Professional Documents
Culture Documents
DISSERTATION
By
****
Dissertation Committee:
Approved By
Professor Gary Maul, Advisor
Vibratory bowl feeders (VBFs) are the most versatile devices used in the feeding of small
engineered parts during the part assembly process. VBF technology is mature; industry
has been using VBFs to great effect for over 30 years. Research has not been lacking;
however, the VBF knowledge base lacks a fundamental understanding of the interactions
between the physical characteristics of a part and the various design parameters of a VBF
with the least amount of instability and a maximized effective travel distance. The
objective of this research was to explore these interactions and develop a fundamental
understanding of the principals involved. An experimental apparatus that allowed for the
rapid modification of various VBF parameters was designed, built and used in three
empirical studies. The results of these studies were verified in a traditional commercial
VBF. Classical impact mechanics and high speed digital recordings of part micro-
motions were used to develop a narrative of the pertinent factors involved and their
ii
DEDICATION
And to:
Tracey, whose love, support and tolerance lightened my burden and brought true
happiness.
iii
ACKNOWLEDGMENTS
his intellectual support and encouragement which made this work possible, and
for his patience in correcting both my stylistic and technical errors. I am greatly
thankful to Professors Jerald Brevick and Dave Farson for their insightful
Finally, I want to express my thanks to John, Eric, Bob and Mary, who taught me
iv
VITA
FIELDS OF STUDY
v
TABLE OF CONTENTS
Page
Abstract………………………………………………………………………………. ii
Dedication…………………………………………………………………………..... iii
Acknowledgments…..……………………………………………………………....... iv
Vita…………………………………………………………………………………… v
List of Figures.……………………………………………………………………….. ix
List of Symbols………………………………………………………………………. xi
Chapters:
1.1 Introduction…………………………………………..………………………. 1
1.2 Overview of Vibratory Bowl Feeders and Related General Literature……… 4
1.3 Part Kinetics and System Modeling……………………………………….…. 12
1.4 Natural Resting Aspects of Parts…………………………………………….. 19
1.5 Orienting Devices and Tooling………………………………………………. 24
1.6 Programmable VBF Research- Flexible Parts Feeding……………………… 35
1.7 Waveform Research………………………………………………………….. 48
1.8 Research Objectives and Organization………………………………………. 54
2. Experimental Organization
2.1 The Experimental Apparatus………………………………………………… 55
2.2 The Experimental System……………………………………………………. 59
2.3 Experimental Parts…………………………………………………………… 61
2.4 Experimental Methodology………………………………………………….. 62
vi
3.5 Statistical Analysis…………………………………………………………… 73
4. Study I
5. Study II
6. Verification Study………………………………………………………………. 82
7. Analysis…………………………………………………………………………. 85
7.1 Introduction…………………………………………………………………... 85
7.2 Trajectory Analysis…………………………………………………………... 85
7.3 Impact Analysis……………………………………………………………… 89
APPENDICES
APPENDIX A……………………………………………………………………. 105
APPENDIX B……………………………………………………………………. 107
APPENDIX C……………………………………………………………………. 109
APPENDIX D……………………………………………………………………. 115
APPENDIX E……………………………………………………………………. 119
APPENDIX F…………………………………………………………………….. 121
APPENDIX G……………………………………………………………………. 123
APPENDIX H……………………………………………………………………. 125
APPENDIX I…………………………………………………………………….. 127
APPENDIX J…………………………………………………………………….. 131
APPENDIX K…………………………………………………………………… 138
APPENDIX L……………………………………………………………………. 140
APPENDIX M…………………………………………………………………… 143
Bibliography…………………………………………………………………………..146
vii
LIST OF TABLES
Table Page
viii
LIST OF FIGURES
Figure Page
ix
A Infrared Transmitter and Receiver Circuits……………………………… 107
C.1 Angle Measurement VI…………………………………………………...111
C.2 Electromagnet Control VI…………………………...…………………....112
C.3 Infrared Gate Feed Rate Measurement VI……………………………….. 113
C.4 ADXL250 Calibration VI………………………………………………... 114
C.5 Acceleration and Frequency Verification VI….………………….………115
E FV vs. Vibration Frequency, Pilot Study…………………………………121
F Response Surface Regression, Pilot Study………………………………. 123
H Study I Regression Analysis…………………………………………….. 127
J.1 Study II: Regression Analysis 2.4 g Normal Acceleration………...…….. 133
J.2 Study II: Regression Analysis 2.64 g Normal Acceleration………...…… 134
J.3 Study II: Regression Analysis 2.95 g Normal Acceleration………….….. 135
J.4 Study II: Regression Analysis 3.11 g Normal 136
Acceleration………..……..
J.5 Study II: Regression Analysis 3.45 g Normal Acceleration…….....…….. 137
J.6 Study II: Response Surface Regression………………………………….. 138
M Parallelepiped……………………………………………………………..144
x
LIST OF SYMBOLS
Symbol Meaning
xi
CHAPTER 1
1.1 Introduction
The advent of assembly line technology irrevocably changed the world. Previously, the
complete assembly of a product was carried out by a single operator. For most skilled
trades such as clock making or gunsmithing, this operator also manufactured the
individual component parts of the assembly. This meant that each operator had to master
many independent skills; resulting in long apprenticeships, and low overall production
volumes. The ‘free market’ ideal pioneered by Adam Smith could not be implemented
until a means of satisfying demand could be met. Adam Smith would not live to see the
world stand in awe of Eli Whitney’s interchangeable musket locks, or Oliver Evan’s
conception of conveying materials from one place to another without manual effort, but
his ideals inspired them, and the dozens of pioneers that followed. The introduction of
machines removed human skill from the manufacturing equation; a great number of high
quality nearly identical parts could be made quickly and inexpensively. No longer would
the market be strictly ruled by supply. Eventually, with the brilliant efforts of men like
Elihu Root, Fredrick Winslow Taylor and Henry Ford, the assembly process would be
polished. The ideal of “divided work multiplies the output” took hold; assembly work
1
was reduced to very basic operations that required minimal training to produce high
efficiency output.
Indeed, it might be argued that these advances were too successful. Large assembly
Workers were expected to keep up with increasingly faster machines, industrial accidents
were common place with no recourse for those injured or killed, OSHA did not exist. As
the Victorian era came to a close, special machines for assembly were beginning to make
which early on, were most likely to be humans, by parts feeders. The advent of, even
primitive, automation: improved part quality; increased productivity through faster cycle
times; increased yield (less rework and scrap); reduced labor costs, and improved worker
safety [1]. Some four decades later, industrial manipulators and automated work cells
were developed in order to increase the flexibility of automated assembly. With the
advent of the computer revolution, improved robotic control and integrated automated
With these modern assembly methods came the means to analyze their effectiveness.
Various statistical methods for quality control were quickly developed. These methods
continuously enhanced the manufacturing process, fine tuning and optimizing every
aspect of production. If the processes could be tightly controlled, higher quality, lower
cost, and more uniform products could be produced to satiate the world’s ever growing
2
Specialized assembly machines are generally not cost effective for small-to-medium
production operations. Instead, flexible automatic assembly systems are used. These
systems have the advantage of re-programmability, when one product has completed its
production cycle, the machine can be set-up to run the next. An automated flexible
more part feeding devices, and a part transfer system. Reprogramming the system usually
means altering the configuration of the manipulator and the parts feeders within the
feeding system. Since the next production run usually consists of a similar product;
generally, only the feeder system is retooled. Re-tooling for alternate production can
consume considerable financial and temporal resources; hence, great intellectual effort
A parts feeder can be defined as any device that receives a number of randomly oriented
parts at its input and delivers parts in a unique orientation at its output [2]. In general,
parts feeders can be divided into two categories based on the underlying motive force:
nature and are employed in long production run scenarios. Vibratory feeders tend to be
generic in nature and are further broken down into linear and bowl configurations (VBF).
Examples of non-vibratory feeders include: center board, rotary hook, stationary hook,
slotted wheel, and external gate hopper feeders. Both types play a key role in many
industrial automated assembly systems. The VBF is currently considered the most
versatile hopper feeding device for feeding small parts in assembly manufacturing. Due
to high tooling costs and long down times needed for retooling, traditionally equipped
3
VBFs are implemented in medium to high production operations. New designs are
constantly being proposed to reduce or eliminate the high costs of retooling and increase
the overall flexibility of the system. One third of the cost of an assembly system can be
A typical VBF, depicted in figure 1.1, consists of a bowl with helical track running along
the inside wall. The angle of incline of the track generally does not exceed 4°. The bowl
is attached to a heavy base by either three or four inclined leaf springs. An electromagnet,
sometimes two, placed vertically beneath the bowl and secured to the base periodically
pulls the bowl vertically downwards. The heavy base rests on support feet, usually made
of rubber and designed to isolate the vibrations of the VBF from the rest of the
equipment. The electromagnet does not act directly on the bowl, but rather on an iron
attachment mounted horizontally under the bowl. In the case of duel electromagnets, the
There is no limit as to the material the bowl may be constructed of; typical examples
include aluminum, food grade stainless steel and polymer composite. Due to the
constraint of the inclined leaf springs, the bowl vibration has two components, a linear
causes the parts on the track to experience a vibrational force at an angle greater than that
of the track. The effect of the vibratory motion causes the parts to climb up the track to
the outlet at the top of the bowl. The movements of the track are small compared with the
4
length of the suspension springs and the diameter of the bowl; thus, on a local scale, the
movement of any point on the track may be assumed linear. Traditional VBFs are
typically equipped with passive orienting devices arranged in sequence along the track of
the bowl. All parts leaving the bowl will be oriented in a geometrically similar manner;
any part not oriented in the required manor is rejected. Any part which falls off the track
or is rejected by an orienting device returns to the bottom of the bowl, and would
5
Traditionally, VBFs derive power from SCR (silicon controlled rectifier) controlled AC
sources. The SCR acts like a diode, in that it allows current to flow in only one direction,
but differs from the diode in that it will only permit the current to flow after it has been
switched on. Once it has been gated on, and the current is flowing, the only way it can be
turned off is to expose the SCR to negative current. Typically, this occurs at the crossover
point (180°) between the positive and negative segments of the AC waveform. This
cancels the original gating command. SCR control may be either half or full-wave
Varying the point in each cycle at which the SCR is fired controls the power level for the
electromagnet. The later in the cycle the gating occurs, the lower the energy to the
decrease in the kinetic energy transferred to the part. Recent innovations in VBF drives
include the use of switching mode power supplies. This method generates many small
square waves, which approximate a sine wave. This method does not depend on the line
Parts in a VBF travel by either sliding or hopping motions. When sliding, as the bowl
rises and twists forward the friction between the bowl and the part is such that both move
together. In the second half of the feeding cycle, when the bowl descends and rotates
away from the outlet, the friction between the part and the bowl becomes sufficiently low
that the part slides over the track, advancing the part. In hopping, the electromagnetic
force is so high that the normal reaction between the part and the track becomes zero; the
6
bowl’s downward acceleration exceeds that of gravity and the part experiences free-fall,
Vibratory motion gives the impression that parts are conveyed forward with an almost
constant conveying velocity; however, this is not true. This motion is, in actuality, a
repeats with the frequency of the drive. The detailed types of motion have been described
and analyzed in the literature [3,5,6]. In short, the part starts to slide forward at some
instant when the track nears the upper limit of motion. If there is no hopping mode,
forward sliding continues until the lower limit of motion is reached; at which point the
part may either remain stationary relative to the track, or slide backward until the cycle is
In order to maximize the feed rate of a traditional VBF, a number of parameters are
considered during the design phase. The conveying velocity of a part moving along the
track of a VBF depends on the: amplitude of the track vibration ao; track inclination angle
θ; vibration angle ψ; effective coefficient of friction between the part and the track (both
static µs, and dynamic µd); total mass of the parts in the bowl; vibration frequency of the
bowl ƒ; stiffness of the leaf springs ks; physical characteristics of the part; position of the
part within the track, and number, sequence and type of orienting devices. A brief
discussion of some of these parameters is to follow, however, in [3, 5, 6], these factors
7
Boothroyd [3] predicted that for given conditions and for constant track acceleration, the
ƒ vm = constant (1.2.1)
One consequence of this result is that for high conveying velocities and hence high feed
rates, it is desirable to use as low a frequency as practicable. However, since the track
acceleration must be kept constant, this result means a corresponding increase in track
amplitude. The mechanical problems of connecting the feeder to a stationary machine
imposes a lower limit on the frequency [5,7].
Empirical confirmation of these predictions appears in the same works. Figure 1.2 is a
graph of the effect of vibration angle, normal track acceleration and frequency on
conveying velocity. A brief mathematical explanation of this can be found in section 4.1.
Figure 1.2: Effect of Vibration Angle, Normal Track Acceleration and Frequency on
Conveying Velocity. (After Boothroyd [5])
8
Generally, an increase in track acceleration produces an increase in conveying velocity.
to some critical value, different for each part, the hopping takes on an unstable erratic
It has been shown that the conveying velocity is sensitive to changes in vibration angle
[5,7]. For any given part there exists an optimum value of vibration angle, dependant
primarily on the coefficient of friction of the part and the angle of inclination of the track.
General trends indicate that faster feed rates are obtained with smaller vibration angles.
Smaller track angles produce greater feed rates; a zero degree track angle produces the
greatest feed rates. However, the mechanical design of a bowl feeder necessitates a
positive track angle of three or four degrees in order to raise the parts to the bowl outlet.
Track angles greater than five or six degrees generally produce unsatisfactory results.
In vibratory feeding, the practical range of coefficient of friction of both the parts and the
track. In general, for practical values of track acceleration, an increase in friction leads to
an increase in feed rate. A common practice in industry is to line the track with rubber,
VBFs are well described in the literature. An early work by Povidaylo [8] steps through
the design calculations necessary for the construction of a VBF. This topic is revisited;
9
using similar analytical constructs and employing modern devices equipped with
those investigating their use, Smith [10] examines the basic factors to consider in the
selection of a VBF for a particular application. Similar works considering all forms of
vibratory feeders can be found in [8,11,12]. In [13], Redford examines in detail the
traditional VBF, the factors effecting performance, and inherent systemic weaknesses. As
Many papers have been presented outlining the possible applications of vibratory feeding
in industry. Typical examples include [14,15], where examples of VBF usage range from
pickle slicing machines to bottle cap presenters. In [16], the authors report on a
modification to the traditional VBF that tosses and twists parts to separate them. This
allows parts like springs and hooks, which easily congregate into bunches and become
VBFs. Schroer [17] examines the results of a series of experiments using custom tooled
VBFs for presenting non-standard electronic parts to a robot for insertion on printed
circuit boards. The parts proved difficult, but not impossible, to feed at adequate feed
rates. Four classes of part jamming phenomenon were described with their rates of
occurrence. It was determined that not all parts could be adequately processed using a
VBF, especially those parts with delicate components. A methodology was developed to
10
Improving feeding efficiency has been the objective of many research efforts. By 1960,
Pavidaylo [8] had identified the optimum VBF operating conditions, and had assembled
these into a set of design equations: to ensure maximum productivity; efficiency; part
movement stability, and the lowest sensitivity of these characteristics to the magnitude of
the friction coefficient. It was determined that the most effective working conditions are
those in which the parts move in a series of minute jumps, coming in contact with the
track only to receive an impulse for the following jump. This effort laid the groundwork
for further studies. Akhmechet et al. [18] revisited this issue with a VBF of improved
design. A variable-angle driver for use with a traditional VBF is examined in [19]. This
system is based on an elliptical throw; the bowl does not return on the same path as it
moved forward on, but returns on a lower level. The bowl moves away from the part as it
returns, rather than throwing the part away from the track. The result is extremely high
feeding efficiencies. The parts are in contact with the bowl only during the forward
motion. This method also tends to transport parts individually, leading to part
segregation. This work led to the developments in [7], described later. Rulh [20]
examines various methods of tweaking an installed VBF with simple, inexpensive, and
The literature provides a host of new designs based on the traditional VBF. In [22], a
simple VBF is developed to feed and orient cylindrical parts with a length-to-diameter
ratio close to one. This system consists of a sensor-less VBF with a flat bottom and no
helical track. The tooling is composed of a”discharge hole” through which parts, only in
11
the proper orientation, may fall. This hole is connected to the rest of the assembly system
via a delivery chute. Any part that fails to travel down the “discharge hole” simply
recycles back into the system and attempts the passage at a later time. The bowl, due to
simplicity, can be machined out of a single piece of stock. This versatility eliminates the
need for liners and allows for the bowl to be crafted out of virtually any material. In [23],
a bowl feeder is presented that replaces the single (or double) electromagnetic actuator
with a piezoelectric actuator located on each leaf spring. Three (or four) independent
actuators allow for the potential of increased driving signal flexibility. A VBF with micro
perforations arranged uniformly over the area of the helical track to reduce the
attenuation system to a traditional VBF to reduce excess mechanical vibration and noise
[25].
