Vous êtes sur la page 1sur 138

ADAMS/Insight Training Guide

Release 2003

Visit us at: www.mscsoftware.com

The information in this document is furnished for informational use only, may be revised from time to time, and
should not be construed as a commitment by MSC.Software Corporation. MSC.Software Corporation assumes no
responsibility or liability for any errors or inaccuracies that may appear in this document.
Copyright Information
This document contains proprietary and copyrighted information. MSC.Software Corporation permits licensees of
MSC.ADAMS software products to print out or copy this document or portions thereof solely for internal use in
connection with the licensed software. No part of this document may be copied for any other purpose or distributed or
translated into any other language without the prior written permission of MSC.Software Corporation.
Copyright 2003 MSC.Software Corporation. All rights reserved. Printed in the United States of America.
ADAMS, EASY5, MSC, MSC., MSC.ADAMS, MSC.EASY5, and all product names in the MSC.ADAMS Product
Line are trademarks or registered trademarks of MSC.Software Corporation and/or its subsidiaries.
NASTRAN is a registered trademark of the National Aeronautics Space Administration. MSC.Nastran is an enhanced
proprietary version developed and maintained by MSC.Software Corporation. All other trademarks are the property of
their respective owners.
Government Use
Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR 12.212
(Commercial Computer Software) and DFARS 227.7202 (Commercial Computer Software and Commercial
Computer Software Documentation), as applicable.



About MSC.Software 6
Course Overview 7
Getting Help 8

Collaboration with ADAMS/Insight 12

MSC.ADAMS Product Solutions 13
DOE Theory and Response Surfaces 14
Recent Design Advances 15
Workshop 1Collaboration with ADAMS/Insight 19

Defining DOE Factors 26

Specifying DOE Responses 28
What are Response Surfaces? 29
Workshop 2Understanding Factors and Responses 31

Outdated: Parametrics by Hand 46

Nominal Models 47
Creating Factors in MSC.ADAMS 48
Creating Responses in MSC.ADAMS 49
Workshop 3Parametrics 50

Identifying What You Want to Learn About 66

Screening DOEs for Important Factors 67
Fitting Response Surfaces and Publishing Data 68
Workshop 4Classical DOE 69


Equation Order 78
Exporting and Importing 80
The Work Space Column Calculator 81
Analyzing Results 82
Workshop 5Results Interpretation 87

Overview 102
Templates and Execution 103
Process 104
Workshop 6Using the ASCII Conduit 105

Batching Simulation Runs 116

Suppressing Chart Updates 118
Composite Responses 119
Tying Factors 122
Workshop 7Advanced Topics 123

Answer Key for Workshop 1

Answer Key for Workshop 2
Answer Key for Workshop 3
Answer Key for Workshop 4
Answer Key for Workshop 5
Answer Key for Workshop 7



:(/&20( 72 $'$06,16,*+7

Welcome to ADAMS/Insight training. In this course, you learn how to use

DOE techniques to intelligently sample your design space and make rapid
predictions about the performance of other models within the design space.
Sharing this information in a dynamic, Web-based manner completes the
intelligent design study. Along the way youll learn about the basics of DOE
techniques, parametrics in the MSC.ADAMS products, classical DOE
methods, and strategies for analyzing your results.


About MSC.Software, 6

Course Overview, 7

Getting Help, 8







Or contact your local support center





Welcome to ADAMS/Insight


Response = 3 + 7X1 + X2 + 4X1X2 + 5X22

-5 <= X1 <= 5
-5 <= X2 <= 5

Welcome to ADAMS/Insight

To access the online help:

Press F1 when a dialog box is active

From the Help menu, select ADAMS/Insight Help


While working in an ADAMS/Insight dialog box, you can press F1 to display online
help specific to that dialog box.

Once the online help is displayed, you can also search for any terms you are looking
for or browse through the index or table of contents.

Table of contents for

selected tab

Demonstrate the online help. Press F1.

Read the FAQ for online help.

Welcome to ADAMS/Insight

To receive more consistent, targeted news and information, go to:
http://my.adams.com/cgi-bin/myadams.cgi, a Web personalization site for MSC.ADAMS
users. Some of the news channels this site provides are:

Case studies - Practical application stories

Company news - MSC.Software corporate and financial information

Events - Seminars, user conferences, and trade shows

Product alerts - Known problems, workarounds, and Service Packs

To find your local support center, go to:

To read the Service Level Agreement, go to:


Go to: http://support.adams.com/kb
For a quick tour of the knowledge base, go to:

Welcome to ADAMS/Insight



To join the community of MSC.Software users, go to:


Select MSC.ADAMS to view the MSC.ADAMS discussions.

Welcome to ADAMS/Insight

Explain the VPD Community tool (previously known as the ASK List) and guide the students to register during class.

Use ADAMS/Insight to investigate the relationship between the landing gear link
lengths (the design variables) and the storage clearance and torque requirements
(the design objectives). Make the results accessible to others in a Web page


Collaboration with ADAMS/Insight, 12

MSC.ADAMS Product Solutions, 13

DOE Theory and Response Surfaces, 14

Recent Design Advances, 15

Workshop 1Collaboration with ADAMS/Insight, 19


Retracted - bad

Retracted - good


Use ADAMS/Insight to explore the design space of a virtual prototype and optimize the design. Share the results
with colleagues and supervisors.



Cut design cycle time

Predict behavior without

physical tests

Get more information up-front,

learn about designs earlier

Have an easy way to share


Achieve efficient collaboration

between organizations







We think of the pyramid as a structured solution for the common problems outlined above; starting with a cost-saving
base of virtual prototyping, a natural progression is to have a rapid manner to evaluate
different virtual prototypes. This is all great, but an intelligent way to choose prototypes, then process and distribute
information is key.


The base is ADAMS/Solver, required for solving equations of motion of mechanical

system. Custom components are incorporated for elements such as tires and flexible

On top of ADAMS/Solver is ADAMS/View, the parametric engine to create

ADAMS/Solver models. Customized ADAMS/View products enable analysis of
common machine groups, such as automobiles.

ADAMS/Insight drives the parametric engines intelligently, helps interpret results,

and simplifies sharing information.




Minimize the number of design evaluations
using DOE theory to discretize the design
space intelligently.

Fit results of trial evaluations by assuming
design surface continuity, and gain the
ability to rapidly predict performance of
any model within the design space.



Virtual Prototyping




Now lets talk about the three levels of the pyramid and the design trends in the recent years.




Focus on developing ONE model of

ONE design.

Evaluation of design changes required

modeling expertise; could be time

Evaluation of design change was







Preprocessor developed for specific

mechanical systems.

Analysts maintained model


Standard modeling practices

leveraged throughout organization.

Attribute specialists no longer needed

to be MSC.ADAMS experts to get job

Faster A/B comparisons.







Analyst quickly evaluates design


Analyst quickly communicates to


Analyst enables collaboration within



This workshop takes about one hour to complete.

Use ADAMS/Insight to investigate the relationship between the landing gear link lengths (the
design variables) and the storage clearance and torque requirements (the design objectives).
Make the results accessible to others in a Web page format.
Retracted - bad


Retracted - good

Here you will start ADAMS/Insight, import the model, and run a simulation.

Change to the directory exercise_dir/mod_01_introduction.

Start ADAMS/View from the command line as follows:


Windows: adams03 aview ru-s i

adams03 -c aview ru-s i

Import the model file landing_gear.cmd.



Make sure no other databases exist before starting class. Before using a database for the first time in the class, make
a new database.


Run the simulation script retract_gear to see an example of the retraction exercise.

To run the model using this simulation script, from the Simulate menu, select Scripted

Enter retract_gear in the Simulation Script Name text box.

Select the Play tool

The simulation stops after a few seconds, when a sensor fires.


From the Simulate menu, point to ADAMS/Insight, and then select Display.

Enter lg_experiment.xml.

Select OK.
ADAMS/Insight opens. You will learn about the elements of the ADAMS/Insight interface
in later modules. For now, note what happens when you expand the Factors node, and then
the Inclusions node in the treeview. These are the design variables that you can alter in the
model. Now, expand the Responses node, and then the Inclusions node. These are the
design objectives, or Responses, that you want to monitor as the variables are changed.
Next, youll run through some simple design tasks.

Here you will use ADAMS/Insight to check the design space quickly through a process called
a perimeter study. The perimeter study evaluates your model at the extents of the design space,
giving you an indication of how robust the model is within the chosen design space.

In the treeview, select the + in front of Design, and then select Specification.
The Design Specification form appears.




Complete the Specification form as shown next:

Select Apply.

Select the Generate Workspace tool

ADAMS/Insight creates a matrix of model configurations representing models on the

perimeter of the design space.

In the treeview, select Work Space Review to verify the model configurations. Select Lower
link mount or Upper link length in the chart at the bottom of the main window to see how
many times each design variable value appears in the matrix of model configurations. In
this case, each variable appears once at the lower, middle, and upper range of the variable




Select the Run Simulations tool

Now that ADAMS/Insight knows what model configurations need to be evaluated for the
perimeter study, it has enough information to go back to ADAMS/View and evaluate all of
the model configurations.
ADAMS/View opens and the model runs in several configurations. If there are problems
with the model in any of the configurations, you will see warnings and error messages. In
this experiment, you should just see the warnings that occur when the sensor fires.

When the simulations are complete, close the Information box and select OK in the Mdi
Insight Display dialog box.

ADAMS/Insight displays the results data.


To view the results data, in the treeview, expand the Design node, and select Work Space.
The chart shows the model configurations that were evaluated and the values for each of
the torque and clearance design objectives. These values were completed based on the
results of the previous ADAMS/View simulations.

Did all of the models evaluated run and give values for Maximum Torque and
Wingbox Clearance?

What is the difference between the highest and lowest Maximum Torque values?

The perimeter study is intended to be a quick way to run the model at the low, middle, and high
values of the design variables, giving a quick indication of how robust the design space is. If
you notice any failures during the simulation process, or any design objective values appear odd
in the Work Space chart, now is the time to reconsider your model and the design space extents.

One of the main functions of ADAMS/Insight is generating response surfaces. A response
surface is simply a polynomial fit between your inputs (the design variables in the model that
can be changed) and the outputs (the design objectives that you are monitoring as the inputs
change). A polynomial fit between the inputs and outputs has many benefits, including:


The ability to rapidly look up any solution within the design space (simple
polynomial evaluation at any valid input value).

Rapid analysis of design space shape; essential for sensitivity analysis studies.

Then, after the response surface has been fit to the input/output relationship in this design
problem, perform an optimization on this surface. You will do a simple optimization on the
surface such that the model has:

The maximum wingbox clearance.

No regard for the maximum torque requirement.


Return to ADAMS/View by closing the ADAMS/Insight window:

From the File menu, select Close ADAMS/Insight Window.

Display the experiment lg_experiment_rs:

From the Simulate menu, point to ADAMS/Insight, and then select Display.

Enter lg_experiment_rs.xml.

Select OK.

Notice that there is an Analysis node containing a Model_1 node that represents the response
surface fit to the previous design problem.

From the Tools menu, select Optimization.

