Systems and Computing

ISBN: 9783319079943

Año de publicación: 2014

Volumen: 299

Páginas: 165-174

Tipo: Capítulo de Libro

DOI: 10.1007/978-3-319-07995-0_17 SCOPUS: 2-s2.0-84927713351 WoS: WOS:000343754200017 GOOGLE SCHOLAR

Resumen

Double-Row Tapered Roller Bearings are mechanical systems widely used in vehicles for the transmission of high load and moderate rotation speeds. These kinds of bearings are designed to withstand high contact stresses on their raceways, which are usually quantified using numerical methods such as the finite element method (FEM). This method has recently been widely used for designing mechanical systems, but has the disadvantage of requiring a high computational cost. The myriad of possible combinations of operating loads on the bearing (preload, radial load, axial load and torque) makes it much harder to calculate the distribution of these contact stresses. This paper shows the results of several regression models built using different Data Mining (DM) techniques that model and optimize the contact ratio obtained from the contact stresses in the outer raceway in Double-Row Tapered Roller Bearings. Firstly, a representative three-dimensional Finite Element (FE) model was generated according to the material properties, geometries and mechanical contacts of all parts which make up the bearing. Subsequently, a design of experiments (DoE) was performed considering four inputs (preload, radial load, axial load and torque), which were simulated in the FE model. Based on the contact stresses obtained from the FE simulations at different operating loads (inputs), a group of regression models (using linear regression (LR), quadratic regression (QR), isotonic regression (IR), Gaussian processes (GP), artificial neural networks (ANN), support vector machines (SVM) and regression trees (RT)) were built to predict the contact ratio which acts on the bearing. Finally, the best combination of operating loads were achieved by applying evolutionary optimization techniques based on Genetic Algorithms (GA) on the best regression models previously obtained. The optimization of the bearing was achieved when the radial loads obtained were the maximum value while the contact ratios were close to 25%. © Springer International Publishing Switzerland 2014.