The motion of a part in a VBF, with its accompanying dynamics, is adequately described
in the traditional literature [5,6,26,27,28] etc. Many research efforts have attempted to
broaden this base of knowledge; in some cases simplifying the analysis for quick
resolution, in others complicating it, to allow for a complete analysis. The advantages of
modeling an entire system have been long known. The ability to simulate a process
before a prototype is built has obvious economic and intellectual advantages. Prior to the
wide availability of computing resources most of this type of work relied heavily on
12
Much of the earlier work was also strictly limited to sinusoidal acceleration cycles. Early
researchers of note include: Berry [27,28], who examined the effect of stick-slip motion
with a modified coefficient of friction; Booth [29], who pioneered the use of digital
[30], who examined the conveying velocity of a part and the characteristics of the modes
A pivotal early work was that of de Cock [31]. The basic theory of VBF operation was
examined, modeled and verified experimentally. It was discovered that the vibrations of a
VBF are almost purely harmonic, despite the electromagnet drive. Investigations into
spring design were performed, resulting in model design equations. The importance of
material transfer rate was recognized; efforts were undertaken to identify the dependent
variables and how they should be chosen to obtain the optimum result. A nomogram
composed of: coefficient of friction, track angle, and frequency verses feed rate was
assembled to aid developers with design parameter calculations. It was determined that
for any part, optimum operation occurs when the part just touches the track once in every
period, and then only when the track just reaches its maximum velocity in the upward
direction.
Winkler presented a pair of papers examining the various aspects of vibratory conveying
[32,33]. The first examines several possible conveyor designs: the “sealskin” conveyor;
the “jerk” conveyor; an inclined track conveyor, and a conveyor with out-of-phase
motion. The inclined vibratory track conveyor, where the part is limited to sliding
motions, was chosen as the most promising and a detailed analysis is developed. Due to
13
the difficulty of analytically predicting the performance of the track with a sinusoidal
velocity profile, a triangular velocity profile was introduced. The results of this
simplification were compared with simulated results for a sinusoidal profile run on an
analog computer. The second paper repeats the original research on inclined vibratory
tracks, but extends the scope of the research to include hopping motions. A similar
analysis can be found in [34], however, no simplification of the velocity profile is used.
An extension of this work provides practical solutions to the equations of motion, as well
as a look into single particle theory applied to beds of granular material [35]. Square
later work using a similar simulation methodology, and the square wave approximation,
track, using analog, digital and hybrid computer analyses [38,39]. His work considered
mechanically generated by four bar linkages. The aim of this research was the
between different parameters of a feeding system. Detailed flow charts are given which
describe the stable states and transitions of the system. An analog circuit is constructed to
simulate the system, and compared to a similar digital simulation. The analog computer
was found to be faster and more versatile when applied to the sinusoidal case, but nearly
14
In any vibratory feeder system there are two types of objects in motion, the feeder and the
part. Okabe et al. examine the motion of a part on a vibratory conveyor with a simplified
dynamic model [40]. For simplicity, the part is modeled as a particle which does not lose
contact with the vibrating surface at any time. The theoretical model includes:
acceleration of the vibrating surface; friction angle between the vibrating surface and the
particle; inclination angle of the surface, and the direction of linear vibration. The types
of motion are classified into seven primary modes: (i) no primary slip between the
particle and the vibrating surface; (ii) the particle slides forward intermittently; (iii) the
particle slide backward intermittently, and (iv-vii) the particle slides forwards and
determines into which exact class this type of motion falls. Later experimental
investigation with an inclined vibrating plate and a rectangular prismatic part confirmed
A general computer model for predicting the performance of vibrating conveyors was
developed [41]. The computer model enabled vibrating conveyors with reciprocating,
circular or elliptical vibrations to be simulated. The basis of the model is a single particle
mass analysis using bulk material properties as the input parameters. The performance of
the vibrating conveyor can be predicted by measuring the coefficients of static and
dynamic friction of the bulk material on the conveyer. Four modes of motion are
considered: stick, slide, hop, and the transitions from one mode to another. An
variable amplitude and frequency was constructed and tested. For the examined materials
15
the model proved valid. The results of the experimentation were compiled into contour
natural frequencies of a VBF [42]. A lumped parameter approach was used, whereby a
rigid bowl and rigid base are separated by three banks of flexible leaf springs. The bowl
is considered to have four degrees of freedom: vertical and rotational translation about Z;
for the base are identically defined. Bending of the leaf springs was modeled using the
Euler-Bernoulli beam theory. The masses of the bowl and the base are assumed to be
distributed symmetrically about the polar Z axis. The stiffness of the three leaf springs
are assumed identical. The system is assumed to be perfectly isolated from the outside
world. To verify the results from the numerical model, an experimental analysis of a
‘real-world’ VBF was undertaken. In order that it may be excited as a free structure, the
VBF was suspended from a rigid framework by low stiffness springs. A nontraditional
vibrator using a thin wire to transfer vibrations was used to eliminate any bending
constraints between the vibrator and the structure. Piezoelectric sensors were used for
force and velocity measurements. It was found that at higher frequencies of excitation the
lumped representation is not valid. The model was found to be over constrained; the
assumption that the base and the bowl are rigid does not result in a completely viable
model.
methods to evaluate system parameters [43,44]. Six degrees of freedom were considered:
16
vertical deflection of the base; twisting of the base; deflection of the leaf springs, and
their derivatives. The bowl and base are assumed to be perfectly rigid bodies with their
motions constrained to vertical displacements and twisting about the vertical axis. The
leaf springs are assumed to be without mass; their actual mass being distributed between
the base and the bowl. In addition, the deflection of the leaf springs is assumed to consist
solely of bending; tension and compression are ignored. Euler’s method was used to
determine the steady state motion of the bowl. Part motion was calculated using Taylor
experimentation with the driving electromagnet. The model was validated experimentally
vibratory feeder [45]. Incorporated into the dynamic simulation are various system and
frequency, and coefficient of friction. It is assumed that the part never leaves the track at
any time. The program is capable of generating graphically the position and velocity
values of the moving parts at small time increments, and to arrive at the average feed rate.
Theoretical test values compared with empirical results were found to be in general
agreement. The prime contribution of this work is a program flow chart which could be
The dynamic analysis and model development of an inclined linear vibratory feeder was
presented in [46]. In considering the various operating parameters that could affect the
conveying velocity of a part, five were chosen: angle and amplitude of vibration;
17
coefficient of friction; track angle, and operating frequency. The track was assumed to
vibrate with simple harmonic motion. The states of motion cover both contact (sliding)
considered: non-sliding, slow-sliding, and fast-sliding. The dynamic model was first
simulated on a personal computer, and then experiments were conducted to verify the
validity of the system model. Empirical data indicated that the model adequately
predicted the mean conveying velocity of the feeder. Variations in the coefficient of
friction and the inclination angle of the track were found to have profound influence on
In VBFs, the deformation of the spring is a complicated sequence of events. The exact
calculation of spring constant is difficult. Okabe and Yokoyama present a simple method
to calculate the natural frequency of a VBF for both fixed and floating mountings [47].
To make this possible, limiting assumptions about the deformation of the leaf springs are
made. The deformation is limited to the thickness, width, and torsion directions. These
deformations are independent of one another, lacking any geometrical constraint, so that
the total deformation can be calculated by means of the vector addition of each
deformation. This greatly eases the calculation of the spring constant. For the natural
frequency of the fixed system, an equation with a single degree of freedom is developed;
investigation concluded that the theoretical values are good when the width of the spring
is relatively small. If the width becomes large, the experimental values are smaller than
18
the theoretical values. This is attributed to micro-slip and insufficient rigidity of the
clamping parts.
developed that encompasses the geometry of the parts, track and orienting device, as well
as the main physical effects, including: multiple impacts with friction, and the interaction
with a vibrating track. The model is based on a structure variant multi-body system with
unilateral constraints and Coulomb friction. The result is a set of differential equations
with inequality constraints. Parts are assumed to be rigid bodies with two-dimensional
contact surfaces. This allows for any geometric object to be modeled, as the outline of the
contact surface is a flat plane. A simulation tool is developed that allows for the
the orienting devices; influence of different base devices; stability, and the influence of
different parameters on each other, as well as on the overall process. Three simulation
exercises are illustrated to prove the effectiveness of the tool; no empirical verification is
provided.
Automatic assembly machines require parts to be correctly oriented before the assembly
process can be undertaken. This usually necessitates the employment of suitable orienting
devices in the feeder that supplies parts to the automatic work head. Two classes of
orienting devices exist, passive and active. Passive devices reject parts which are
19
incorrectly oriented; while, active devices physically reorient the parts in the desired
manner.
Knowledge of the ways in which a part will naturally feed (a component of the natural
resting aspect) when placed at random in the feeder bowl is essential to the efficient
design of orienting devices. If, for example, a particular part has two possible natural
resting aspects, one feeds the part correctly while the other causes assembly failure, either
a passive system could be used to reject the improper orientation back into the bowl, or
an active system could be used to reorient the part into the desired resting aspect. The
natural respecting aspects of a part must therefore form the basis of any research dealing
with the feeding and orienting of parts, and is essential in any analysis of the performance
Boothroyd [49] studied the statistical distributions of the natural resting aspects for
various classes of parts. Basic part shapes consisting of regular prismatic and cylindrical
geometries were selected. The probability that these parts will come to rest in their
various natural resting aspects when placed in the bowl of a vibratory feeder was
analyzed and verified experimentally. Two conditions are examined the static solution
and the dynamic solution. The static solution consists of the probability of the part
tending to fall into a particular natural resting aspect immediately on impact with the
surface. The dynamic solution takes into account the bouncing and rolling of the part
after the initial impact. It was found necessary to assign cylindrical parts an empirical
factor to complete the analysis successfully. The authors believe that in real world
20
Further investigation by Boothroyd [50], revealed that it is possible to predict the
behavior of both regular prismatic parts and cylinders without requiring an empirical
factor. The additional factor of surface type was included in the analysis: soft rubber like
surfaces, where, on impact, a corner of the part depresses the surface, adding a horizontal
component to the motion and hard resilient surfaces where deformation does not occur,
making the horizontal component negligible in effect. The probability that a part will
attain a particular natural resting aspect was determined to be a function of two factors:
the energy barrier tending to prevent a change and the amount of energy possessed by the
Ngoi [51,52] presented an alternative for analyzing the probabilities of the natural resting
aspects of a part. This method utilized the concept of the energy envelop in conjunction
with the previously described energy barrier. The energy barrier method requires tedious
plotting of the changes of location of the centroid. The energy envelops are constructed
in a CAD program solid modeler, from which changes in centroid location are easily
determined. The resulting probability values for both regular prismatic parts and
cylinders agreed with the previous experimental data of Boothroyd and collaborators. The
Investigators still sought a simpler method to analyze the natural resting aspects of a part.
A hypothesis that the probability of a part coming to rest on a particular aspect is directly
proportional to the magnitude of its centroid solid angle [53] and inversely proportional
to the height of its centroid from the aspect in question was examined [54]. The centroid
solid angle was determined for any given geometry by using a CAD solid modeler. The
21
hypothesized method was applied to rectangular prismatic parts and cylinders. The results
were found to agree well with earlier published data using the energy barrier and energy
envelop methods. These results were also verified [55] for square and rectangular prisms
on a hard surface. In both cases, the Experimental data agreed well with previous centroid
Utilizing the same method, Ngoi et al. [56] examined the natural resting aspects of
complex parts. Previously, complex geometries were decomposed into simpler regular
prismatic and cylindrical shapes and then analyzed. These analyses provided only
approximations, which under working conditions could prove detrimental to the proper
symmetrical (T-shaped) complex parts verified the validity of the hypothesis. A similar
work [57] used the centroid solid angle method to analyze the natural resting aspects of
parts with displaced centers of gravity. Previously, only parts whose center of gravity
resided at their geometric centers had been examined in detail. The difficulties lie in the
fact that empirical ‘compensating’ factors have to be generated for each part, a tedious
and time consuming exercise. The experimental results confirmed the validity of the
Validation of the existing hypotheses on predicting the natural resting aspects of parts has
been conducted by application of the “drop test”. The “drop test” consists of repeatedly
dropping parts from a certain height onto a surface in order to analyze the natural resting
behavior of the parts. Work has been conducted by Ngoi et al. [58] to determine the
accuracy of the “drop test” in determining the validity of these hypotheses when applied
22
to a VBF. Accurate experiments were conducted in a laboratory to examine the
aspect ratios by varying the vibration amplitude levels and surface textures of the
vibratory bowl. These results compared favorably to those of Boothroyd, Ngoi, and
hypotheses validation. In addition, it was discovered that vibration amplitude had little
This final conclusion is directly conflicting with Rosario [59], who conducted an
feeder. A series of experiments were conducted to generate a set of empirical data used to
investigate how complex parts behave in a vibratory bowl feeder, and to establish a
relationship between the orientation of the parts and the vibration amplitude of the bowl
feeder. Three amplitude levels and four electrical connecter types were used. The
empirical data suggested two primary conclusions: (1) the tendency for a part to rest in a
certain position is directly influenced by the vibration amplitude of the feeder, and (2)
when parts are allowed to recirculate in the feeder; the probability distribution could
greatly differ from the distribution of parts that exit the feeder with no recirculation.
Active orientation devices and vibration amplitude changing controls are suggested as
methods of efficient part feeding once the proper connector orientation is identified.
23
1.5 Orienting Devices and Tooling
In any automatic assembly machine, the parts must be fed to the workheads correctly
oriented and in discrete units. The nature of the VBF allows for both of these
requirements to be fulfilled. In standard operation, the VBF generally discretizes the parts
without intervention; while, the flexibility of the VBFs design allows for a wide array of
orienting devices. The devices employed to ensure that only correctly oriented parts are
fed to the workhead fall into two groups: those that are incorporated within the parts
feeder, which are usually referred to as in-bowl tooling, and those that are fitted to the
transfer system between the feeder and the work head. Passive tooling refers to orienting
devices that operate on the principle of rejection. Only those parts that are fed correctly
oriented pass through the device; the remainder fall back into the bowl. The rejected parts
are then re-fed and make further attempts to pass through the orienting devices. While
simple and generally inexpensive, this method of orienting parts has a detrimental effect
on the feed rate. For example, if six possible orientations need to be tooled for, the feed
rate of the tooled VBF will be, on average, one-sixth that of the untooled VBF. Active
tooling refers to orienting devices that actively reorient parts into the proper orientation,
generally without rejection. While costlier and technically more sophisticated this option
offers the advantage of increased feed rate, as few to none of the parts are recirculated.
Delicate parts that may not survive a trip back to the bottom of the bowl often require
active orienting devices. Out of bowl tooling is usually of the active type, as there is no
24
Active and passive tooling can be further classified into dimension orienting devices and
feature orienting devices. Dimension orienting devices use the dimensional differences
within the basic shape of a part to reject an incorrect orientation of the part. An example
would be a cut-away placed on the VBF track that would drop a rectangular part back
into the bowl if it happened to be fed with the longer dimension perpendicular to the bowl
wall. Feature orienting devices utilize a certain feature of the part to reject or reorient it.
A slot cut into the above mentioned rectangular part could be used to ensure a proper
keyed orientation. Orienting devices are often used in series to obtain the desired effect.
The literature covers many of these basic orienting devices, their implementation, and
feeders, as well as VBFs, Boothroyd [63] introduces a novel coding system for use with
feeding and orienting small parts. This system uses the basic and underlying geometric
features of a part to describe its overall shape; assigning values developed empirically to
properties such as symmetry, rotation, thickness and length. These values are then
compiled into coded numbers that provide a reference as too how difficult the part will be
Early examinations of orienting devices for VBFs generally involved passive devices.
Boothroyd and Murch [64] performed a statistical analysis of the performance of a vee
cutout orienting device installed on a VBF track. The results of the analysis allowed for
the design and implementation of the vee orienting device without the need for traditional
trial and error procedures, while also retaining a high level of efficiency. The experiment
was conducted in such a manor as to allow for the substitution of most other types of
25
passive orienting devices; creating the potential for the development of an empirical
database to aid designers in the choice of components for assembly, and in the design of
VBF orienting devices and systems. This goal was partially realized in the “Handbook for
Feeding and Orienting Techniques for Small Parts” [26,63], where the complete analysis
of some 30 basic types of passive orienting devices is described. In addition to VBFs, this
work covers the design and tooling of a vast array of parts feeding devices, including but
not limited to: bladed wheel, centrifugal, reciprocating tube, stationary hook, centerboard,
and rotary disc hoppers. Murch and Poli [65] exemplified one of these orienting devices
(edge riser I) for feeding and orienting rectangular prisms. The purpose of this work was
to describe in detail the procedure used to obtain the data presented in [63], providing a
Boothroyd and Murch [66] designed a technique using matrices to calculate the
device would be the product of that matrix. To determine the overall efficiency for a
series of devices, each device’s matrix is first multiplied by an initial distribution matrix,
and then the product of all the matrices is found. The resulting number is the efficiency of
the combined system. Similarly, Yeralan and Chang [67] described three generalized
manipulations.
26
Sethi and Sriskandarajah [68] developed and proved theorems to order the orienting
developed to select the optimum devices for a series system for either one part or m-part
families. These methods proved time consuming as optimal solutions were found by
exhaustive searches. Later efforts by Sethi [69], in an attempt to improve earlier heuristic
procedures, provided new analytical methods for developing serial systems with two
orientations for single and multiple part families. A practical example of using this
heuristic model with the tooling of a VBF, for use with cup shaped parts, is provided.
Parallel configurations are briefly examined. Jaumard et al. [70] also built on Yeralan’s
work to examine the design parameters and ordering of series part-orienting devices.