Assign weighting factors for the objectives that you want to satisfy:

Maximum Torque:

Weight of 0

Wingbox Clearance:

Weight of 1

Set Op in the Design Objectives column (the optimization criteria) to Max for the
Wingbox Clearance.

Select Run.
The optimization should finish quickly.

What configuration of the model yields the maximum clearance, without regard for the
torque Objective?



Since a response surface is simply a polynomial fit of the outputs to the inputs, it is easy to share
this information with others who dont know how to run MSC.ADAMS products. Response
surface data can be created in a dynamic Web page format from ADAMS/Insight, facilitating
in the sharing of the experimental data. Next, you publish a dynamic Web page from
ADAMS/Insight so that others can test model configurations within the design space, without
running the MSC.ADAMS products.

Select the Export to web, slk, etc. tool

Leave the default name in the File Save text box, and select Save.
Now the results can be viewed and tested from within a Javascript-enabled Web browser.

Open lg_experiment_rs.htm in a Web browser.

The Web page lets you alter the factors, or design variables, within the design space extents.
As you change the variables, the responses, or design objectives, change according to the
current location on the response surface. Assuming that the response surface fit was good,
these values indicate the performance of your MSC.ADAMS model without having to run

Assuming that your boss is not an MSC.ADAMS user and just wants the answers, what
values would he or she get for the torque and wingbox clearance objectives if he or she
set the design variables at 58 cm for the Upper link length and 39 cm for the Lower link mount?

You used ADAMS/Insight to verify the design space that you were investigating. Then, a
response surface was fit to the design problem where you could rapidly optimize for different
criteria. Once you found the design configuration(s) that you were after, you published the
response surface data to a dynamic Web page so that others could share the results.



Change pump characteristics to realize maximum flow output. Publish factorto-response relationship to a Web page.


Defining DOE Factors, 26

Specifying DOE Responses, 28

What are Response Surfaces?, 29

Workshop 2Understanding Factors and Responses, 31



Factors influence the behavior of the mechanical system.

Factors are also often referred to as:

Design variables

System inputs

Input parameters

Changing R, the radius factor, directly affects the output of the pumping system.



Factors and Responses


Factors must have limited bounds that represent the range of possible values.
ADAMS/Insight can specify factors using:

Absolute maximum and minimum values

Nominal value +/- relative value

Nominal value +/- percent relative value

max and

Nominal +/relative:

Nominal +/percent:

Lower limit


15 - 5

15 - 15 * 33.3%

Upper limit


15 + 5

15 + 15 * 33.3%

Factors can be either continuous or discrete.

Continuous: factors that can have all values within the upper and lower bounds.

Discrete: factors that can have only specified values within a range (for example,
car tires are available only in specified widths).

Factors and Responses



Are measurements of the system performance, or yield.

Are the quantities that youre interested in, asking questions such as:

What is the maximum yield for the given factor ranges?

How stable is a response value for some set of factors?

Are usually influenced by the factors.

The piston stroke volume is dependent on the piston radius squared. Varying the piston radius
yields the volume-to-radius relationship shown below. We can now express this as an analytical

Responses are defined indirectly in the MSC.ADAMS products using entities such as:


Result sets


Function values

Macro values

This is covered in detail in Parametrics on page 45.


Factors and Responses

It says that factors usually influence responses because sometimes they dont; thats one of the things ADAMS/
Insight is good for (the effects Pareto diagrams in the Web page show which factors influence which responses.)


DOE techniques look to find a relationship between the factors and responses in your

The data from the DOE trial runs is used to fit an algebraic relationship between the
factors and response. This relationship is called a Response Surface.

A simple example of this is the pump stroke volume-to-radius plot shown on page 28.
A slightly more complex response surface might show how two parameters affect a
system objective, looking something like this:

A response surface equation relates the response value to the factors, typically using
either a linear relationship, an interaction relationship, or a quadratic relationship.

Factors and Responses



This analytical relationship means that what-if studies are simple and quick for
further combinations of factors. As the design space is continuously approximated by
the response surface equations, you dont need to do any physical (or virtual) testing
to determine a response value for a different set of factors.

The response surface equations provide continuous knowledge of the

design space.


Factors and Responses

Emphasize that this is one of the main goals of ADAMS/Insight -- generation of response surfaces for rapid what-if
studies and ease of collaboration.

This workshop takes about one hour to complete.

Change pump characteristics to maximize flow output. Publish factor-to-response relationship
to a Web page.

First you will start ADAMS/View and import the model.

Start ADAMS/View from the working directory exercise_dir/mod_02_factors_and_responses

(where exercise_dir is the directory where your exercise files are installed).

Import the model wobble_pump.cmd.

Factors and Responses


Now, you will review the factors and responses, and then simulate the model.
In ADAMS/View there are two kinds of modeling entities that can be factors in
ADAMS/Insight. They are:

Design variables

Construction points

Design variables and construction points are used to parameterize a model in ADAMS/View,
so when ADAMS/Insight is given a model from ADAMS/View, it knows to look for these
particular parametric entities.

From the Build menu, point to Design Variable, and then select Modify.

Which of the design variables likely doesnt affect the dynamics of the pump?

There were no construction points used in this model. This is a bit unusual as most models use
construction points to parameterize key geometry locations or implement symmetry
relationships. To see a listing of construction points in a model, from the Tools menu,
select Table Editor. Then, select Points on the bottom as the filter criteria.

ADAMS/Insight looks for design objectives when ADAMS/View gives it a model. A design
objective is a single value that is derived from simulation results, such as the maximum value
of a reaction force or the last value of a velocity measurement.

From the Simulate menu, point to Design Objective, and then select Modify. Then, expand the
wobble_pump node.

Look at the piston volume objective. List the five kinds of entities that can be used to
define an objective in ADAMS/View:
___________, ____________, ____________, ____________, ____________


Factors and Responses


Note the objective obj_average_drive_torque. Apart from its oddly long name, it differs
from the other objectives in that it was defined based on a REQUEST. Workshop 3
Parametrics on page 50 covers the ways of making different kinds of objectives in

You want to ensure that our design variables (potential factors) influence the objectives
(potential responses) before you perform a DOE. First, you must simulate the model to generate
a baseline dataset from which to work.
Having a simulation script that runs the model is good practice, particularly if you need to string
together analyses (for example, you always do a static first, then run your dynamic simulation).

From the Simulate menu, point to Simulation Script, and then select New.

Specify the following parameters for the script:


Script Name:


Script Type:

Simple Run

End Time:

2 seconds



Simulation Type:

Transient - Dynamic

Clear Start at equilibrium

Select OK.
To run the model using this simulation script, from the Simulate menu, select Scripted

In the Simulation Script Name text box, enter run_pump.

Select the Play tool.

The pump simulation runs. Now there are results that you can use to evaluate the objectives
in the model.

Factors and Responses


There is no graphical way to view objective values in ADAMS/View unless youre explicitly
performing a DOE, design study, or optimization. To see objective values, you must use the
ADAMS/View command line.

Open the command line window by pressing F3.

On the command line, enter opti?

The command-completion functionality should activate when you press the question mark
(?) key, producing the result optimize.

Continue typing a few letters and using the question mark key until you get optimize
objective evaluate analysis_name = on the line. Now, press the ? key again and the only
possible choice, Last_Run, should appear.
You can see how the ? key on the command line either fills in valid command language,
prompts you for command syntax completion, or lists the valid entities after an equal sign.
The final string on the command line should look like the following:
optimize objective evaluate analysis_name = Last_Run
objective_name = obj_piston_volume

Press Enter to write the current value of the objective to the command window.
Using this technique you'll evaluate the three objective values and fill in the first column in
the chart below. Then, you'll manually change the model twice in the next step and reevaluate the objective values to complete the chart.


value (270,
41, 300):

Baseline with
higher shaft
speed (270, 41,

Baseline with
smaller radius
(270, 38, 300):


Before performing a DOE, make sure that the design variables (potential factors) influence
the objectives (potential responses) in our model.

Factors and Responses


Change the baseline variable set from:

piston_length = 270 mm

piston_radius = 41 mm

shaft_speed = 300 deg/s

to two new variables sets:

Baseline values+ (shaft_speed = 330 deg/s)

Baseline values + (piston_radius = 38 mm)

The objective values should respond accordingly.


Complete the chart on page 34.

Reset the variables to the baseline values before doing the DOE, next.

Now you will use ADAMS/Insight to fit a response surface to the relationship between the three
objectives and two of the design variables.

Transfer the current model from ADAMS/View to ADAMS/Insight by doing the


From the Simulate menu, point to ADAMS/Insight, and then select Export.

Specify the following:





Simulation script: use the simulation script that you wrote previously.

To see the current scripts in the model, from the Simulate menu, point to Simulation Script,
and then select Modify.

Factors and Responses



Select OK to launch ADAMS/Insight.

The treeview in ADAMS/Insight should correspond to the possible factors and responses
that ADAMS/Insight found in your ADAMS/View model.

Expand the Factors node, then the Candidates node, and then the wobble_pump.

Why do you think the bearing_color design variable didnt appear as a potential factor?
Factors arent included in the DOE until theyre promoted from the Candidates list to the
Inclusions list. To promote a candidate factor, select the factor name in the list then click
the Promote tool as shown below.

Promote the piston_radius and shaft_speed factors to the Inclusions list. Note that the
design variable characteristics of each factor have been retained from the ADAMS/View
model (the piston_radius factor is continuous, having a relative delta type; the
shaft_speed is also continuous, but has a relative percent delta type as specified by the
ADAMS/View design variable).

Note: For more information on factor and response specification methods, press F1 and

review the online help.


Factors and Responses


Change the default abbreviation for the factors as follows:

piston_radius: radius

shaft_speed: speed

Using the same procedures used earlier, promote the following responses to the
Inclusions list under the Responses node:



Change the abbreviations as follows:

obj_average_drive_torque: torque

obj_total_pump_output: pump_output

After defining the factors and responses you are going to look at, you now need to specify the
kind of study youre going to do. Do you want to determine which variables most affect the
different outcomes and do a Screening study? Or, do you want to determine the equations that
describe the input/output relation by doing a Response Surface fit?
This is a small experiment with only a few factors and responses. You are confident that each
factor influences at least one of the responses, so you will fit a response surface to the input and
output entities.

Select the Design Specification tool

Note: For more information on the entries in the toolbar, press F1 and review the online


Under Objective, select DOE Response Surface.

Set Model to Linear. This means that you want to fit a linear relationship between the
responses and factors so that you can write an equation relating them as follows:

Response_1 = a1 * Factor_1 + a2 * Factor_2

Response_2 = b1 * Factor_1 + b2 * Factor_2

Factors and Responses


You know that the total pump output varies with the piston radius squared
(cylinder volume = PI * cylinder_length * cylinder_radius**2), so youre simplifying
things assuming a linear relationship. The results will show whether or not this is a valid

Select Apply.

ADAMS/Insight now knows the kind of response surface being fit, as well as boundary
information about the factors in the model. Using this information, ADAMS/Insight determines
what trials (experiments) to run so that enough information is obtained to calculate the response
surface equations.
Each experiment is called a trial; a sequence of trials with factor values is called a project

Select the Generate Workspace tool

Select the Design Space and Work Space items under the Design node. They should display
as follows:

These tables indicate that four trials need to be evaluated with the factor values (as shown
in the Work Space table) in order to fit a linear response surface between the factors and

What is the significance of the -1 and 1 values in the Design Space table?

