Ijme Vol 11 n1 Fall2010

7/30/2019 Ijme Vol 11 n1 Fall2010

1/99

7/30/2019 Ijme Vol 11 n1 Fall2010

2/99

Finally, we are talking about AFFORDABLE PRICE

QUALITY, and CUSTOMIZATION. And we are talk

ing about each of them at the same time.

By the way, you don't have to be a computer

geek to use our systems. We have a couple of

them, and they will do all the technical mumbo

jumbo for you. We also have a few select people

from academics like you, and they know what

you do. You just relax and enjoy our systems.

If you are still reading this ad, chances are youare interested in our systems or services. So

visit us at www.acamedics.com. While you are

there, check the names of our clients as well

Most of them are quite familiar, but the list is too

long to include here.

f you are not using the Internet, please skip

this ad and enjoy the rest of the journal.

The average burden of reading this ad is

two minutes, which may add two months

of productivity to your academic life. Seriously!

Whether you organize a conference, publish a

journal, or serve on a committee to collect and

review applications, you can use the Internet to

make your life easier. We are not talking about

emails. We are talking about state-of-the-art onlinesystems to collect submissions, assign them to

reviewers, and finally make a decision about each

submission. We are talking about value-added ser-

vices, such as payment and registration systems

to collect registration fees online. We are talking

about digital document services, such as proceed-

ings CD/DVD development and duplication, or

creating professional-looking digital documents.

TIMETOCHANGETHEWAYACADEMICSWORKS?

TRYACAMEDICS!

Acamedics.com 44 Strawberry Hill Ave Suite 7B, Stamford, CT 06932 Phone: 203.554.4748 [email protected]

I

7/30/2019 Ijme Vol 11 n1 Fall2010

3/99

INTERNATIONAL JOURNAL OF MODERN ENGINEERING

INTERNATIONAL JOURNAL OF MODERN ENGINEERING

The INTERNATIONAL JOURNAL OF MODERN ENGINEERING (IJME) is

an independent, not-for-profit publication, which aims to provide the engineering

community with a resource and forum for scholarly expression and reflection.

IJME is published twice annually (Fall and Spring issues) and includes peer-

reviewed articles, book and software reviews, editorials, and commentary that con-

tribute to our understanding of the issues, problems, and research associated with

engineering and related fields. The journal encourages the submission of manu-

scripts from private, public, and academic sectors. The views expressed are those

of the authors and do not necessarily reflect the opinions of IJME or its editors.

EDITORIAL OFFICE:

Mark Rajai, Ph.D.Editor-in-ChiefOffice: (818) 677-2167Email: [email protected]. of Manufacturing SystemsEngineering & ManagementCalifornia State UniversityNorthridge18111Nordhoff StreetNorthridge, CA 91330-8332

THE INTERNATIONAL JOURNAL OF MODERN ENGINEERING EDITORS

Editor-in-Chief:

Mark Rajai

California State University-Northridge

Associate Editors:

Alok Verma

Old Dominion University

Li Tan

Purdue University North Central

Production Editor:

Philip Weinsier

Bowling Green State University-Firelands

Subscription Editor:

Morteza Sadat-Hossieny

Northern Kentucky University

Financial Editor:

Li Tan


Executive Editor:

Sohail Anwar

Penn State University

Manuscript Editor:

Philip Weinsier

Bowling Green State University-Firelands

Copy Editors:

Victor J. Gallardo

University of Houston

Li Tan


Publishers:

Jerry Waite

University of Houston

Hisham Alnajjar

University of Hartford

Web Administrator:

Saeed Namyar

Namyar Computer Solutions

7/30/2019 Ijme Vol 11 n1 Fall2010

4/99

INTERNATIONAL JOURNAL OF MODERN ENGINEERING |VOLUME 11,NUMBER 1,FALL/WINTER 2010

TABLE OF CONTENTS

Editor's Note: Upcoming IAJC-ASEE Joint International Conference ................. ........... ........... .......... ........... .......... ........... ...... 3

Philip Weinsier, IJME Manuscript Editor

Second Order task Specifications in the Geometric design of Spatial Mechanical Linkages ...................................................... 5Nina P. Robson, Texas A&M University; J. Michael McCarthy, University of California, Irvine

Application of Six Sigma to Gear Box Manufacturing .......... ........... .......... ........... .......... ........... .......... ........... .......... ........... ...... 12

Kambiz Farahmand, North Dakota State University; Jose Victorino Marquez Grajales; Mohsen Hamidi, North

Dakota State University

Microstepping Control of Stepper Motors with Data Interpolation and Direct Voltage Control........ ........... .......... ........... ...... 20

Dong-Hee Lee, Kyungsung University; Shiyoung Lee, the Penn State University Berks Campus; Tae-Hyun Won,

Dongeui Institute of Techology

An Enhanced Framework for Multi-module Embedded Reconfigurable Systems .............. .......... ........... .......... ........... .......... ... 27

Muhammad Z. Hasan, Texas A & M University; Timothy Davis,Texas A & M University;Troy Kensinger,Texas A &

M University;Sotirios G. Ziavras, New Jersey Institute of Technology

Sound Source Localization Employing Polar Directivity Patterns of Bidirectional Microphones ............................................ 35

Vijay Varada, The University of Toledo; Hong Wang, The University of Toledo; Vijay Devabhaktuni, The University

of Toledo

Modeling and Validation of Autonomous Rendezvous and Docking of Air Bearing Vehicles .......... .......... ........... .......... .......... 45

Amir Mobasher, Alabama A&M University; Paul Shiue; Christian Brothers University; Hossein Jamshidi, Alabama

A&M University; Zhengtao Deng, Alabama A&M University

Case Study of an Adaptive Automated Health Insurance Fraud Auditor .................................................................................. 52

Fletcher Lu, University of Ontario Institute of Technology

A Neural Oscillator Model for Tinnitus and its Management by Sound Therapy...................................................................... 58

Hirofumi Nagashino, The University of Tokushima; Kenichi Fujimoto, The University of Tokushima; Yohsuke Ki-nouchi, The University of Tokushima; Ali A. Danesh, Florida Atlantic University; Abhijit S. Pandya, Florida Atlan-

tic University

Hydroelectric Plant Design as a Synthesis Tool for an Engineering Technology Curriculum......... .......... ........... .......... .......... 67

Henry Foust, Nicholls State University; George Watt, Nicholls State University

Computation of Shockwave Structures in Weakly Ionized Gases by Solving Burnett and Modified

Rankine-Hugoniot Equations .......... .......... ........... .......... ........... .......... ........... .......... ........... .......... ........... .......... ........... .......... ... 74

Xiaoqing Qian, Z.T. Deng,Alabama A&M University

Interactive Data Visualization and Analysis for Mobile-Phone Performance Evaluation ........... ........... .......... ........... .......... ... 82

Yongsuk Lee, Florida Atlantic University (FAU), Xingquan Zhu, FAU, Abhijit Pandya, FAU, and Sam Hsu, FAU

Instructions for Authors: Manuscript Requirements........ .......... ........... .......... ........... .......... ........... ........... .......... ........... .......... . 95

7/30/2019 Ijme Vol 11 n1 Fall2010

5/99

Editors Note: Upcoming IAJC-ASEE Joint International Conference 3

EDITOR'S NOTE:UPCOMING IAJC-ASEE

JOINT INTERNATIONAL CONFERENCE

Philip Weinsier, IJME Manuscript Editor

IAJC-ASEE 2011Joint International Conference

The editors and staff at IAJC would like to thank you, ourreaders, for your continued support and look forward to see-

ing you at the upcoming IAJC conference. For this thirdbiennial IAJC conference, we will be partnering with theAmerican Society for Engineering Education (ASEE). Thisevent will be held at the University of Hartford, CT, April29-30, 2011, and is sponsored by IAJC, ASEE and IEEE(the Institute of Electrical and Electronic Engineers).

The IAJC-ASEE Conference Committee is pleased to in-vite faculty, students, researchers, engineers, and practition-ers to present their latest accomplishments and innovationsin all areas of engineering, engineering technology, math,science and related technologies.

Presentation papers selected from the conference will beconsidered for publication in one of the three IAJC journalsor other affiliate journals. Oftentimes, these papers, alongwith manuscripts submitted at-large, are reviewed and pub-lished in less than half the time of other journals. Please re-fer to the publishing details at the back of this journal, orvisit us at www.iajc.org, where you can also read any of ourpreviously published journal issues, as well as obtain infor-mation on chapters, membership and benefits, and journals.

IAJC Welcomes Three New AffiliateJournals

IAJC, the parent organization of the International Journalof Modern Engineering (IJME), theInternational Journal ofEngineering Research and Innovation (IJERI) and the Tech-nology Interface International Journal (TIIJ), is a first-of-its-kind, pioneering organization acting as a global, multi-layered umbrella consortium of academic journals, confe-rences, organizations, and individuals committed to advanc-

ing excellence in all aspects of education related to engineer-ing and technology. IAJC is fast becoming the association ofchoice for many researchers and faculty due to its high stan-dards, personal attention, fast-track publishing, biennialIAJC conferences, and its diversity of journals.

In 2010, IAJC accepted the Technology Interface Interna-tional Journal as the third official, IAJC-owned journal. Alsowelcomed to the growing list of affiliate journals are theInternational Journal of Engineering (IJE), theInternationalJournal of Industrial Engineering Computations (IJIEC) andtheInternational Transaction Journal of Engineering, Man-agement, & Applied Sciences & Technologies (ITJEMAST).With three official IAJC-owned journals and 10 affiliatejournals, authors now have a venue for publishing workacross a broad range of topics.

Current Issue of IJME

The acceptance rates for IJME range from about 20-45%.This issue saw an abundance of quality papers; thus, theacceptance rate for this issue was roughly 45%. And, due tothe hard work of the IJME editorial review board, I am con-fident that you will appreciate the articles published here.IJME, along with IJERI and TIIJ, are available online(www.ijme.us , www.ijeri.org & www.tiij.org) and in print.