However, where Sethi selected the orienting devices in a sequential manor, this work
emphasized precedence selection. Three Heuristic methods are developed and simulated
in Fortran-77 for a VBF with three orienting devices: step, scallop cutout and ledge for a
cup-shaped part. Efficiencies are calculated and found reasonable when compared with
For even the simplest parts, manually designing part orienting systems can quickly
investigations were launched into the development of methods that would automate the
design of part orienting systems. Natarajan [71] presented an algorithmic approach to the
automated design of two simple types of part orienting devices: a stepped belt and a
panhandler. The belt is defined as a device for orienting two dimensional polygonal
objects that are infinite in the third dimension, while the pan handler is for planar
27
polygonal objects. This work attempts to decompose ill-defined physical problems into
the computational domain for analysis. For the examined cases, the results proved
form of parts feeder, it is unclear how effective this method would be when applied to a
Natarajan later expands this examination by classifying parts feeders as either orienters or
filters within an algorithmically defined boundary [72]. First, a formal statement of the
parts feeder design problem as it pertains to the orientation and presentation of parts is
given. “Given an object O and two sets of orientations T0(Θ0) and Tf for it, design a set of
obstacles Tfeeder(Θ0β) that forces the object to move from any orientation in T0(Θ0) to
some orientation in Tf.” The generic algorithm proposed is as follows: starting with a
geometric representation of the part which has to be fed; a range of initial orientations of
the part that the feeder must handle, and the desired final orientation of the part as it exits
the feeder, a feeder design algorithm should come up with a suitable feeder and the
detailed setup data for it, along with performance estimates and other technical
predictions. Three devices were considered for this work. The belt and panhandler from
the previous work are classified as orienters. In the filters class, an inclined track
vibratory feeder (approximating a VBF) equipped with passive orienting devices such as
pins and cut-outs is studied for use with planar polygonal parts. As before, the problem is
first made precise by making suitable assumptions and then abstracted to the
computational domain. Also as before, the focus of this work was on the combinatorial
28
Christiansen et al. [73] developed a genetic algorithm specifically tailored for the design
of gates on VBF tracks. The highly parallel nature of the genetic algorithm allows this
method to supercede earlier exhaustive search methods in terms of speed, while still
the stochastic matrices defined by Murch and Boothroyd [66]. Screw like fasteners and
cup shaped parts were used to test and compare the system to earlier work. Results
combined with the recent advent of inexpensive computing power, has led to the
development of an advanced algorithm specifically for use with VBFs to design trap
filters for n-sided polygonal parts [74]. In this work, parts are described by the number of
vertices, the center of mass, and the number of possible orientations. Three types of traps
are considered: gap, canyon, and slot. A general framework is also provided for the
custom application of the algorithm to other trap types. Parts are filtered only in 2-D,
difficulties arose with 3-D filtering, namely parts getting stuck halfway falling. Small
tolerance variations in parts and positional uncertainly associated with vibratory feeding
algorithm and the analysis of two addition trap types: balconies and bridges. Agarwal et
al. [76] take a minimalist approach to the development of a geometric algorithm to design
presented. The simpler data structure reduces the time to solution. It is shown that if a
29
feeder exists that supports one particular pose of a part the algorithm will eventually find
statically tooled parts feeders can be found in [77]. Parts feeders that do not use sensing
information to accomplish the task of orienting a part are considered: traditional VBFs -
the design of passive traps is considered; simple parallel jaw grippers - push and squeeze
functions; conveyor belts - fence design, with the goal of finding the shortest possible
design that will result in an acceptable orientation, and for tilted plates, examinations of
the problems associated with pushing parts in 3D space in combination with fence design.
The algorithms take the description of the part geometry as input and output a sequence
of actions that moves the part from its unknown initial pose into the desired unique final
pose. For each part feeder considered, the classes of orientable parts are determined,
algorithms for synthesizing the sequences of actions are derived, and system limitations
are provided.
Berkowitz and Canny [78] presented a tool based on dynamic simulation for doing
parameter enumeration, analysis, and Markov model building of tooling in a VBF. The
VBF was simulated as if the track had been unwound and laid linearly at an angle equal
to the average angle of inclination along the helical track. The VBF is characterized by a
small set of parameters, thus, allowing for a large range of parameter values and
simulation of two parts feeder designs indicate that dynamic simulation may be a feasible
method for determining the proper configuration of orienting devices. In [79], the effects
30
of the dynamic simulation tool are compared to a real design. Results reflected reality
with a high degree of accuracy; however, it was found that minor inaccuracies arose in
the simulation because of the simplifications in the dynamics model employed by the
Attempts to computerize the design tools outlined in Boothroyd’s methodology [63] have
through an interactive ‘query, answer and reason’ session. These systems made use of
expert knowledge to generate part codes for the selection of suitable feeding and
orienting devices for use in automatic assembly. Tapadia and Henderson [83] combined
the UMASS coding system of [63] with feature recognition and extraction together with a
rule based expert system. Li and Huang [84] used feature based geometric modeling to
automatically generate the code. Lo and Dick [85] used CAD solid modeling techniques
to generate part orienting tracks in a limited non-analytical fashion. While functional for
the purpose of generating part codes, these methods lack significant analytical
capabilities and with one exception the ability to link part codes and orienting devices
seamlessly.
Yeong and Warren developed a methodology for parts feeder design combining previous
design efforts into a framework for feeder design [86]. From [72], the problem statement
on design specifications and constraints was borrowed, but modified to include time
allowed to orient parts and the maximum forces that a part can tolerate. Selection of
translation and rotations about the X, Y, and Z axis in mathematical context. To reduce
31
the complexity of series configuration of orienting devices, a sequence of feeding zones,
or zones, responsible for a single reorientation process, the task of designing complex
sequences of tools is made manageable. In order to fully characterize the feeding device,
the physical parameters of each zone, such as: barrier angle, step height, length,
recommend using published analytical models for whichever feeder the designer is
improve the zone designs. To illustrate the validity of the methodology, a brief case
study is provided. A belt feeder with a step tool is used to orient a rectangular prismatic
A comprehensive expert system that linked part features directly with an orientation plan
was developed by La Broy [87]. The system accepts user-drawn views of parts of an
assembly, from either the internal drawing package or from a variety of commercial CAD
systems, evaluates the design for efficiency based on the principles of Product Design for
Assembly, and specifies the details of the machinery needed to feed and orient parts at
high speed (usually VBF based). The strength of this system lies in the comprehensive
Ou-Yang and Maul [33] developed a system that directly coupled the CAD representation
geometrical and topological properties of the part and selecting appropriate orientation
devices for efficient feeding. This analysis package incorporates: near convex shape
32
computation; basic shape analysis; symmetric analysis; feature identification and
analysis; stable orientation analysis, and orientation device analysis into a comprehensive
design tool. Early stage part modification allows complete design flexibility. However,
this system has constraints; part geometry is limited to rectangular and rotational parts. In
addition, only one feature from each part face may be considered.
presented by Lim et al. [88] and later, with improved functionality by Tan et al. [89]. The
traditional CAD representation model was deemed too complex to represent geometric
features and replaced with a system of three-dimensional spatial matrices called the
‘Spatial Representation’ [90]. A knowledge based expert system was employed for the
selection of the orienting devices through the use of a unique internal coding system
similar to, but more comprehensive than, Boothroyd’s [63]. Although this system lacks
the analytical capabilities of Ou-Yang and Maul, the limitations of this system are based
Manufactured parts are rarely identical. Part tolerance introduces a certain amount of
shape uncertainty, yet most orienting devices are designed with the ideal part in mind.
The main operational effect of shape uncertainty on parts orienting is the introduction of
nondeterminism arising from variations in part geometry. Akella [91] examines the
design of orienting devices that account for part shape uncertainty and proves that non-
to generate reliable sensor-based and sensor-less orienting plans for a class of shape
uncertain parts.
33
For part feeding devices other than VBFs, the orienting of parts has been a brisk topic of
research. Erdman and Mason [92] were able to eliminate configuration uncertainty of a
part by repeatedly tilting a tray. A vibrating plate was used to position and orient parts
[93]. A number of researchers dealt with the general design of a sequence of passive
fences placed over a conveyor to automatically orient a sliding part [94, 95, 96, 97]. A
complete set of algorithms for designing a sequence of passive curved fences to orient
polygonal parts was developed in [98,99]. Active part manipulation systems utilizing a
sequence of fences to push parts at different angles was examined in [100]. Rules
predicting the rotational direction of pushed objects formed the basis of this research
[101]. A system using a single, movable, one degree of freedom fence oriented parts on a
implemented by using a tilted tray and a tactile probe [104]. Later, sensor-based orienting
plans were generated automatically for tray tilting and grasping by a parallel jaw [105].
Orientability and diameter sensing part recognizability are defined using a parallel-jaw
gripper with polygonal parts in [106]. In [107], a new type of feeder is proposed that
orients polygonal parts with elevated edges by pull operations with an overhead finger. A
general comparison of sensor-less and sensor-based orienting plans showed that simple
sensors reduce the number of orienting steps [108,109]. Redford et al. presented a broad
survey of feeding and orienting systems for presentation of parts to a multi-arm assembly
robot [110].
34
1.6 Programmable VBF Research- Flexible Parts Feeding
Dedication to single part geometry has relegated traditional fixed methods of parts
increased in capacity and complexity, the need for dynamic orientation of parts and
increased flexibility in assembly systems was soon recognized [111, 112, 113, 114, 115,
116]. With the introduction of the personal computer and reasonably fast machine vision
systems, new hopes for the development of limitless assembly techniques arose [117].
Over the past 20 years researchers have examined a number of different approaches to
overcome the problems associated with dedicated feeders. The objective has been to
reduce the amount of special purpose equipment associated with feeding a particular
component while simultaneously reducing the time required to reconfigure the system for
different component types. The result of these labors has been generically termed the
‘flexible feeder”.
component types by either mechanical or sensor based means. The principle of flexible
feeding attempts to eliminate or reduce the costs associated with dedicated systems, the
resulting feeder is not limited to one component type. Costs associated with parts feeding
can be dramatically reduced as the cost of the feeder is no longer tied to the life of a
single product, but rather amortized over the life of the feeder. In addition, Flexible
feeding methods allow for dynamic product modification; which in the past would have
35
Few attempts were made at flexible parts feeding without some programmable
tracks. The VBF was replaced with a linear vibrator equipped with orienting tracks
machined with CNC milling centers. To automate NC code generation and enable
software package was developed. The user would design a track by first selecting, from a
library of functions, the required sequence of orientations and then by entering the
physical dimensions of the part. The program would generate NC code to machine the
feeder track into modular sections of equal mass; the number of section being dependent
on the complexity of the orienting scheme. The cost of this system still proved to be high,
as different components still required the physical exchange and re-tooling of track
sections.
conjunction with a part classification scheme based on Boothroyd’s coding system [63].
One part family, the headed part, was chosen to demonstrate the programmable tooling
concept. Six tools were required for the orientation of the headed part; the wiper, pressure
break, narrow track, retaining rail, slotted track and hold down. Each tool consisted of a
lead screw carriage with a dial to count the number of turns applied. With the machine
off, a part is placed at each tool in the desired position. The lead screw is then
manipulated into the required setting and locked into place. In this way, each tool can be
independently programmed and the ‘program’ saved for later reference. While simple and
36
easily programmed, this system, once implemented, is generally limited to a single part
family, incurring re-tooling costs if the need to feed alternate parts families arises. Later
work on this concept included an adaptive control algorithm that learned the proper
the presence and passage of parts [120]. Each tool in sequence receives preset
information from the data gathered from the previous tool. The original tooling plan for
headed parts was used to determine the effectiveness of the algorithm. The testing of the
adaptive algorithm was simulated by manually executing the steps of the control
algorithm on a programmable feeder to determine the tool settings for a variety of parts.
Programming times were found to be rapid; the effectiveness of the algorithm was
proven.
University [121]. The basis of this system was a simple, efficient, and inexpensive sensor
package mounted to the outlet of a VBF. In place of a camera, a grid of sixteen fiber optic
light sensors with overhead lighting was used. The sensor array interfaced to a personal
computer which first memorized and then comparatively recognized part silhouettes.
information in the silhouette patterns. An air-jet mounted in the VBF wall immediately
before the outlet was used to return incorrectly oriented parts back into the bowl. The use
of an air jet as tooling provided many advantages. An air jet is passive, no complex
algorithms are required for control; non-contact, fragile parts are less likely to be
damaged; selective, response time is quick to allow correctly oriented parts to pass; and
37
common, compressed air is readily and inexpensively available in most manufacturing
environments. The sixteen-bit vision sensor, although simple and inexpensive, was well
suited for simple part orientations. The PSR concept faced low throughput issues due to
the nature of passive tooling systems. Stacked, overlapping or contiguous parts created
irresolvable difficulties for the relatively low resolution detection system. In addition,
Later work at The Ohio State University improved upon the first PSR system [122,123].
The VBF could now handle a sequence of geometrically different parts. A personal
computer system was used to store the recognized images, recall them in a random access
manor, and control an air jet based on the results of a comparative analysis. A pre-
Programming the system consisted of placing a part in the desired orientation in the
sensor window, turning a control switch to “Read” and pushing a “Set” button. Any
number of parts may be entered in this fashion, limited only by the physical memory of
the computer. A script could then be generated instructing the controller to accept parts in
a certain order. An algorithm for predicting the cycle time for any sequence of parts was
provided. This allowed for the mixing of many different types of parts in a single VBF
system, greatly reducing the number of feeders required in an assembly operation and
allowing for greater overall flexibility in the manufacturing process. Like all systems
singularly employing passive tooling, this system suffered from low throughput.
Further development of the PSR system at The Ohio State University [2,124] involved a
new computer based 3-D sensing strategy and a faster 16-bit single board computer
38
system. Three primary concerns were addressed: the problem of contiguous and
overlapping parts; the computational resolution of parts into discrete units, and the speed
of recognition and decision making. Two groups of eight fiber-optic sensors placed
parts move by the sensor arrays, two silhouettes are recorded, one from each array, and
compared with the silhouettes of the correct orientations previously stored in computer
memory. If they don’t match, the air jet returns the part back into the bowl. Following
analysis the algorithm automaticly loops, awaiting the next part. The vertical set of
optical-fibers is used to detect overlapping parts; while contiguous parts are detected
system, for example, with rectangular prismatic parts, it is often not possible to detect
where one part ends and the other begins. In addition, variations in part velocity inherent
in VBF systems seem to occasionally confuse the algorithm. For most geometrics,
however, the developed algorithm deals successfully with overlapping and contiguous
parts. Like all PSR systems before, this system suffered from throughput issues. With the
system became apparent; limited sensor precision resulting from only eight fiber-optic
If an air jet stream could knock a part back into a VBF, could not a lesser, more precisely
directed stream change the orientation of that part? The logical extension of the work
with passive air jet tooling is the analysis of active air jet tooling. Jaksic [87, 125, 126,
127] designed, built, tested and analyzed a VBF equipped with a set of active
39
programmable air jet tools in combination with a passive air jet tool (air-barrier). This
system could quickly readjust operating parameters to successfully feed a variety of parts
and eliminate jamming. A PSR, identical in basic function to that in [2, 124], with the
exception of an improved computer system, was used. Three active tools were studied to
control: yaw, pitch, and roll of a part. A mathematical model of part behavior during
reorientation was developed for each tool type. This model was incorporated into a
computer algorithm to supply an air jet pulse to reorient parts. Experiments with
prisms were used to empirically validate this model. By incorporating active tooling the
parts, difficult to feed in conventionally tooled VBFs, are handled without breakage as
they are not regularly returned to the bowl. Research empirically examining the forces in
the air jets used by this system can be found in [128]. Additionally, a closer examination
of part reorientation modeling in a plane orthogonal [129] and parallel [130] to the track
A modular approach for the design of passive orienting devices in a VBF was proposed
using modular orienting devices (MODS), it was hoped that the function of a VBF could
be rendered flexible without the need for sensors and algorithmic control. Each MOD
MODS can be sequenced in any manner and easily interchanged, depending on the
40
proposed. Twelve MOD sections were designed and tested for effectiveness. A series of
thirteen parts, nine rotational and four prismatic, were tested to determine the viability of
the modular approach. All thirteen parts were successfully oriented using combinations of
the ten prototype sections. A MOD sequence design methodology was not provided; the
authors used trial and error to determine the proper sequences for the test parts.
Cronshaw et al. [132,133] developed a flexible assembly system using a vibratory bowl
feeder combined with a vision system. The VBF is the first stage in the part orienting
process. The VBF outputs a part in a stable orientation onto a belt which transfers the part
travel velocity as the part passes the sensors. The sensors are composed of a
semiconductor line scan camera with fiber optic inputs which form 2D binary images of
the part by repeated rapid scans. The camera and fiber optics are easily configurable to
allow for the examination of different part geometry envelops. A microcomputer analyzes
the image and determines if the part is acceptable in both orientation and type.
Programming the system involves the use of a TV screen and a light pen. This system
represents an early adaptation of active selective tooling coupled with a vision system.
The in-bowl tooling is still inflexible, prone to jamming and re-tooling expenses.
