Fatigue Design and Evaluation Committee |
FD&E HomeSearchEvents CalendarShort CoursesMinutes, Presentations, AgendasContact InfoATV ProjectWeld ProjectBolt ProjectDivisionsData OnlineLinksSteering Commitee
|
SAE Fatigue Design and Evaluation Committee Meeting Micro Minutes
April 4
& 5, 2000, Cedar Falls (Waterloo), IA Host: Ray
Thompson - John Deere Product Engineering Center Tuesday
8:00am Main Committee Meeting Announcements Phil
Dindinger opened the meeting and announced that SAE has initiated a new "SAE
Strategic Alliance." Committee
participants from non -SSA companies will no longer be assessed a participation
fee. However, there is a mailing
list fee if you are not from an "SSA" company. Retired participants
are staff exempted. Note that SAE is switching to entirely electronic
distribution of minutes and meeting notices. You will need to pull the minutes
and meeting notices off the SAE website. Please include your latest e-mail
address on the registration form. Local
Arrangements Ray
Thompson introduced Don Funsteen, Manager of Product Verification at John Deere
Product Engineering Center. Don
said the SAE FD&E Committee work is well recognized at John Deere Product
Engineering Center. Their
work in product verification and validation utilize the tools of this committee.
They are focused on doing a better job by using the analytical side of fatigue
analysis to build their products right the first time, use full tractor
evaluations at test sites across the country, and look forward to the progress
of work by this committee. Ray
announced a social hour starts at 6:00 p.m. today and a tour of the John Deere
tractor works is scheduled for 9:30 am on Wednesday. We will return at noon for lunch with planning sessions
starting at 1:00 p.m. The
instrumented ATV, just returned from a load measurement exercise at Sundown Ski
Resort, was on display in the lobby. Report on
Meeting with JSAE Fatigue and Reliability Committee: Phil
reported that the division chairmen met with representatives of the Japan SAE to
discuss mutual activities and interests at the recent SAE Congress in Detroit,
and are considering ways to work together. Their group meets six times per year
and consists of 26 members (17 from their auto companies, four and from material
suppliers, four university professors and one from a national lab.)
They have concentrated on evaluating fatigue of both mig and spot welds,
and are very interested in digital prototype optimizations. The JSAE group
presented two papers on fatigue of welds at the SAE Congress.
Andrew Whalen has discussed our ATV test project with them. If you are interested in participating with this group
contact, Russ Chernenkoff or Chris
Leser. Education Ralph
Stephens announced the SAE FD&E/U of Iowa Fatigue Concepts in Design
Short Course will again be held at the MSU Management Education Center in
Troy, Michigan, August 1-4, 2000. Steve Haeg has replaced Gail Leese on the
faculty this year. Ralph will send
out course notices. Please make the
course information available to your colleagues. H.O. Fuchs
Student Travel Award: Gavin Horn,
from the University of Illinois, presented his work on "Nondestructive
Residual Life Estimation for Damaged Fiber Reinforced Polyurethane and SRIM
Epoxy." Gavin has worked with
Peter Kurath, and acknowledged previous work by Ed Stanley and Ali Fatemi at the
U. of Toledo, and support fom the National Science Foundation and Tri State
Machining Company. Thermoelastic
stress analysis is based on an effect discovered by Lord Kelvin that the
temperature of a material changes in cyclic loading in proportion to the change
in stress. These small changes in
temperature are detected with infrared cameras. The sample is painted flat black to eliminate reflected hot
spots. A cyclically varying load is
applied to the specimen, and a thermal image of the component is obtained.
He defined a Modified Stress Concentration Factor as MSCF
= (Tlocal /(T far field and
Modified Stress as (mod
= MSCF x applied) The test
materials were polyurethane /fiberglass composite with about 9%(?) glass fiber,
and an SRIM epoxy/fiberglass composite with 40% glass fiber. The samples were
impacted on a Dynatup impact tower. Grayfield scanned images of the test
specimens were obtained after impacts of 6 joules up to 27 joules and
thermoelastic stress analysis showed progression of cracks.
The SRIM samples were tested at two levels , 17 Joules and 25.8 joules.