How do these match up to the factor values as shown in the Work Space table?


Factors and Responses

ADAMS/Insight displays the design position of all the model configurations that it is going to
evaluate. This allows you to inspect all of the models that will be evaluated.

From the Data menu, point to Simulation, and then select Preview.
ADAMS/Insight closes and ADAMS/View displays.

To pause after each frame, select Yes in the Warning box that appears.
The first model configuration displays.

Look at the piston and cylinder radii as you select Next frame in the Warning box. The
changes are subtle, but noticeable.

After all the frames are displayed, return to ADAMS/Insight:

From the Simulate menu, point to ADAMS/Insight, and then select Display.

The name of the experiment should appear in the dialog box.

Select OK.

The model configuration changes for this model werent too exciting, as the only thing that
visibly changes is the radius (you cant see the speed function changing in the design
configuration). For a larger model with many geometric factors, you can see how this could
be useful in verifying what is about to be done.

Factors and Responses


The Work Space table on the previous page didnt have torque and pump_output values for each
of the trials. To get the response values for each trial, you must run the model in ADAMS/View
(or ADAMS/Car, ADAMS/Chassis, and so on) to see the performance at each configuration.

To evaluate the trials, select the Run Simulations tool

ADAMS/Insight closes and ADAMS/View displays, running the model with the four
different configurations.

Here you will view your results and publish them to a Web page.

Return to ADAMS/Insight by doing the following:

From the Simulate menu, point to ADAMS/Insight, and then select Display.

Specify exp_wobble as the experiment.

Select OK.

In the treeview, expand the Design node, and then select Work Space.
Notice that all of the response values are now filled in.

Press the Fit Results tool on the toolbar.

A new node named Model_1 appears in the treeview under the Analysis node.

To review the statistics for the fit, select Model_1 in the treeview. In the two columns at
the bottom, note the following:

Summary Rules of Thumb: The yellow and red indicators show that the linear fit

between the design variables and the outputs is poor.

Torque Fit: The red indicators mean that a linear response surface isn't the best

choice for correlating the design variables to this design objective.

Pump Output Fit: The green indicators mean that there is some meaning in

relating the pump output to the design variables using a linear relationship.


Factors and Responses

You will learn more about the statistics in Results Interpretation on page 77, but for now,
you can make a simple change to your design to get additional information.
Since the linear fit assumes no interaction between the two design variables being studied,
you can change the assumed relationship to include the interactions between the variables.

From the Tools menu, point to Refine Model, and then select Change Order.

Select Interactions, and then select OK.

Note that the Fit data for each response now displays many green (good) indicators. This
indicates that an Interactions fit was likely more appropriate for this combination of design
variables and design objectives.

Change the model order back to Linear.

Factors and Responses



Press the Export to Web tool on the toolbar.

Enter the name of the HTML file you want to create.

Select Save.
You can now open your results file in a Web browser.

Now you will use a Web page to verify the response surface fit.

Verify that the response values on the Web page change as the factors are varied.

Fill in the table below with the response surface values when the model is in the baseline,
higher speed, and smaller radius configurations that were defined in the table on page 34.






with higher








Considering the differences shown above, do you think that a linear model was a good
choice? _____________

What were the four choices for the response-surface model fit?




Factors and Responses


Check the Stats and Effects boxes on the Web page. The effect of each factor on each
response is shown as a Pareto chart. The speed should have a large effect on the drive
torque response and the piston radius should have about half the effect of the speed on the
pump output response.

This module considered the differences between factors and responses. A typical goal when
running a DOE is to fit a response surface between the factors and responses so that physical or
virtual testing runs need not be done; an analytical relationship is used to predict the system
response for other configurations.
Some of the key elements that need further exploration are:

Setting up parameterization in your model

Interpreting the statistical results

Using different model fits for more accurate prediction

Reducing the number of significant factors through screening

Factors and Responses


To calculate the pump output, we assumed that the pump only discharges flow in one stroke
direction. To determine when the pump was discharging, a state variable (piston_velocity) was
created, having a positive velocity value when the piston moves into the cylinder.
MSC.ADAMS can integrate values in time using the DIFF statement, so the differential pump
volume needs to be written as a differential with respect to time.

V = volume

v = velocity


Total volume = integral (dV)

but dV = A dx, where

dx = piston differential displacement

A = piston area

dx = v dt

dV = A v dt

Therefore, the differential pump output can be written as A v dt. This value is continuously
calculated in the state variable piston_volume_differential. This is then used in a DIFF statement
(diff_integrated_volume) to calculate the pump output as a function of time.


Factors and Responses

Use the parametrics in ADAMS/View to investigate how to prepare a model
for an ADAMS/Insight DOE.


Outdated: Parametrics by Hand, 46

Nominal Models, 47

Creating Factors in MSC.ADAMS, 48

Creating Responses in MSC.ADAMS, 49

Workshop 3Parametrics, 50


Note that this process is tough and quite error-prone.

Without parametric preprocessors, the dataset model needs to be changed by hand, simulated,
and the results exported and accounted for.



Export and verify




It is assumed that a DOE study begins from a robust nominal model.

A robust model is one that has defined solutions throughout the design space,
meaning no ADAMS/Solver failures and physically realistic model configurations.

Most models are robust within a range of the variables being studied (that is, the
design space). The extent of the range is determined by the model. You should start
with a good idea of what these limits are.

The DOE method essentially tweaks the model about some operating point, so the
operating point needs to be within a robust design space.



Note the fourth point: DOE techniques tweak the model, so its very important that the model is robust within the
design space.

Factors are typically the objects in an ADAMS/View model that are defined parametrically.
These are:

Construction points: These are inherently parametric, made to be varied and have
other entities update as they change.

Design variables: Similar to construction points in that these elements define the
parametrics in your model. When you alter a design variable, anything referencing it
is also updated.

As in the earlier pump radius example, changing R directly affects the output of the pumping



Responses are the system quantities being monitored. These are defined as objective entities
within ADAMS/View.
Objectives can be defined by the following:

Measure characteristic

REQUEST characteristic

Result set component characteristic

ADAMS/View function

ADAMS/View macro and variable definition

An objective might calculate the maximum reaction in a joint or a derived quantity that
MSC.ADAMS doesnt automatically provide for you.



This workshop takes about two hours to complete.

Use the parametrics in ADAMS/View to prepare a model for use in an ADAMS/Insight DOE

First you will start ADAMS/View.

Change to the exercise_dir/mod_03_parametrics directory (where exercise_dir is the directory

where your exercise files are installed).

Start ADAMS/View from the command line as follows:



Windows: adams03 aview ru-s i

adams03 -c aview ru-s i

Import the model, car_suspension.cmd.


Next you will run several simulations.

Simulate the model for 1 second and 50 steps.

The model goes through a simple parallel wheel travel maneuver. The purpose of a
simulation like this is to verify the kinematics of the suspension system. Of particular
interest is the orientation of the wheel as it translates up and down.

Changes in the model geometry affect the wheel orientation-to-displacement relationship.

Verify this by doing two simulations with different mechanism geometries.

Run a dynamic simulation with the default model for 2 seconds and 100 output time steps.

Inspect the simulation results in ADAMS/PostProcessor by pressing F8 and then plotting

the x-component (U2) of the TCC (pseudo toe caster camber angle) user-defined

Return to ADAMS/View (press F8 in ADAMS/PostProcessor).

From the Tools menu, select Table Editor.

Modify the suspension geometry by filtering the table editor on Points (in the lower left
corner of the window), and change the x-location of hpl_tierod_outer from 417 to 422.




Simulate the model again and overlay the new plot with the original from step 2, above.
Your plot should look something like the one shown next.

This verifies how easy it is to make modeling changes in the parametric ADAMS/View
environment. Next you will repeat this manually to get a feel for how difficult this process is
without a parametric modeling tool.



To get an idea of how a DOE study might have been conducted in the past, you will alter the
model by hand. You will then explore the design space for the mechanism to determine settings
for a baseline, or nominal, model from which to start a future DOE.
Performing the process manually requires an ADAMS/Solver dataset model (.adm) file. For
each configuration of the model that must be analyzed, you will alter this dataset file and run it
through ADAMS/Solver to generate results. You can then analyze the resulting set of output
files for all of the configurations.

Reset the tierod location to have an x-value of 417.

Write out a dataset .adm model file that ADAMS/Solver can understand by doing the

From the File menu, select Export.

Select ADAMS/Solver Data Set as the file type.

Keep the default file and model names and select OK.

Run the dataset file through ADAMS/Solver to generate results for the original model.
The file run_default.acf has commands to load the model file named car_suspension.adm,
run the simulation for 2 seconds and 100 steps, and save the results with a filename prefix
of default_results.
Ensure youve written the .adm file out as car_suspension.adm and then simulate from the
command line as follows:


Windows: adams03 ru-s run_default

adams03 -c ru-s run_default

You should now have results files with .out, .gra, .msg, and .req extensions.

Open the .adm file in a text editor. You will change the model geometry by editing the
.adm file and running the simulation again.


When the students get to this point, make sure to emphasize that its tough and error-prone to do this by hand.


There are two markers in the model that are affected by the location of the point
hpl_tierod_outer in the ADAMS/View model. These are MARKER/18 on PART/6 (the left
tierod) and MARKER/17 on the left knuckle part. When you changed the location of
hpl_tierod_outer, these two markers changed locations due to the parametrics that have
been set up. The ADAMS/Solver dataset doesnt contain any parametrics; to mimic a
hardpoint change that you would make in ADAMS/View, change the marker locations in
the .adm file.

Change the x-location of MARKER/17 and MARKER/18 from 417 to 418.

MARKER/17, for example, would become: MARKER/17, PART=8, QP = 418, -750, 330

Save this modified dataset file as car_suspension1.adm. Change the x-location to 419, 420,
and 421, saving the files as car_suspension2.adm, car_suspension3.adm, and
car_suspension4.adm, respectively. Each of these will be simulated, yielding 4 sets of
ADAMS/Solver output files that you can analyze.

Run the run_all.acf file through ADAMS/Solver using the same commands as in step 3 on
page 53.

Did all of the simulations run successfully? ________________________

To investigate this, look at the output in the .msg files. There shouldnt be any errors or
warnings for these models.

Do the variable values affect the probability of a simulation running successfully?


10 Should a simulation run for any variable value, or should there be a range limiting the

variable values? __________________________________________

At this point, if you were really doing a DOE by hand, you would likely import the results file
data into a statistical package of some sort and set up the relationships of interest. We wont go
into this level of detail at this time.
As you can see, performing parametrics by hand is quite a chore.



You can run a design, or sweep, study in ADAMS/Insight to see what happens when you run a
factor linearly through a range of values. This is often done when initially exploring the design
space of a model to help understand how a design variable influences a particular aspect of the

Import a new copy of car_suspension.cmd, deleting the existing database.