Suzuki and Kohno [134] utilized a system of multilevel untooled VBFs capable of
delivering parts to a vision system for orientation by a robotic arm. Parts to be examined
were fed singularly into a scanning area, where they are forced against orienting device to
assure a consistent datum plane. The vision system would then scan the part to determine
if it needed to be reoriented. If the current orientation of the part had been identified as
41
unsuitable, a robot arm was used to reorient the part. Flexibility was limited to part
shapes within a particular size and stable resting plane; reprogramming was required if a
part that fell out of this range was to be fed. The main advantage of this system is the
elimination of the special purpose bowl tooling costs associated with the traditional VBF.
In [135], Swift and Dewhurst examine a VBF fitted with a laser scanner to orient small
parts with slight asymmetry, which would be difficult to orient via other means. This
system is programmable and capable of handling a wide variety of part types. This
technique is based on measuring light intensities scattered from the surface of the part.
These measurements were then used to identify the part orientation and compare it to
Pherson et al. developed a programmable parts feeder based on a double belt system
[136]. The feeder consisted of an inclined feeding belt, a horizontal return belt and a
components on the return belt. The programmability of the system was provided by a
series of optical reflection sensors. These sensors detect the orientation signature of a part
and compare the results with reference signatures stored in a computer database. Based
on the signature, various orientation blades were used in sequence to manipulate the
A similar system designed by Zenger and Dewhurst simplified this design by utilizing a
single array of fiber-optic sensors to detect the signature of correctly oriented parts [137].
Those parts that did not express the correct orientation were rejected by means of a
42
solenoid pusher to re-circulate. While this approach simplified the system design,
minimizing costs in tooling and time spent on set-up and programming, it does not
passively or actively re-orient components. Orientation is set only by chance as the part
travels up the track towards the sensors, and reoriented by chance, if it is rejected back to
the bottom of the bowl to recirculate. A part classification system for this type of feeder
was developed. Parts are classified as stable and non-stable; where those classified as
stable have one or more stable orientations when traveling along the feed belt.
With the advent of inexpensive computing and sensing equipment, researchers revisited
the double belt feeder system with the goal of providing a low-cost and flexible parts
feeding system that is capable of feeding complex geometries at feed rate between 30 and
60 parts per minute [138]. Low cost, high resolution, pattern recognition sensors
combined with a simple, efficient controller and low cost active tooling is used to
upgrade the previous feeder into a design with significant increases in efficiency and cost
effectiveness. It was realized that the earlier proposed component classification scheme
was insufficient for the needs of the double belt feeder; correctly oriented components are
also dependent on the manor in which they interact in the feeder, not just the number of
stable orientations. In addition to stable and non-stable, parts were also classified as semi-
stable; different sides of the same part may exhibit varied stability. A comprehensive
analysis of feed-rate errors and systemic limitations for double belt type feeders is also
presented.
A flexible feeding system that utilizes multiple ‘cartridge like’ bins to feed different parts
is presented in [139]. Programmable mechanical guides are used to discretize and orient
43
parts. Several parts are dumped onto a vibrating platform. The parts then hop onto two
central conveyors; adjustable bars are used to separate overlapping parts. Adjustable
wedge shaped guides create two columns of parts, in single file, along each of the two
conveyors. Return conveyors recirculate any parts which happen to fall off of the belt.
After parts reach an inspection station at the end of the belt, a high-speed machine vision
system is used to determine the orientation of the parts. With a mirror placed at 45ْ to the
inspection station, two views of the part are simultaneously analyzed. A robotic arm
picks the part if it happens to be in the appropriate configuration, if not, the belt returns
the part to the feed bin and the process repeats. The number of part types which may be
Hill and Sword [140] developed a flexible feeder using a vision system and turntable to
orient small non-rotational parts. Parts are initially fed onto a translucent turntable. A
back light was use to provide illumination through the table to enable a vision system to
see the profile of the part and identify its orientation. Following identification of part
orientation one of two possible scripts would be executed. The turntable would either
rotate to the left, rejecting the part for recirculation, or rotate to the right, allowing the
part to proceed onto a programmable elevating platform whose height determined into
which orientation the part would finally tumble. Hill later applied a vision system and air-
jet tooling to a VBF [141]. Parts moving along a track enter a view window. When a part
reaches the end of the view window the camera takes a picture. An air jet known as a
‘gate-jet’ is turned on, preventing other parts from entering the camera viewing area. The
picture is analyzed with a computer to determine the state and orientation of the part. If
44
the part happens to be in the desired state and orientation, the second ‘return-jet’ is turned
off long enough for the part to pass. If the ‘return-jet’ is not turned off, the part is
knocked back to the bottom of the bowl for recirculation. In either case, the ‘gate-jet’ is
For the frictionless gripper studied in [103, 142], Goldberg et al. [143] presented an
algorithm to find stochastically optimal parts feeding plans for n-sided polygonal parts
for incorporation into a flexible parts feeding system. Two classes of actions were
studied, the squeeze action and the push-grasp action, where the part is pushed by one
jaw prior to grasping. The algorithm was implemented with a linear bearing and a Lord
parallel-jaw gripper attached to a PUMA arm. Two rectangular parts were examined to
rapid set-up of assembly lines using this system, Goldberg built a simulator to provide
visualizations and realistic timing estimates of feeder throughput [144]. The simulator
takes a CAD model of the part as well as the current position of the robot, conveyor belts,
and cameras as input. An estimated feeder throughput dependent on overall part arrival
rate, the distribution of stable states in which the parts arrive, and the time required to
carry parts from the feeder conveyor belt is provided. A description of how robot cycle
times may be estimated using statistical sampling is also provided. Experimental results
Branicky et al. [102] introduced the modeling of a complete feeder system based on the
statistical probabilities of the feeder sub-systems and on the probabilities of parts resting
45
in stable orientations. The system studied was a conveyer belt and robot-arm vision based
selective feeder described in [145]. This work differs from previous modeling efforts by
offering a high-level statistical equivalent of the physical system that can be used to
strategies and high-level simulation and analysis. The vision system and conveyor system
were modeled independently by the Poisson process, while the parts retrieval system was
determined by empirical testing. This data was then combined into a generalized semi-
Markov process model of the entire flexible feeder system. Three physical test cases were
modeled: a single part, multiple parts fed at once, and multiple parts fed in a specific
order. The model proved consistent with previously collected empirical data.
Industry has driven many of the latest innovations in flexible feeding. Sony, with APOS,
was first to reach market with a viable flexible feeding system [146]. Originally meant to
feed electronic parts, this system utilizes vibratory dynamics to orient parts in large
pallets of part nests. Parts to be oriented are poured from a hopper onto the inclined
pallet. The pallet is subsequently vibrated with three degrees of freedom, two linear and
one rotational. The pallets are designed to help guide the parts into part-shaped cavities.
Over several seconds a high proportion of the parts fall into the nests; unnested parts are
swept away and recirculated. Specially tooled pallets are required for each part, as well as
a pallet transport system to transfer the palleted parts from the feeder to the assembly
station where robot arms utilize the oriented parts for assembly.
46
Adept Technologies inc. [147] has developed a flexible feeder composed of three primary
components: a robot arm, a part circulation unit, and a vision system. Parts are loaded
into a lifting bucket which then tips them onto a feed conveyor. To deal with the
possibility of overlapping parts, the feed conveyer feeds the parts onto a second belt
running at a higher velocity. This velocity differential is usually all that is required to
discretize the parts to be examined. The second belt is then stopped to allow a vision
system to scan and identify part orientations. The robot arm is then instructed to pick the
parts identified as having the correct orientation; parts may be oriented around the
vertical axis if required. Those that are not identified as having the correct orientation or a
The Polyfeed system from MIKRON Assembly Technology is a hybrid of the Adept
system and a linear vibratory feeder [148]. The feeder consists of a compact part
conveying system, a vision system, and a robot arm. The basic operation of the system is
the same as the Adept feeder. A camera and lighting system moves over the inspection
area and identifies components lying in the chosen orientation. After a certain number of
components have been identified, a robot arm is instructed to pick them up. Once all the
components with recognizable orientation have been processed, the inspection area
vibrates in an attempt to re-orient the remaining parts; the identification and picking
process begins again. This process continues until the vision system is unable to identify
any further parts in the inspection area, at which point, a new batch of parts is delivered.
System cost is the primary drawback of these new flexible feeding systems; they all
require a dedicated robot cell resulting in large, complex, and expensive machinery.
47
1.7 Waveform Research
feeder track. This generally results in the inclusion of inefficient part motions in a feeding
cycle. Typically these involve backward sliding or unstable bouncing type motions,
producing inefficiencies which result in the reduction of feed rate; increasing the cost of
the standard sinusoidally driven parts feeder. A number of unique examples are
bristled ciliated track [149]. Yokoyama and Okabe equipped a standard vibratory
conveyor with a non-linear damper that advantageously altered the sinusoidal waveform
[150]. In an attempt to provide a quieter less disruptive feeding system, Singh replaced
conveyor with a 1-hp motor eccentrically driving a four bar linkage [151].
Okabe et al. [152] examined vibratory feeding by non-sinusoidal vibration. The velocity
associated with the generalized vibration wave form is approximated by six connected
straight lines. Five equations called ‘distortion factors’ were defined in order to derive the
optimum vibration wave form and the optimum conveying conditions. The types of
motion of the part were classified into seven modes by the combinations of positive slip,
negative slip, and stick relative to the surface. It was determined that the ‘noise’
generated by the hopping of parts would hinder the fundamental analysis, hence, part
motion normal to the vibration surface is not considered. The experimental set-up
48
moving coil type vibrator. It was found that if a suitable wave form and proper vibrating
conditions are chosen, the instantaneous velocity of the particle never takes a negative
value and the mean conveying velocity becomes large relative to conventional vibratory
feeder designs. If the amplitude of the waveform is increased, it was found that the
conveying velocity was increased, bounded only by the limitations of the hardware. The
limited nature of the experimental setup prevented the exhaustive examination of all
that such a system would result in higher feed rates. A research effort, similar to [152],
vibration, was described in [154]. The results of the theoretical analysis indicate that
In [155], the motion of a part on an oscillating track having both parallel and normal
analysis is limited to the case where the part does not lose contact with the track at any
point during the cycle. To simplify calculations, acceleration cycles in both parallel and
normal directions are assumed to be step functions. An attempt is made to determine the
important factors of track design and asses quantitatively their influence on feeding
49
improved feed rates over traditional parallelly excited tracks. A rational design
normally excited track was presented in [156]. A vibratory conveyor excited by two in-
phase non-sinusoidal waveforms, normal and parallel with the track, was examined.
These waveforms were composed of the first two harmonics of motion. The analysis of
part motion was based on single-particle dynamics. Types of part motion considered
include: positive slip, negative slip, stick, and hopping. To study the performance of this
vibratory conveyor, an exhaustive parametric study was carried out for 81 waveform
combinations. Analysis was not limited to a completely horizontal track, various angles
of inclination were considered. The study revealed that non-sinusoidal excitation gives
mean conveying velocities higher than that of a conventionally driven feeder. Also,
where conventional feeders have difficulty feeding parts when the track is configured at a
high angle of inclination and low coefficient of friction [157], it was found that the non-
Quaid developed a miniature mobile parts feeder resembling a small VBF with no bowl
reservoir [158]. A specially shaped feed tray is rigidly attached to a planar linear motor.
The tray has an annular feed path for parts, a sloped ramp section, and a flat plateau
motion of the parts. When the bulk parts are loaded at the bottom of the ramp, parts
slowly climb the ramp, but only near the outside edge, resulting in a single file line. Once
in the plateau section, the parts speed up and spread out. They continue to move around
50
the plateau, where an overhead vision system can be used to identify parts in the correct
orientation. Incorrectly oriented parts pass over the drop-off and return to the pile of bulk
parts. This feeder exploits the unique capabilities of a closed-loop planer linear motor to
reorient, segregate, and position parts using only horizontal vibrations. Large planar
motions are also possible, allowing a single parts feeder to serve multiple overhead
robots. The design of the feeder utilizes few critical physical dimensions; allowing
different parts to be fed with only software changes. Multiple waveforms are examined
for their suitability for this application, two are chosen as the most promising: the stick-
slip waveform and the Coulomb pump waveform. Dynamic models of the feeder system
using these waveforms are generated and verified experimentally. No provision for
selecting the appropriate waveform for a given part was provided. In addition, the feeding
As an extension of [13], Redford and Boothroyd [7] examine a VBF driven by out-of-
phase parallel and normal components of vibration. A theoretical analysis was performed
which allowed predictions to be made of the motion of a part on a track which vibrates
with simple harmonic motion with or without a phase difference between the normal and
concluded that traditional VBF drives, where the normal and parallel motions are in
phase, express serious limitations. It was found that part movement on the track will
generally be achieved by the pushing action of one part on another while traveling around
the bottom of the bowl. The mean velocity of parts was found to be seriously affected by
changes in the coefficient of friction between the component and the track. By
51
introducing an appropriate phase difference between the normal and parallel motions of
the track, high conveying velocities, independent of the coefficient of friction between
the component and the track, could be achieved. A method was developed to use these
phase difference to design a VBF such that the conveying velocity of parts on the bowl
track is greater than that of parts traveling around the bottom of the bowl, eliminating the
pushing phenomenon. Since the parallel and normal components of motion can be altered
independently, it was shown that feed rates could be altered by adjusting the parallel
component only, while the normal component can be set at a stable fixed level. Also, the
system automatically senses the natural frequency of the VBF and generates signals equal
to this value, maximizing feeding efficiency. This system is totally independent of the
main power source frequency. Compensation for changes in mass and relaxation of the
springs is also provided. This allows for increased feed rates without excessive bouncing.
Unlike the traditional VBF, the drive unit of this design is not dedicated to any particular
bowl. The driver permits the interchange of bowls on standard drive units; hence,
eliminating the requirement of bowl tuning. The capability to change the phase of the
driver permits the conversion of the drive unit from left-hand feed to right-hand feed, or
vice versa. Feed direction may also be reversed at any time to clear jams or reject
improperly oriented parts. Operating at the natural frequency of the system also allows
52
Han et al. designed a nearly identical system utilizing either traditional electromagnets or
two groups of piezoactuators to separately generate the parallel and normal components
of motion [159]. A systems model of this type of VBF was developed. Since the vertical
linear vibration and the horizontal vibration are decoupled; a model could be generated
where each of the components of motion are treated independently. Since mechanically
this system is a modular structure with one system superimposed on the other, it was
presumed that a systems model could be produced accordingly. Each system was
actuation. Empirical data was collected and compared to the proposed systems model,
Mahdavian et al. studied the effect of various shapes of driving input signals on the feed
rate of a VBF [160]. A new design for the driving unit of a VBF was presented. A
computer was interfaced to the VBF driving unit to digitally control amplitude,
frequency, and driving signal. A software interface was written to allow for the rapid
frequencies and amplitudes, were studied. It was found that the square wave input signal
drives a VBF at a considerably greater feed rate than either the triangular or sinusoidal
waveforms. Studies were also made of the sensitivity to the loss of mass in the bowl as
the parts feed out under various driving waveforms. The square wave was shown to be
53
least sensitive to the changing mass of the system; keeping the feed rate nearly constant
VBF technology is mature; industry has been using VBFs to great effect for over 30
years. Research has not been lacking; altered driving input waveforms (DIWs) have been
examined, empirical and theoretical studies on the nature of part resting aspects have
been conducted, new and novel designs have and are being considered. However, a
deficiency exists in our understanding of the nature of VBFs. It had been shown that
optimum performance, defined as motion with the least amount of instability and a
maximized effective travel distance, occurs when the part just touches the track once in
every period, and then only when the track just reaches its maximum velocity in the
upward direction [31]. The VBF knowledge base lacks a fundamental understanding of
the interactions between the physical characteristics of a part and the various design
That is, the generalized optimum performance for a given generic part is known, trial and
error adjusting of VBF parameters can be used to laboriously obtain this optimum value
for any specific part, but a link relating the geometry of a part to VBF parameters for
optimal performance is not available. Many authors neglect part geometry as significant
in determining the optimum feed rates of parts, preferring to model parts as simple point
masses [44, 42]. Recent research has indicated that part geometry has a large influence on
optimum feed rate [161]. This research suggested that a non-symmetrical part, when fed
54
in different orientations have different feed rates. The primary focus of this research is to
The objective of this project is to design and build an apparatus simulating a VBF to test
and analyze the relationship between part geometry, VBF parameters and performance.
The experimental portion of the project will be conducted in three phases. The first phase
will be a pilot study. Two parts that are of different geometry and that express different
feed rates but are of the same mass, surface finish, material, surface area and volume will
be selected. These parts will be run on the experimental apparatus at varying VBF
parameter settings to determine if these settings influence the observed feed rate
difference. The second phase of the project will be an in-depth examination of the VBF
parameters determined to be significant on the feed rate difference, additional parts will
be manufactured if necessary. The third phase will involve verification of the results in a
commercial VBF.