The lower impact level resulted in cracks just starting and at the higher level,
the impactor passed through the specimen. Fatigue
tests were run in a tension/tension mode at 2hz (with R=0.1). John Deere, U. of
Illinois, U. of Iowa, and U. of North Dakota fatigue data on polyurethane
composites showed wide scatter, but when the data was plotted against Modified
Stress Concentration Factor the data aligned nicely in a curve similar to the
baseline material, with some divergence at lower cycle levels due to heating
effects. Fatigue results for the
epoxy composite showed large scatter with applied stress, but damage lines up
with modified stress concentration factor. However, the fatigue results did not
show as good correlation as the polyurethane composite. Thermoelastic stress analysis can show inherent cracking and
delamination. Machining can cause
delaminating, matrix fiber to chip out, fiber breakage, and matrix overheating. Measured
machining damage to SRIM epoxy composites for a 6.3 mm diameter hole made from a
standard point drill (cost = $1) after 1000 holes were drilled were: heavy chip
out, spalling, and a rough surface inside the hole. The carbide brad point drill
(cost =$33) produced very clean holes with slight chip out, and the abrasive
water jet with 80mesh garnet produced heavy delamination and rounded edges.
Results from Thermal stress analysis and fatigue test results were Med. SCF
50% Prob. of Failure SEM Image
Standard Point Drill, 3.19, 288,000, fibers at edge Brad Point Carbide, 3.23, 271,000, clean cuts Abr. Water Jet,
3.89, 187,000, small cracks within 2mm of edge, of
hole Component
Testing Division Paul
Lubinski opened the meeting and the October 1999 minutes were approved. ATV
Structural Analysis Dan Klann
reported on the progress of structural analysis on the ATV project. Dan needs
loads information before attempting serious stress analysis.
U. of Illinois students built the original model in Hypermesh 2.0. Dan is
concerned about the mesh density and faceting, and wants to remesh.
He converted the model to Patran, but lost the connectivity information. Dan is concerned that the model needs refinement, and the
mesh is no longer associated with the geometry information. These problems should take one to two days to clean up. Dan
proposes to analyze just the frame first and needs the loads from a test
measurement or simulation. Mary
Wickham has talked to Darrel Socie about tests at the U. of Illinois. ATV Multi
Body Dynamics Simulation Ric
Mouseau has been developing a simple, easy to use, dynamics model to predict
durability loads information and support Adams and DADS simulation efforts. This
model uses the codes "Autosim" which creates source code and very
efficient simulations, and "CarSimEd" which interfaces the dynamics
model to the plotter and animation and executes in Windows. The model is a 24 d.o.f. rigid body model with revolute
joints and force elements and no suspension bushings.
The initial tire model consists of vertical inputs.
Inputs are vehicle speed and left and right road profiles. Ric showed the
results of a single impact event. He wants to clean up the model and ATV
parameter sets, implement a tire model lookup table, enveloping tire model,
bushings and ball joints, add additional acceleration outputs, post on the
website for people to evaluate, and eventually develop loads for structural
analysis. This might be an ideal model to investigate driver characteristics. ATV Data
Acquisition Gary
Mauritzon showed video tape of some data acquisition that took place recently at
Sundown Ski area. They recorded 56 channels of information including four wheel
forces, steering wheel position, suspension position, straingages, spindle F/A
accelerations, and lateral accelerations. Note that the electrician usually
carries a toolbox with him that is heavier than this instrumentation.
Most runs were made on dry land, but the ski area still had some patches
of snow. Four drivers, Al (at) 185 lb,
G.S. (at) 245 lb, G.M. and D.H. were used. Runs were made on and off the seat, one
aggressive run up and down the hill and one run on the snow patchs.
The ATV was also run over a flat bumpy area at the technical center. Bill
Nybeck has offered to analyze the data tape.