International Review Board

IJME is steered by IAJCs distinguished Board of Direc-tors and is supported by an international review board con-sisting of prominent individuals representing many well-

known universities, colleges, and corporations in the UnitedStates and abroad. To maintain this high-quality journal,manuscripts that appear in the Articles section have beensubjected to a rigorous review process. This includes blindreviews by three or more members of the international edi-torial review boardwith expertise in a directly relatedfieldfollowed by a detailed review by the journal editors.

7/30/2019 Ijme Vol 11 n1 Fall2010

6/99

4 INTERNATIONAL JOURNAL OF MODERN ENGINEERING |VOLUME 11,NUMBER 1,FALL/WINTER 2010

Acknowledgment

Listed here are the members of the editorial board, whodevoted countless hours to the review of the many manu-scripts that were submitted for publication. Manuscript re-views require insight into the content, technical expertiserelated to the subject matter, and a professional background

in statistical tools and measures. Furthermore, revised manu-scripts typically are returned to the same reviewers for asecond review, as they already have an intimate knowledgeof the work. So I would like to take this opportunity to thankall of the members of the review board.

Editorial Review Board Members

If you are interested in becoming a member of the IJMEeditorial review board, go to the IJME web site (Submis-sions page) and send mePhilip Weinsier, Manuscript Edi-toran email.

Mehmet Bahadir Murray State UniversityRendong Bai Eastern Illinois UniversityBoris Blyukher Indiana State University (IN)Walter Buchanan Texas A&M University (TX)Isaac Chang Cal Poly State University SLO (CA)Hans Chapman Morehead State University (KY)Rigoberto Chinchilla Eastern Illinois University (IL)Raj Chowdhury Kent State University (OH)Michael Coffman Southern Illinois University (IL)Kanchan Das East Carolina University (NC)Brad Deken Southeast Missouri State University (MO)Dave Dillon North Carolina A&T State UniversityMehran Elahi Elizabeth City State University (NC)Ahmed Elsawy Tennessee Tech University (TN)

Bob English Indiana State University (IN)Rasoul Esfahani DeVry University, USAFereshteh Fatehi North Carolina A&T State U. (NC)Verna Fitzsimmons Kent State University (OH)Richard Freeman US Coast Guard Academy (CT)Vladimir Genis Drexel University (PA)Liping Guo Northern Illinois University (IL)Rita Hawkins Missouri State University (MO)Youcef Himri Safety Engineer in SONELGAZ AlgeriaXiaobing Hou Central Connecticut State University (CT)Shelton Houston University of Louisiana at Lafayette (LA)Luke Huang University of North Dakota (ND)Dave Hunter Western Illinois University (IL)Pete Hylton Indiana University Purdue (IN)Anwar Jawad University of Technology Baghdad IRAQ

Rex Kanu Ball State University (IN)Khurram Kazi Acadiaoptronics (MD)Ognjen Kuljaca Alcorn State University (MS)Chakresh Kumar Uttar Pradesh Technology Uni. (INDIA)

Zaki Kuruppalil Ohio University (OH)Ronald Land Penn State University (PA)Jay Lee Purdue University Calumet (IN)Shiyoung Lee Penn State University Berks (PA)Soo-Yen Lee Central Michigan University (MI)Chao Li Florida A&M University (FL)Jimmy Linn Eastern Carolina University (NC)G.H. Massiha University of Louisiana (LA)

Jim Mayrose Buffalo State College (NY)Thomas McDonald Eastern Illinois University (IL)David Melton Eastern Illinois University (IL)Sam Mryyan Excelsior College (NY)Wilson Naik University of Hyderabad (INDIA)Arun Nambiar California State U.Fresno (CA)Argie Nichols University Arkansas Fort Smith (AR)Hamed Niroumand University Teknologi MalaysiaTroy Ollison University of Central Missouri (MO)Reynaldo M Pablo Jr. Indiana University - Purdue (IN)Basile Panoutsopoulous United States NavyJose Pena Purdue University Calumet (MI)Karl Perusich Purdue University (IN)Thongchai Phairoh Virginia State University (VA)John Rajadas Arizona State University (AZ)

Desire Rasolomampionona Warsaw U. of Technology(POLAND)

Mulchand Rathod Wayne State University (MI)Sangram Redkar Arizona State University-Poly (AZ)Michael Reynolds University Arkansas Fort Smith (AR)Mehdi Safari Isfahan University of Technology (IRAN)Anca Sala Baker College (MI)Hiral Shah St. Cloud State UniversityEhsan Sheybani Virginia State University (VA)Carl Spezia Southern Illinois University (IL)Randy Stein Ferris State University (MI)Adam Stienecker Ohio Northern University (OH)Pingping Sun IBMJalal Taheri Bostan Abad Islamic Azad U. (IRAN)Li Tan Purdue University North Central (IN)

Li-Shiang Tsay North Carolina Ag & Tech State (NC)Philip Waldrop Georgia Southern University (GA)Liangmo Wang Nanjing University of Science &Tech-

nology (CHINA)Faruk Yildiz Sam Houston State University (TX)Emin Yilmaz U. of Maryland Eastern Shore (MD)Pao-Chiang Yuan Jackson State University (MS)Jinwen Zhu Missouri Western State U. (MO)

7/30/2019 Ijme Vol 11 n1 Fall2010

7/99

SECOND ORDER TASK SPECIFICATIONS IN THE GEOMETRIC DESIGN OF SPATIAL MECHANICAL LINKAGES 5

Second Order Task Specifications in the Geometric

Design of Spatial Mechanical Linkages

Nina P. Robson, Texas A&M University; J. Michael McCarthy, University of California, Irvine

AbstractThis paper builds on the authors planar kinematic

synthesis for contact task specifications and formulates thekinematic specification of the synthesis problem for spatialopen-serial chains in which a desired acceleration of the end-effector is specified.

Applications of this research focus on the design of spa-tial linkages to maintain specified local motion. A recentlydeveloped failure recovery strategy of a general six-degree-of-freedom TRS robotic arm is discussed and some experi-mental set up and tests of the proposed failure recovery arepresented. The authors also briefly show another possibleapplication of the geometric design of linkages using accele-ration task specifications. It combines the second-order ef-fects of the task with the particular kinematics of the chain toyield free parameters that allow for more than one system toaccomplish one and the same task.

Introduction

This study considered the synthesis of spatial chains toguide an end-effector through a number of multiply sepa-rated positions [1], [2]. The kinematic specification is anumber of task positions with specified end-effector veloci-

ties and accelerations. The goal of this study was to obtainall of the solutions to a given task specification in order todesign mechanical linkages that could move the end-effectorsmoothly through the specified task. Research in the synthe-sis of serial chains to achieve acceleration requirements islimited. It is primarily found in the synthesis theory for pla-nar RR chains, and the work by Chen and Roth [3] for spa-tial chains. The use of second-order effects first appeared inthe analysis of grasping in a work by Hanafusa and Asada[4], where planar objects are grasped with three elastic rods.

Cai and Roth [5] and Montana [6] developed an expres-sion for the velocity of the point of contact between two

rigid bodies that includes the curvature of the contact bodies.Second-order contact kinematics for regular contacts such assurface-surface, curve-curve, curve-surface and vertex-surface are formulated in a unified framework in the recentwork of Park et al. [7], extending Montana's first-order con-tact kinematics for surface-surface contact only. Sarkar et al.[8] develop an expression for the acceleration of the contactpoint between two contacting bodies.

Second-order considerations have also appeared in workby Trinkle [9] in the study of stability of frictionless polyhe-dral objects in the presence of gravity. The mobility of bo-dies in contact has been studied using first-order theoriesthat are based on notions of instantaneous force and veloci-ties [10]. For example, Ohwovoriole and Roth [10] describethe relative motion of contacting bodies in terms of ScrewTheory, which is a first-order theory. Using first-order no-tions, Reuleaux [11], Mishra et al. [12] and Markenskoff etal. [13], derive bounds on the number of frictionless pointcontacts required for force closure, which is one means ofimmobilizing an object. However, first-order theories areinadequate in practice. The source of deficiency is that the

relative mobility of an object in contact with finger bodies isnot an infinitesimal notion but a local one. One must consid-er the local motions of the object, not the tangential aspectsof the motions, as employed by the first-order theories.

Rimon and Burdick [14], [15] show that accelerationproperties of movement can be used to effectively constraina rigid body for part-fixturing and grasping applications. Inprevious work by the authors, planar synthesis [16] was pre-sented as a technique for deriving geometric constraints onposition, velocity and acceleration from contact and curva-ture task requirements. These constraints yielded design eq-uations that can be solved to determine the dimensions of the

serial chain.

In this current study, the authors briefly present this planarapproach, expand on the spatial approach [17], [18] andpresent some of the applications of second-order task speci-fications for the geometric design of spatial linkages.

Geometric Design of Planar Me-chanical Linkages with Task Acce-leration Specifications

Assume that the planar task consists of positioning an end-

effector of a robot arm at a start and a finish position Mj ,j=1,, n, such that in these positions there are prescribedvelocities and accelerations. Let the movement of a rigidbody be defined by the parameterized set of 3 x 3 homoge-neous transforms [T(t)]=[R(t), d(t)] constructed from a rota-tion matrix, R(t), and translation vector d(t). A point p fixed

7/30/2019 Ijme Vol 11 n1 Fall2010

8/99

6 INTERNATIONAL

in the moving body traces a trajectory P(t)nate frame F such that:

or

The goal is to determine the movement of tdefined by [T(t)].

The movement of M relative to a world fraity of a reference position, defined by t=0,by the Taylor series expansion,

The matrices [Tj

0 ], [ ] and [ ] are def

tion, velocity and acceleration of the end-cinity of each task position Mj. Thereforhas the trajectory P(t) defined by the equati

Let p=[ ]-1P, which yields

where

are the planar velocity and planar accelwhich are defined by the end-effector velotion specifications in the vicinity of some tj=1, ..., n.