At the conclusion of the experimental portion of the project, the results will be discussed
and the data analyzed. In the final chapter, the research contributions will be summarized,
conclusions derived, and areas of further research indicated. In the appendices complete
55
CHAPTER 2
EXPERIMENTAL ORGANIZATION
For this project it was determined that track angle, vibration angle, frequency and
acceleration amplitude (normal to track) are the most pertinent VBF parameters to
commercial VBF allows for the alteration of driving frequency and acceleration
amplitude, it does not allow for the simple or rapid alteration of parameters such as track
angle and vibration angle. An apparatus (figure 2.1, figure 2.2) was designed and built
that simulates a traditional VBF and allows for the simple and expedient alteration of
Figure 2.1 illustrates the conceptual design of the experimental apparatus; figure 2.2
displays the as built model. To change the vibration angle; the ‘platform’ is moved to the
desired angle, in 5û increments, and secured with bolts through the ‘securing bolt holes’.
The track angle is set by loosening the ‘cross bar’ nut, setting the track at the desired
angle and then retightening the nut. To ensure that the proper angle has been set, an
ADXL202 [162] accelerometer is used to determine the angle of tilt relative to the earth.
The ADXL202 was calibrated on a surface plate which had been leveled to within
55
.000050” with a master precision level. The calibrated accelerometer had an accuracy of
+/- .10û. Frequency and amplitude are controlled by directly manipulating the driving
waveform to the electromagnet. Both values are verified through the use of an ADXL250
[163] accelerometer mounted as shown in figure 2.2. The electromagnet chosen to drive
the apparatus was the Duramagnetics ER2-201 [164]. Its compact size, high holding
Two tracks were constructed; one of acrylic and another of aluminum. Both tracks have
slots and holes drilled to mount infra-red (IR) LED gates used to measure the feed-rate of
the parts. Once a part breaks the first IR beam, a timer is started, upon breaking the
second beam it is stopped. The distance between the gates is known, the feed rate is
simply derived by diving the distance traveled by the time required to travel that distance.
IR Gate Slot
Track
C ross Bar
Ele ctromagne t
56
In addition, the plastic track has a 2û side-to-side incline to enhance the simulation of a
VBF. Each track was carefully polished to ensure consistent surface finish. Figure 2.3
and figure 2.4 are measurements of the damped natural frequency of the system with each
track mounted. Table 2.1 is a summary of the response properties of the apparatus. The
system mass is defined as the moving mass and is comprised of all moving parts. The
acrylic track was chosen for the experiment; the lower mass of the acrylic track allows
for a greater damped natural frequency, thereby reducing the size of the power supply
required to drive the system in the desired ranges. In addition, the acrylic track more
Track
IR Transmitter
ADXL250
Accelerometer
IR Receiver
Cross Bar
Electromagnet
Support Spring
Platform
57
Track Moving Mass Damped Natural Damping Natural Spring
ounces(grams) Frequency (hz) Ratio Frequency (hz) Constant
Lbs/in
Plastic 6.57(186.53) 43 hz .198 43.86 80.27
Aluminum 9.38(266.07) 36 hz .188 36.65 79.94
(a) (b)
Figure 2.3
(a) System Impulse Response with Plastic Track
(b) FFT of Impulse Waveform (43hz max)
(a) (b)
Figure 2.4
(a) System Impulse Response with Aluminum Track
(b) FFT of Impulse Waveform (36hz max)
58
2.2 The Experimental System
2.5, figure 2.6 is a photograph of the actual system. Frequency and amplitude were
controlled with a solid state relay (SSR) and a National Instruments 1620E [165] data
acquisition system (DAQ). The input of the SSR was connected to the timer/counter
output of the DAQ; the output connected a 48v DC power supply to the electromagnet of
the experimental apparatus. The DAQ generated a pulse train which in turn activated the
SSR. The frequency of the pulse train determines the frequency of vibration, the duty
cycle of the pulse train determines the amplitude of vibration. Amplitude and frequency
were verified with an ADXL250 accelerometer attached as indicated in figure 2.2; the
output of the accelerometer was feed into a second National Instruments 1620E DAQ.
Two DAQ systems were required as the 1620E is not capable of reliably providing output
while receiving input; however, it is capable of receiving inputs from multiple sources
information and as such was also responsible for receiving the IR gate signals for
determining feed rate and the pulse width modulated output of the ADXL202E for
determining track angle. National Instruments Labview 6.1i [165] was chosen as the
interface software for data collection and electromagnet control. Detailed design
information can be found in the appendices: Appendix A contains the circuit diagram of
the IR gate interface circuit; Appendix B contains a detailed equipment list; Appendix C
diagrams the National Instruments LabView programs used to drive the electromagnet
59
Figure 2.5: Block Diagram Overview of Experimental System
60
5v Sensor Power Supply SC 2345 #1
48v DC Power Supply
IR Gate Interface Circuit DAQ #2
DAQ #1
DAQ
Computer One
Experimental ADXL202 DAQ
SSR
SC 2345 #1 Apparatus Tilt Sensor Computer Two
Initial observations suggested that length was one of the key geometrical factors to
consider. With this in mind, two parts of different length were constructed for the pilot
study. To reduce the number of possible variable interactions the parts would have equal
bottom surface areas, surface finishes, mass and height. It was not possible to keep the
width the same and have different lengths while maintaining equivalent bottom surface
areas. Nor was it possible to have equal side surface areas while changing length and
keeping the heights the same. All parts were rough cut on a vertical milling machine and
61
finished on a Browne & Sharpe surface grinder. Careful attention was paid to the final
grinding process to ensure equivalent surface finishes on all parts. Hot rolled steel was
chosen as the construction material because it was relatively available and easy to
machine.
At the conclusion of the pilot study, it was determined that additional experimental parts
would be required. Four additional pairs of parts were manufactured out of hot rolled
steel. Two pairs altered length while maintaining the widths of the original parts the
others altered width while maintaining length. The new parts preserved the mass of the
original parts; however, since the volume had to remain identical neither the bottom or
side surface areas nor height would be the same as the originals. To improve the
manufacturing process a Sodic EDM machine was used to rough cut the parts; this
resulted in a significantly improved finished product. The same careful attention was paid
to the final finishing process to ensure identical surface finishes. Table 2.2 summarizes
the physical properties of experimental parts. The target mass for all parts was 18.44 gr.
(.65 oz), however, due to inaccuracies in the manufacturing process some variation is to
developed and enforced. This general policy was followed for every phase of data taking
involving the experimental apparatus. First, the desired vibration angle was set. Next, the
track angle was set. Since this variable is the most difficult to set and has the greatest
62
probability of user error; three independent readings of the ADXL202E tilt sensor were
taken to ensure that the track angle had been correctly adjusted. Frequency and amplitude
were then set with software controls; however, to determine if the desired result ensued,
the values were verified through the use of the ADXL250 accelerometer. Necessary
adjustments were made, usually to the duty cycle, and the verification process was
repeated. Once the frequency and amplitude satisfied the requirements, the IR gates were
checked for proper operation. Often, before the first trial of a session, the trimming
potentiometers of the IR driving circuit would need adjustment. After the experimental
system was shown to function properly, data collection began. Should the apparatus need
to be powered down, to either change settings or conclude the day’s activities, this
The verification data set required a different procedure. The only variable adjusted was
verify the desired amplitude. The IR gates could not be remounted; a different procedure
was required to determine feed rate. A Kodak Kodacrome [166] high speed digital
camera and a paper grid were used for this purpose. Motions of the part were recorded
over the length of this grid. To determine the feed rate, the length of the grid was dived
by the time required to traverse it. While fundamentally identical to the previously
employed method, the errors introduced by manually collecting data required that at least
63
Part # Length Width Height Mass
in. In. in. gram(ounce)
1. (original) 1.004 .748 .195 18.52(.653)
2. (original) 1.502 .496 .195 18.35(.647)
3 1.167 .750 .165 18.40(.649)
4 1.750 .500 .165 18.40(.649)
5 .833 .750 .232 18.48(.651)
6 1.250 .500 .232 18.48(.651)
7 1.500 .249 .385 18.35(.647)
8 1.001 .374 .385 18.38(.648)
9 1.500 .749 .128 18.37(.647)
10 .999 1.124 .128 18.35(.647)
64
CHAPTER 3
Preliminary investigations had suggested that length would be the critical geometric
factor. A pilot study using parts one and two was conducted in an attempt to formalize
this relationship and determine the influence, if any, of vibration angle, frequency, and
track angle on feed rate. It was decided that equation 1.2.1 would be used with part #1 to
80 Hz, vibration angle of 30û, and a track angle of 3û a reasonable feed rate was chosen;
this feed rate would be used for this frequency and track angle combination for all values
of vibration angle. Using equation 1.2.1, the feed rates were determined at all other
frequency settings for a track angle of 3û. In each case, for each vibration angle, the duty
cycle of the electromagnet driving waveform was adjusted until the desired feed rate was
obtained; the resulting normal acceleration was recorded. In this way, by fixing the
constant in equation 1.2.1, part #1 would have the same feed rate for a given frequency if
track angle was fixed at 3û but the vibration angle varied. For all other combinations of
parts, vibration frequency, and track angle, given a particular vibration angle; the
amplitude found to drive part #1 at a constant feed-rate would be used. Thus, a base line
65
Base-line feed rates and normal accelerations for frequency and vibration angle variations
are indicated in table 3.1. The factors and levels utilized for this experiment are
summarized in table 3.2. Averaged experimental results are indicated in tables 3.3-3.5,
Boothroyd [5,7], if vibration angle was increased, the normal acceleration required to
maintain 1.2.1 at a constant value must also be increased; this is displayed in figure 3.1
Table 3.1: Part #1: base-line feed rates for fixed 3û track angle
66
Frequency Part #1 Part #2
Hz. (in./s.) (in./s.)
2û 3û 3.5û 2û 3û 3.5û
40 2.04 1.521 1.250 2.437 1.816 1.508
60 1.381 1.012 0.826 1.656 1.216 0.987
80 1.075 0.790 0.647 1.305 0.956 0.789
Table 3.3: Averaged Feed Rates (in./s.) for 20û Vibration Angle
Table 3.4: Averaged Feed Rates (in/s) for 30û Vibration Angle
Table 3.5: Averaged Feed Rates (in/s) for 40û Vibration Angle
67
4
3.5
Figure 3.2 illustrates feed rate verses track angle for given settings of vibration angle and
frequency for both parts. In all cases, increasing track angle decreases feed rate. This
correlates well with Boothroyd [5,7]. Moreover, the feed rate difference between the two
parts does not appear to be effected by alterations in track angle. Figure 3.3 graphs the
percentage difference in feed rate between the two parts against track angle. In all cases,
68
3 3
2.5 2.5
1 1
0.5 0.5
1.5 2 2.5 3 3.5 4 1.5 2 2.5 3 3.5 4
40 Hz Driving Frequency 40 Hz Driving Frequency
Track Angle (deg.) Track Angle (deg.)
40û Vib ration Angle 30û Vib ration Angle
3 1.9
1.7
2.5
Feed Rate (in./s.)
1.9 1.9
1.7 1.7
Feed Rate (in./s.)
0.9 0.9
0.7 0.7
0.5 0.5
1.5 2 2.5 3 3.5 4 1.5 2 2.5 3 3.5 4
60 Hz Driving Frequency 60 Hz Driving Frequency
Track Angle (deg.) Track Angle (deg.)
30û Vib ration Angle 20û Vib ration Angle
1.4 1.4
1.3 1.3
1.2 1.2
Feed Rate (in./s.)
1.1 1.1
1 40û P.1. 1 30û P.1.
0.9 40û P.2. 0.9 30û P.2.
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
1.5 2 2.5 3 3.5 4 1.5 2 2.5 3 3.5 4
80 Hz Driving Frequency 80 Hz Driving Frequency
Track Angle (deg.) Track Angle (deg.)
40û Vib ration Angle 30û Vib ration Angle
1.4
1.3
1.2
Feed Rate (in./s.)
1.1
1 20û P.1.
0.9 20û P.2.
0.8
0.7
0.6
0.5
1.5 2 2.5 3 3.5 4
80 Hz Driving Frequency
Track Angle (deg.)
20û Vib ration Angle
69
22.5 22.5
20 20
Percentage Difference
Percentage Difference
17.5 17.5
15 15
20û vib. 20û vib.
12.5 12.5
30û vib. 30û vib.
10 10
40û vib. 40û vib.
7.5 7.5
5 5
2.5 2.5
0 0
1.5 2 2.5 3 3.5 4 1.5 2 2.5 3 3.5 4
Track Angle Track Angle
40 Hz Driving Frequency 60 Hz Driving Frequency
25
22.5
Percentage Difference
20
17.5
15 20û vib.
12.5 30û vib.
10 40û vib.
7.5
5
2.5
0
1.5 2 2.5 3 3.5 4
Track Angle
80 Hz Driving Frequency
Figure 3.4 is a graph of feed rate verses vibration frequency. It is important to note that at
all experimental settings; the general trends follow those of the base-line configuration.
Figure 3.5 graphs the percentage difference of feed rate between parts #1 and #2 verses
frequency for all experimental settings. No significant deviations outside of the range of
experimental error are observed. APPENDIX E contains graphs displaying the constant
in equation 1.2.1 verses frequency for all experimental settings. For each part at a given
track angle and frequency, the constant remains nearly unchanged. The observed errors
are well within those reported by Boothroyd [5,7]. This verifies the assumed premise
stated in section 3.1 regarding the use of equation 1.2.1 as the basis for comparison of
70
3 3
Feed Rate (in./s.) 2.5 P.1.2û 2.5 P.1.2û
0 0
35 40 45 50 55 60 65 70 75 80 85 35 40 45 50 55 60 65 70 75 80 85
Frequency (Hz) 20û Vibration Angle Frequency (Hz) 30û Vibration Angle
2.5 P.1.2û
Feed Rate (in./s.)
2 P.2.2û
P.1.3û
1.5
P.2.3û
1 P.1.3.5û
0.5 P.2.3.5û
0
35 40 45 50 55 60 65 70 75 80 85
Frequency (Hz) 40û Vibration Angle
25 25
Percentage Difference
Percentage Difference
20 20
15 2û Track 15 2û Track
3û Track 3û Track
10 3.5û Track 10 3.5û Track
5 5
0 0
35 40 45 50 55 60 65 70 75 80 85 35 40 45 50 55 60 65 70 75 80 85
Frequency Frequency
20û Track Angle 30û Track Angle
25
Percentage Difference
20
15 2û Track
3û Track
10 3.5û Track
0
35 40 45 50 55 60 65 70 75 80 85
Frequency
40û Track Angle
71
3.4 The Effect of Vibration Angle
Feed rate verses vibration angle for a given part at given settings is indicated in figure
3.6. The base-line configuration, P.1.3, was determined so as to allow part #1 to have the
same feed rate over all ranges of vibration angle for a given vibration frequency and 3û
track angle. For increased vibration angles, the normal acceleration had to be increased,
as indicated in figure 3.1.1, to allow for the desired feed rates. The remaining curves
represent experimental data run at the base-line settings for amplitude. This indicates that
for all other combinations of parts and settings, the general trend is to follow the base-
line. Figure 3.7 graphs the percentage difference in feed rate between the two parts
against vibration angle. In all cases, the difference remains approximately constant and
3 1.8
1.6
2.5 P.1.2û P.1.2û
1.4
Feed Rate (in./s.)
1.4
1.2
P.1.2û
Feed Rate (in./s.)
1 P.2.2û
0.8 P.1.3û
0.6 P.2.3û
0.4 P.1.3.5û
P.2.3.5û
0.2
0
15 20 25 30 35 40 45
Vibration Angle (deg.) 80 Hz. Vibration Frequency
72
25 22.5
22.5 20
Percentage Difference
Percentage Difference
20 17.5
17.5 15
15 2û Track. 2û Track.
12.5
12.5 3û Track 3û Track
10
10 3.5û Track 3.5û Track
7.5 7.5
5 5
2.5 2.5
0 0
10 20 30 40 50 10 20 30 40 50
Vibration Angle (deg.) Vibration Angle (deg.)
80 Hz Driving Frequency 60 Hz Driving Frequency
22.5
20
Percentage Difference
17.5
15
2û Track.
12.5
3û Track
10
3.5û Track
7.5
5
2.5
0
10 20 30 40 50
Vibration Angle (deg.)
60 Hz Driving Frequency
The data was placed into a design of experiments matrix and analyzed. The results are
indicated in APPENDIX F. The percentage difference (PD) of feed rate between the two
Initial examinations of the p-values indicate that vibration and track angle are most likely
small coefficient for frequency, an adjusted R2 value of 47.6% and the data presented in
section 3.2-3,4; the most likely conclusion is that the factors of frequency, track angle,
and vibration angle have an insignificant effect on the percentage difference between the
feed rates of the two parts within the bounds of this experiment.
73
CHAPTER 4
STUDY I
4.1 Study I
In the previous experiment, it was determined that the percentage difference in feed rate
between parts one and two was insensitive to changes in track angle, vibration angle, and
frequency. It was decided to manufacture additional parts that varied length and width,
and explore variations in feed rates at fixed frequency, track angle and vibration angle.