A question was asked: "What's next?" Despike, dedrift,
annotate, store on website, fatigue analysis of strain gage data, analyze effect
of different driver weights, store video, and determine FEA loads are some of
the tasks. The vehicle goes back to Technical
Presentation Mark
Prezlawski discussed "Load History Editing for Analytical and Experimental
Durability Assessments." (Mark announced that Kurt Munson who has also
contributed to this project could not be present because of the birth of
daughter, Lauren.) Currently
load histories are measured from and to everywhere, for example
from field to rig tests, or for use as input data for finite element
models. We often need to simplify
and establish continuity, and
control over raw and user data. Raw
data analysis manipulates data with
filters, mean removal, and re-scaling, establishes baseline fatigue damage for raw data, and selects raw events to
simulate. For example one test
evaluated eighteen strain gages over eight events for fatigue damage. They removed lower damage events through peak
/valley extraction (no frequency
content) peak/valley slicing maintaining synchronization across channels (still no frequency content)
and then rainflow cycle counting
(which reduces high count low damage cycles.) Frequency
correlated damage calculates damage as a function of frequency.
This highlights frequency regions of high damage content, and has an
impact on test development by comparing rig responses and influencing fixture
designs. If you remove a portion of
the frequency band does the fatigue life change?
Band pass filter, calculate damage and compare life. If rig only goes up
to 30 Hz you may miss 50 Hz. components. Manual
deletion of "smooth sections" is time consuming and not very
scientific. Damage editing can be
accomplished by Rainflow cycle counting the entire file.
Determine load at one second time slices and evaluate damage. Use damage
from critical locations. Evaluate % damage retained or % time reduction. For
example, one original test required 225 day versus 96% damage retained in a
12-day durability test. For parts with multiple input loads establish damage
from multiple combinations of input loads.
Look at all possible combinations, assess for every few degrees of
rotation. Refer to Vladimir Ogarovic's work. The question for a multiaxial part
is can we load uniaxially and
obtain reasonable results? One
final technique is based on finite element analysis where you take customer
accelerated tests and eliminate lower stressed nodes and with correlation find
an equivalent accelerated test. R. S.
questioned the consideration of corrosion effects? G.Glinka. stated that in the offshore industry editing out
small cycles is dangerous. R.L.said
that work on corrosion effects on fatigue has increased particularly on aging
aircraft. Also, editing out frequency content can cause problems.
Minor changes to design can cause major changes to response frequencies.
Standardized procedures still need to apply some degree of engineering
judgement. Fatigue
Life Prediction Division Chin-Chan
Chu opened the meeting and the October 1999 minutes were approved.
Al Conle has agreed to serve as vice chairman of the Fatigue Life
Prediction Division. Charles
Sieck discussed "Differences of Variability and Uncertainty."
Variability does not drive uncertainty. Deterministic
analysis takes discrete values and calculates discrete answers, where
probabilistic analysis uses probability distributions to calculate probability
of failures. Considering fixed
costs and variable costs there is some optimum number of loads to measure.
If we just looked at number of loads measured versus cost, it would
indicate we should measure no loads. If we consider sources of uncertainty like
variations of inputs, customer applications, material strengths, and
manufacturing in our calculations and put these variations into a Monte Carlo
simulation, we might determine that four or five load measurements would be
optimum. It is very rare that we get data based on 100 test samples. If we
required 90% confidence the optimum sample size might require measuring loads on
ten samples. In order to
characterize uncertainty we could say that we have a mean of 300 and a standard
deviatio Russ
Chernenkoff discussed updates on "Effects of High Mean Stress Study." Last
meeting Chin-Chan Chu proposed a project to determine the effects of high mean
stress on fatigue. They did a literature search and started tests at Ford on
gray iron. Some earlier work on
investment cast A356 T6 samples from Mike Mitchell includes some constant
amplitude tests results. They conducted some variable amplitude tests with high
mean stress and periodic compressive overloads, as developed by Dave DuQuesnay
for his MS thesis. They looked at
strain amplitude at 10000 cycles and Miners rule where D total = D small cycle +
D overload then looked at equivalent cycles to failure. There is not much data
available on the effects of high mean stress. Chin-Chan Chu would like to see
more comparison tests run with strain histories and material properties
available on the website. Al Conle
gave a brief update on "Standard
file formats for Load Histories and Material Properties."