Export the model to ADAMS/Insight, specifying the following:

Experiment: sweep_test


specify current model name



Promote only the hpl_tierod_outer.x factor.

Promote both candidate responses to be inclusions.

Set the Design Specification as follows:

Sweep Study


Number of Runs: 5

Select Apply.

Build the workspace and run the models.

Return to ADAMS/Insight by displaying the experiment sweep_test.

Review the workspace.

Look at the Terms Display for each of the responses in the Regression list. Note the
coefficient values for the tierod x-location for each response.

10 Is it reasonable that the right toe angle is not dependent on the left tierod location?




Any kind of design study, be it a DOE, optimization, variable sweep, and so on, requires bounds
for the variables that are being investigated. A response surface fit, for example, assumes that
the factor-to-response relationship can be expressed as a polynomial of some order. This means
that solutions exist for each combination of variables within the design space being considered.
A nominal model is one that represents a typical solution within the design space, and is also
robust when changes are made to it, meaning that it still solves in ADAMS/Solver correctly and
the results make physical sense.

Using the Table Editor as in the first section of the module, verify that you can alter
the location of the hpl_tierod_outer hard point by +/- 10 units in any direction, and the
model will solve without problems.

Only certain entities inside of ADAMS/View are automatically identified by ADAMS/Insight
as potential factors. These are construction points (or hard points) and design variables. Other
entities, such as marker locations or geometry attributes, arent inherently parametric, so they
arent considered factors in ADAMS/Insight. If a model doesnt contain the parametrics that
you need, you can always add the required parametric relationships.

Import a new copy of car_suspension.cmd, deleting the existing database.

Run a simulation for 2 seconds and 100 steps to create a simulation script in the

Export the model to ADAMS/Insight, specifying the current model and Last_Sim as the
simulation script.

How many potential factors are reported by ADAMS/Insight? ________

All of the entities are children of which body? __________________

Close the ADAMS/Insight window, returning to ADAMS/View. Open the Table Editor
(From the Tools menu, select Table Editor) and filter on Points in the lower left corner.

How many construction point entities are in the suspension model?





Does this correspond to the number of factors reported in ADAMS/Insight?


Delete the construction point hpl_tierod_outer as follows:

Click on the toolbox pointer to clear any selections in ADAMS/View.

From the Edit menu, select Delete.

Select the construction point under the ground part.

Select Delete All But Expressions.

10 How many potential factors are there now that one of the construction points is gone?


Return to ADAMS/View and add a design variable to the model. Create a design variable
that specifies the tire thickness as follows:

Right-click on the tire geometry and modify the wheel cylinder.

Right-click in the Length text box, select Parameterize, and then select Create Design

This creates a design variable having the current length value of the cylinder.

From the Build menu, point to Design Variable, and then select Modify. Change the
standard value of the design variable and the wheel width should update accordingly.

As design variables form the basis of parametrics within ADAMS/View, ADAMS/Insight

knows to include these as potential factors.
12 Verify this by exporting the model to ADAMS/Insight again. The number of potential

factors should have increased by one.



ADAMS/Insight looks for all of the defined design objectives within an ADAMS/View model
to determine candidate responses. The objective entity can be created in a number of different
ways within ADAMS/View. You can view these methods in the Modify Design Objective
dialog box.

Modify the toe_left_REQ objective in the model by doing the following:

From the Simulate menu, point to Design Objective, and then select Modify.

As in the previous module, list the five ways in which a design objective can be defined:
1. __________________________

2. __________________________

3. __________________________

4. __________________________

5. __________________________

Because a proper objective formulation is important in correctly defining your DOE study,
youll create one objective of each type for this model. Before creating the objectives, you need
a method to evaluate an objective value.

If a simulation doesnt exist, run one to generate a results set.

Press F3 to open a command line in ADAMS/View.

To evaluate the objective toe_left_REQ, enter the following:

optimize objective evaluate objective_name = toe_left_REQ
Hint: Use the command completion features discussed in Factors and Responses on

page 25.



You can create an objective based on the following:

Creating objectives based on measures

Creating objectives based on result set component and existing result set component

Creating objectives based on ADAMS/View functions

Creating objectives based on ADAMS/View variable and macro

Objectives based on measures are the simplest kind of objectives to create: make a measure
object in ADAMS/View, then create an objective and reference the measure object. Select from
properties of the measure such as maximum value, last value, average value, and so on.
In this section youll create an objective of this type that measures the maximum displacement
magnitude in the upper left ball joint, joint jl_uca_ball.

Right-click on the ball joint jl_uca_ball and create a measure on the marker
gel_upper_control_arm.marker_10 that measures the displacement magnitude. Name the
measure meas_marker_displacement.

From the Simulate menu, point to Design Objective, and then select New. Create an objective
with the following parameters:

Name: obj_max_displacement

Definition: Measure

Measure: meas_marker_displacement

Value: max. value during simulation

Rerun the simulation and evaluate this objective on the command line to verify that it
works as expected.

What is the maximum displacement value? ___________________________________



These are simple objectives to create as well. Instead of getting data from a measure object, the
objective gets data from a REQUEST or an existing result set. The existing results set method
is a newer way to create objectives. It can access the same data as REQUESTs, but it gives you
the advantage of browsing through a results set in ADAMS/View, looking for the component
you want.
The existing objectives in the model, toe_left_REQ and toe_right_REQ, are based on

In the Result Set Component text box, specify the REQUEST name.


Right-click in the text box, and select Result Set Component.

Use the Browse option to navigate through the result set objects to specify the result you

These are more powerful, but somewhat difficult to create. An ADAMS/View function can
operate on the data returned from a simulation, accessing the rich functions available in
ADAMS/View (see the guide, Using the ADAMS/View Function Builder, for a reference of all
available functions). You would use an ADAMS/View function to calculate a quantity that isnt
a minimum, maximum, last, or average value that is provided with the REQUEST- and
measure-based techniques.
For example, you use a function if you want to calculate a quantity in the frequency domain. A
Fast Fourier Transform (FFT) can be done on a dataset using the ADAMS/View function
FFTMAG(). Because the FFT must be done on a complete dataset, the calculation must be done
after a simulation is run, which is the case with an ADAMS/View function-based objective.



To get the maximum magnitude in the frequency domain of a joint angle measure named
JointAngle, for example, the ADAMS/View expression language would be:
AmpMax = MAX(FFTMAG(JointAngle, COLS(JointAngle)))

where the variable AmpMax stores the result.

Youll create a similar objective in the model that calculates the maximum amplitude in the
frequency spectrum for a measure of the upper control arm angle at the revolute joint. While not
the most exciting measurement, it is simple to visualize.

Create a measure named meas_uca_angle that tracks the joint angle in the revolute joint

Create an objective named obj_max_freq_mag based on an ADAMS/View function.

You need to create the ADAMS/View function now.

In the Function text box, right-click and create a new function.

The Function Builder appears.

Create the function by specifying the following:


MAX(FFTMAG(meas_uca_angle, COLS(meas_uca_angle)))



Passing analysis as the argument to the function is essential because the objective calls the
function and passes into it the analysis object containing all results from the last simulation.
Without this analysis object, the function wouldnt know from which analysis to extract the
meas_uca_angle data.




Test the function by selecting Evaluate. Name the function func_max_fft_mag and select OK.
The Create Design Objective dialog box should now look as shown next.

Select OK to create the objective.

Ensure that there are simulation results in the model, then test the objective by evaluating
it on the command line.

This is probably the most powerful way to create an objective, as macros allow for a lot of
flexibility in calculating quantities. While an ADAMS/View function is a single function
expression that can get long and complicated, a macro can contain lines of computations in the
same manner that a FORTRAN or C subroutine does. Macros are, therefore, more readable than
an ADAMS/View function expression.
In addition, macros can directly interact with the ADAMS/View environment. For example, a
macro could run the model for a bit longer if the simulation didnt reach a desired configuration
within the simulation time. For more information on creating macros, see Automating Your
Work Using Macros, in the guide, Customizing ADAMS/View.
To illustrate this, youll create an objective based on a macro that does the same computation as
the ADAMS/View function example did.




Create a new objective named obj_max_freq_mag_macro that is defined by an

ADAMS/View variable and macro.

Because there is no macro created yet, right-click on the Macro text box and select Create.

Name the macro mac_find_max_freq_mag and enter commands that are equivalent to the
ADAMS/View expression presented previously. Heres what the macro commands could
look like, splitting up the expression and adding comments to make things more readable:
! Macro to find max magnitude in frequency spectrum of
! measure named meas_uca_angle
! Create and initialize a variable that is the FFT of the measure:
variable set variable_name = the_fft &
real_value = (eval(FFTMAG(meas_uca_angle, COLS(meas_uca_angle))))
! Create and initialize a variable that contains the max value of above FFT:
variable set variable_name = max_mag real_value = (eval(max(the_fft)))
! To finish up, copy this value into variable named var_max_mag, as
! the objective expects:
variable set variable_name = var_max_mag real_value = (eval(max_mag))

Comment lines in macros start with exclamation points, with the exception of the very first line
which defines the type and name of any arguments passed into the macro.




Specify the variable name that the objective uses. This must be the same name as the
result variable, var_max_mag. The Create Design Objective dialog box should now have
these parameters:

Again, check the objective value on the command line. It should return the same result as
the ADAMS/View function objective definition.

Export the model to ADAMS/Insight to check that the new objective appears as expected.

There are several ways to create objectives in ADAMS/View. The higher complexity in
ADAMS/View functions and macros offers powerful flexibility for creating very specific
measures. Standard MSC.ADAMS quantities are very easy to capture using the Measure and
Request methods.



Use a classical DOE approach to represent a complex mechanical system as a
Web page model for others to use.


Identifying What You Want to Learn About, 66

Screening DOEs for Important Factors, 67

Fitting Response Surfaces and Publishing Data, 68

Workshop 4Classical DOE, 69



Any DOE study needs to have clearly defined objectives.

As the designer, you must identify the qualities you are trying to affect and identify
which measurables indicate these qualities.

You may need to set up new objectives and modify existing ones to ensure that the
quantities being measured are correct.

Dont try to create a single, all-encompassing objective. You can create several
different basic objectives, then later create a combined objective with weighting

The loads department may be interested in bearing reactions while the motor expert may only
be interested in drive torque requirements. You could create an objective for each of these to
satisfy both groups.


Its important to stop and think about what youre really interested in before starting a DOE study.

Classical DOEs


For any model with a large number of factors, it is very difficult to gauge which
factors affect which responses.

Confounding (factors interacting with one another) confuses the issue even more,
making design sweep study results misleading.

A screening DOE simply queries the system about an operating point and relates the
response changes to factor changes. This roots out the big player factors, letting you
intelligently reduce the size of your design space.

Large design space

(many factors)

Screening DOE

Factor effect

Reduced design

Classical DOEs


Once youve identified what youre interested in (the responses), you typically need to cut down the
number of factors that contribute to your responses, therefore the screening study. This just identifies the major
players in the factors.


Once a minimum set of factors has been identified, a response surface can be fit.

The response surface captures the factor-response relationship to a specified order:


Interactions (linear + term interaction)





Sharing the response surface data is easy in a Web page format.