Track angle was fixed at 2.82û; this was chosen to represent the average of those track
angles tested in the pilot study. Vibration angle was set at 30û to represent the median
value of those used during the pilot study. Normal acceleration was set at 3.04 g in an
attempt to get as close as possible to 3 g; the normal acceleration value used during the
pilot study for 30û vibration angle. The driving frequency was fixed at 80 Hz. This value
was selected in an attempt to reduce the amount of time required to balance the
experimental apparatus and to increase the overall lifespan of the equipment. The logic is
as follows: equation set 4.1.1 represents the displacement, velocity and acceleration of a
74
amplitude; putting less stress on the system, thereby reducing the number of times the
system must be rebalanced and increasing the overall lifespan by reducing wear.
x = A sin ω
*
x = Aω cos ω (4.1.1)
**
x = − Aω 2 sin ω
All of the parts indicated in table 2.2 were used; for convenience this table is repeated in
4.1. Table 4.2 indicates the average feed rated; complete experimental data in available in
APPENDIX G. Figure 4.1 is a graph of length verses feed rate, initial observation
indicates that length is a strong factor in determining the feed rate difference. Figure 4.2,
a graph of feed rate verses width, tells us nothing. Figure 4.3, a graph of feed rate verses
height is similar in this regard; however, a pattern appears to be evident, but no definitive
74
Length Width Height Mass
Part # in. In. in. gram(ounce)
1. (original) 1.004 .748 .195 18.52(.653)
2. (original) 1.502 .496 .195 18.35(.647)
3 1.167 .750 .165 18.40(.649)
4 1.750 .500 .165 18.40(.649)
5 .833 .750 .232 18.48(.651)
6 1.250 .500 .232 18.48(.651)
7 1.500 .249 .385 18.35(.647)
8 1.001 .374 .385 18.38(.648)
9 1.500 .749 .128 18.37(.647)
10 .999 1.124 .128 18.35(.647)
75
1.1
1.05
1
1.1
1.05
1
Feed Rate (in./s.)
0.95
0.9
0.85
0.8
0.75
0.7
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Width (in.)
76
=
1.1
1.05
1
The data was processed by means of regression analysis. The results are indicated in
APPENDIX H. The percentage difference (PD) of feed rate between parts #2-9 and part
#1, the original base-line part, was chosen as the response variable. The results are in
general agreement with the initial observations regarding the graphs. Length has a p-
value of 0.000 and a large coefficient; suggesting a strong association with the observed
feed rate difference. Width has a p-value of .792 and a small coefficient value; suggesting
however, it has a p-value of .495 and a coefficient that is greater than width’s by a factor
of 6.28, which does not inherently exclude it; this will be discussed in further detail in
chapter 7. With an adjusted R2 value of 99.0% and the resultant p values, the most likely
length, while width appears to be insignificant within the bounds of this experiment.
Height is most likely insignificant; however, there are some indications to the contrary.
77
CHAPTER 5
STUDY II
5.1 Study II
In the previous experiments, it was determined that the percentage difference in feed rates
between parts was insensitive to track angle, vibration angle, and frequency. Further, it
was determined that length contributes significantly to this difference. Width was shown
to be most likely insignificant, while no definitive conclusion may be drawn about height;
it may or may not be significant; if it is significant, the contribution is small; this will be
normal acceleration on the observed feed rate difference between parts. Track angle was
fixed at 2.96û, frequency at 80 Hz and vibration angle at 30û for the reasons stated in
section 4.1. All of the parts indicated in table 2.2 were used; this table is repeated in 5.1
for clarity. Normal accelerations of 2.4g, 2.64g, 2.95g, 3.11g and 3.45g were used: 2.4g
represents the lowest normal acceleration that still moves parts, while 3.45g is the largest
Complete experimental data can be found in APPENDIX I. Table 5.2 is a summary of the
averaged data. Figure 5.1 is a graph of feed rate verses normal amplitude. The data
indicates that the increase in feed rate is approximately linear for each part, until a certain
78
value of normal acceleration is reached, after which the graph takes on a slightly non-
linear characteristic. In addition, the percentage difference in feed rate increases as the
amplitude is increased. Figure 5.2 graphs feed rate vs. length. It is immediately apparent
that the percentage difference in feed rate is dependant on normal acceleration; the lower
the normal acceleration, the smaller the difference between parts. At the higher end of the
normal acceleration/length spectrum, the graph expresses the slightly nonlinear nature
N. Accel. Part 2 3 4 5 6 7 8 9 10
(g) #1 in./s. in./s. in./s. in./s. in./s. in./s. in./s. in./s. in./s.
2.4 0.500 0.51 0.502 0.514 0.492 0.504 0.511 0.505 0.507 0.496
2.64 0.636 0.701 0.657 0.737 0.617 0.673 0.71 0.647 0.693 0.628
2.95 0.803 0.953 0.852 1.024 0.743 0.878 0.963 0.816 0.949 0.793
3.11 0.89 1.080 0.952 1.140 0.820 0.989 1.092 0.909 1.072 0.879
3.45 1.044 1.283 1.144 1.349 0.959 1.184 1.295 1.069 1.274 1.036
1.6
1.4
1.2
Feed Rate (in./s.)
2.4g
1 2.64g
0.8 2.95g
0.6 3.11g
0.4 3.45g
0.2
0
0.65 0.8 0.95 1.1 1.25 1.4 1.55 1.7 1.85
Length (in.)
80
5.2 Statistical Analysis
The data was processed by means of regression, and response surface regression analysis.
The results are indicated in APPENDIX J. The percentage difference (PD) of feed rate
between parts #2-9 and part #1, the original base-line part, was chosen as the response
variable. The results of the regression analysis agree with those of the previous
experiment. The response surface regression results indicate that the linear terms of
length, width, and height, are insignificant and that normal acceleration may or may not
be significant with a p-value of .284. Examining the squared terms suggests that only
normal acceleration is significant. The interaction terms indicate that length and normal
acceleration interact in a strong and meaningful manor with a p-value of 0.000. This
result agrees well with the data; length affects the observed percentage difference in feed
rates between parts. The degree to which this occurs however, is strictly determined by
amplitude; the greater the amplitude, the greater the observed difference.
81
CHAPTER 6
VERIFICATION STUDY
The purpose of the verification study was not repeat earlier experiments in full, but rather
same trends observed while using the experimental apparatus would be evident under
real-world conditions. An Automation Devices [167] VBF was chosen for the verification
study. Three parts and three normal accelerations were chosen. Previous experimentation
confirmed that the interaction of length and amplitude was the primary contributing
factor to the observed percentage differences in feed rate. Lengths of .833, 1.25 and 1.75
were chosen as they represented the widest possible distribution of lengths. A normal
acceleration of 1.68g was chosen as it was the lowest possible setting still capable of
feeding parts; 2.49g represented the highest possible setting of acceleration that reliably
fed parts; 1.99g represented a median value. The averaged results are indicated in table
6.1; a complete data set is available in APPENDIX K. Figure 6.1 and 6.2 display the feed
rate verses length and normal acceleration amplitude. The results, in general, agree with
the findings obtained while using the experimental apparatus. At the lowest possible
normal acceleration amplitude, the differences in feed rate are small, as this amplitude is
increased, the percentage difference increases, at high values, the difference saturates.
Differences in slope can be attributed to systemic differences between the two systems.
82
Normal Acceleration Part #5 Part #6 Part #4
(g) (in./s.) (in./s.) (in./s.)
1.68 2.910 3.05 3.23
1.99 3.259 3.7 4.23
2.49 3.980 4.7 5.18
5.500
5.000
Feed Rate (in./s.)
4.500
1.68g
4.000
1.99g
3.500
2.49g
3.000
2.500
2.000
0.750 1.000 1.250 1.500 1.750 2.000
Length (in.)
83
5.500
5.000
Feed Rate (in./s.)
4.500
.833 in
4.000
1.25 in
3.500
1.75 in
3.000
2.500
2.000
1.5 1.7 1.9 2.1 2.3 2.5 2.7
Normal Acceleration (g)
84
CHAPTER 7
ANALYSIS
7.1 Introduction
The purpose of this chapter is to analyze the VBF and part interactions in order to
describe the trends observed in the experimental data. 2-D classical impact mechanics
equations will be developed to describe the part and VBF movements during a single
launch, flight and impact cycle. These equations will then be extended and incorporated
into a simulation to determine this effect over multiple cycles. Figures 7.1 and 7.2 depict
a typical sequence of motions of the longest part, part #4, and the shortest part, part #5.
Under ideal conditions, and ignoring air friction, Winkler [168, 169] indicates that the
condition for free flight on the track of a VBF requires that the normal track acceleration
exceed the gravitational constant g. After achieving maximum velocity, the VBF leaf
springs cause the bowl to decelerate. The part begins to slide as it does not decelerate at
the same rate as the bowl because it is not attached to the bowl. Boothroyd [5] indicates
An µ s + tan θ
> (7.2.1)
g n cotψ + µ s
85
Figure 7.1: Sequence of Motions for 1.75” Part at 1.68 g Normal Acceleration
86
Figure 7.2: Sequence of Motions for .833” Part at 1.68 g Normal Acceleration
87
Assuming a reference plane with the normal axis perpendicular to the track and a tangent
axis parallel to it, the normal track velocity at the point of separation will be used as a
reasonable estimation of the normal velocity of the part at separation. Calculating the
value of 7.2 .1 indicates that the difference in tangential velocity of the track from the
point the part starts to slide and the point its separates from the track is between 1.5-2%.
Thus, a reasonable assumption for tangential velocity of the part at separation is the
tangential velocity of the track at separation. From the point of maximum normal
velocity, the normal force continuously decreases until it exceeds that of gravity and the
part separates from the track and follows a course prescribed by projectile motion.
Thomas [44] indicates that wall friction is an insignificant factor, thus it is ignored.
A typical graph of the normal displacement of the track and the part for the initial cycle is
Direct observation of the motions of the parts indicates that a torque is applied to the
parts at the moment of separation (figure 7.1.1,7.1.2). This torque induces a momentary
angular acceleration over a small time period, and hence imparts a component of angular
velocity about the axis of length during the flight trajectory. The exact magnitude of this
torque is unknown, but it is most likely very small. For these simulations, a torque of .1
this is that the resulting impact on the track is oblique regardless of the magnitude of the
torque. The point of contact thus experiences not only a normal and tangential impulse,
but also an impulsive torque. The source of this torque is revealed when the nature of the
motions of the leaf springs in the VBF and the configuration of the coil springs in the
88
experimental apparatus is examined. After the electromagnet in a VBF releases the bowl,
the individual leaf springs follow the path of an arc. In the experimental apparatus; under
compression, the coil springs act as a solid cylinder, uncompressed, they ‘lean’ forward,
providing a small torque as the parts are launched. Figure 7.2.1 is an estimation of the
path of all the parts, since the initial angular velocity is very small, this will not be so for
subsequent impacts.
0.01
Y Displacement (inches)
0.008
0.006
0.004
0.002 Track
0
-0.002 Part
-0.004
-0.006
-0.008
-0.01
0 0.0025 0.005 0.0075 0.01 0.0125 0.015
Time (seconds)
The methods of classical impact mechanics with friction presented by Wang and Mason
[170] and repeated in Stronge [171] are used for the analysis of the impact between the
89
The sudden, short-term encounter between two colliding bodies is a very complicated
event. The major characteristics are the very brief duration and large magnitudes of the
forces generated. The vast complexity of the process leads to serious and significant
body assumption and Coulomb’s friction law, the process is significantly simplified.
The primary simplifying assumption in rigid body impacts is the notion of set
deformation histories. All collisions follow the same basic timeline consisting of two
distinct periods: the period of compression which extends from the instant of collision to
the point of maximum compression and the period of restitution, lasting until the point of
impulsive interactive forces with finite forces, such as gravity, being ignored. Since the
process is assumed to be near instantaneous, the linear and angular velocities of the
bodies have discontinuous changes. The near instantaneous change also indicates that no
The motion of the part and VBF are assumed to be those of 2-D planer rigid bodies with
Newton’s law of restitution by virtue of its apparent superiority [170,171]. The VBF
bowl is aluminum and the parts steel, hence a coefficient of restitution of .9 [172] and a
coefficient of dynamic friction of .47 [173] is chosen. Since a moment about the length is
applied during part separation from the track, a point to plane oblique impact is assumed
90
for all impacts. The motion of the VBF is assumed not to change after impact as the mass
A frame of reference is attached at the contact point with the y axis normal to the
boundary and points into the part and the x axis pointing up the track. The center of mass
of the parts relative to this frame is considered (x , y). The initial linear velocities of the
⋅ ⋅ ⋅ ⋅
part and the track are labeled ( x o , y o ) and ( x ot , y ot ) respectively. Based on these
⋅
assumptions, the part will also have an initial angular velocity; θ o . hence, the initial
⋅ ⋅ ⋅
C o = ( y o + θ o x) − y ot (7.3.1)
⋅ ⋅ ⋅
S o = ( x o − θ o y ) − x ot (7.3.2)
At any instant during the impact, the motion of the part is governed by the linear and
⋅ ⋅
m( x − xo ) = Px
⋅ ⋅
m( y − y o ) = Py (7.3.3)
⋅ ⋅
I (θ − θ o ) = − Px y + Py x
91
Where I is the moment of inertia of the part about the rotating axis and m is the mass and
Px and Py are tangential and normal components of impulse. There is an analogous set of
equations for the effect of the impulse on the VBF, however, these will not be considered
As the collision progresses to its conclusion, the accumulated impulse will change the
relative sliding and compression velocities. By combining the dynamics and kinematics
equations presented in [170] and removing those terms related to the VBF, since mVBF>>m
and IVBF >> I , the final relative velocities can be expressed in terms of the accumulated
impulse:
S = S o + B1 Px − B3 Py
(7.3.4)
C = C o − B3 Px + B2 Py
where
1 y2
B1 = +
m I
1 x2
B2 = + (7.3.5)
m I
xy
B3 =
I
Under this system, collisions are classified by three criteria: whether the collisions
stopped sliding or not; if so, whether it entered a sticking mode or a reverse sliding mode;
and whether it changed modes during the compression restitution phase of the impact.
92
between the part and track, µ, and the following quantities derived from Routh’s impact
Pd = sS o ( B2 + sµB3 )
Pq = −C o ( µB1 + sB3 )
B3
µs = − (7.3.6)
B1
So
s=
So
Where s is the sign of the initial sliding velocity So, µs the friction necessary to prevent
sliding. Pq and Pd are constants derived from Routh’s method, refer to APPENDIX L for
a more detailed explanation. Once the collision mode has been determined, the resulting
impulses are given by Wang and Mason [170] and repeated below.
93
• Sliding
Px = − sµPy
Co (7.3.7)
Py = −(1 + e)
B2 + sµB3
• Restitution Sticking
B3 Py − S o
Px =
B1
(7.2.8)
Co
Py = −(1 + e)
B2 + sµB3
• Compression Sticking
B3 Py − So
Px =
B1
(7.3.9)
B C + B3 So
Py = −(1 + e) 1 o
B1B2 − B32
2S o
Px = sµ[ Py − ]
B3 + sµB1
(7.3.10)
Co
Py = −(1 + e)
B2 + sµB3
2S o
Px = sµ[ Py − ]
B3 + sµB1
(7.3.11)
1+ e 2sµB3 S o
Py = − [C o + ]
B2 − sµB3 B3 + sµB1
94
The following is an analysis of the impact between the VBF track and the shortest part
(part #5), for a normal acceleration of 2.49g. The impact is assumed oblique and to occur
on the leading edge of the part. The mass moment of inertia I for the axis passing through
1
I= m(a 2 + b 2 ) = .040oz * in 2 (7.3.12)
12
Where ‘a’ is the height and ‘b’ is the length. Table 7.2 summarizes the relevant velocities
and positions at the point of impact. Table 7.3 summarizes the part property constants
and Routh’s quantities. From these it can be determined that Pd>(1+e)Pq and that µ < |µs|;
which places the mode of impact as sliding. Table 7.4 indicates the normal and
contact, angular velocity, and the velocities of the center of mass. Table 7.5 summarizes
Examining the results indicate that as length is increased the angular velocity on rebound
is decreased. The normal and tangential velocities also differ; as the length increases, the
tangential velocity increases while normal velocity decreases. When paired with
observations of high speed digital footage of part motions, an insight into the relationship
between length and feed rate begins to materialize; the shorter the part, the higher the
angular velocity on rebound because of the smaller moment of inertia about the axis
through the length. This higher angular velocity creates instabilities in subsequent
interactions between the part and VBF. This is clearly illustrated in figure 7.4, where the
95
shortest part is clearly seen to undergo unstable inefficient motions, wasting kinetic
energy on actions not directly related to positive feed rate. The long part on the other
hand, figure 7.1, tends too rebound, just as the short part, but continues on its path with
few instable perturbations. While this chaotic behavior is not wholly described by simple
2-D planer impact analysis, simulation of multiple cycles could prove insightful.