Try the routine to create a digital curve for use in simple analysis
tools that is on the website at www.fde.uwaterloo.ca. Road Load
Data Acquisition Division Christoph
Leser opened the meeting and the October 1999 minutes were approved. For
comments on proposed Rainflow Counting standard contact Mark Pompetzki. The time
history format issue is closed. Contact Andrew Whalen or Al Conle for comments
on ASCII data for the web page. Chris
Leser reported on "Modeling of Non-stationary Variance in Vehicle Loading
Histories for Fatigue Loading." His objective is to find a concise fatigue
load distribution. His motivation includes storage reduction, improved FEA
capability, for test machines, monitoring, and to concentrate work on major
events. For stationary load
histories, the data reduction scheme includes a full time history recording in
the time domain, a stochastic stationary generally linear model.
A scheme for the non-stationary load history the scheme would include
counting methods such as Markov, or rainflow methods, deterministic nonlinear
models and autoregressive random characterizations.
Comments: Lee Tucker stated it would be valuable to be able to simplify
and condense real road load data and determine what are the characteristics of a
load history? How do we model a bump as a discrete event, and then what do we do
with it? Run an instrumented vehicle over a measured surface? We don't know the
road profile of the "Sundown R. L.: We
have a similar problem except load measurements are expensive and time
consuming. One company monitored
aircraft loads extensively for various events and calculated exceedance curves,
and monitered these curves over a long period of time. L. T.: There are two
types of loads: deterministic and vibration or resonance. The question is are we
making the right decisions? The customer controls changes to the operating
environments, and make the machines work harder, or as hard as possible and
still comfortable. T.C.:Is it a function of this group to figure out how to
compress the data used in the future? A.P.: Why
not use Kurt Munson's RPC test data for initial correlation? What happens in the
first stages of data acquisition? Why do we measure and what happens? D.K: Can
we run the vehicle over "steps" test track? What is the size of the
steps? This vehicle may not follow standard procedures. Discrete events should
be easier, and offer an incremental advance over D. Socies initial work. There
may be a suitable track available in Moline. Material
Properties Division Ed Lu
opened the meeting and the October 1999 minutes were approved. John Bonnen has
agreed to take the position of Chairman of the Material Properties Division. ATV
Project Support Phil
Dindinger reported on "Chemistry of the ATV Frame": The material was
characterized as SAE 1015 with very clean and very fine grain microstructure.
The tubing contained .08% carbon, .67% manganese, .03% copper, and .03%
chromium. Technical
Session Peter
Kurath presented " Nuances of Approximate Notch Methods ." Plastic
zone stress calculations from FEA models are usually handled with Neuber or
Glinka relations for small scale yielding or with Hoffman-Seeger for net section
yielding. For a plate with a center
hole and w/t = 50 look at plane strain and plane stress.
For a notched shaft with tension past yielding and bulk yielding, the
basic conclusion is that you can sometimes extrapolate a little farther. The
methods usually stop at bulk yielding. Nonproportional
loading is very path dependent. There
are significant differences between Seeger and FEA results for a "bowtie
path." Greg
Glinka reported on "Fatigue Crack Growth Through Residual Stress Fields
(Shot Peening)" In order to calculate Stress Intensity Function for an
application we need to calculate "Weight Functions" and "Green's
Functions." Find stress distribution without the cracks and superimpose for
Mode I loading. Calculate stress distribution and apply that stress distribution
to the crack surface. Gregg discussed an example of a crack in a thick walled
cylinder (a cannon.) FEA results
compared very well with his methods. Structural
Analysis Division Zheng Xian
Bai opened the meeting and the October 1999 minutes were approved. Technical
Session Pingsha
Dong, from Battelle, discussed " Mig-welded Structural System Fatigue
Simulations." Pingsha works on weld performance and processing in a
"Virtual Welding Lab" on property heat flow, weldability, fatigue and
fracture. He said there are many
fatigue related design issues with today's light weight tubular construction
including mig-welds on thin gage automotive structures, and hydro-forming
induced prior stress and strain conditions.