For example, the design parameters for a

are the coordinates B=(Bx, By) of the fixednates P1=(Px, Py) of the moving pivot whenis in the first position, and the length R oftask position the moving pivot Pj is constrdistance R from B, so we have,

OURNAL OF MODERN ENGINEERING |VOLUME 11,NUMB

in a fixed coordi-

e end-effector as

e F in the vicin-can be expressed

ined by the posi-

ffector in the vi-, a point p in Mn

eration matrices,ity and accelera-ask positions Mj,

planar RR chain

pivot, the coordi-the floating link

the link. In eachined to lie at the

The first and second derivative ofvelocity constraint equation

and the acceleration constraint equa

In order to determine the five desigequations are required. Choosing onbe the first and using the relative dis

= [ ][ ]-1 allow one to defin

the moving pivot as follows:

It is now possible to substitute Pj in

These are the position design equatthe 3 x 3 identity matrix. From ou

velocity matrix, we have Pj = [

Pj into (8), we obtain the velocity de

From our definition of the 3 x 3 acc

( Pj )= [j][D1j]P

1 and su

(9) yields

where j=1,., n. These are the ations. Thus, for each of the n tasvelocity and acceleration design eqing form:

The algebraic solution to the setfor an RR chain is presented in Mof five position synthesis and appliethe design equations (14).

Tj

0

ER 1,FALL/WINTER 2010

his equation provide the

ion

parameters, five designe of the task positions toplacement matrices [D1j]

e coordinates Pj taken by

equation (7) to obtain

ions. Notice that [D11] isr definition of the 3 x 3

][D1j]P1 and substituting

sign equations

leration matrix, we have

bstituting Pj in equation

celeration design equa-positions, the position,

uations have the follow-

f four bilinear equationsCarthy [19] for the cases without any changes to

7/30/2019 Ijme Vol 11 n1 Fall2010

9/99

SECOND ORDER TASK SPECIFICATIONS IN TH

Geometric Design of Spatiical Linkages with Task ASpecifications

Assume that the spatial task consists

end-effector of a robot arm at a start and a fj=1,...,n, such that in these positions thevelocities and accelerations. The rotation afine the orientation of the moving body into be (j, j, j), representing the longitroll angles that position the z-axis of the moj-th position. Thus, the rotation matrix [Aj] i

where [X(.)], [Y(.)], and [Z(.)] represent rx, y and z axes, respectively. Using this conotation dj=(dx,j, dy,j, dz,j), the position data

as the 4 x 4 homogeneous transform

where sin(.) and cos(.) are noted with s(.)tively. Let the movement of the task frameworld frame F be defined by the 4 x 4 hoform [K(t)], and consider its Taylor seriesvicinity of both start and finish positions, su

The matrices [ ], [ ] and [ ] are de

tion, velocity and acceleration of the end-cinity of the two task positions Mj. A poinframe has the trajectory P

j(t) in the fixed fcinity of a task position Mj (see Figure 1), gtion

E GEOMETRIC DESIGN OF SPATIAL MECHANICAL LINKAG

l Mechan-cceleration

f positioning an

inish position Mj

,e are prescribedngles used to de-space are chosende, latitude, andving frame in thes given by

tations about thenvention, and thecan be expressed

and c(.), respec-M relative to theogeneous trans-

expansion in thech that

ined by the posi-

ffector in the vi- p in the moving

rame F in the vi-iven by the equa-

Figure 1. A spatial 6R serial chain wit

M frames

This equation can be rewritten by su

obtain the relative transformation

The kinematic specification consisplacements and the associated ang[Vj ]=[Wj, vj], j=1, ..., n in start and

The dot denotes derivatives with rethe 4 x 4 spatial velocity matrix [ j

where wj=(wx,j, wy,j, wz,j) is the anv

j=(vx,j, vy,j, vz,j) is the linear velocition. Assuming the acceleration prodefined at the j-th position, yields to

In order to define the 4 x 4 acceler

troduce the 4 x 4 matrix construct

aj], to obtain

where j denotes the position in whiare defined.

S 7

h its fixed F and moving

bstituting p=[ ]-1

Pj to

ts of set of spatial dis-lar and linear velocitiesfinish positions, where

spect to time. From this,] is given by

ular velocity vector andty vector at the jth posi-perties of the motion are:

ation matrix [j], we in-

d from (22), [ j]=[j,

h the acceleration terms

7/30/2019 Ijme Vol 11 n1 Fall2010

10/99

8 INTERNATIONAL

Spatial Synthesis Applicati

The spatial synthesis example is a part oforts to explore new, efficient methods ffault-tolerant robot manipulators, as welplanning techniques. Particularly, the aut

non-redundant general six-degree-of-freemanipulator (see Figure 2), mounted on ais fault-tolerant with respect to the originalconsisting of second-order specifications,joints fails and is locked in place.

Figure 2. The TRS arm is a general six-degre

chain configured so that the first pair of revolsect at right angles forming T- oint (also kno

the last three joints intersect in a point to defi

wrist

The Dentavit-Hartenberg parameters [20listed in Table 1. The task data is presentconsists of two positions with velocity, dposition and velocity and acceleration spesecond position.

Table 1. Denavit-Hartentberg parameters f

Table 2. Task data for planning movement

Figure 3 shows the TRS arm moving throtask. The trajectory is determined from theusing a fifth-degree polynomial interpolatio[22].


ns

f the authors ef-or the design of

as novel task-ors examined a

om TRS robotovable platformly specified task,

after one of its

of freedom serial

ute joints inter-n as U-joint) and

ne a spherical

] of the arm areed in Table 2. Itfined in the firstcifications in the

or the TRS arm

of the TRS arm

ugh the specifiedjoint parametersn following [21],

Figure 3. The world frame F, the loca

base of the arm, the moving pivot p o

In the following sections, the autstrategy for the six-degree-of-freedthe originally specified task in the c

Arm Actuator Failures

The recovery strategy is based onthe arm base so that the point B atS2 of the TRS can be placed wherplane parallel to the X-Y plane ofpoint B lies at the origin of the fixe-140 mm at all times, i.e. the armthe X-Y plane. The authors assumegrasp the tool frame where necessarP could be positioned in F wherestrategy reconfigures the arm-platfoof freedom that exist in the systemmovement. Thus, the recovery plantifying values for the reconfigurati-140) and P=(Px, Py, Pz) that ensur

the TRS arm can achieve the speciflar arm joint failure, shown belowtions combine with the specified talynomial equations for the reconfigplatform arm system. Solutions totained numerically using the polynuation sofware PHC [23]. The mgured arm is determined by solvinfailure model in each of the taskjoint trajectory interpolation to guidthe prescribed task [14].

Assume that the actuator of join

controls the shoulder azimuth angfailed and that the brakes have beestant angle 1.The remaining actuatea parallel RRS chain that can posiperpendicular to the horizontal axisGy, Gz) to this plane and a positionPy, Pz) are identified, then the coord


tion of fixed pivot B in the

the TRS arm

ors present the recoverym TRS arm, to achievese of an actuator failure.

the ability to repositionhe intersection of S1 and

needed in a horizontalthe world frame F. Thecoordinate system, Bz=

base can move freely inthat the TRS arm could

y so that the wrist centernecessary. The proposedrm system using degreesut are locked during armis achieved by first iden-n parameters B=(Bx, By,that the end-effector of

ied task for each particu-. These constraint equa-k to provide a set of po-ration parameters of thethese equations are ob-

omial homotopy contin-vement of each reconfi-

the inverse kinematicsositions, and then usingits end-effector through

S1, in Figure 2, which

e, of the TRS arm hasn set to maintain a con-d joints of the TRS formion the wrist in a plane

. Once a normal G=( Gx,f the wrist center P=(Px,

inates of the base pivot B

7/30/2019 Ijme Vol 11 n1 Fall2010

11/99

SECOND ORDER TASK SPECIFICATIONS IN TH

can be computed to reposition the base oallow the arm to guide the tool frame throtask, despite the S1joint failure [24]. The pof design equations for the parallel RRSbilinear quadratic equations and one lineafive unknowns r=(Gx, Gy, Px, Py, Pz). The tsystem is 24 = 16, which is small enough

nate the variables and obtain a univariategree six. The solution, corresponding to faigiven in Table 3.

Table 3. The reconfiguration parameters fo

Figure 4(a) shows the reconfigured platand the trajectory generated to guide thethe original task with an S1joint failure.

(a) (b)

Figure 4. (a) The reconfigured platform arm

actuator failure. (b) The reconfigured platfor

an S2 actuator failure. (c) The reconfigured pl

tem for an S3 actuator failure

Next, the authors considered the case inof the second joint S2 of the TRS arm, wshoulder elevation angle, fails and that theset to maintain 2 at a constant value. Theof the arm in Figure 2 form a perpendiculacan locate the wrist center on a circular toru

polynomial equations that define the recometers r=(Bx, By, Px, Py, Pz) that allow thetem to complete the task despite the failurfive quartic polynomials has a total degrThe real solution, corresponding to shouldure at 2 = 0

o, is listed in Table 4.

E GEOMETRIC DESIGN OF SPATIAL MECHANICAL LINKAG

f the platform tough the specifiedlynomial systemconsists of four

r equation in theotal degree of theto directly elimi-

olynomial of de-lure at 1= 90

o, is

a failed S1joint

orm arm systemRS arm through

(c)

ystem for an S1

arm system for

atform arm sys-

hich the actuatorhich controls thebrakes have beenremaining joints

r RRS chain thats. We obtain five

nfiguration para-latform arm sys-

e. The system ofe of 45 = 1024.er elevation fail-

Table 4. The reconfiguration param

Figure 4(b) shows the reconfigurthe crippled TRS arm with an S2original task. If the elbow actuato

arm in Figure 2 fails then we assumthat 3 has a constant value. The re

form a TS chain that can position t

on a sphere, with a radius R, aboradius R is defined by the link leng3, and is equal to

where the value of3 is determinedthe failed actuator. As in the prevreconfiguration parameters r=(Bx,the arm to perform the original task

polynomial system consists of fiveunknowns r and has a total degreesolution, corresponding to elbow fR= 400 mm, is given in Table 5.

