automated cam profile designer and optimizer

15
www.ricardo.com © Ricardo plc 2009 RD.08/######.# Automated Cam Profile Designer and Optimizer Ricardo Software European User Conference 2010 Farzin Montazersadgh Ben Playfoot

Upload: kaesar

Post on 17-May-2017

231 views

Category:

Documents


8 download

TRANSCRIPT

Page 1: Automated Cam Profile Designer and Optimizer

www.ricardo.com© Ricardo plc 2009RD.08/######.#

Automated Cam Profile Designer and OptimizerRicardo Software European User Conference 2010

Farzin MontazersadghBen Playfoot

Page 2: Automated Cam Profile Designer and Optimizer

2© Ricardo plc 2009RD.08/######.#

Project Objectives and Challenges

Objective– Automated optimization of valve event based on input lift and duration– Consider kinematic and dynamic indicators of control and durability– Reduce cam profile design cycle time– Maintain acceptable run times

Challenge– Significant number of potential input variables and levels

• 3 levels of 23 variables = 94x109 possible combinations!– Significant number of engineering criteria to meet

• Kinematic and dynamic guidelines at multiple speeds– Manual approach relies on engineering expertise and is difficult to

automate• Experienced based modification to shape of lift, velocity, acceleration

and jerk curves

Page 3: Automated Cam Profile Designer and Optimizer

3© Ricardo plc 2009RD.08/######.#

Cam Profile Design Methods

Multi-Polynomial approach– Traditional– Defined by 6 opening & 6 closing

5th order polynomials– Direct control of mathematical

constraints– Limited flexibility

Spline approach– Relatively new– Lift, velocity, acceleration

manipulated by click-and-drag– Interactive tool– Greater flexibility

Multi-polynomial approach chosen for optimization since user has direct access to mathematical constraints

Page 4: Automated Cam Profile Designer and Optimizer

4© Ricardo plc 2009RD.08/######.#

Excel Macro Development

Cam profile optimization is problematic since there are millions of potential designs

Even using a smart DOE (e.g. Latin Hypercube) thousands of runs are necessary to understand the potential cam design options

Objectives for developing excel macro– Reduce processing time– Easily find extremes or sensible boundaries for the DOE

Unexpected benefits– Useful training tool to understand the sensitivity of the cam profile to each

parameter• Just scrolling a sliding bar to see the affect of each parameter

Page 5: Automated Cam Profile Designer and Optimizer

5© Ricardo plc 2009RD.08/######.#

Excel Macro Snap Shot

Unacceptable AcceptableUnacceptable

Screening checks for sensible profile

Input variables to check sensitivity

Page 6: Automated Cam Profile Designer and Optimizer

6© Ricardo plc 2009RD.08/######.#

Automated Kinematic and Dynamic Optimization

1) DOE of valve profile coefficients

Screening of infeasible designs

2) Coarse optimization based on kinematic results

3) Refined optimization based on kinematic &

dynamic results

A variety of methods involving DOE and Optimization routines have been investigated

The best method uses the following approach– Step 1 – DOE using the Excel macro that filters to identify hundreds of potential profiles– Step 2 – Short Optimization to rank all feasible DOE profiles based on kinematic and

dynamic results– Step 3 – Final optimization to refine top ten profiles based on kinematic and dynamic

results

Final results highly depend on initial starting points, therefore a DOE is needed to provide multiple starting point for the optimization

Case studies showed about 5000 DOE profiles provide enough resolution to explore the design space, this cuts the DOE runtime to about 1.5 hours

Page 7: Automated Cam Profile Designer and Optimizer

7© Ricardo plc 2009RD.08/######.#

DOE of potential profiles with wide ranges for each

parameter

Acceptable acceleration

profile?

Run kinematic valvetrain model

Kinematic parameters meet

guidelines?

Change parameters within a certain range

Acceptable acceleration

profile?

Run kinematic valvetrain model

Kinematic parameters meet

guidelines?

Change parameters within a certain range

Acceptable acceleration

profile?

Run dynamic valvetrain model

Is LAI and Limiting Speed

maximized?