This allows others to enter factor values and see the response results immediately.

Classical DOEs

A response surface is used to relate the inputs (factors) and outputs (responses). It is easy to share with others and
does quick calculations.

This workshop takes about one hour to complete.

Your supervisor and your co-worker want to test different specifications on your pump model,
but dont use MSC.ADAMS regularly. Use a classical DOE process to create a response surface
model in a Web page that lets them easily investigate their respective design problems.

First, you will start ADAMS/View and import the model.

Change to the exercise_dir/mod_04_classic_doe directory.

Start ADAMS/View from the command line, by doing one of the following:


adams03 -c aview ru-s i


adams03 aview ru-s i

Classical DOEs



Import the model file named wobble_pump.cmd.

This is the model from Factors and Responses on page 25, with additional design
variables and objectives.

Run the simulation script run_pump to generate results and verify the pump motion.

Verify the outputs by browsing through the plots in ADAMS/PostProcessor. Note the
new measure main_bearing_axial_load.

Return to ADAMS/View and look at the design variables in the model. There should be
new variables named:




and automatically created variables under the spring elements.

The first step is to determine which responses relate to which design goals. Look through the
objective definitions in the model and select one for each of the design tasks specified by your
co-workers. Write the objective name in the blanks below.
Design Tasks:

Best Objective:

Supervisor is trying to size a motor to drive

the pump, so is interested in torque/power

Supervisor should use objective:

Co-worker is studying stresses on the base
part and sizing the bearing needed for the
main shaft.

Co-worker should use objective:

For reference, you would choose obj_average_drive_torque to satisfy the supervisors design
requirements and obj_max_bearing_load for the co-workers task.


Classical DOEs

To provide your supervisor and co-worker with a viable design tool, youll have to pare down
the factor list. You could simply perform a response surface fit of the two responses with respect
to all of the factors, but a better way is to perform a screening DOE first.
Before doing a screening DOE, investigate the model and try to determine which factors will
affect the responses. In the exercise below, draw a line from each response on the left to each
of the two factors that you think will affect the response the most. One has been done for you.
Youll verify the predictions in the screening study next.






Now perform a screening DOE to test your predictions. A screening DOE is generally used to
identify the major contributors to the responses. It simply perturbs the model (evaluates at the
design space extents) and records the responses. It can show you which factors have the biggest
effect and which have relatively little effect so that you can perform a response surface fit with
a reduced set of factors.

Classical DOEs



Export the model to ADAMS/Insight.

Look at the candidate factors.

There should be six factors corresponding to the design variables and another 12
corresponding to the spring elements. The spring elements are a special kind of
parameterization that ADAMS/View uses internally and arent meant for public use. For
this reason, you will simply ignore these candidates and focus on the design variable ones

Promote all six design variable candidates to the inclusions list. Go through the list and
change the abbreviation for each to the factor name.

Promote the two responses chosen earlier (obj_average_drive_torque and

obj_max_bearing_load) and specify proper abbreviations.

Set the Design Specification with the following parameters:

DOE Screening (2 Level)


Note: For more information on Design Specification parameters, press F1 and review

the online help.


Look at the Design Type choices. How many runs are needed to do a Full Factorial
experiment? _________________

How many are needed for a D-Optimal experiment? _____________________

Select D-Optimal and enter 32 for the number of runs.

The D-Optimal method makes assumptions about the design space, enabling it to run fewer
evaluations than a full factorial design. Note that you can specify the number of runs within
a range of values. Youll use roughly 1.5 times the minimum number needed for a
D-Optimal study.

Select Apply.

10 Generate the Work Space, then run the simulations.

This takes a little while.



Fit the results in ADAMS/Insight, and export the results to a Web page. Name the Web
page screening_study.htm.
Classical DOEs

12 Look at the results in the ADAMS/Insight window.

The results will be covered in more depth in the next module, but for now look at the Rules
Summary node under Summary. The icons indicate that the model fit is good, but the

coefficient terms that were determined need to be looked at more closely.

13 Open the screening_study.htm file in a Web browser.

The next module goes through all of the results presented. For now youre only interested
in the Web page results.
14 In the upper window, select Effects checkbox.

This creates Pareto diagrams of the relative factor effects, giving a visual idea of which
factors are most important for each response.
15 What are the top two factors that effect obj_average_drive_torque?

__________________________, _______________________________
16 What are the top two factors that effect obj_max_bearing_load?

__________________________, _______________________________
17 To pare down the size of the design space, remove two of the factors to leave four

candidate factors. Which factors should stay?

_________________, _________________, _________________, _________________

Classical DOEs


Now you can perform a response surface fit using the four factors and two responses.

Save the screening study as screening_study.xml, so you can return to the results later if you
need to.

To speed up the evaluations, go back to ADAMS/View and make sure that the execution
display is set to Never update.

Export a new ADAMS/Insight experiment named rs_interactions.

Promote the four factors and two responses, remembering to give them meaningful

Select the Design Specification tool to create a new design. This time, use the parameters:

DOE Response Surface


Look at the number of runs for a Full Factorial design. This should be far less than seen
The Full Factorial runs every permutation of the factors and typically is a lot larger than the
other schemes. In this case, you likely have the time to run all 16 trials, so choose Full

Run the trials then return to the ADAMS/Insight window by displaying the experiment.
Fit the results to generate the statistics.

Look at the results under the Analysis node.

Under Summary, the Rules Summary node shows that the fit and the residuals are good, but
the terms need consideration. This is covered more in the next module, but it essentially
means that the fitted results correlate well with the experimental data, but ADAMS/Insight
doesnt have a lot of confidence in the exact form of the response surface (the coefficient

Look at the entries for each of the two responses. The Residuals node presents a table with
the actual and estimated values. Compare the differences for each response.

10 On average, which response has higher residual values?


Classical DOEs


Look at the response magnitudes and compare these to the residual values. The residuals
shouldnt be too different as a percent error value.

12 Publish the Web page for your co-workers to investigate.


Is one of the predicted response values worse than the other?


An interactions model was used to fit the results. What do the interaction terms
represent? ______________________________

If you changed the model order to a quadratic model, would there be more or less trials
needed? ______________________________

If you have time, create an experiment named rs_quadratic that performs a quadratic model fit
on the data. Use the D-Optimal model with about half of the number of runs that a Full Factorial
would need.

Do the term values seem to improve with this higher order (and more expensive) model?

Are the residuals better or worse?


Classical DOEs




Classical DOEs

Use ADAMS/Insight to design an experiment to discover a known response.
Use visual cues to refine the experiment.


Equation Order, 78

Exporting and Importing, 80

The Work Space Column Calculator, 81

Analyzing Results, 82

Workshop 5Results Interpretation, 87

Response = 3 + 7X1 + X2 + 4X1X2 + 5X22

-5 <= X1 <= 5
-5 <= X2 <= 5


ADAMS/Insight lets you model your system as a fit of the responses with respect to the factors.
The fits are:





You can easily visualize Linear, Quadratic, and Cubic least-squares regression fits for 1 factor
and 1 response. The following are examples of these fits:




Results Interpretation

The order of fit, corresponding equations, and minimum number of data points (trials) are
shown below:
Fit Type:




A + BX1 + CX2


A + BX1 + CX2 + DX1X2


A + BX1 + CX2 + DX1X2 + EX12 + FX22


A + BX1 + CX2 + DX1X2 + EX12 + FX22


Results Interpretation


+ HX2X1 + IX1 + JX2



ADAMS/Insight lets you export a design space to a file as comma-separated values

(csv). You can then fill in the response values using another tool, such as a
spreadsheet or custom application.

This is typically used for physical DOEs, such as the tuning of a physical prototype at
a testing track. This isnt usually applicable if youre using ADAMS/Insight solely
with the other MSC.ADAMS products.



After filling in the response values, you can then import a csv file into
ADAMS/Insight as a work space. You can perform the fitting on the data that you
entered for the responses.

Results Interpretation

The Work Space Column Calculator is a built-in tool for manipulating factor or response
column data. It is used when you want to scale a factors column value, or create a user-defined
response value from a simple algebraic expression.
It is easy to assign a known response value:
Response = 3 + 7X1 + X2 + 4X1X2 + 5X2^2

Or, scale a factor by a constant:

Factor_1 = Factor_1 * 2

The expression language is standard Python syntax. Press F1 in the calculator window for more
information on:

Python syntax


Variables available in your expressions

Results Interpretation



ADAMS/Insight provides graphical indicators that appear next to computed statistics

of interest. The indicators are defined as follows:
Entity needs investigation
Entity isnt necessarily bad, but
could use some consideration
Entity is likely appropriate



ADAMS/Insight compares the assumed solution to the actual data and provides a
wealth of statistical information. Refer to the ADAMS/Insight online help for
definitions of the terms, including:

The model fit

Term significance


Residual values

Results Interpretation


Provides a quick overview of the three main indicators for each response in the
model. Indicates important quantities to focus on for each response.


How do the predicted values in the model compare with the actual data?

R-squared and R-squared-adjusted indicate the variance in the predicted results versus
the real data. A score of 1 is a perfect fit.

Regression Significance (P) is the probability that the fitted model has no useful
terms. Small values of P (less than 0.02) indicate that the fit does have useful terms.

R/V: Range-to-Variance values greater than 10 mean that the fit likely predicts values
well. R/V scores less than 4 mean that values may not have high confidence.

Results Interpretation



One dataset, two regression fits. The statistics back up the visual effect.

The linear response fit to the dataset

is fine, but ADAMS/Insight flags
the R-squared value as suspicious.


The quadratic response fit to the

dataset is much better, earning very
high scores for all indicators.
Visually, the fit is much better.

Results Interpretation

Term significances estimate the probability that the term is insignificant in the calculated fit.
Very low scores mean that the term is highly significant in the fit.
The example below is the quadratic fit from page 84. Note that the constant term and factor_01
squared are most significant. Is it possible for a valid fit without including the second (factor_1)

These quickly show the influence of each run on the fit. They can help identify outlier results
that can be removed before the model is re-fit.

The first trial appears to affect the fit

more than the others, hence the
caution indicator.

Results Interpretation


Summarizes the fitted model characteristics. For each response, it shows the coefficient
(multiplier) for each term used in the fit.
Each row has the term definition
(constant, factor, factor*factor,
factor squared, and so on) with the
coefficient in front of each factor
coming from the fit process.

This means:
Response_2 = 3.2567 * 1 1.1103 *X1 1.4707*X2 + 3.1864*X3 + 1.421 *X1*X2 +
0.68597*X1*X3 + 1.0783*X2*X3
Response_1 = 3*X1 + 4*X2 + 5*X1*X


Results Interpretation

This workshop takes about two hours to complete.

Use ADAMS/Insight to design an experiment to discover a known response. Use visual cues
and the ADAMS/Insight online help when considering statistics.
Defined Response = 3 + 7X1 + X2 + 4X1X2 + 5X22

-5 <= X1 <= 5
-5 <= X2 <= 5

First you will start ADAMS/Insight.

Change to the exercise_dir/mod_05_results_interpretation directory.