⋅ 4.09 in/sec
xo
⋅ -2.93 in/sec
yo
⋅ 0.012 in/sec
θo
⋅ 2.538in/sec
x ot
⋅ 1.129 in/sec
y ot
x 0.421 in
y -0.094 in
So 1.335 in/sec
Co -4.141 in/sec
B1 1.755063
B2 5.925428
B3 -0.98036
Pd 7.221222
Pq -0.64455
µs 0.558591
s 1
96
Py 1.441
Px -0.677
S 1.545
C 3.730
⋅ ⋅
4.866 in/sec
y f −θ f x
⋅ ⋅
4.098 in/sec
x f +θ f y
⋅
12.857 rad/sec
θf
⋅ -0.558 in/sec
yf
⋅ 2.886 in/sec
xf
Mode of Impact ⋅
⋅ ⋅
y f (rad/sec) x f (rad/sec)
θ f (rad/sec)
Part #5 Unidirectional Sliding 12.857 -0.558 2.886
(L=.833”)
Part #1 Unidirectional Sliding 11.022 -0.693 2.913
(L=1”)
Part #3 Unidirectional Sliding 9.618 -0.779 2.915
(L=1.167”)
Part #6 Unidirectional Sliding 8.857 -0.715 2.918
(L=1.25”)
Part #2 Unidirectional Sliding 7.449 -0.789 2.932
(L=1.5”)
Part #4 Unidirectional Sliding 6.532 -0.863 2.945
(L=1.75”)
Table 7.5: Summary of First Impact Results for Parts of Varying Length at An=2.49
97
A simulation of equations 7.3.1 to 7.3.12 was written in Microsoft Visual Basic. All
parameter assumptions from the single point simulation were carried over into the multi
point simulation. Figure 7.5 is a typical graph of the part and track normal displacement
over the time period examined. The ‘lead’ line represents the position of the leading point
while the ‘trail’ line represents the position of the trailing point. A collision is resolved
anytime either of these points intersects with the track. Tables 7.6 and 7.7, summarize
the simulation results for amplitudes of 1.99, 2.49 g normal acceleration. The lower
setting of 1.68g was not simulated because the simulation accounts for impacts only, and
The results are reasonable; they indicate a relative increase in feed rate from shorter to
longer parts. While the feed rates do not reflect those of the actual VBF, or of the
percentage differences observed they do fit the observed data in the basic trends. Longer
parts have higher resistance to angular accelerations, thus for parts of the same mass, the
longer the part the more stable it will be in its interactions with the VBF. Higher stability
allows for more efficient feeding, less energy is wasted on nonproductive motions.
Height is a factor; it has a squared term in the moment of inertia equation. However, for
the geometric envelope of the parts examined, the height is small, and varies by small
amounts between the various length parts which have different heights. This verifies the
earlier suspicion that height was a significant factor, even though the statistical analysis
was unclear in this regard. This position is supported by figures 7.6 and 7.7; graphs of
feed rate verses moment of inertia about length for the data presented in Study II and the
verification study. The larger the moment of inertia about length, the greater the feed rate;
98
the interaction between the moment of inertia and normal acceleration amplitude is also
apparent.
The high speed digital video indicates that increasing normal acceleration amplitude
decreasing normal acceleration amplitude leads to less significant differences. 2-D planer
impact analysis verifies the basic underlying causation but does not adequately describe
this phenomenon. Further video analysis indicates that the majority of these inefficiencies
99
Figure 7.4: Instabilities of the Shorter Part: 1.68 g
100
track Lead Trail
4.00E-02
1.6
1.4
Feed Rate (in./sec)
1.2 2.4g
2.64g
1
2.95g
0.8
3.11g
0.6 3.45g
0.4
0.2
0.025 0.05 0.075 0.1 0.125 0.15 0.175
Moment of Inertia (oz/in^2)
101
5.5
5
Feed Rate (in./sec)
4.5
1.68g
4
1.99g
3.5
2.49g
3
2.5
2
0 0.025 0.05 0.075 0.1 0.125 0.15 0.175 0.2
Moment of Inertia (oz./in^2)
102
CHAPTER 8
8.1 Summary
In this study, parts with different geometric properties, but identical physical properties
were empirically examined in conjunction with VBF parameters to determine the effect,
if any, of geometry on feed rate and the interactions that these effects may have with
VBF parameters. Ten different parts were constructed and four VBF parameters were
An experimental apparatus was designed and built by which VBF parameters may be
empirical studies were conducted using this apparatus at the full range of its capabilities.
A commercial VBF was used to verify the results of these studies. High speed digital
recordings of the part motions in the VBF were taken to aid in analysis.
Statistical methods and classical impact analysis were used to provide an insight into the
quantitative analysis of the data; a conclusion on the nature of one aspect of the
interaction of part geometry and VBF parameters for the studied class of parts was made.
103
8.2 Research Contributions and Conclusions
The major research contribution of this dissertation is the description of the interaction
between part geometry, VBF parameters and feed rate for the examined class of parts. Up
until this point, the phenomenon by which parts of uniform physical properties such as
mass, surface finish and material could feed at different rates when placed or configured
in different geometric orientations in a VBF was not completely understood. It has been
shown that a primary contributing factor involved in this phenomenon is the moment of
inertia about the axis perpendicular to length; the larger this moment, the higher the
stability, the greater the feed rate. This conclusion is supported by Bauer ‘s [174] data.
The results of this work generate many interesting questions. Much of the current
research in modern VBF technology is centered on tooling and part orientation. What
then would the effect of part geometry be on this? If part stability is an important factor
in determining feed rate, then most certainly it is an important factor in the interaction of
a part with passive and active tooling elements. This research could also be extended to
consider parts outside of the class of parts studied; perhaps a long-term and systematic
empirical effort to determine the boundaries of this phenomenon and to explore the
relevance of these issues in a more practical industrial setting. Considering the continual
simulation of a VBF accounting for stability and the effect of micro-motions would be of
104
APPENDIX A
105
106
APPENDIX B
107
Item Part Number Quantity
Data Acquisition Box National Instruments 6020E 2
Signal Conditioning Box National Instrument SC 2345 2
+/- 10v Analog Input Module SCC-AI03 2
Digital Input Module SCC-DI01 2
Solid State Relay CRYDOM DC60S7 1
Research Computer #1 Pentium III 1 Gz 1
Research Computer #2 Sony VAIO Laptop 1
Power Supplies 24 v American Standard 2
108
APPENDIX C
LABVIEW VIs
109
Figure C.1: Angle Measurement VI
110
Figure C.2: Electromagnet Control VI
111
Figure C.3: Infrared Gate Feed Rate Measurement VI
112
Figure C.4: ADXL250 Calibration VI
113
Figure C.5: Acceleration and Frequency Verification VI
114
APPENDIX D
115
Part #1 Part #2 Part #1 Part #2 Part #1 Part #2
2û 2û 3û 3û 3.5û 3.5û
80 Hz.
Trial #1 1.099 1.308 0.800 0.976 0.632 0.754
#2 1.080 1.308 0.800 0.978 0.696 0.799
#3 1.035 1.269 0.808 0.955 0.639 0.809
#4 1.094 1.323 0.783 0.941 0.669 0.828
#5 1.068 1.318 0.762 0.932 0.600 0.755
Average 1.0752 1.3052 0.7906 0.9564 0.6472 0.789
60 Hz.
Trial #1 1.338 1.644 0.975 1.233 0.874 1.034
#2 1.43 1.677 1.054 1.235 0.800 0.939
#3 1.335 1.682 0.977 1.201 0.858 1.013
#4 1.395 1.645 1.016 1.21 0.782 0.956
#5 1.410 1.634 1.041 1.205 0.816 0.993
Average 1.3816 1.6564 1.0126 1.2168 0.826 0.987
40 Hz.
Trial #1 1.978 2.337 1.479 1.783 1.272 1.542
#2 2.142 2.587 1.448 1.691 1.298 1.557
#3 1.999 2.35 1.541 1.873 1.243 1.482
#4 2.024 2.37 1.595 1.928 1.196 1.453
#5 2.096 2.544 1.545 1.813 1.243 1.506
Average 2.047 2.437 1.521 1.817 1.250 1.508
116
Part #1 Part #2 Part #1 Part #2 Part #1 Part #2
2û 2û 3û 3û 3.5û 3.5û
80 Hz.
Trial #1 1.033 1.226 0.768 0.931 0.606 0.751
#2 1.015 1.225 0.766 0.946 0.656 0.808
#3 1.082 1.344 0.771 0.942 0.603 0.739
#4 1.031 1.259 0.766 0.942 0.655 0.797
#5 1.054 1.309 0.760 0.949 0.585 0.710
Average 1.043 1.273 0.7662 0.942 0.621 0.761
60 Hz.
Trial #1 1.344 1.694 0.990 1.259 0.862 0.943
#2 1.418 1.635 1.050 1.237 0.808 0.928
#3 1.373 1.622 0.993 1.233 0.824 0.978
#4 1.387 1.626 1.028 1.201 0.755 0.983
#5 1.395 1.651 1.026 1.229 0.778 0.969
Average 1.3834 1.6456 1.0174 1.2318 0.8054 0.9602
40 Hz.
Trial #1 1.977 2.321 1.505 1.754 1.242 1.452
#2 2.147 2.480 1.457 1.765 1.221 1.452
#3 1.988 2.337 1.522 1.771 1.192 1.392
#4 1.995 2.407 1.564 1.900 1.132 1.347
#5 2.094 2.546 1.521 1.814 1.211 1.474
Average 2.0402 2.4182 1.5138 1.8008 1.1996 1.424
117
Part #1 Part #2 Part #1 Part #2 Part #1 Part #2
2û 2û 3û 3û 3.5û 3.5û
80 Hz.
Trial #1 1.086 1.275 0.763 0.948 0.635 0.756
#2 1.070 1.290 0.803 0.990 0.675 0.843
#3 1.131 1.292 0.812 0.973 0.607 0.728
#4 1.061 1.313 0.770 0.928 0.684 0.847
#5 1.063 1.349 0.798 0.947 0.618 0.742
Average 1.0822 1.3038 0.7892 0.9572 0.6438 0.7832
60 Hz.
Trial #1 1.342 1.675 1.01 1.202 0.897 1.070
#2 1.437 1.648 1.052 1.252 0.821 0.970
#3 1.383 1.702 0.999 1.242 0.864 1.015
#4 1.389 1.707 1.029 1.244 0.810 0.967
#5 1.411 1.633 1.032 1.240 0.826 1.016
Average 1.3924 1.673 1.0244 1.236 0.8436 1.0076
40 Hz.
Trial #1 1.992 2.316 1.514 1.857 1.198 1.448
#2 2.187 2.685 1.464 1.790 1.265 1.499
#3 2.007 2.364 1.547 1.820 1.240 1.517
#4 2.026 2.370 1.571 1.820 1.228 1.426
#5 2.087 2.491 1.534 1.882 1.234 1.512
Average 2.0598 2.4452 1.526 1.8338 1.233 1.4804
118
APPENDIX E
119
120 120
FV (Hz*in./s.)
FV (Hz*in./s.)
80 P.2.2û 80 P.2.2û
P.1.3û P.1.3û
60 60
P.2.3û P.2.3û
40 P.1.3.5û 40 P.1.3.5û
20 P.2.3.5û 20 P.2.3.5û
0 0
35 40 45 50 55 60 65 70 75 80 85 35 40 45 50 55 60 65 70 75 80 85
Frequency (Hz) 20û Vibration Angle Frequency (Hz) 30û Vibration Angle
120
100 P.1.2û
FV (Hz*in./s.)
80 P.2.2û
P.1.3û
60
P.2.3û
40 P.1.3.5û
20 P.2.3.5û
0
35 40 45 50 55 60 65 70 75 80 85
Frequency (Hz) 40û Vibration Angle
120
APPENDIX F
121
Response Surface Regression
122
APPENDIX G
STUDY I DATA
123
Part #1 #2 #3 #4 5 #6 #7 #8 #9 #10
Trial #1 0.820 0.988 0.872 1.042 0.766 0.900 1.006 0.833 0.982 0.805
2 0.827 0.992 0.879 1.037 0.756 0.931 0.974 0.842 1.002 0.808
3 0.828 0.992 0.865 1.056 0.768 0.936 0.974 0.850 0.998 0.804
4 0.825 0.975 0.892 1.036 0.746 0.936 0.990 0.851 0.988 0.822
5 0.827 0.986 0.862 1.059 0.753 0.903 1.014 0.859 0.996 0.810
6 0.835 0.997 0.884 1.063 0.762 0.907 1.019 0.848 0.990 0.829
7 0.838 0.996 0.863 1.036 0.755 0.930 1.007 0.821 0.981 0.815
8 0.804 1.005 0.865 1.052 0.766 0.903 1.011 0.848 0.997 0.823
9 0.802 0.986 0.892 1.063 0.757 0.923 0.995 0.832 0.965 0.820
10 0.825 0.994 0.860 1.063 0.762 0.907 0.993 0.841 0.975 0.807
11 0.835 0.999 0.874 1.036 0.746 0.931 1.017 0.832 0.985 0.825
12 0.814 1.014 0.891 1.042 0.737 0.904 1.023 0.816 0.985 0.829
13 0.820 0.984 0.889 1.050 0.753 0.922 1.003 0.857 0.971 0.801
14 0.812 0.972 0.872 1.058 0.780 0.915 1.015 0.826 0.975 0.799
15 0.805 0.984 0.887 1.047 0.754 0.934 1.022 0.827 0.965 0.810
16 0.837 1.014 0.881 1.044 0.771 0.907 0.992 0.834 0.993 0.822
17 0.828 1.012 0.891 1.062 0.749 0.909 0.982 0.836 1.002 0.801
18 0.806 0.984 0.886 1.055 0.751 0.904 1.024 0.860 0.972 0.801
19 0.839 0.996 0.886 1.063 0.739 0.922 0.989 0.817 0.984 0.807
20 0.830 1.004 0.872 1.051 0.779 0.932 1.021 0.856 1.000 0.825
21 0.804 0.995 0.861 1.037 0.753 0.907 1.010 0.854 0.971 0.828
22 0.832 0.978 0.893 1.054 0.771 0.929 0.985 0.849 0.991 0.810
23 0.824 0.993 0.887 1.047 0.768 0.922 1.005 0.824 0.976 0.820
24 0.832 0.982 0.891 1.037 0.734 0.917 1.016 0.838 0.970 0.824
25 0.825 1.012 0.865 1.055 0.772 0.891 1.006 0.820 0.987 0.815
Ave. 0.8230 0.9934 0.8784 1.0498 0.7579 0.9169 1.0037 0.8388 0.9840 0.8144
124
APPENDIX H
125
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 1275.07 425.02 296.25 0.000
Residual Error 6 8.61 1.43
Total 9 1283.68
Source DF Seq SS
L 1 1265.44
W 1 8.88
H 1 0.76
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
10 1.75 27.616 29.540 0.788 -1.924 -2.13R
126
APPENDIX I
STUDY II DATA
127
Trial Part #1 #2 #3 #4 #5 #6 #7 #8 #9 #10
#1 1.036 1.277 1.133 1.35 0.942 1.216 1.312 1.101 1.246 1.017
2 1.024 1.262 1.179 1.386 0.978 1.202 1.285 1.069 1.273 1.033
3 1.060 1.286 1.121 1.361 0.955 1.199 1.295 1.051 1.285 1.012
4 1.042 1.280 1.147 1.371 0.941 1.155 1.318 1.089 1.271 1.019
5 1.044 1.286 1.120 1.324 0.984 1.192 1.293 1.069 1.297 1.054
6 1.077 1.300 1.145 1.351 0.970 1.168 1.288 1.090 1.275 1.036
7 1.063 1.306 1.135 1.341 0.949 1.175 1.272 1.059 1.259 1.046
8 1.022 1.310 1.163 1.342 0.974 1.167 1.272 1.054 1.302 1.043
9 1.036 1.253 1.145 1.341 0.940 1.184 1.305 1.054 1.256 1.067
10 1.044 1.274 1.153 1.329 0.963 1.191 1.316 1.059 1.277 1.038
Ave. 1.044 1.283 1.144 1.349 0.959 1.184 1.295 1.069 1.274 1.036
Ave. 0.89 1.080 0.952 1.140 0.820 0.989 1.0925 0.909 1.072 0.879
128
Trial Part #2 #3 #4 #5 #6 #7 #8 #9 #10
#1
#1 0.803 0.964 0.859 1.018 0.746 0.881 0.951 0.818 0.949 0.783
2 0.811 0.956 0.842 1.038 0.739 0.889 0.977 0.805 0.954 0.807
3 0.805 0.945 0.841 1.034 0.738 0.870 0.960 0.823 0.942 0.783
4 0.793 0.938 0.854 1.022 0.729 0.887 0.973 0.825 0.945 0.803
5 0.807 0.962 0.844 1.017 0.752 0.865 0.961 0.804 0.967 0.793
6 0.800 0.944 0.863 1.014 0.755 0.873 0.970 0.829 0.961 0.784
7 0.793 0.906 0.838 1.030 0.738 0.878 0.958 0.799 0.938 0.777
8 0.815 0.963 0.864 1.037 0.754 0.891 0.975 0.823 0.940 0.803
9 0.815 0.943 0.853 1.020 0.732 0.868 0.949 0.824 0.947 0.805
10 0.794 0.958 0.863 1.014 0.752 0.886 0.965 0.813 0.947 0.794
Ave. 0.803 0.953 0.852 1.024 0.743 0.878 0.963 0.816 0.949 0.793
Ave. 0.636 0.701 0.657 0.737 0.617 0.673 0.710 0.647 0.693 0.628
129
Trial Part #1 #2 #3 #4 #5 #6 #7 #8 #9 #10
#1 0.499 0.511 0.502 0.518 0.494 0.507 0.515 0.502 0.511 0.494
2 0.501 0.509 0.501 0.516 0.498 0.508 0.505 0.512 0.511 0.494
3 0.501 0.511 0.505 0.512 0.497 0.502 0.512 0.500 0.507 0.497
4 0.502 0.512 0.495 0.514 0.486 0.498 0.510 0.503 0.506 0.502
5 0.494 0.506 0.506 0.513 0.489 0.510 0.519 0.504 0.506 0.491
6 0.504 0.504 0.496 0.513 0.490 0.500 0.512 0.508 0.509 0.504
7 0.502 0.517 0.499 0.511 0.489 0.506 0.513 0.510 0.508 0.496
8 0.494 0.511 0.506 0.507 0.489 0.506 0.506 0.506 0.501 0.492
9 0.505 0.504 0.504 0.519 0.490 0.505 0.512 0.509 0.514 0.496
10 0.504 0.515 0.506 0.519 0.498 0.504 0.513 0.500 0.505 0.501
Ave. 0.500 0.510 0.502 0.514 0.492 0.504 0.511 0.505 0.507 0.496
130
APPENDIX J
131
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 16.0653 5.3551 30.33 0.001
Residual Error 6 1.0594 0.1766
Total 9 17.1246
Source DF Seq SS
L 1 14.5741
W 1 1.4497
H 1 0.0415
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
5 0.83 1.600 0.909 0.248 0.691 2.04R
132
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 345.36 115.12 772.88 0.000
Residual Error 6 0.89 0.15
Total 9 346.25
Source DF Seq SS
L 1 337.55
W 1 7.77
H 1 0.04
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
4 1.75 15.881 -15.243 0.254 -0.638 -2.19R
133
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 1158.39 386.13 1428.87 0.000
Residual Error 6 1.62 0.27
Total 9 1160.01
Source DF Seq SS
L 1 1152.72
W 1 5.08
H 1 0.59
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
5 0.83 7.472 6.543 0.306 0.929 2.21R
134
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 1335.44 445.15 305.92 0.000
Residual Error 6 8.73 1.46
Total 9 1344.17
Source DF Seq SS
L 1 1324.04
W 1 10.19
H 1 1.21
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
4 1.75 -28.090 -30.176 0.793 2.087 2.30R
135
Regression Analysis
Analysis of Variance
Source DF SS MS F P
Regression 3 1447.66 482.55 126.51 0.000
Residual Error 6 22.89 3.81
Total 9 1470.55
Source DF Seq SS
L 1 1438.23
W 1 9.30
H 1 0.13
Unusual Observations
Obs L PD Fit StDev Fit Residual St Resid
4 1.75 -29.215 -32.215 1.284 3.001 2.04R
136
Response Surface Regression
137
APPENDIX K
VERIFICATION DATA
138
1.68 g 1.99 g 2.49 g
Part #5 6 4 5 6 4 5 6 4
0.272 0.261 0.227 0.24 0.205 0.184 0.185 0.151 0.142
0.254 0.264 0.223 0.243 0.213 0.188 0.170 0.147 0.156
0.24 0.228 0.226 0.221 0.211 0.183 0.192 0.149 0.147
0.27 0.243 0.219 0.22 0.206 0.168 0.192 0.151 0.15
0.273 0.258 0.239 0.217 0.207 0.189 0.205 0.174 0.147
0.268 0.257 0.235 0.24 0.193 0.184 0.177 0.160 0.155
0.241 0.264 0.248 0.221 0.206 0.166 0.189 0.167 0.138
0.238 0.252 0.233 0.231 0.199 0.188 0.182 0.167 0.144
0.257 0.224 0.241 0.211 0.194 0.177 0.198 0.154 0.138
0.275 0.241 0.233 0.233 0.189 0.172 0.172 0.161 0.15
0.26 0.230 0.246 0.245 0.211 0.172 0.200 0.164 0.141
0.264 0.232 0.232 0.217 0.200 0.182 0.174 0.170 0.134
0.26 0.240 0.239 0.227 0.210 0.168 0.173 0.144 0.139
0.234 0.223 0.216 0.248 0.191 0.172 0.188 0.152 0.147
0.246 0.256 0.218 0.232 0.196 0.175 0.172 0.157 0.137
0.235 0.237 0.225 0.238 0.212 0.169 0.181 0.170 0.151
0.267 0.228 0.217 0.226 0.193 0.169 0.189 0.155 0.139
0.253 0.235 0.244 0.23 0.212 0.180 0.203 0.163 0.147
0.262 0.248 0.237 0.216 0.214 0.174 0.184 0.148 0.138
0.272 0.223 0.227 0.245 0.212 0.174 0.194 0.170 0.145
0.275 0.252 0.244 0.212 0.212 0.175 0.201 0.174 0.143
0.257 0.262 0.219 0.213 0.197 0.185 0.190 0.149 0.149
0.263 0.265 0.247 0.248 0.188 0.179 0.182 0.171 0.151
0.235 0.250 0.236 0.234 0.187 0.186 0.206 0.159 0.155
0.265 0.257 0.223 0.244 0.206 0.169 0.201 0.162 0.135
0.25744 0.245 0.232 0.230 0.203 0.177 0.188 0.160 0.145 Av.