Table 5. The reconfiguration param

Figure 4(c) shows the reconfiguand the trajectory generated to guidjoint failure to achieve the originallface Mobility Platform (Gears LL

arm, integrated using Single-BoarInstruments) are used for the experthe proposed strategy are currentlyRobotics Lab at Texas A&M (see Fi

Figure 5. Experimental set up. Arm i

Figure 6 (a) shows the healthy arthrough a task consisting of secFigure 6(b) shows the new locationing p pivots of the arm, after an e

S 9

eters for a failed S2 joint

ed rover arm system foroint failure through the

r of joint S3 of the TRS

e its brakes can be set soaining joints of the arm

e wrist center p=[ ]P

t the base point B. Theths a23 and a34, the angle

from the joint sensor ofious cases, we seek the

y, Px, Py, Pz) that allow, despite the failure. The

uadratic equations in thef 25 = 32 [25]. The realilure at 3 = 68.56

o, i.e.

eters for a failed S3 joint

ed platform-arm systemthe TRS with the elbow

y specified task. A Sur-), a Lynxmotion robot

RIO 9632 (Nationalimental set up. Tests ofperformed in the Spacegure 5).

stowed position

holding a tool, movingnd-order specifications.of the fixed B and mov-lbow failure. The closer

7/30/2019 Ijme Vol 11 n1 Fall2010

12/99

10 INTERNATIONAL

look shows how the re-grasping ability ofhad allowed for the tool to be grasped at a d

(a)

(b)

Figure 6. (a) The healthy arm moving throu

New locations for the base pivot B and the m

been obtained in order for the crippled TS

originally specified task despite the elbow fail

Finally, Figure 8 is a recent result from

sign mechanical linkages, constrained tocoordinates for the fixed pivot B and the spath trajectory.

Figure 8. The TS chain (Bp1) and the perpend

(Bp2) move smoothly through the second ord

The animation shows that both TS and plinkages satisfy the second-order task

move smoothly throughout the task.

Summary

Formulation of the kinematic specificatisis of spatial kinematic chains with specifition was presented. Applications have focu


the end-effectorifferent location.

gh the a task. (b)

ving pivot p have

rm to obtain the

ure

ur efforts to de-have the same

ame end-effector

icular RRS chain

r task

rpendicular RRSpecification and

n for the synthe-ed task accelera-sed on exploring

new strategies for the failure recoveof-freedom manipulators. The lastsecond-order effects of the task witics of the chain to yield free paramthan one system to accomplish the s

References[1] D. Tesar and J. W. Sparks,

cept of Five Multiply Separanar Motion,J. Mechanisms,

[2] H. J, Dowler, J. Duffy, andlised Study of Four and FiPositions in Spherical KiMach. Theory, 1978, 13:409-

[3] Chen P. and Roth B., DesFinitely and Infinitesimally Sthesis of Binary Links and CASME Journal of Engineeri

Vol. 91: 209-219.[4] Hanafusa H., and Asada H.,

a Robot Hand with Elastic FSymp. Ind. Robots, 1977, pp.

[5] Cai, C.C., and Roth B., OnRigid Body with a Point Coon Robotics and Automation,

[6] Montana, D.J., The KineGrasp, Int. Journal Robot.N0. 3, pp. 17-25.

[7] Park, J., Chung W. and YoContact Kinematics for ReConference on Intelligent

2005, pp.1723-1729.

[8] Sarkar N., Yun S., and Kumtrol of 3D Rolling Contacts ition,IEEE Int. Conf. on Ro1993, pp. 978-983.

[9] Trinkle J.C., On the StabiVelocity of Grasped FrictiTrans. Robotics and Autom

560-572.[10] Ohwovoriole M. S. and Rot

Screw Theory, J. Mech. Dpp. 725-735.

[11] Reuleaux F., The KinematicNew York: Dover.

[12] Mishra B., Schwarz J.T., aExistence and Synthesis ofGrips,Algorithmica, 1987,

[13] Markenskoff X., Ni L. anThe Geometry of Graspinsearch, 1990, vol. 9,pp. 61-7


ry of general six-degree-example combines the

h the particular kinemat-ters that allow for more

ame task.

The Generalized Con-ed Positions in Copla-1968, 3(1), 25-33.D. Tesar, A Genera-e Multiply Separated

nematicsII, Mech.435.ign Equations for theeparated Position Syn-mbined Link Chains,g for Industry, 1969,

Stable Prehension byingers, Proc. 7-th Int.384-389.he Spatial Motion of atact,IEEE Int. Conf.1987, pp. 686-695.atics of Contact andesearch, 1988, vol. 7,

um Y., Second Orderular Contacts," IEEE

Robots and Systems,

ar V., Dynamic Con-n Two Arm Manipula-otics and Automation,

ity and Instantaneousnless Objects, IEEEtion, 1992, vol.8, pp.

B., An Extension ofsign, 1981, Vol. 103.

s of Machinery, 1963,

d Sharir M., On theMulti-finger Positiveol. 2, pp. 541-558.Papadimitriou C.H.,

, Int. J. Robot. Re-.

7/30/2019 Ijme Vol 11 n1 Fall2010

13/99

SECOND ORDER TASK SPECIFICATIONS IN THE GEOMETRIC DESIGN OF SPATIAL MECHANICAL LINKAGES 11

[14] E. Rimon and Burdick J., A Configuration SpaceAnalysis of Bodies in Contact - I. 1-st Order Mobil-ity, Mechanism and Machine Theory, 1995,Vol.30(6) : 897-912.

[15] E. Rimon and Burdick J., A Configuration SpaceAnalysis of Bodies in Contact - II. 2-nd Order Mo-bility, Mechanism and Machine Theory, 1995,

Vol. 30(6): 913-928.[16] N. Robson and McCarthy J. M., Kinematic Syn-

thesis with Contact Direction and Curvature Con-straints on the Workpiece, ASME Int. Design En-gineering Conference, 2007.

[17] N. Patarinsky Robson and J. M. McCarthy, Syn-thesis of a Spatial SS Serial Chain for a PrescribedAcceleration Task, Proc. IFToMM, The 12thWorld Congress on Mechanism and Machine

Science, June 18-21, 2007, Besancon, France.[18] N. Patarinsky Robson, J. M. McCarthy and I. Tu-

mer, Applications of the Geometric Design of Me-chanical Linkages with Task Acceleration Specifi-

cations, ASME Int. Design Eng. Tech. Conf., 2009,August 30-Sept.2, San Diego, CA.

[19] McCarthy, J.M., Geometric Design of Linkages,Springer-Verlag, 2000, New York.

[20] Hartenberg, R. S., and Denavit, J., Kinematic Syn-thesis of Linkages, 1964, McGraw-Hill, New York.

[21] Craig, J., Introduction to Robotics, Mechanics andControl, 1989, Addison - Wesley Publishing Co.

[22] N. Patarinsky Robson, J. M. McCarthy and I. Tu-mer, 2009, Exploring New Strategies for FailureRecovery of Crippled Robot Manipulators,ASME/IFToMM Int. Conf. On Reconfigurable

Mech. and Robots (ReMAR09), June 22-24, Lon-

don, UK.[23] Verschelde, J., and Haegemans, A., "The GBQ-Algorithm for Constructing Start Systems of Homo-topies for Polynomial Systems", SIAM Soc. Ind.Appl. Math. J. Numer. Anal., 1993, 30(2), pp. 583 594.

[24] N. Patarinsky Robson, J. M. McCarthy and I. Tu-mer, Failure Recovery Planning for an ArmMounted on an Exploratory Rover, 2009, IEEETransactions on Robotics, 25(6), pp. 1448-1453.

[25] N. Patarinsky Robson, J. M. McCarthy and I. Tu-mer, The Algebraic Synthesis of a Spatial TSChain for a Prescribed Acceleration Task, Me-chanism and Machine Theory, 2008, 43: pp.1268-

1280.

Biographies

NINA ROBSON is an assistant professor in Manufactur-ing and Mechanical Engineering Technology at Texas A&MUniversity. Her research is in robotics, kinematics of motion,and biomechanics. Her e-mail address is

[email protected].

J. M. MCCARTHY is a professor in Mechanical Engi-neering at University of California, Irvine. His research is indesign of mechanical systems, computer aided design, andkinematic theory of spatial motion. He can be reached [email protected].

7/30/2019 Ijme Vol 11 n1 Fall2010

14/99


APPLICATION OF SIX SIGMA TO GEAR BOX

MANUFACTURINGKambiz Farahmand, North Dakota State University; Jose Victorino Marquez Grajales; Mohsen Hamidi, North Dakota State University

Abstract

This study applies Six Sigma to optimize the keyway cut-ting operation on a shaft of a gear box produced by HortonAutomatics Company. Since the operation had not been op-timized, many problems had been raised in the assembly ofkeys with shafts and a considerable number of scrap partshad been produced which deteriorated profitability of thecompany. The main problem was that the keyway width didnot allow keys to be assembled into the shaft tightly. Tosolve the problem and optimize the process, DMAIC (De-fine, Measure, Analyze, Improve, and Control) methodologyof Six Sigma was applied. In this paper, the steps of DMAICto optimize the operation are presented and illustrated. In the

Define phase, the problem was defined and the specific op-eration of the manufacturing process was recognized by thePFMEA (Process Failure Mode Effects Analysis) technique.In the Measure phase, the operation was studied and theprocess capability was measured. In the Analyze phase, thefactors that affect the keyway width were identified by acause-and-effect analysis. In the Improve phase, the bestcombination of the levels of the factors was determined byusing the Design of Experiments (DOE) technique and thebest combination was applied. Finally, in the Control phase,some recommendations were given in order to keep theprocess in good condition.

IntroductionIn statistical terms, reaching Six Sigma means that the

process or product performs with almost no defects, but thereal message of Six Sigma goes beyond statistics [1]. SixSigma is a philosophy of managing that focuses on eliminat-ing defects through practices that emphasize understanding,measuring, and improving processes [2]. The focus of SixSigma is reducing variability in key product/service qualitycharacteristics to the level at which failure or defects areextremely unlikely [3]. The model of a Six Sigma processassumes that if the process is centered at the target and thenearest specification limit is six standard deviations from the

mean, the process will operate at the 3.4 parts-per-milliondefect level [3]. Six Sigma was heavily inspired by six pre-ceding decades of quality improvement methodologies suchas quality control, TQM, and Zero Defects [4], [5]. Motorolafirst made Six Sigma popular in the 1980s, AlliedSignal em-braced it in the early 1990s and then General Electric madeit the most popular management philosophy in history [6].