Start ADAMS/Insight standalone from the command line as follows:


Windows: adams03 ainsight

adams03 -c ainsight

Youll use ADAMS/Insight to design an experiment aimed at discovering this governing



Look at the response that has been defined. This equation is the real system that
ADAMS/Insight will model. What is the highest order term in this equation?

Results Interpretation



ADAMS/Insight can fit response surfaces to your system that are linear, interactions,
quadratic, or cubic. Referring to the general form of each equation presented earlier and
the order of the defined response above, circle the model below that is likely the best fit
for this experiment:





You will build a full factorial RSM experiment using two factors and a quadratic model to
capture the squared term effects in the response definition equation.

Create a new experiment in ADAMS/Insight:


From the File menu, select New.

Add two factors to the experiment by doing one of the following:

From the Define menu, select Factor.

From the toolbar, select the Add factor tool

Results Interpretation


Modify the factors as follows:


leave default

Abbreviation: give meaningful name (factor_1, factor_2)



Delta Type:



-5, 5

Leave all other settings at defaults.


Select Apply to save your changes.

Create a response by doing one of the following:

From the Define menu, select Response.

From the toolbar, select the Add response tool

Assign a meaningful abbreviation (response_1) to your response. Leave all other default

Select Apply.

Build the experiment by doing one of the following:

Press the Design specification tool.

From the Define menu, point to Experiment Design, and then select Set Design

Define the experiment as follows:

DOE Response Surface


Full Factorial

Standard run order

No simulation script

10 Select Apply.

Select Generate Work Space.

Results Interpretation


Now that the experiment is built, the response values must be filled in. To do this you can export
a workspace from ADAMS/Insight as a comma-separated value (csv) file that can be read in to
most spreadsheet applications, or use the workspace column calculator that is built into
In this module, we will use the workspace column calculator in ADAMS/Insight to enter the
formula for the known response.
This calculator lets you enter simple expressions for any column of the workspace. So, you can
use it to:

Calculate response values

Scale or change factor values

The instructions for exporting a workspace in the form of a .csv file are included for reference

From the Tools menu, select Work Space Column Calculator.

The calculator appears as shown next:

From the Column to Compute menu, select Response_1.

In the Expression text box, enter the following:

3 + 7*factor_1 + factor_2 + 4*factor_1*factor_2 + 5*factor_2**2


Results Interpretation


Select OK.
Your workspace should now be complete and look as shown next.

Following are the instructions for exporting a workspace in the form of a .csv file. They are
included for reference only, as we used the workspace calculator instead. If you had data coming
from a different source, or the workspace calculator didnt suit your needs, you can use the
following steps to export the incomplete workspace and then import the completed workspace.

Export the incomplete workspace as follows:

From the File menu, point to Export, and then select Work Space.

Assign a name, such as workspace_export. ADAMS/Insight automatically appends the

.csv extension.

Import this new .csv file into a spreadsheet application and fill in the response column
using the formula from the problem definition:
Response = 3 + 7X1 + X2 + 4X1X2 + 5X22

Save the completed spreadsheet as workspace_completed.csv and import the workspace into
ADAMS/Insight (from the File menu, point to Import, and then select Work Space).

Results Interpretation


Before analyzing the results, the fit must be done on the work space.

Fit the model to the data using the Fit results tool on the toolbar.

Now you can analyze the results by looking at the model analysis summary.

Under Analysis, select the Model node.

Select the response from the Regression list in the form.

These statistics indicate the model fit to the data. The green dots for this model indicate that
the predicted equations match the test data very well. You will use the online help to
interpret the meaning of the various indicators.


From the Help menu, select ADAMS/Insight Help.

The online help for ADAMS/Insight appears.

From the top left area of the window, select Index.

A list of ADAMS/Insight terms appears.

Select a term and view its definition.

You can also search the entire ADAMS/Insight help by selecting the Search tab, and
entering a word (or group of words) in the search text box.
Now you will step through some of the more commonly used statistics and see if they make
sense in the context of the perfect example problem.


Results Interpretation


Expand the Analysis node, and select Model_01.

In the form, under Regression, select Summary.

Under Display, select Rules-of-thumb summary.

Though we havent talked about these items yet, does it seem reasonable that the Fit and
Residuals items have green lights for good?


Inspect the Fit statistics by selecting the Goodness-of-fit item. All of the statistics presented
indicate how well the model fits the measured data. Open the ADAMS/Insight glossary
and look up the definition of R-squared (the R2 statistic under Goodness-of-fit). Does the
model fit the experimental data well?

Next look at the Adjusted R-squared definition in the ADAMS/Insight glossary. The
present model should have a value of 1, indicating a perfect fit. If you had fit a linear
model and found poor R2 and R2adjusted statistics, you could go back and re-fit a
quadratic model to your results.

Results Interpretation



The P statistic indicates whether or not a regression or term is insignificant. In the model
summary context, the P statistic refers to whether or not the regression fit is insignificant
(see Probability and Regression Significance in the ADAMS/Insight glossary for more
information). What is the value for the present model? ____________________

Does this mean that there is a high or low probability that the model fit is a good one for
this experiment (circle one)?


The R/V or Range-to-variance ratio, indicates how well the model predicts values at the
data points. The ADAMS/Insight glossary indicates that a high value for the
R/V ratio is what? ________________

The statistics presented in this node indicate which terms of your model are useful and which
arent. This means that if youve used a quadratic model and all of the squared terms related to
your factors are insignificant, then perhaps you should reduce the order of your model (to an
interactions model, or a linear one).
Before looking at the significance values, investigate the terms displayed. Remember the
general form of the model that you have fit in this experiment. The single response is to be
approximated as a quadratic response surface equation in the two unknowns (factors).
Therefore, ADAMS/Insight fits the trial run data to an quadratic equation of the form:
Response = A + BX1 + CX2 + DX1X2 + EX12 + FX22
where the least-squares fitting process determines the A, B, C, D, E, and F coefficients. Take a
look at the description of each line in the Term column.


Results Interpretation


Inspect the statistics by selecting the Term significances entry.

Does it make sense that there are 6 terms? _______________

How many terms were actually specified in the problem statement equation?
As stated in the ADAMS/Insight glossary, the Term Significance values under response_1
show the probability that the particular term does not affect the response.

Is it proper that the terms with very small values have green (good) icons?

Is it proper that the only term having a bad icon (red with an X) is one of the squared
terms? (Refer to the response definition equation specified earlier.) Why is the squared
term bad?

To emphasize the above, suppose the response definition equation that was initially
specified had looked like this:
Response = 3 + 7X1 + X2 + 4X1X2
and you had fit the same quadratic model.
Which of the terms in the Significances table would likely have had bad icons?

Results Interpretation



Scroll down to the Term coefficients entry. Consider the assumed response surface equation:
Response = A + BX1 + CX2 + DX1X2 + EX12 + FX22
and the response definition equation:
Response = 3 + 7X1 + X2 + 4X1X2 + 5X22
Its quite clear from this table the term coefficients that ADAMS/Insight has determined for
the unknowns (A, B, C, D, E, F) in the quadratic model.

All of the coefficients are correct (the constant term is 3, the X1 term is 7, and so on)
except for the X12 term. Shouldnt this term be zero?


Results Interpretation

The last item that youll look at under the Summary node is Residuals. This column is a quick
way to verify that your response surface is accurate, showing the difference between the actual
trial value and the value as predicted by the response surface equations. These values are all
essentially zero for this experiment, as expected.
Each response has a Terms node underneath it that quickly summarizes the important points
discussed previously.

Under Regression, select response_1.

Under Display, select Terms. You can quickly see the:

Coefficient for each term

Probability that the term is not insignificant (icons show this at a glance)

Definition of the term

Other information, such as Standard Error and T test scores

The Fit entry gives model statistics at a glance. The R2 and R2adjusted values are likely of
the most interest; you can review the other statistics if the R2 and R2adjusted values dont
look good.
The difference between the least squares predicted response surface and the actual model
value is shown under the Residuals node for each of the trial runs that was performed.

Look at the values under the Residuals entry to ensure that the residual (error) values are
about the same for each trial.
An even distribution of error throughout all the trial runs is likely a sign of a reasonable
response surface fit to the design space. If only a few of the runs contained most of the error,
perhaps your design space is too complex for the order of model that youre trying to fit.
The ADAMS/Insight glossary defines the condition number, warning against results that
present large condition numbers.

Results Interpretation


The simple example that youve been looking at presents all the ideal statistics (R2 = 1,
P = 1.58E-30, and so on) that likely wont be found in a real experiment. What can you do if the
statistics indicate problems with your model? You can refine your model by doing the

From the Tools menu, point to Refine Model.

Notice that you have four choices:

Remove Outliers: Lets you completely remove a single trial from the results. This is
more applicable in a physical DOE when you have inadvertently altered other
factors and want to remove the trial.

Remove Terms: Lets you remove a term from the fit. Perhaps your fit has an
interactions term, but there is no actual confounding in the experiment. The
interactions term has a high probability of not affecting the response (the red
circle with black X), so you want to remove this term completely from the fit

Transform Response: Lets you scale your results.

Change Order: Lets you change the entire model order. This does the same thing as

the Remove Terms function, changing the assumed equation of the response
Now refine the existing model by getting rid of the bad term in the response equation.


Referring back to the Terms node under response_1, which term could be removed from the
assumed response equation? ______________________________________

Remove the X12 term from the model. Before doing this, write down the value of P in the
Goodness-of-fit node. ___________________

Remove the X12 factor:

From the Tools menu, point to Refine Model, and then select Remove Terms.

Select factor_12.

Results Interpretation


What is the new P-value for the model? __________________________

Note how the factor_12 row in the Terms node has been removed.

The response surface equation and the defined response now have the exact same terms so the
model fit is even better. What if you had originally chosen an interactions fit for the model? How
good of a fit could you achieve with a linear model for this problem? This can be investigated
as well using the Refinements node.

Put the factor_12 term back into the model:

From the Tools menu, point to Refine Model, and then select Remove Terms.

Clear the selection of factor_12.

Select OK.

Note how the factor_12 row reappears.


From the Tools menu, point to Refine Model, and then select Change Order.

Select Interactions.

Select OK.


Investigate the Fit and Terms nodes under response_01. How well does the interactions
model approximate the response? __________________

Do the term coefficients look correct? __________________

Which term changed the most in the new fit?


A linear model isnt appropriate for fitting the current design space. What if the factor
range was reduced to a smaller interval such as -0.5 <= X <= 0.5. Could a linear
response surface be used on a smaller portion of the design space?

Results Interpretation




Results Interpretation

Use ADAMS/Insight and ASCII conduit with ADAMS/Solver simulation files
that represent a bungy-jump event.


Overview, 102

Templates and Execution, 103

Process, 104

Workshop 6Using the ASCII Conduit, 105




Lets you parameterize generic model files from:

Other software packages

Custom, in-house software applications

Customizations of MSC.ADAMS products that you have developed

Brings ADAMS/Insight functionality to general analysis domain, extending its

capabilities beyond MSC.ADAMS products.

Following is a sample MSC.EASY5 analysis input file. Note the PARAMETER

VALUES section containing several design variable candidates.

Using the ASCII Conduit


Start command: adams03 ainsight ascg

Stores file templates that you can parameterize using the interface.