He showed a modified fatigue design rule approach that considered nominal
stress and defined a "structural stress" and extraction method that
reduces 16 design curves to two. . He also discussed a fatigue resistant joint
design and welding procedure optimization. Technology
Session (Wednesday Morning, April 5,2000) Alice Popescu-Gatlan and Merlin Green
discussed "Alternative Load Measurement Methods" The problem is to
move fatigue analysis into the product development process. Fatigue life
prediction requires loads information. ADV/RPC requires dynamic loads.
Complicated machines like a crawler mainframe have seven inputs per side. Some,
like lift cylinders and pivot pins are easy to strain gage. Cross-arms and final
drive housings are more complicated and require more structural analysis to turn
into transducers. Refer to M. Wickham's PhD thesis. This includes running an FEA
model to determine gage locations, writing out strain tensors for each element.
Ideally, choose one gage location for each load-case.
They use an optimization procedure to calculate gage locations and angles
to locate strain gages. Accurate
location of gages is very important, gage angle should be within 15 sec. of arc.
Calibration is done with unit loads applied one at a time in same direction as
FEA. After a final matrix is created a calibration check measures cross-talk. If
cross-talk is unacceptable the system is redone.
Som C.S. Do
you need to know the magnitude of loads? We
apply unit loads. Vertical loads are relative to jacking up tractor, F/A are
relative to a tractor pull. Pete
Kosmicki (and D. Klann) discussed " FEA Analysis of a Rig Test. " They
are trying to do more rig type test simulations with multiple load channels FEA,
with the objective of predicting when their structure will fail.
They have repeated tests experimentally with brittle lacquer and strain
gages and compared to FEA based results for verification.
The procedure includes: determine the component and principal stresses
for each of the three directions under unit loads. Modify sensitivities to
convert to strain histories, and adjust for "big loop" loading factor.
Calculate damage including "big loop" damage and a "net "
damage. They concluded that
accounting for "big loop" cycles helps to more accurately position
smaller hysteresis loops and provide better damage estimates. Ray
Thompson , "Understanding Product Use in a World Wide Market." Ray
discussed his involvement in product use validation process that includes an
overview of one tractor program with a 10 bottom plow in the United Kingdom, and
another tractor program in Turkmenistan with a 6 bottom plow.
Load histories were collected from customer with known heavy usage. These
histories produced high loads on chassis, axles, and hitches.
These loads are then used in analytical assessments for new designs
before hardware is built. Loads are used in durability tests for multiaxial
variable amplitude test to verify product durability. Product usage factors:
It is very important to know how customer actually uses the products,
effects of changes in farming practices, and usage in new markets.
The example from the U.K. had received complaints about the hitch, which
was used with an implement that was unique to the U.K.
Test planning included making sure of access to the implement, tractors
and the regular operators. They built the transducers and shipped to the site.
The U.K. plow is on Turkmenistan
is part of the former Soviet Union, independent since 1991,and is located north
of Iran. The Turkmenistan project
was a new market for John Deere with unknown operator conditions, and various
complaints causing a need to understand customer usage. Travel accommodations
required an invitation from Turkmenistan government, a visa was obtained upon
arrival at airport, U.S. dollars only in new bills, visitors were not allowed to
drive, few hotels available, they were cautions about the food and water, and
two interpreters were required. Ray
video taped inspections and operation of the tractors. Strain gage chemicals
were not allowed because of hazardous material requirements, so substitute
chemicals were improvised, vodka and superglue.
In summary understanding product usage is very important to structural
durability. They need to stay in touch with new applications and conditions,
when load measurements are required, and they need a database to share
information with design and test engineers. Divisional
Planning Sessions were held concurrently in two meeting rooms. The chairman and
the task will coordinate the future work efforts of these five Divisions group
leaders of the respective divisions. Please contact the chairman with any
questions or needs. Documentation of future work planned will be included in the
full minutes to be distributed before the nest meeting. Future
Meeting Schedule Fall 2000:
(October 9-11) University of Illinois, Host Darrel Socie Spring
2001: Detroit MI, Host Zheng Xain Bai GM
Fall 2001:
University of Toledo, Host Ali Fatemi Spring
2002: Detroit, MI - Host Russ Chernenkoff Respectfully
submittted by: John Hakala - SAE FD&E Vice Chairperson |