Six Sigma efforts target three main areas of improvingcustomer satisfaction, reducing cycle time, and reducing

defects [1]. Improvements in these areas usually representdramatic cost savings to businesses, as well as opportunitiesto retain customers, capture new markets, and build a reputa-tion for top-performing products and services [1]. Unlikemindless cost-cutting programs which reduce value andquality, Six Sigma identifies and eliminates costs which pro-vide no value to customers, or waste costs [7]. Basu andWright [8] listed some real benefits from the adoption of SixSigma: for example, in 1997 Citibank undertook a Six Sig-ma initiative and after just three years it was reported thatdefects had reduced by ten times; General Electric reportedthat $300 million invested in 1997 in Six Sigma deliveredbetween $400 million and $500 million in savings, with ad-

ditional incremental margins of $100 to $200 million; andWipro Corporation in India says that two years after startingin 1999, defects were reduced to such an extent as to realizea gain of eight times over the investment in Six Sigma.

Six Sigma employs a well-structured program methodolo-gy; namely, Define, Measure, Analyze, Improve and Control(DMAIC) or Define, Measure, Analyze, Design and Vali-date/Verify (DMADV) [9]. DMAIC is used for projectsaimed at improving an existing business process andDMADV is used for projects aimed at creating new productor process designs [10]. The five steps of DMAIC are asfollows [8]. Figure 1 shows a flow diagram of the steps of

DMAIC.

1) Define opportunities: This is done through identifying,prioritizing, and selecting the right projects.

2) Measure performance of the projects and process para-meters.

3) Analyze opportunities: Opportunities are analyzed byidentifying key causes and process determinants.

4) Improve performance: This is achieved by changing theprocess so as to optimize performance.

5) Control performance: This is essential if gains are to bemaintained.

In the following sections, the steps of DMAIC applied tothe manufacturing process of a gear box produced by HortonAutomatics Company are described. Since 1960, when theydeveloped the first automatic sliding door in America, Hor-ton Automatics Company has been designing and manufac-turing automatic systems such as automatic sliding, swing-ing, folding, and security revolving doors, service windows,

7/30/2019 Ijme Vol 11 n1 Fall2010

15/99

APPLICATION OF SIX SIGMA TO GEAR BOX

presence and motion detection systems, AMover (APM) transit doors, and vehicle do

Figure 1. Flow Diagram of DMAIC

Define Phase

The automatic swinging door uses theorder to open and close the door. The gthree shafts and three gears. Each gear is aby the employment of a key. The most comis flat key, as shown in Figure 2. On the shclassified according to the process by whicure 3 shows three of the most common ke

function of the key is to transmit torque froa shaft (see Figure 4).

In one part of the production line, geamanufactured and in the other part, gears,other components, such as casings and swibled together. One of the main subassembli

ANUFACTURING

utomated Peopler operators.

000 gear box inear box includesttached to a shaftmon type of keysft, the keyway is

h it is made. Fig-ways. The main

a component to

rs and shafts areshafts, keys, andtches, are assem-es of the product

is the C7113 subassembly, which ina C7017 gear. There were manyC7113 subassembly. The main pronot fit in the keyway and this madescrap parts which had to be reworke

Figure 2. A Flat Key (w: Width, h: H

Figure 3. Three Types of

Figure 4. Keys Role for Transmitting

The primary purpose of the Defithe team focuses on the right thingall of the problems in the manufa

shaft, PFMEA (Process Failure Moapplied. The main processes on tprocess, performed by a hobbing mshaft and a cutting process, performcalled key machine, to cut the kPFMEA for these two processes

13

ludes a C7034 shaft andailures occurring in thelem was that the key didconsiderable number of

d or discarded.

ight, l: Length) [11]

Keyways [11]

Torque to a Shaft [11]

e phase is to ensure that[9]. In order to identify

cturing processes of the

e Effects Analysis) washe shaft are a hobbingchine, to cut teeth on theed by a milling machineyway on the shaft. Theis shown in Figure 5.

7/30/2019 Ijme Vol 11 n1 Fall2010

16/99


Through PFMEA, the following items were identified ordetermined for each process:

Potential failure modes. Potential effects of failure and severity (SEV) rankings of

the consequences of failure; the severity ranking is basedon a relative scale ranging from 1 to 10 [12].

Potential causes of failure and occurrence (OCC) rankingsbased on how frequently the cause of the failure is likelyto occur; the occurrence ranking scale is based on a rela-tive scale from 1 to 10 [12].

Process controls for detecting the failure and detection(DET) rankings based on the chances the failure will bedetected prior to the customer finding it; the detectionranking scale is on a relative scale from 1 to 10 [12].

The overall risk of each failure which is called Risk Priori-ty Number (RPN): RPN = SEV OCC DET; the RPN(ranging from 1 to 1000) is used to prioritize all potentialfailures to decide upon actions leading to risk reduction.

Recommended actions to reduce the risk (RPN).

The PFMEA in Figure 5 indicates that the key machineprocess, which is performed by a milling machine, has the

highest RPN. Based on the observations in this study, it wasfound that the method of cutting the keyway on the shaft hadnot been studied and validated by the plant and there wasmuch confusion in the way to cut the keyway on the shaft. Avertical milling machine was used to cut the keyway. Theoperator frequently changed the machine setup and generat-ed different dimensions in the width of the keyway; conse-

quently, many keys could not be assembled to the shaftstightly and properly due to incorrect size of the keyway.

Measure Phase

In this phase, a process capability analysis was performed. Asample of 30 pieces was taken and Minitab software wasused to do the analysis. Cpk index is a process capability in-dex. Equation (1) shows the formula used for calculatingCpk. Usually, Cpkshould be above 1.33

= 3,3min

LSLMeanMeanUSLCpk (1)

Figure 5. PFMEA Chart

7/30/2019 Ijme Vol 11 n1 Fall2010

17/99

APPLICATION OF SIX SIGMA TO GEAR BOX MANUFACTURING 15

In Figure 6, the results of the analysis are shown. Thewidth of the keyway should be 0.188 (-0.000, +0.002) inch-es; LSL (Lower Spec Limit) is 0.188 inches and USL (UpperSpec Limit) is 0.190 inches. Cpk of the process was 0.31,indicating that the process was in a bad condition. As seen inFigure 6, the sample mean was larger than the expectedmean (0.189), which means that the keyway width was

usually larger than the width that is ideal for assembling thekey onto the shaft. Also, some widths were outside the specrange, less than LSL or more than USL, and as a result manykeys could not be assembled onto the shafts tightly andproperly due to the incorrect width of the keyway.

Figure 6. Process Capability Analysis

Analysis Phase

In order to find out the possible causes that affect thekeyway cutting process, a fish-bone diagram was developed,as shown in Figure 7. The possible causes are divided insetup, operator, end milling, and machine. For setup, thevariables to control are the center of cutter and the feed rate.The center of the cutter does not affect the keyway width,but the effect of the feed rate was unknown and the operatorwas confused about the suitable feed rate for the operation.The effect of the feed rate, then, had to be investigated. Foroperator, this is a factor that can be controlled with operatortraining. For end milling, there are three variables to con-trol: wear out, the material of the cutter, and the number offlutes in the cutter. Wear out can cause imperfections in thekeyway and there was no estimated time to change the cut-ter; it was only changed when it was broken. To control thiscause, the supervisor was told to monitor the start date of a

new cutter and how many pieces could be made with thesame cutter. The types of cutters used were either carbide orhigh-speed steel, which are quite similar and do not affectthe keyway width. Basically, the selection of the cutting tool(carbide or high-speed steel) was determined not to be a con-tributing factor to the incorrect dimension in the keyway.Also, there were two types of cutters: 2-flute cutters and 4-

flute cutters. The effect of the number of flutes was un-known and had to be investigated. For the machine factor,the variables to control are: the holder of the shaft, machinecleaning, and the head machine. The holder of the shaftshould be locked but it is dependent on the operator. Clean-ing is done by the operator but there was not a maintenanceprogram. The head machine is an important factor but it is

dependent on the operator; as long as it is locked and 90degrees to the arm, it does not create any problems.

Therefore, the factors for which the effects were unknownwere identified to be the number of flutes (2 or 4) in the cut-ter and the feed rate. The feed rate was determined by theRPM (Revolutions per Minute) of the milling machine. Therotational speed of the machine can be 1750 RPM or 2720RPM. Basically, to improve the process, the effects of thenumber of flutes of the cutter and the speed of the machinehad to be investigated.

Improve Phase

Design of Experiments (DOE) is a technique for examin-ing controlled changes of input factors and the observationof resulting changes in outputs, i.e., the response to inputchanges [8]. DOE was applied in order to determine whichfactor(s) have major effects on the width of the keyway andwhat combination of factors levels gives the best result. Asmentioned earlier, there are two factors, number of flutesand RPM, each having two levels. Thus, the experimentaldesign is called a factorial design. In statistics, a factorialexperiment is an experiment whose design consists of two ormore factors, each with discrete possible values or levels,and whose experimental units take on all possible combina-

tions of these levels across all such factors. Such an experi-ment allows one to study the effect of each factor on theresponse variable, as well as the effects of interactions be-tween factors on the response variable. For the vast majorityof factorial experiments, each factor has only two levels.Table 1 shows the data gathered for the analysis. In eachcombination of the levels of the factors, there are ten replica-tions.

The statistical hypothesis is as follows:

43210 : ===H

jiH :1 ; For at least one pair of i and j (i, j=1,2,3,4)

where:

:1 Width mean with 2 flutes and low RPM (1750 rpm)

:2 Width mean with 4 flutes and low RPM (1750 rpm)

:3 Width mean with 2 flutes and high RPM (2720 rpm)

:4 Width mean with 4 flutes and high RPM (2720 rpm)

7/30/2019 Ijme Vol 11 n1 Fall2010

18/99

7/30/2019 Ijme Vol 11 n1 Fall2010

19/99

7/30/2019 Ijme Vol 11 n1 Fall2010

20/99


of Six Sigma was applied to solve the problem. In the De-fine phase, PFMEA was applied to distinguish the high-priority area of the process to focus on. The cutting opera-tion performed by a vertical milling machine got the highestRPN and so the study focused on this area. In the Measurephase, the keyway cutting process on the shaft was evaluatedand depicted by using a process capability measure of Cpk.