Allows you to specify pre- and post-run steps, and other customizable features.

The following example template file is for a standard ADAMS/Insight analysis.

The original data input has been replaced with

parametric information for ADAMS/Insight.

Using the ASCII Conduit



Generate nominal model files for model and analysis.

Open ASCII Editor and parameterize input files.

Save experiment and export to ADAMS/Insight.

Run ADAMS/Insight.

Model files


ASCII Editor


Using the ASCII Conduit

This workshop takes about one hour to complete.

Use ADAMS/Insight and ASCII conduit with ADAMS/Solver simulation files that represent a
bungy-jump event.

Here you will create your working directory and copy the necessary files.

Create a working directory called ain_asc_adm. This directory will contain all of the files
for this tutorial.

Copy the following files from <install_dir>/ainsight/examples to the newly created working



From a command prompt in the ain_asc_adm directory, copy the two MSC.ADAMS
simulation files by executing the following commands:

On Windows:

copy ain_tut_101_asc_adm.acf bungy.acf

copy ain_tut_101_asc_adm.adm bungy.adm


cp ain_tut_101_asc_adm.acf bungy.acf

cp ain_tut_101_asc_adm.adm bungy.adm

Using the ASCII Conduit



Ensure that the simulation runs by submitting the bungy.acf to ADAMS/Solver by

entering the following:

On Windows: adams03 bungy.acf

On UNIX: mdi -c ru-s i bungy.acf exit

Here, you will create an ADAMS/Insight ASC system. You will first start the
MSC.ADAMS/ASC editor and import two text files, bungy.adm and bungy.acf. You create an
ASC template by importing a text file(s) into an ASC system. In this tutorial, you import two
text files, which results in two ASC templates. (An ADAMS/Insight ASC system can have
many ASC templates.)
Once youve created these ASC templates, you will use the ADAMS/Insight ASC editor to
parameterize the ASC templates by annotating them. This process involves delineating regions
of text in the templates. In this way, you can identify parts of the template that will subsequently
be replaced. This substitution process occurs when variants of the nominal system are
automatically generated as part of a design study or DOE.


Using the ASCII Conduit


From the command prompt, start the ADAMS/Insight ASC Editor from the working
directory as follows:

On Windows: adams ainsight -ascg

On UNIX: mdi -c ainsight -ascg exit

From the File menu, select Import, or select the Import text file toolbar button.

Select the ASCII files bungy.acf and bungy.adm by holding the Ctrl key and clicking each

Select Open.

Annotate the two ASCII templates by identifying the regions that will be substituted, as

Highlight the text that is to be substituted.

Associate a variable with the highlighted text.

For example, if you have a string such as, PART/02 , MASS = 160.0, CM = 0203 and you
want to alter the mass of the part, you would first highlight the numeric value of 160.0,
and then right-click and assign the highlighted text to a variable (by either creating a new
variable, or referencing an existing variable displayed in the shortcut menu). When
creating a new variable, you can define the following:

Name: Descriptive name of the variable.

Format: Controls how the value will be printed. The convention follows the C
printf() convention of %d for integer, %f for float, %e scientific, and %s for string.
On UNIX, use the man printf command to get more information on this numeric

formatting convention. In most cases, you dont need to modify the default value.

Value: Default value that was originally highlighted.

Description: Optional supplemental information regarding the particular variable.

Note: The double curly brace delimiters '{{' '}}' that appear in the text file are the

default delimiters. You can change them on a template-by-template basis in

the template properties.

From the Template menu, select bungy.adm.

Highlight the numeric value of the mass of the jumper (160.0).

Using the ASCII Conduit



Right-click and select Create.

The Create Variable dialog box appears.

Change the value of the Name text box from F_05 to mass.

10 Leave the default values in the remaining text boxes.


Select OK.
The corresponding text is modified as follows:
PART/02 , MASS = 160.0, CM = 0203

PART/02 , MASS = {{mass=160.000000}}, CM = 0203
12 Highlight the numeric value of the function of the free-length variable.
13 Right-click and select Create.
14 In the Name text box, enter free_length.
15 Select OK.
16 Select the IC value of the free-length variable.
17 Right-click and select Replace.
18 Pick the variable free_length.

The following shows how the corresponding text was modified:

From: VARIABLE/01, IC=100.0, FUNCTION=100.0
To: VARIABLE/01, IC={{free_length}}, FUNCTION={{free_length=100.000000}}
19 Create another variable for the stiffness by repeating the steps above, using the following

From: VARIABLE/02, IC=4.0, FUNCTION=4.0
To: VARIABLE/02, IC={{stiffness}}, FUNCTION={{stiffness=4.0}}


Using the ASCII Conduit

20 To parameterize the title of the simulation, add the predefined variable of ascTrialName to

the title description so that the first line of bungy.adm ASC Template looks as follows:
Bungy Jump (ADAMS/Insight ASCII Conduit) {{ascTrialName}}__bungy.adm
Note: Be sure that there are two underscores between the closing curly brace and


The parameterization of the bungy.adm ASC template is complete. Now, you

parameterize the analysis names specified in the bungy.acf ASC template (first you open
the appropriate ASC template and then parameterize it).
21 From the Template menu, select bungy.acf.
22 Parameterize the first two lines of the bungy.acf ASC template so they look like the

Note: Be sure that there are two underscores between the closing curly brace and

23 Specify ASC template properties specific to the bungy.acf ASC template: from the Edit

menu, select Template Properties.

The Template Properties dialog box appears.
24 Now you will specify how the simulations will be run. Do this by completing the

following text boxes:

On Windows:

Execution Prefix: adams03 ru-s


Execution Prefix: mdi -c ru-stan i

Execution Postfix: exit

Using the ASCII Conduit


25 Enter the following Python dictionary definition in the Post Operations (Dict) text box:

{total_length:plts:max:req_1:3, max_acc:plts:max:req_1:4}

This string specifies what happens after the simulations are complete. Specifically, what
parts of the simulation results files will be interrogated or how the postprocessing will
Note: You can see an example of this string on a commented-out line in the bungy.adm

26 Select OK.

Now you will specify some optional ASC system attributes, such as a name and
description. This can be helpful for future reference.
27 From the File menu, select ASC Properties.

The ASC System Properties dialog box appears.

28 Complete the text boxes as follows:

Name: Jump

Description: Bungy jump tutorial example

29 To save the ASC system properties, select OK.

30 Save the ASC system to disk.

From the File menu, select Save As.

In the File name text box, enter j.

Select Save.

This creates a file on disk called j.xml. This is the ASC system with the two ASC
templates. You can view this file using a text editor or a browser.
31 From the File menu select Export Experiment.

This automatically generates an ADAMS/Insight experiment with the factors and

responses you defined in the ASC system. The default ADAMS/Insight experiment file is
called j_exp.xml.
Note: Look in the window that you used to start the ADAMS/Insight ASC editor for

warning messages. Make corrections as necessary.


Using the ASCII Conduit

Here you will create your experiment.

Start ADAMS/Insight from the ASCII Conduit by selecting the



Your experiment will automatically open in ADAMS/Insight.


Select the desired investigation strategy:

In the treeview, expand Design, and then select Specification.

In the Design Specification form, select Full Factorial as the Design Type.

Generate the workspace.

You should now have a workspace with eight trials.

Select the Workspace Column Calculator

(ASC-MER) file.

tool to write the ASC Multi-Event Regulator

The *_mer.py file is generated in the current directory. This Python script file works with
the ASC system file, and can be used to build, run, and post-process the ASC systems
defined in the _mer.py file. You can run it from the command line using the
MSC.ADAMS Python commands, or from the ASC system editor using the toolbar
buttons diff, run one, and run all.
Note: For more information on the options available with the _mer.py file, execute one

of the following commands:

On Windows: ADAMS03 python j_mer.py -h

On UNIX: mdi -c python j_mer.py -h exit

Select Yes when youre prompted to process the ASC system file j.xml.

Click OK.

From the File menu, select Exit, being sure to save your experiment.

Using the ASCII Conduit


Here you will run the ASC Multi-Event Regulator (_mer.py) file, test the configuration, and run

Open a command prompt window and change to the ain_asc_adm working directory.
Note: To get help, enter one of the following:

On Windows: adams python j_exp_mer.py -h

On UNIX: mdi -c python j_exp_mer.py -h exit

To test the configuration and only run the first trial, enter one of the following

On Windows: adams python j_exp_mer.py

On UNIX: mdi -c python j_exp_mer.py exit

The results of this operation are placed in a subdirectory with the default prefix of tst.

To run all trials, enter one of the following:

On Windows: adams python j_exp_mer.py -t

On UNIX: mdi -c python j_exp_mer.py -t exit

To check that the results files were created, go to the tst_dir subdirectory and view its
Note: You can also use the toolbar icons to run the trials.


Using the ASCII Conduit

Here you will import the results into ADAMS/Insight and then review them.

Start ADAMS/Insight from the ASCII Conduit by selecting the



From the File menu, point to Import, and then select Results.

Select the file tst_dir/tst_pos.csv.

To review the workspace matrix, in the treeview, expand Design and then select Work

Fit the results.

Export the model to an .htm file.

Review the model in a Web browser.

Using the ASCII Conduit




Using the ASCII Conduit

Evaluate failure-prone model simulations in batch mode and investigate the
actual toe-curve shape as a response. Tie factors for efficiency.


Batching Simulation Runs, 116

Suppressing Chart Updates, 118

Composite Responses, 119

Tying Factors, 122

Workshop 7Advanced Topics, 123


DOE trials that fail in the ADAMS/View simulation phase will prevent subsequent runs from
being evaluated. There are two ways to get around this:

Instead of evaluating the DOE trials within ADAMS/View, you can use ADAMS/View to
generate a sequence of .adm files and an .acf that runs the trials on the command line.
Simulation failures dont affect the .acf script file.




Advanced Topics

The other option is to specify that ADAMS/View use external Solver when simulating. This
means that ADAMS/View manages the simulation runs, but simulation failures in a single trial
dont halt the entire evaluation process.

Advanced Topics


By default, ADAMS/View (and other MSC.ADAMS products) plots response and factor values
for each trial. For models with many trials, this may take some time, as ADAMS/View must
overlay many curves in the strip-chart displays.
To speed up this process, you can turn off this plotting feature using the ADAMS/View
Command Navigator. Make these changes before transferring to ADAMS/Insight:


From the Tools menu, select Command Navigator.

Expand simulation, then multi_run, and then select set.

Set Chart Objectives and Chart Variables to no.

Advanced Topics


A typical response is a scalar value. In other words, it measures a single quantity.

Maximum value, minimum value, and average value are all examples of scalar
quantities derived from time-history data. Graphically, this might look like the figure
below where a maximum force value is being determined:

What if you want to capture information about the shape of the curve? What if you
need a mathematical description of how much curvature one data trace has with
respect to another? Composite responses allow you to do a simple polynomial fit to
mathematically capture the shape of a curve.

Examples where you might require a mathematical description of a data trace include
toe, caster, and camber curves in automobile suspensions, mechanism trajectories for
collision avoidance, and velocity profile matching.