10 optimized profiles for the user

to choose from

Time constraint for each input

Final optimization loop

Yes

Yes

NoNo

YesNo

No

Yes

Yes No

Yes

No

Top

10 p

rofil

es a

re p

asse

d to

the

next

ste

p

DYNAMIC

KINEMATIC

KINEMATIC

Cam Profile Design and Optimization Flowchart

Page 8: Automated Cam Profile Designer and Optimizer

8© Ricardo plc 2009RD.08/######.#

i-SIGHT Integrated Model Snap Shot

Smart DOE (Latin Hypercube)

Short optimizationRun kinematic and dynamic on all feasible DOE results

Final optimizationRun kinematic and dynamic on top 10 from short optimization

Plotting final results

Reading screened profiles from DOE

Defining local boundary conditions

Kinematic run with VALKIN

Dynamic run with VALDYN

Saving all optimized profile

Optimization loop on each potential profile

Page 9: Automated Cam Profile Designer and Optimizer

9© Ricardo plc 2009RD.08/######.#

Optimization Objectives and Constraints

Objective and constrains could change based on user needs and targets

Common objectives/constraints:– Max contact stress– Oil film thickness (flat follower)– Min and max radius of curvature– Spring cover factor– Lift Area Integral (LAI)– Dynamic limiting speeds– Area under dynamic summary plots

Ricardo Guideline for Cam Lifter Separation

Ricardo Guideline for Seating Velocity

Cam Contact Stress

Page 10: Automated Cam Profile Designer and Optimizer

10© Ricardo plc 2009RD.08/######.#

Typical Dynamic Summary Plots

Graphs showing improvement in dynamic performance

Page 11: Automated Cam Profile Designer and Optimizer

11© Ricardo plc 2009RD.08/######.#

Typical Dynamic Plots at a Specific Engine Speed

Dynamic response of original and optimized valvetrains at 5500 rpm

Page 12: Automated Cam Profile Designer and Optimizer

12© Ricardo plc 2009RD.08/######.#

Case Study 1 – Pushrod Valvetrain Example 1

Objective: Increase Lift Area Integral (LAI) and cam/follower dynamic limiting engine speed Final results LAI Dynamic Limiting Engine Speed

– Original (heavily optimized) 0.4937 6710 rpm– Optimized_1 0.5011 (1.5%) 6745 rpm

Manual optimization of this profile was carried out and estimated time to complete is 1-2 weeks

Cam optimization program designed and optimized the profile in about 35 hours, with about 8 hours engineer time

Case study indicates at least 80% reduction in engineers’ time and a marginally improved cam profile

Valve acceleration (No Lash)

Cam/Follower Separation

Ricardo Guideline for Seating Velocity

6710 rpm

6745 rpm

Page 13: Automated Cam Profile Designer and Optimizer

13© Ricardo plc 2009RD.08/######.#

Case Study 2 – Pushrod Valvetrain Example 2

Objective: Increase LAI and cam/follower dynamic limiting speed

Final results LAI Dynamic Limiting Engine Speed– Original 0.5337 5350 rpm– Optimized_1 0.5649 (+6%) 5500 rpm– Optimized_2 0.5594 (+5%) 5550 rpm

Valve acceleration (No Lash)Cam/Follower Separation

5350 rpm

Ricardo Guideline for Cam/Tapper Separation

5500 rpm

Page 14: Automated Cam Profile Designer and Optimizer

14© Ricardo plc 2009RD.08/######.#

Case Study 4 – Pushrod Valvetrain Example 3

Objective: Increase LAI and keep original dynamic limiting speed

Final results LAI (above base circle)– Original 0.4333– Optimized_1 0.4519 (+5%)– Optimized_2 0.4407 (+2%)

Benchmark Target for Spring Surge

Spring Surge

Page 15: Automated Cam Profile Designer and Optimizer

15© Ricardo plc 2009RD.08/######.#

Conclusions and Further Development

Conclusions– One case study indicates about 80% reduction in engineers’ time, further

experience on future projects will help to determine real improvement in efficiency

– Designing cam profile and its optimization using this method frees up more time for the analyst to carry out studies such as parametric studies on geometry, stiffness, damping, HLA parameters etc.

– Small improvements were achieved on previously optimized profiles– Significant improvements were achieved on two non-optimized examples – A typical optimization run is about 35 hours and longer run times could be

used to investigate further optimization– Run time could be reduced even further by using VALDYN multiple run

function which uses multiple CPUs

Further developments– Design and optimize an asymmetric valve profile– Design and optimize valve spring