As expected, the mean of the sample was larger than theanticipated mean (according to standard spec). In the Ana-lyze phase, the cause-and-effect analysis was performed torecognize the main causes of the problem and the main fac-tors that affect the keyway width and their levels as well. Inthe Improve phase, DOE was used to find the best combina-tion of the levels of the factors. A factorial design analysiswith two factors (number of flutes of the cutter and RPM ofthe milling machine) at two levels (2 and 4 flutes, 1750 and2720 rpm) was performed. The results of the experimentprovided by Minitab software showed that there was a sig-

nificant difference between the means of four possible com-binations of the factors levels. Both factors have significanteffects and the most significant factor affecting the keywaywidth was the number of cutter flutes. The best combinationof factor levels is having a 2-flute cutter and an RPM of2720, which would provide a 27.2 in/min feed rate. Finally,in the Control phase, some recommendations were given in

order to keep the process in good condition. The main rec-ommendation is to trace the usage time of cutters becauseafter a certain amount of time, a cutter cannot perform well.During this study, cooperation among engineers and opera-tors and their excellent teamwork in the PFMEA analysis,process capability analysis, cause-and-effect analysis, andDOE were valuable. Actually, Six Sigma methodology pro-duced a team that together would optimize the process andrecognize and remove quality problems that had affectedcustomer satisfaction and profitability of the company.

Appendix 1. Minitab Results of the Factorial Design

Analysis of Variance for WidthSource DF Seq SS Adj SS Adj MS F PMain Effects 2 0.00037645 0.00037645 0.00018823 74.54 0.0002-Way Interactions 1 0.00000003 0.00000003 0.00000003 0.01 0.921Residual Error 36 0.00009090 0.00009090 0.00000253

Pure Error 36 0.00009090 0.00009090 0.00000253Total 39 0.00046738

Estimated Effects and Coefficients for WidthTerm Effect Coef SE Coef T P

Constant 0.191375 0.000251 761.70 0.000FLUTE 0.005850 0.002925 0.000251 11.64 0.000RPM -0.001850 -0.000925 0.000251 -3.68 0.001FLUTE*RPM 0.000050 0.000025 0.000251 0.10 0.921

Least Squares Means for WidthMean SE Mean

FLUTE2 0.1884 0.0003554 0.1943 0.000355RPM1750 0.1923 0.000355

2720 0.1904 0.000355

7/30/2019 Ijme Vol 11 n1 Fall2010

21/99

APPLICATION OF SIX SIGMA TO GEAR BOX MANUFACTURING 19

References

[1] Pande, P. S. & Holpp, L. (2002). What is six sigma?.McGraw-Hill.

[2] Brue, G. (2002). Six Sigma for managers. New York:McGraw-Hill.

[3] Montgomery, D. C. (2010). A modern framework forachieving enterprise excellence.International Journalof Lean Six Sigma, 1(1), 56-65.

[4] Stamatis, D. H. (2004). Six Sigma Fundamentals: AComplete Guide to the System, Methods, and Tools.New York: Productivity Press.

[5] Montgomery, D. C. (2009). Statistical Quality Con-trol: A Modern Introduction (6 ed.). Hoboken, NewJersey: John Wiley & Sons.

[6] Eckes, G. (2003). Six sigma for everyone. Hoboken,New Jersey: John Wiley & Sons.

[7] Pyzdek, T. (2003). The Six Sigma handbook: a com-plete guide for green belts, black belts, and managers

at all levels. New York: McGraw-Hill.[8] Basu, R., & Wright, J. N. (2003). Quality beyond Six

Sigma. Boston: Butterworth-Heinemann.[9] Ray, S., & Das, P. (2010). Six Sigma project selection

methodology.International Journal of Lean Six Sig-ma, 1(4), 293-309.

[10] De Feo, J. A. and Barnard W. W. (2005). JURANInstitute's Six Sigma Breakthrough and Beyond:

Quality Performance Breakthrough Methods. TataMcGraw-Hill Publishing Company Limited.

[11] Technical drawings of Horton Automatics.[12] 10 Steps to Conduct a PFMEA. Resource Engineer-

ing, Inc.http://www.qualitytrainingportal.com/resources/fmea/

fmea_10step_pfmea.htm, retrieved December 3,2010.

Biographies

KAMBIZ FARAHMAND is currently a professor anddepartment head in the Industrial and Manufacturing Engi-neering department at North Dakota State University. Dr.Farahmand has more than 30 years of experience as an engi-neer, manager, and educator. He is a registered ProfessionalEngineer in North Dakota and Texas and past president ofIIE Coastal Bend Chapter. Dr. Farahmand may be reached [email protected]

MOHSEN HAMIDI is currently a doctoral student andinstructor in the Industrial and Manufacturing Engineeringdepartment at North Dakota State University. Mr. Hamidimay be reached at [email protected]

7/30/2019 Ijme Vol 11 n1 Fall2010

22/99


MICROSTEPPING CONTROL OF STEPPER MOTORS

WITH DATA INTERPOLATION AND DIRECT

VOLTAGE CONTROLDong-Hee Lee, Kyungsung University; Shiyoung Lee, the Penn State University Berks Campus;

Tae-Hyun Won, Dongeui Institute of Technology

Abstract

This paper presents a novel approach for microsteppingcontrol of a low-power stepper motor (SM) for anautomotive dashboard instrument application with a directconnection of a low-cost microprocessor. The microsteppingoperation of the SM is very important to indicatinginstrument applications, for example, meters for fuel,battery, speed, oil, and engine revolution. The proposedsystem uses pulse-width modulated output and digital outputpins of a microprocessor for low-cost implementation. In

order to perform a smooth positioning operation for anindicator application, a simple low-pass filter (LPF) andmodified sine data table are introduced.

The modified sine data table is produced by injectedharmonics to reduce the low-frequency harmonics of the SMphase voltage. Also proposed are the S-curve function forthe smooth position reference and the amplitude control ofthe motor phase voltage depending on motor accelerationand deceleration speeds. The S-curve function provides asmooth response while maintaining high acceleration. Theproposed voltage amplitude controller can change the motorphase voltage according to the motor speed to compensatethe back-electromotive force (EMF) at high speed and toreduce the torque ripple in the low-speed region. Theproposed system is implemented with a low-cost, 8-bitmicroprocessor without any external memory and powerdevices. The effectiveness of the proposed control scheme isempirically verified by a practical automotive dashboardinstrument system.

Introduction

The SM is widely used in an open-loop position controlsystem for its inherent stepping-position operation characte-ristics without any feedback loop [1-3]. Today, low-power

and small-size SMs are extensively adapted as indicatinginstruments on the dashboards of automotive vehicles [4-6].In an indicating instrument system, the SM controller shouldperform positioning smoothly. In order to achieve smoothpositioning, a microstepping mode using a pulse-widthmodulation (PWM) approach is an excellent choice [3], [7-9]. Some applications use the multi-phase SM for smoothoperation; however, the drive system of the multi-phase SM

is more complex than a conventional two-phase one [10-15].In this paper, a low-cost dashboard indicating system is de-veloped using a low-power SM which is directly connectedto a microprocessor.

The proposed system uses the PWM output and the digitaloutput pin of a microprocessor for each motor phasewinding. The phase current can flow from the PWM pin tothe digital pin during the positive voltage region, and it canflow in the opposite direction from the high active digital pinto the PWM pin in the negative voltage region.

In order to perform smooth positioning, a simple LPF anda modified sine data table are introduced to reduce the low-frequency harmonics of the motor phase voltage. The mod-ified sine data table is produced by injected harmonics toreduce the harmonics of the motor phase voltage. Furtherproposed in this paper are the S-curve function for smoothposition reference and amplitude control of phase voltage inconnection with acceleration and deceleration. The S-curvefunction can generate the position reference with respect tothe measured frequency for smooth indication with the SM.

The voltage amplitude controller can produce enoughphase voltage to compensate for the back-EMF when the

motor operates with high speed in the fast acceleration anddeceleration region. Similarly, the amplitude of the phasevoltage is decreased to reduce the torque ripple in the lowacceleration and deceleration region from the low speed andlow back-EMF of the motor.

The proposed system is designed with a low-cost, 8-bitmicroprocessor without any external memory and powerdevice. The effectiveness of the proposed control scheme isverified with a practical automobile dashboard system. Theexperimental results show the smooth positioning of thedashboard indicator system.

Conventional MicrosteppingOperation

Figure 1(a) shows a conventional H-bridge circuit for anSM and its bipolar switching method for the microsteppingoperation. Different from the conventional full-step and half-step operation modes, the microstepping operation can indi-

7/30/2019 Ijme Vol 11 n1 Fall2010

23/99

MICROSTEPPING CONTROL OF STEPPER MOTORS WITH DATA INTERPOLATION AND DIRECT VOLTAGE CONTROL 21

SM

Q1

Q2

Q3

Q4

Vdc

Vo

0

PWM

PWM

cossin

as m e

bs m e

i Ii I

= =

cos

sin

ma T m e

mb T m e

T K I

T K I

=

=

( )2 2m ma mb T mT T T K I = + =

cate the precise rotating angle, using sinusoidal phase vol-tage with PWM technology, as shown in Figure 1(b).

The output torque of the SM can be derived by the phasecurrent and vector summation of each phase torque [7]. Thephase current equation can be summarized as follows:

(1)

where, ias and ibs are phase currents of phase A and phase

B. Im denotes the current per phase. e is the electrical posi-tion of an SM.

The resulting torque generated by the corresponding phas-es is derived by

(2) (2)

where KT is the torque constant of the motor [7].

The total torque of the motor can be derived by the vectorsummation of the torque phases as follows:

(3)

In order to produce sinusoidal phase voltage and phasecurrent, the conventional methods use a complex digital-to-analog converter (DAC) and a comparator circuit for PWMswitching. The phase current is limited for stable operation.

In this paper, a simple microstepping operation scheme forthe SM is directly connected to a low-cost microprocessor,and a passive LPF is introduced and applied to a dashboardindicator. The proposed scheme uses a simple modified sinetable for reduction of low-frequency harmonics.

