Mft Crankshaft Proe Ansys Avl

download Mft Crankshaft Proe Ansys Avl

of 16

Transcript of Mft Crankshaft Proe Ansys Avl

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    1/16

    Robust Design Crankshaft OptimizationRobust Design Crankshaft Optimization

    A multi-disciplinary real application case

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    2/16

    Model definition and parameterisationModel definition and parameterisation

    Engine crankshaft, dynamically coupled with 5 radial bearings and 1axial bearing

    The geometry of the crankshaft (web shape) is defined andparameterised by ProEngineer

    10 geometric variables, that are assigned to the web side (Beziercurves) and to fillet radius

    Rup

    Bezup

    RmidRdown

    Bezdown

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    3/16

    Modal Analysis by ANSYS WorkbenchModal Analysis by ANSYS Workbench

    The first natural frequency is computed byANSYS Workbench (~150 Hz with free-free BC)

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    4/16

    EquivalentEquivalent discretizeddiscretized model by AVLmodel by AVL AutoshaftAutoshaft

    AVL-Autoshaft AVL-Excite

    The crankshaft is discretized by finite volumes: the equivalent model of masspoint is computed

    AVL-Excite uses this model to calculate the forces (inertial, active and

    reactive) acting on the crankshaft and in particular in correspondence of thebearings

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    5/16

    Dynamical coupling with bearings by AVLDynamical coupling with bearings by AVL--

    ExciteExcite

    The equivalent model of crankshaft is coupled with the bearings

    The general data (bearing type, friction coefficients, oil pressure andviscosity, roughness and clearance, etc.) are specified

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    6/16

    Hydrodynamic analysis of bearings by AVLHydrodynamic analysis of bearings by AVL

    ExciteExcite

    The fitness of the bearings isquantified by the evaluation of threefunctions, calculated by a not-

    stationary hydrodynamic analysisof the bearings in a complete cycle

    The complete cycle analysis isrequired since the contact point (and

    thus risk of damage) in the bearing isnot fixed during a cycle

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    7/16

    Definition of stochastic variables andDefinition of stochastic variables and

    optimisation objectivesoptimisation objectives

    Oil film pressure is considered as deterministic variable [1-10bar]

    10 geometric deterministic variables (web shape) Bearing clearance and roughness are stochastic variables

    (due to manifacturing tolerances): defined by a fixed mean

    value and a standard deviation

    D fi i i f i i i bj i

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    8/16

    Definition of optimisation objectivesDefinition of optimisation objectives

    (deterministic and stochastic)(deterministic and stochastic)

    Obj.1: minimize crankshaft mass (DETERMINISTIC)

    Obj.2: maximize first natural mode frequency (DETERMINISTIC)

    Obj.3: minimize mean of maximumoil filmpressure (STOCHASTIC)

    Obj.4 minimize mean of oil filmpowerloss (STOCHASTIC)

    Obj.5 maximize mean of oil film thickness (STOCHASTIC)

    Constr. on standard deviation of oil film pressure, loss andthickness to be lower than original (STOCHASTIC)

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    9/16

    Design parameters

    Input variables

    10 geometric var for crankshaft, oil pressure, 10stochastic var for bearings

    Output variables

    Mass, frequency, oil film

    power loss, pressure and thickness

    Design goals

    Minimize mass, mean of power loss and pressure

    Maximize frequency and mean of thickness

    Optimizations set-up data

    Sampling phase:

    Original configuration nr initial individuals: 1

    Exploration phase:

    MOGT scheduler 6 Simplex iterations for 4 players steps

    Constraints

    Standard deviation of oil filmoutput variables less than

    original

    modeFRONTIER workflow setupmodeFRONTIER workflow setup

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    10/16

    Color=Powerloss

    Diameter=Max pressure

    46 designs of the total 180 computed by MOGT belong toPareto front (not-dominated)

    The numbers indicatePareto front designs

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    11/16

    Pareto front designs are compared relatively to the 5 objectives

    Original and best design 156 (arbirarychosen) are indicated in green

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    12/16

    objective constraint

    Mass

    [mm3

    ]

    Frequency

    [Hz]Mean

    pressure

    [bar]

    Mean film

    thickness

    [mm]

    Mean loss

    [W]

    pressure

    [bar]

    film

    thickness

    [mm]

    loss

    [W]

    original 3.77 106 150 79.3 1.174E-5 628 6.77 1.48E-7 64.3

    best 3.54 106 164 77.0 1.175E-5 612 6.77 1.48E-7 64.3

    original best

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    13/16

    Comparison between original and optimized oil film pressure

    original best

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    14/16

    Values of clearance for 3 bearings vs power loss:only one bearing indicates a inverse linearity

    The distribution of clearancevalues is normal such as defined

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    15/16

    Similar conclusions for roughness vs power loss

    modeFRONTIER postmodeFRONTIER post--processingprocessing

  • 7/29/2019 Mft Crankshaft Proe Ansys Avl

    16/16

    ConclusionConclusion

    The basic definition ofRobust Design has beenformulated

    The necessity of Robust Design has been

    clarified by two examples in different fields

    The application ofMulti Objective Game Theoryalgorithmhas revealed particularly efficient forthe solution of these kind of applications