Dimensional Variation in Automotive Body Assembly Student: Timothy Ian Matuszyk Academic supervisory...

Post on 21-Dec-2015

215 views 0 download

Tags:

transcript

Dimensional Variation in Automotive Body Assembly

Student:Timothy Ian Matuszyk

Academic supervisory panel:

Prof. Michael Cardew-HallDr. Bernard F. RolfeDr. Paul Compston

Funding:

Australian Research Council Linkage Grant (#LP0560908)

Industry Partner: Ford of Australia

Territory front-cross #10922

Front Cross member

Front Cross member & Fender

Front Cross / Fender / Hood

Improving manufacturing processes

“In the future sustainable competitive advantage will depend more on new process technologies

and less on new product technologies”

(Thurow 1992)

Continuous quality improvement benefits

Higher quality assemblies, Less warranty concerns, Reduced launch time

Rigid vs Non-rigid assembly

Takezawa (1980) first showed that the additive theorem of variance does not hold for non-rigid assemblies, and that variation was in fact absorbable.

Rigid assembly Non-rigid assembly

21 hhH

h22

2

2

1 hhH

h1

H

4.111 22 H

1

2L 4L 5L

17L

38L 19L

8L7L

37L 36L

32L

33L

29L

30L

31L

27L28L

23R

22R

42L

21R

26L

14L

10L

11L41L

16L

43L40L

39L

Note: Points and locators mirrored on opposite side

Clamp/rest -

Rest -

1

2L 4L 5L

17L

38L 19L

8L7L

37L 36L

32L

33L

29L

30L

31L

27L28L

23R

22R

42L

21R

26L

14L

10L

11L41L

16L

43L40L

39L

Note: Points and locators mirrored on opposite side

Clamp/rest -

Rest -

Assembly x 9

Component D x 9 Component C x 9

Component A1/A2 x 9

Component B1/B2 x 9

• Observe and compare variation levels in components & assembly (9 samples)

• 38 points & 22 holes measured in final assembly

Initialstudy

Industry study findings Looked at production assembly issues & identified areas of investigation,

which included: Cases of variation levels decreasing over the assembly process Consistent positional shifts of holes from components to assembly

Lower Variation

Higher Variation

FE Assembly models

A way of simulating process variation stack-up. Linear models are fast but lack

accuracy. Non-linear models are more

accurate but are slow and suffer from convergence issues.

Thermo-mechanical approaches add even more complexity.

New data analysis possibilities

Optical co-ordinate measuring machines have allowed for quick and detailed inspection.

Shape characterization

Regression modelling of responses

Machine learning to deal with large data sets.

Aims

This project aims to identify:How component variation propagates through

an assembly processWhich process changes can reduce overall

variability in assemblies

Experimental vs FEM

•Actual process provides the best data

•Rapid prototyping

•Easy dimensional inspection

ADVANTAGES

•Time consuming

•Resources

•Model assumptions = less accuracy

DISADVANTAGES

How does part variation translate to assembly variation?

Assembly shape?Bow

Bow and

Spring-back

Twist

Do different processes affect final assembly variability?

Comparison of final assembly shapes for 3 different clamp sequences given the same input part variability (bow in the hat).

Data reduction and patterns

Component

shapes

Assembly

shapes

Key steps

1. Understanding/modelling variation transmission.

2. Structured experimentation to identify the variation of alternative processes.

3. Classifying component shapes into groups that share the same optimal process.

An adaptable assembly process

Imagine a process that can measure input components

and select the optimal assembly approach for

minimized variability in the final assembly.

In-line OCMM