(a) H-bridge circuit per phase of a SM

(b) Bipolar switching for microstepping operation

Figure 1. H-bridge circuit and bipolar switching for

microstepping operation

Proposed Direct MicrosteppingOperation

PWM and Digital Output with a Passive FilterAlso presented here is a direct microstepping operation

scheme using PWM and the digital output of a low-cost mi-croprocessor is proposed. In the front-end of the phasewinding, a simple passive LPF is connected to reduce cur-rent ripple from PWM switching. The SM is designed forlow-power consumption operating at 5V. The phase currentis under 20mA, so the motor can be directly operated by themicroprocessor output pin without an H-bridge converter oramplifier.

Figure 2(a) shows the passive LPF circuits for each phaseand pin connection between the microprocessor and the SM.In Figure 2(b), PWM 1 and DO 1 are output pins of themicroprocessor. In the positive voltage region, phase currentcan be generated by the Low output of the digital pin andpositive PWM output as shown in Figure 2(b). In the nega-tive voltage region, phase current can be generated by theHigh output of the digital pin and the negative PWM out-put.

The pulse duty ratio is controlled by the sine table accord-ing to the reference position of the SM and amplitude gainaccording to acceleration and deceleration. The digital out-put signal is changed according to the reference position.The digital output is kept at the Low signal for 0 to 180

electrical degrees, and High signal for 180 to 360 electricaldegrees.

7/30/2019 Ijme Vol 11 n1 Fall2010

24/99


1

2

1 2

0

s cc

s

o s s cc

V d V

V

V V V d V

=

=

= =

( )1

2

1 2

1s cc

s cc

o s s cc

V d V

V V

V V V d V

=

=

= =

vo

io

vo

io

DO 1

PWM 1

0

PWM 1 DO 1

ARf Rf

Cf Cf

Vo

Vs1 Vs2

(a) Passive filter and pin connection per phase

(b) Signal output of PWM and digital output for sinu-soidal waveform

Figure 2. The proposed phase connection and signal waveforms

The phase current flows from the PWM pin to the digitalpin in the positive region, and flows from the digital pin tothe PWM pin in the negative region. In the positive regionfrom 0 to 180 electrical degrees, the terminal and phase vol-tages can be derived without the filter as follows:

(4)

where dis the duty ratio of the PWM and Vcc is the controlvoltage of the microprocessor.

In the negative region, from 180 to 360 electrical degrees,the terminal and phase voltages can be derived as follows:

(5)

From these relationships, the phase voltage can be con-trolled by the PWM duty ratio.

In order to supply a sinusoidal phase voltage, the internalmemory of the microprocessor is used for the sine table.Figure 3(a) shows the output phase voltage and current withthe passive LPF in conventional 8-bit PWM data.

As shown in Figure 3(b), the conventional sine data pro-duces low-frequency harmonics in phase voltage and cur-rent, such as 3rd and 5th harmonics. The switching harmonics

has a high-frequency component which slightly affects theposition error. But, the low-frequency harmonics can pro-duce an additional position error in the microstepping opera-tion. In order to reduce the low-frequency harmonics, amodified sine table for 8-bit PWM signal is used. The mod-ified sine table is generated with an additional harmonicinjection to reduce the 3rd and 5th harmonic frequencies in

the output voltage and current as follows:

[ ] 128 sin( ) sin(3 ) sin(5 )PWM a b = + + (6)

where a and b are the injected harmonic coefficients of the3rd and 5th harmonic frequencies, and where 6 and 4 are usedfor a and b respectively.

Figure 4(a) shows reduced low-frequency harmonics inthe phase voltage and current with modified sine data. Bycomparison with Figure 3, the modified sine data with in-jected harmonics can reject the low-frequency harmonics.This pure sinusoidal phase current can produce a constant

torque in any position.

(a) Output voltage and current per phase

(b) FFT analysis of output voltage and currentFigure 3. Output characteristics of a conventional sine table

7/30/2019 Ijme Vol 11 n1 Fall2010

25/99


vo

io

vo

io

S curveFunction

FrequencyDetector

Tacopulse

internalcounter

actf reff

1

s

aK

PositionCalculator

ref

Modified

Sine Table

DO1

x PWM1

LPF

LPF

SM

DO1

PWM1

DO2

PWM2

(a) Output voltage and current per phase

(b) FFT analysis of output voltage and currentFigure 4. Output characteristics of the modified sine table

Voltage Controller

Smooth torque control is essential in order to reduce the

vibration of an indicator. In a conventional open-loop posi-tion controller of an SM, simple acceleration and decelera-tion curves are used without any current control. The con-ventional method is very simple, but the same torque at adifferent speed can cause indicator vibration.

In this paper, a simple voltage control scheme to adjust theoutput torque of an SM according to motor speed is pro-posed. The variable voltage can change the phase currentand output torque. The practical voltage controller is imple-mented by a PWM duty ratio control in order to change thePWM duty cycle by the position and motor speed and isgiven by

D = Ka PWM[ref] (7)

whereD is the duty ratio of the PWM data. PWM[ref] is thePWM data determined according to the reference position asdescribed in equation (6). Ka is the proportional gain accord-ing to the motor speed and is lower than 1 as follows:

Ka= lim( ref/base ) (8)

Figure 5(a) shows the proposed control scheme for thedashboard indicator. In order to reduce the indicating vibra-tion, the phase voltage is controlled in relation to accelera-tion and deceleration. Ka, shown in Figure 5(b), denotes theamplitude gain of the PWM data according to the accelera-

tion of the position reference. During fast acceleration, theamplitude gain is increased and the duty ratio of the PWMcan be increased to increase the phase voltage.

(a) Proposed control scheme for a dashboard indicator

(b) SM connectionFigure 5. The proposed control scheme of a dashboard

indicator using a low-power SM

For fast acceleration, the speed and the back-EMF of theSM should be increased with respect to the motor speed. Inorder to compensate the back-EMF in relation to increasingspeed, phase voltage should be increased. In the proposedcontrol scheme, the amplitude gain can control the phasevoltage according to the motor speed. If the acceleration isdecreased, the amplitude gain is decreased to decrease thephase voltage.

Experimental Results

In order to verify the proposed control scheme, an experi-mental test setup was implemented. A digital controller wasdesigned using the ATmega16 8-bit microprocessor fromATMEL Corporation. The practical indicating instrumentconsists of one SM for indication and a 4-digit liquid-crystaldisplay (LCD) for user display. The SM was directly con-

7/30/2019 Ijme Vol 11 n1 Fall2010

26/99


0.0 200 [ms]

0

20

-20

-40

0

2

-2

[dB]

[V]

Phase Voltage

FFT of Phase Voltage

125[Hz]3rd, 5th, 7th Harmonics

0.0 200 [ms]

0

20

-20

-40

0

2

-2

[dB]

[V]

Phase Voltage

FFT of Phase Voltage

125[Hz]3rd, 5th, 7th Harmonics

0.0 5.0 [sec]

0.0

5.0

5.0

0.0

200

400

0 [Hz]

Pha se-A Vo lta ge

Pha se-B Vo lta ge

Me asu red Fre que ncy

Po siti on Ref ere nce

Ha rmo nic s

0.0 5.0 [sec]

0.0

5.0

5.0

0.0

200

400

0 [Hz]

Pha se- A V olt age

Pha se- B V olt age

Me asu red Fr equ enc y

Po siti on Re fere nce

nected to the PWM output and digital output pins of theATmega16 with a simple passive LPF for each phase. TheSM has an internal gear train with a ratio of 180:1, and themechanical accuracy is 0.3o per step. Figure 6 shows theimplemented experimental test setup for the proposed con-trol scheme.

Figure 6. Experimental system

Figure 7 shows the phase voltage and the fast Fouriertransformation (FFT) analysis of the SM under test withrespect to the conventional sine and the modified sine data toreduce the low-frequency harmonics. As shown in Figure 7,the modified sine data can reduce the low-frequency har-monics with injected harmonics.

Figure 8 shows the experimental results at a 100Hz inputfrequency according to both sine tables. The measured fre-quency, the S-curve position reference, and the motor phasevoltages are displayed in Figure 8. The amplitude of thephase voltage is fixed. The phase voltage waveform has low-frequency harmonics with the conventional sine table, asshown in Figure 8(a). However, the waveforms of the phaseA and B voltages were closer to sinusoidal with the modifiedsine table without low-frequency harmonics, as shown inFigure 8(b).

Figure 9 shows the experimental results of the proposedcontrol scheme at 100Hz and 300Hz input frequencies. Theamplitudes of the phase voltages were changed by accelera-tion and deceleration. Compared with Figure 8(b) with aconstant phase voltage, the amplitude of the phase voltagewas not constant and the value was controlled by the speed,as shown in Figure 9. The low-phase voltage in the low-speed region reduces the phase current and the output tor-que. This is because the load torque of the indicator needle

was almost the same, but the practical torque depends on theacceleration and deceleration torque. This variable phasevoltage according to acceleration and deceleration can re-duce the indicating vibration of the SM.

(a) Conventional sine table

(b) Modified sine tableFigure 7. The phase voltage and FFT analysis

(a) Conventional sine table

(b) Modified sine tableFigure 8. Experimental results at 100Hz input

7/30/2019 Ijme Vol 11 n1 Fall2010

27/99


5. 0 [ s ec]

0. 0

5. 0

5. 0

0. 0

20 0

40 0

0

P h as e- A V ol ta ge

P h as e- B V ol ta ge

M ea su re d F re q ue n cy

P os it io n R ef e re nc e

0.0 5.0 [sec]

0.0

5.0

5.0

0.0

200

400

0 [Hz]

Phase-A Voltage

Phase-B Voltage

Measured Frequen cy

Position Reference

(a) At 100Hz input

(b) At 300Hz inputFigure 9. Experimental results of the proposed control scheme

with two different input frequencies

Conclusions

This paper presents a microstepping operation of an SMwhich was connected directly to a microprocessor for anautomotive dashboard indicator application. In order to im-

plement smooth indicating, the modified sine data table witha simple LPF was used. The FFT analysis showed that themodified sine data can reduce the low-frequency harmonicssignificantly. In addition, an internal S-curve function forproducing a smooth reference position from a detected inputfrequency was designed.