Advanced Topics




A composite response is a group of several values that describe the shape of a curve.

ADAMS/Insight implements composite responses to represent curve shape using

polynomial coefficients. A composite response curve might look like the following,
where the coefficients A, B, C, and D define the curve:

A third-order composite response would be four coefficients that vary with model
configuration, so that each model in the design space has a curve associated with it.

Composite responses are implemented as objectives that return an array of values, not
just a single value.

The built-in POLYFIT() function takes a dependant array, an independent array, and
an order and returns the coefficients of the polynomial fit to the curve.

Advanced Topics


In ADAMS/Insight, composite responses look like four separate responses. Note the
response values r_01(0) through r_01(3) in the following figure:

The Web page now contains dynamic plots that change with the factors:

Advanced Topics



Ties relate common factor values to each another. A tie can be used whenever two
variables vary linearly with respect to each other for all design configurations. Ties
reduce the number of rows in the workspace, therefore reducing the total number of
trials to be evaluated.

One factor can be tied to another if the two vary in a linear manner.
Before tie:


After tie:

A common tie example is vehicle parameters. Changes on the left side of the vehicle
typically need to be tied to changes on the right side to ensure symmetry.

Tied factors display a Tie tab to help you calculate the individual factor values.

Advanced Topics

This workshop takes about one hour to complete.

Run the trial evaluations from the command line instead of in interactive mode. Create a
composite response representing the pseudo toe caster camber curve.

First you will start ADAMS/View and import the model.

Start ADAMS/View from the working directory exercise_dir/mod_06_advanced.

Import the model car_suspension.cmd.

Run the trial evaluations from the command line.

Advanced Topics



Create a simulation script of type ADAMS/Solver Commands that performs a kinematic

simulation for 1 second, 50 steps. Name it wheel_travel.
This will be used later to run the files on the command line.

Run the simulation. Note how the wheel twists as it translates up and down.

Open ADAMS/PostProcessor and plot the toe_angle vs. wheel_height measures. The plot
should look something like this:

Change the location of the construction point HP3 by +/-10 units in any direction.

Rerun the simulation and overlay the new toe curve on the original plot.

Did the curve change? __________________________________________

Here, you will explore how the location of HP3 affects the toe curve in ADAMS/Insight.

Export the model to ADAMS/Insight, specifying wheel_travel as the script and

toe_curve_experiment as the experiment name.

Promote all of the factors under HP3 to inclusions.

Set appropriate abbreviations for the x-, y-, and z-direction factors.


Advanced Topics


Let each factor vary by +/-10 units.

Promote the response obj_max_toe_angle to inclusion. Set an appropriate abbreviation for


Create the Design Specification as follows:

DOE Response Surface


Full Factorial

Generate the work space.

Do not evaluate the trials at this point.

Return to ADAMS/View by closing the ADAMS/Insight window.

Build the .adm and .acf files needed for the simulation by issuing the command:
mdi insight build experiment=toe_curve_experiment

on the command line (assuming toe_curve_experiment is the name of your experiment).

This should write out a sequence of .adm and .acf files. A dialog box appears, telling you
to run the file tst_bat.acf on the command line.
10 Verify that the .adm and .acf files were created by looking in the current working


Does the number of .adm files correspond to the number of trials required in
ADAMS/Insight? _______________________________________


Open a new shell in the directory containing the new .adm and .acf files.

Run the tst_bat.acf file through ADAMS/Solver on the command line.

Several simulation results should scroll by.

View the directory listing. There should now be .req, .res, .msg, and .gra files for each of
the trials.

Return to the ADAMS/View model.

Advanced Topics



Load the results into ADAMS/View using the command line syntax:
mdi insight load experiment=experiment ain_cmd_file=tst_bat.cmd

Fit the data in ADAMS/Insight.

Now, the data is available for standard ADAMS/Insight post-processing of the DOE trials.

Here youll make a new objective in ADAMS/View that will become a composite response in
ADAMS/Insight. You will use an ADAMS/View function called POLYFIT(), which returns
polynomial coefficients of a curve, to represent the toe angle with respect to wheel height.

In ADAMS/View, create a new function (from the Build menu, point to Function, and then
select New).

Set the following properties in the Function Builder:

Name the function func_toe_angle_vs_wheel_height.

Set Type to array, so that the function returns an array of values.

Set Arguments to analysis.

The function will be called by an objective object that youll create later. The objective
object will always pass an analysis set to this function (the last analysis performed).

POLYFIT takes an independent array, a dependant array, and the maximum order of
the polynomial. (For more information about POLYFIT, see the guide, Using the

ADAMS/View Function Builder.)


The composite response should represent toe angle vs. wheel height, so the
independent and dependant quantities are:

Independent: req_toe_angle_vs_wheel_height.f2

Dependent: req_toe_angle_vs_wheel_height.f3

Advanced Topics

The dialog box should now look as shown next:

Create a new objective on the command line. The Objective Builder dialog box doesnt
understand multiple return values from an ADAMS/View function, but the command line
does. Enter the following command (all on one line):
optimize objective create objective_name = obj_composite function_name =
func_toe_angle_vs_wheel_height n_components = 4

Test the objective that you just created by running a model and then evaluating the new
objective on the command line as follows:
optimize objective evaluate objective_name = obj_composite analysis_name = Last_Run

The objective value returned should be an array of four numbers representing the curve

Export the model to ADAMS/Insight as a new experiment.

Inspect the new response in ADAMS/Insight. It has a slightly different property than the
other responses, having a Response Type of User rather than Scalar.

Promote the same location in the factors list as the last time. Promote only the new
response for the responses.

Run a Response Surface, Interactions, Full Factorial DOE on the design.

Advanced Topics



Create the work space, evaluate the trials interactively, and then return to
ADAMS/Insight and fit the data.

10 Look at the resulting workspace. Note how obj_composite has four components

associated with it, as shown in the following image. Each component of the response has
the same statistics associated with it as a scalar response.


Export a Web page representation, and then open the Web page.

12 Select the Plots tool on the Web page.

This creates a plot of the composite response that looks as shown next:

13 Change the factor values, and then select Update Plots to replot the new composite

response. Notice that the curves change dynamically as you alter the Web page.


Advanced Topics

This appendix contains the answers to the questions listed in this training


Answer Key for Workshop 1, 130

Answer Key for Workshop 2, 131

Answer Key for Workshop 3, 133

Answer Key for Workshop 4, 134

Answer Key for Workshop 5, 135

Answer Key for Workshop 7, 137


Step 8, page 22:


About 400,000 units

Step 5, page 23: Upper link length is at the minimum value; lower link mount is at maximum

Step 4, page 24:


obj_max_torque = 4.7763e+05 N-cm

obj_wingbox_clearance = 1.9321e+01 cm

Answer Key

Step 2, page 32: outer_ring_color
Step 2, page 32: measure, result set, existing result set, ADAMS/View function, ADAMS/View

variable and macro

Step 4, page 34:

Baseline value
(270, 41, 300):

Baseline with
higher shaft
(270, 41, 330):

Baseline with
smaller radius
(270, 38, 300):


006 (mm**3)

06 (mm**3)

005 (mm**3)


006 (mm**3)

06 (mm**3)

006 (mm**3)






Step 6, page 35: See chart above.

Step 5, page 36: ADAMS/Insight is smart enough to know that the color variables arent

numeric, so it didnt consider them.

Step 3, page 38: The design space lists the variable extents; -1 is the lowest variable setting, 1

is the highest, and 0 is the nominal value.

Step 4, page 38: The work space evaluates the variables at the setting, so the -1 entry in the

design space should correspond to the variable lower-limit value in the work space.

Answer Key


Step 2, page 42:



Baseline with
higher shaft

Baseline with
smaller radius:















Step 3, page 42: The linear fit doesnt appear to be great. Perhaps a fit with higher order or

interaction terms might provide smaller differences between the predicted and actual responses.
Step 4, page 42: Linear, interactions, quadratic, cubic


Answer Key

Step 8, page 54: Yes.
Step 9, page 54: Yes. Variable values change the model. Some models might not be physically

realistic and wont run.

Step 10, page 54: There should be a range for any variable that your specify. This is a design
consideration that needs to be understood before adjusting the variables.
Step 10, page 55: Yes. The model is not parameterized symmetrically in this case.
Step 4, page 56: 60
Step 5, page 56: ground
Step 7, page 56: 20
Step 8, page 57: Yes. 20 points with 3 components per point equals 60 factors.
Step 10, page 57: 57
Step 2, page 58: Measure, result set, existing result set, ADAMS/View function, ADAMS/

View variable and macro

Step 4, page 59: 926.23 m

Answer Key


Specifying important responses, page 70: Supervisor should use obj_average_drive_torque;
co-worker should use obj_max_bearing_load.
Exercise, page 71: Both responses use shaft_speed and orifice_damping_coefficient.
Step 6, page 72: 64 runs
Step 7, page 72: 22 to 64 runs
Step 15, page 73: shaft_speed, orifice_damping_coefficient
Step 16, page 73: shaft_speed, orifice_damping_coefficient
Step 17, page 73: shaft_speed, orifice_damping_coefficient, piston_radius, full_on_time
Step 10, page 74: The response obj_average_drive_torque has high residual values, but the

residuals need to be compared to the actual values. Comparing relative values for both responses
shows that the residuals arent that different for each.
Step 1, page 75: When you look at the percent error, the predicted values are similar for both

Step 2, page 75: Confounding of the factors; the factors dont affect each response in a simple

linear manner.
Step 3, page 75: More trials are needed for the quadratic model (covered in Workshop 5
Results Interpretation, page 87).
Step 1, page 75: Yes
Step 2, page 75: Better


Answer Key

Step 1, page 87: 2nd order term (5X22)
Step 2, page 88: Quadratic
Step 4, page 93: Yes. Youre fitting a quadratic surface to a real quadratic, so hopefully you can

fit a good surface to the model. If the fit is good, the residuals should be low, therefore, good as
Step 1, page 93: Yes. Should be perfect (a value of 1).
Step 3, page 94: 1.58 E-30
Step 4, page 94: High
Step 5, page 94: Greater than 10
Step 2, page 95: Yes. The quadratic model fitted has six terms.
Step 3, page 95: 5
Step 4, page 95: Yes
Step 5, page 95: The squared term is flagged as bad because ADAMS/Insight determined that

the term is not significant. This is true as this term was zero in the original model.
Step 6, page 95: Both of the squared terms would have bad significances.
Step 2, page 96: The term is nearly zero (-3.553E-16). In computation terms, this is essentially

Step 1, page 98: The X12 term.
Step 2, page 98: 1.58E-30
Step 4, page 99: 3E-40
Step 1, page 99: Not very well at all.
Step 2, page 99: The X1, X2, and interactions terms are actually correct, but the offset term is

huge because its compensating for the missing higher-order terms.

Answer Key


Step 3, page 99: The offset, or steady-state, term.
Step 4, page 99: Likely, yes. Typically, reducing the size of the design space makes linear

approximations more reasonable.


Answer Key

Step 6, page 124: Yes, the curve should change somewhat.
Step 11, page 125: The number of .adm files should be the same as the number of trials.

Answer Key




Answer Key