The amplitude of the phase voltage was controlled by theacceleration and deceleration of the reference position toreduce the indicating vibration. In the proposed controlscheme, the amplitude of the phase voltage could be easilycontrolled by the amplitude gain, and the gain changes theactual duty ratio of the PWM data from the sine table.

The proposed system uses a PWM and digital output pinsfor each phase, allowing any low-cost microprocessor to beused for the dashboard indicator application. From the expe-rimental results, the proposed system does a good job ofcontrolling the SM with regard to the input frequency thateliminates low-frequency vibration.

Acknowledgment

The authors would like to express deep appreciation toKyungsung University, Busan, Korea, for its support incompleting this study.

References[1] P. Acarnley, Stepping Motors: a Guide to Modern

Theory and Practice, 4th ed., IEE ControlEngineering Series 63, ISBN: 0-85296-029-8,Michael Faraday House, 2002.

[2] T. Kenjo and A. Sugawara, Stepping Motors andMicroprocessor Control, 2nd ed., ISBN: 0-19-859385-6, Oxford Clarendon Press, 2003.

[3] A. Bellini, C. Concari, G. Franceschini and A.Toscani, "Mixed-Mode PWM for High-performanceStepping Motors," IEEE Transactions on IndustrialElectronics, Vol. 54, No. 6, pp. 3167-3177, Dec.

2007.[4] Gildas Allain, Micro Stepping: A Smoother Way of

Life Low-Cost Solution for DashboardInstrumentation, Atmel Corporation, October 2008.

[5] Data Instrumentation Technology, Ltd., VID29Series Cluster Stepper Motor Datasheets,www.vid.wellgain.com

[6] MCR Motor, Miniature Stepper MotorMR1107/MR1108 Datasheets, www.mcrmotor.com

[7] A. Astarloa, U. Bidarte and A. Zuloaga,"Reconfigurable Microstepping Control of SMs usingFPGA embedded RAM," in Proc. IEEE IECON, Nov.2003, Vol. 3, pp. 2221-2226, 2003.

[8] G. Baluta, "Microstepping Mode for SM Control,"Signals, Circuits and Systems, 2007. ISSCS 2007.International Symposium, Vol. 2, pp. 1-4, July 2007.

[9] G. Baluta, M. Coteata, "Precision microsteppingsystem for bipolar SM control," Electrical Machinesand Power Electronics, ACEMP '07. InternationalAegean Conference, pp. 291-296, Sept. 2007.

[10] Keun-Ho Rew and Kyung-Soo Kim, A Closed-FormSolution to Asymmetric Motion Profile AllowingAcceleration Manipulation, IEEE Transaction onIndustrial Electronics, Vol. 57, no. 7, pp. 2499-2506,July 2010.

[11] Tzung-Cheng Chen and Yung-Chun Su, HighPerformance Algorithm Realization on FPGA for SMController, SICE Annual Conference 2008, pp.1390-1395, August 2008.

[12] Renesas R8C/25 Application Note, 180 DegreeSinusoidal Motor Control, REU05B0087-0100/Rev.1.00, pp. 1-13, February 2009.

[13] K. S. Chun, H. J. Kim, Y. K. Kwon and S. H. Kang,Design of High Performance 5 Phase Step Motor

7/30/2019 Ijme Vol 11 n1 Fall2010

28/99


Drive System with Current Control Loop, TheTransactions of the Korean Institute of PowerElectronics ,Vol. 11 no.5 , pp. 447-453, 2006.

[14] M. H. Kim, T. Y. Kim, H. K. Ahn and S. K. Park,One-Chip and Control System Design of Low Costfor Micro-stepping Drive of 5-Phase SteppingMotor, The Transactions of the Korean Institute of

Power Electronics , Vol. 9 no.1, pp.88-95, 2004.[15] S. H. Kim, E. W. Lee, D. J. Lee and T. M. Koo,

Microstep Drive of 2 Phase 8 Pole HB Type LinearPulse Motor for Precise Position Control, Thetransactions of the Korean Institute of ElectricalEngineers. B, Vol. 48, no.12 , pp. 671-678, 1999.

Biographies

DONG-HEE LEE was born on Nov. 11, 1970 and re-ceived B.S, M.S., and Ph. D degrees in Electrical Engineer-ing from Busan National University, Busan, Korea, in 1996,1998 and 2001, respectively. He worked as a Senior Re-searcher of Servo R&D Team at OTIS-LG, from 2002 to2005. He has been with Kyungsung University, Pusan, Ko-rea, as an Assistant professor in the Department of Mecha-tronics Engineering since 2005. His major research field isthe Power Electronics and motor control system. Dr. Leemay be reached at [email protected].

SHIYOUNG LEE is currently an Assistant Professor ofElectrical Engineering Technology at The PennsylvaniaState University Berks Campus, Reading, PA. He receivedhis B.S. and M.S. degrees in Electrical Engineering fromInha University, Korea, his M.E.E.E. in Electrical Engineer-ing from the Stevens Tech., Hoboken, NJ, and his Ph.D.

degree in Electrical and Computer Engineering from theVirginia Tech., Blacksburg, VA. He teaches courses in Pro-grammable Logic Controls, Electro-Mechanical Project De-sign, Linear Electronics, and Electric Circuits. His researchinterest is digital control of motor drives and power conver-ters. He is a senior member of IEEE, as well as a member ofASEE, ATMAE, and IJAC. He may be reached [email protected].

TAE-HYUN WON is a professor of Electrical Engineer-ing at Dongeui Institute of Technology, Busan, Korea. Hereceived his B.S., M.S., and Ph.D. degrees in Electrical En-gineering from Busan National University, Korea. He

teaches Automatic Control, Electronic Circuits, and Micro-processor courses. His research interests include the sensor-less control of PMSM, robust control of BLDC motor, andhigh precision and high speed motor control. He is a seniormember of KIEE and KIPE. Professor Won may be reachedat [email protected].

7/30/2019 Ijme Vol 11 n1 Fall2010

29/99

AN ENHANCED FRAMEWORK FOR MULTI-MODULE EMBEDDED RECONFIGURABLE SYSTEMS 27

AN ENHANCED FRAMEWORK FOR MULTI-

MODULE EMBEDDED RECONFIGURABLE

SYSTEMSMuhammad Z. Hasan, Texas A & M University; Timothy Davis, Texas A & M University;Troy Kensinger, Texas A & M University;

Sotirios G. Ziavras, New Jersey Institute of Technology

Abstract

Reconfigurable logic facilitates dynamic adaptation ofhardware and ensures better utilization of hardware space asdesired in embedded applications. Partial reconfiguration ofhardware is a recent trend where a portion of the repro-grammable logic can be altered without affecting other por-tions. Host-based multiple-module reconfigurable hardwarefabric, such as Field Programmable Gate Arrays (FPGAs),can potentially employ partial reconfiguration for embeddedapplications where a FPGA-resident or external host controls

the application execution and reconfiguration. Although thistechnique minimizes area requirements and potential energyrequirements for applications, it may result in a disparity inusage of different reconfigurable modules.

This disparity may cause localized temperature build-upand failure. Moreover, in such a host-based system, a subjec-tive load distribution between the host and the reconfigura-ble module could result in performance improvementthrough parallelism. In this paper, policies are presented thatensure uniform utilization of reconfigurable modules, whileimplementing load-balancing between the host and the re-configurable module for better performance. Experimentalresults involving benchmark kernels with these policiesshow a reduction in disparity of more than 40% of moduleusage as well as improvements in an application executiontime of about 35%, as compared to a reference algorithm. Ingeneral, though, these policies are minimal when comparedwith the execution time for applications.

Introduction

FPGAs contain user-programmable hardware and inter-connections. Thus, the reprogrammable features of FPGAsmake it easy to test, debug, and fine tune hardware designsfor higher performance in follow-up versions. Also, it

enables the hardware implementation of a large design in apiecewise fashion as the complete design may not fit in thesystem. Partial reconfiguration support of current FPGAarchitectures provides support for reconfiguring portions ofthe hardware while the remainder is still in operation [1],[2]. Switching configurations between implementations can

then be fast, as the partial reconfiguration bit-stream may besmaller than the entire device configuration bit-stream.

Embedded systems are currently in virtually all aspects ofeveryday life. They normally are expected to consume smallamounts of power and to occupy few resources. Numerousembedded applications spend substantial time on a fewsoftware kernels [3]. Executing these kernels on customizedhardware could reduce the execution time and energy con-sumption as compared to software realizations [4], [5]. Giv-en reconfigurable hardware, such as FPGAs, a chosen areacould accommodate such kernels exclusively at different

times to conserve resources, thus saving space and possiblypower. Configurations to support kernels can be createdahead of time and stored in a database for future use, facili-tating system adaptability for run-time events. However, thereconfiguration time affects the performance, especially forsmall execution data sets. Also, the reconfiguration processdraws power. To offset the overhead time encountered, vari-ous techniques such as configuration pre-fetching or over-lapping reconfiguration with other tasks must be employed.

Many dynamically reconfigurable systems involve a hostprocessor mainly for control-oriented, less computation-intensive tasks and also for supporting reconfiguration deci-

sions [4], [6-9]. The target of this work was either a singleFPGA embedded with reconfigurable modules or severalindividually reconfigurable FPGAs. It was also shown thatin a system with multiple reconfigurable resources, the dis-parity of usage may be significant under a brute-force policy[10]. In order to overcome such an undesirable effect, theruntime system could keep statistics for the utilization ofeach reconfigurable unit, in either a partially reconfigurablemodule or a complete FPGA. The ones with lower utiliza-tion should be the target for the next kernel implementation.This would not only balance the usage of all the reconfigur-able resources but could also reduce the localization of tem-perature increases in the system for enhanced reliability.

Also, it was implied that when kernel execution on hard-ware, or the host, provides speedup, the whole data set forthat kernel would be processed either on the hardware or onthe host [10]. Instead of processing the whole data set for akernel solely on the host or on the reconfigurable resources,it

Ijme Vol 11 n1 Fall2010

Documents