Feed
2.913 3.058 3.236 3.259 3.702 4.234 3.989 4.700 5.182 Rate
139
APPENDIX L
140
Origin of Pq and Pd:
Routh’s method [175] involves constructing five lines in impulse space along which the
The first two lines are related to the relative normal and tangential velocity of the two
objects. By setting the equation set 7.3.4 to zero, the line of maximum compression and
C o − B3 Px + B2 Py = 0
(K.1)
S o + B1 Px − B3 Py = 0
The line of termination is obtained by setting the first equation in K.1 to -eCo:
(1 + e)C o − B3 Px + B2 Py = 0 (K.2)
This is the line at which the collision will end under Newton’s law of restitution. This
condition will not hold for Poison’s law; this will be discussed later. The fourth line is the
Px = − sµPy (K.3)
As impact progresses through the compression and restitution phases, impulse along the
Py axis will accumulate, changing the normal velocity until the collision is over. Impulse
along the Px axis accumulates due to friction between the part and the track in accordance
with Coulomb Friction. The collision follows a path in impulse space starting at the
141
origin, and initially following the line of limiting friction. The collision continues (in a
sliding mode) along this line until the collision is over or until it crosses the line of
sticking. At that point, the object either sticks (thus following the line of sticking) or
slides in the opposite direction. If the object slides in the opposite direction, the line of
If Newton’s law is used, then the collision terminates when the line of termination is
reached. Poisson’s law however, terminates the impact based on the normal impulse
Point Pq and Pd, represent the intersection points of the line of sticking, line of maximum
compression and the line of limiting friction. Wang and Maison [170] developed a set of
conditions, outlined in table 7.1, that allow for simple comparisons to be made in the
142
APPENDIX M
143
Z
+
b
Y
X
a
Figure M: Parallelepiped
m = ρv
dm = ρdxdydz
I x = 4∫ ( z 2 + y 2 )dm
c a b
2 2 2
I x = 4 ρ ∫ ∫ ∫ ( z 2 + y 2 )dxdydz
−c 0 0
2
z3
I x = 4 ρ [ + y 2 z ]dzdy
3
z3 y3
I x = 4 ρ[ y + z ]dx
3 3
z3 xy 3 z
I x = 4 ρ [ yx + ]
3 3
144
b a b
( ) x( ) 3 ( )
a
Ix = 4 ρ[ 2 x( ) + 2 2 ]
3 2 3
3 3
ba ab
Ix = 4 ρc[ + ]
8* 3* 2 8* 3* 2
b2 + a2
Ix = 4 ρ (abc)[ ]
48
b2 + a2
Ix = 4m[ ]
48
1
Ix = m [a 2 + b 2 ]
12
145
Bibliography
10. F.E. Smith, “Applying Vibratory Bowl-Type Feeders: Part 1 Part, Bowl and
Drive Considerations,” Automation 9, (November 1962): 97-101.
146
14. G. Boothroyd, “Vibratory Feeders Speed handling and Orienting Parts,”
Automation 14, (1967): 86-92.
16. A.N. Kochan, “Feeder automatically separates easily tangled parts,” Design
News Magazine 48, September 20 (1993): 129.
20. R.J. Ruhl, “Increasing parts Feeder Productivity,” Automation 35, (1987): 18-
20.
22. L. Han and W.L. Xu, “A Novel Vibratory Bowl Feeder for Automated
Electronic Production,” Proceedings Of the Japan-U.S.A. Symposium on
Flexible Automation in Otsu, Japan, July 12-15, 1998, 841-843.
23. K. Kaisha, Parts feeding apparatus of the piezoelectric drive type, Toshiba
Corp., US Patent 4795025.
25. Scribner et al., Noise Attenuation System for Vibratory Feeder Bowl, US
Patent 5,414,775.
27. P.E. Berry, “Research on oscillating conveyors and basic theory,” Journal of
Agricultural Engineering Research 3-4, (1958-59).
147
29. J.H. Booth and H. McCallion, “On Predicting the Mean Conveying Velocity
of a Vibratory Conveyor,” Proceedings of. Institution of Mechanical
Engineers 178, no. 20 (1964): 521-538.
31. H.G. De Cock, “Vibratory Feeders,” Phillips Technical Review 24, no. 3
(1962): 84-93.
32. L.M. Rosario, “Expert system to analyze products for ease of assembly,”
Proceedings of NSF Design and Manufacturing Systems Conference in Tempe
(Arizona State University), Arizona, USA, January 8-12, 1990.
33. C. OU-Yang and G.P. Maul, “A computer analysis of orientation devices for
vibratory bowl feeders,” International Journal of Production Research 31, no.
3 (1993): 555-578.
37. S. Daneshmand, System Model and Simulation of the Vibratory Bowl Feeder,
Master’s Thesis, The Ohio State University, 1988.
39. W.M. Mansour, “Analog and Digital Analysis and Synthesis of Oscillatory
Tracks,” ASME Transactions, Journal of Engineering for Industry 94, no 2.
(1972): 488-494.
148
41. N.L. Ng, L.A. Ang and S.C. Chng, “A Computer Model for Vibrating
Conveyors,” Proceedings of the Institute of Mechanical Engineers, Part B,
Management and Engineering Manufacture 200, no. B2 (1986): 123-130.
43. G. Maul and M. Thomas, “A Systems Model and Simulation of the Vibratory
Bowl Feeder,” Journal of Manufacturing Systems 16, no. 5 (1997): 309-314.
44. M.B. Thomas, An Improved Systems Model and Simulation of the Vibratory
Bowl-Feeder, Master’s Thesis, The Ohio State University, 1995.
45. G.H. Lim, “Vibratory feeder motion study using turbo C++ language,”
Advanced Engineering Software 18, no. 1 (1993): 53-59.
46. G. H. Lim, “On the Conveying Velocity of a Vibratory Feeder,” Computers &
Structures 62, no.1 (1997): 197-203.
50. G. Boothroyd and C. Ho, “Natural resting aspects of parts for automatic
handling,” Journal of Engineering for Industry 99, (1977): 314-317.
51. B.K.A. Ngoi et al., “Analyzing the natural resting aspects of a component for
automated assembly,” ImechE Proceeding of the Institution of Mechanical
Engineers 209, (1995): 125-131.
52. S.S.G. LEE et al., A Computational Method for The Prediction of The Natural
Resting Aspects of Small, (Dearborn: SME Technical Publications, 1994).
149
54. B.K.A. Ngoi, L.E.N. Lim and S.S.G. Lee, “Analyzing the Natural Resting
Aspects of a Component by its Centroid Solid Angle,” Journal of Electronics
Manufacturing 5, no. 3 (1995): 193-197.
55. B.K.A. Ngoi, S.W. Lye and J. Chen, “Analyzing the Natural Resting Aspect
of Prism on Hard Surface for Automated Assembly,” International Journal of
Advanced Manufacturing Technology 11, no. 6 (1996): 406-412.
56. B.K.A. Ngoi, S.S.G. Lee and L.E.N. Lim, “Analyzing the Natural Resting
Aspects of a Complex Part,” International Journal of Production Research
33, no. 11 (1995): 3163-3172.
57. B.K.A. Ngoi, L.E.N. Lim and S.S.G. Lee, “Analyzing the Probabilities of the
Natural Resting Aspects of a Component with Displaced Centre of Gravity,”
International Journal of Production Research 33, no. 9 (1995): 2387-2394.
58. B.K.A. Ngoi, L.E.N. Lim and J.T. Ee, “Analysis of Natural Resting Aspects
of Parts in a Vibratory Bowl Feeder - Validation of "Drop Test”,"
International Journal of Advanced Manufacturing Technology 13, no. 4
(1997): 300-310.
62. G. Boothroyd, Assembly Automation and Product Design, (New York: Marcel
Dekker, 1992).
63. G. Boothroyd, C. Poli and L.E. Murch, Handbook of Feeding and Orienting
Techniques for Small Parts, (Amherst, University of Massachusetts, 1977).
65. L.E. Murch and C. Poli, “Analysis of Feeding and Orienting Systems for
Automatic Assembly,” Transactions of the ASME, Journal of Engineering for
Industry 99, (1977): 308-313.
150
66. L.E. Murch and G. Boothroyd, “Predicting the efficiency of parts orienting
systems,” Automation 18, (1971): 55-57.
68. S.P Sethi and C. Sriskandarajah, “Optimal Selection and Ordering of Part-
Orienting Devices: Series Configurations,” Proceedings of the 2nd
International Conference on Production Systems in Le Chesnay Cedex,
France, April, 1-14, 1987.
69. S.P. Sethi, et al., “Heuristic methods for selection and ordering of part-
orienting device,” Operations Research 38, no. 1 (1990).
71. B.K. Natarajan, “An Algorithmic Approach to the Automated Design of Parts
Orienters,” Proceeding of the 27th Annual Symposium on Foundations of
Computer Science in Toronto, Ontario, Canada, October 27-29, 1986: 132-
142.
72. B.K. Natarajan, “Some Paradigms for the Automated Design of Parts Feeder,”
Proceedings of the International Journal of Robotics Research 8, no. 6
(1989).
73. A.D. Christiansen, A.D. Edwards and C.A. Coello, “Automated Design of
Part Feeders using a Genetic Algorithm,” Proceedings of the 1996 IEEE
International Conference on Robotics and Automation in Minneapolis,
Minnesota, USA, April, 1996: 846-851.
74. R.-P. Berretty et al., “Geometric Algorithms for trap design,” Proceedings of
the Symposium of Computational Geometry in Miami Beach, Florida, USA,
June 13-16, 1999: 95-104.
75. R.P. Berretty, K. Goldberg and L. Cheung, “Trap Design for Vibratory Bowl
Feeders,” Proceedings of the IEEE International Conference. on Robotics and
Automation in Detroit, Michigan, USA, May 1999: 2558-2563.
76. P.K. Agarwal, A.D. Collins and J.L. Harer, “Minimal trap design,”
Proceedings of the IEEE International Conference on Robotics and
Automation (ICRA) in Seoul, Korea, May, 2001.
151
77. F. van der Stappen et al., Modeling of Sensor-Based Intelligent Robot Systems,
(Berlin: Springer Verlag, 2001).
78. D.R. Berkowitz and J. Canny, “Designing Parts Feeders Using Dynamic
Simulation,” Proceedings of the IEEE International Conference on Robotics
and Automation in Minneapolis, Minnesota, USA, April, 1996: 1127-1132.
79. D.R. Berkowitzl and J. Canny, “A Comparison of Real and Simulated Designs
for Vibratory Parts Feeding,” Proceedings of the IEEE International
Conference on Robotics and Automation in Albuquerque, New Mexico, USA,
April, 1997.
81. Y.T. Chen and R.E. Young, “PACIES - A Part Code Identification Expert
System,” IIE Transactions 20, 2 (1988): 132-136.
82. H.Y. Lai, “A coding design expert system for automatic assembly,”
Proceedings of the ASME Computers in Engineering Conference in San
Francisco, California, USA, July 31-August 4, 1988: 443-450.
83. R.K Tapadia and M.R. Henderson, “Using a Feature-Based Model for
Automatic Determination of Assembly Handling Codes,” Computers and
Graphic 14, no. 2 (1990): 251-262.
84. R.K. Li and C.L. Huang, “Assembly Code Generation from a Feature-Based
Geometric Model,” International Journal of Production Research 30, no. 3
(1992): 627-646.
85. E.K. Lo and M.D. Dick, “The Solid Modeling of Part Orienting Tracks for
Automated Assembly,” International Journal of Production Research 28, no.
8 (1990): 1513-1525.
86. M.I. Yeong and W.R. De Vries, “A Methodology for Part Feeder Design,”
Annals of the CIRP 43, no. 1 (1994).
87. R. LaBrooy, C. Jiang and P.O. Chan, “Design of an Expert System for
Automatic High-Speed Parts Feeding,” Proceedings of the 6th International
Conference on Manufacturing Engineering in Melbourne, Australia,
November, 1995, 359-367.
152
88. L.E.N. Lim et al., “A Computer-Aided Framework for the Selection and
Sequencing of Orienting Devices for the Vibratory Bowl Feeder,”
International Journal of Production Research 32, no. 11 (1994): 2513-2524.
89. P.S. Tan et al., “A Knowledge-based advisor for the Automatic Selection and
Sequencing of orienting Devices for Vibratory Feeding,” Engineering
Applications of Artificial Intelligence 8, no. 1 (1995): 1-13.
93. K.F. Bohinger, V. Bhatt and K.Y. Goldberg, “Sensor-less Manipulation Using
transverse Vibrations of a Plate,” Proceedings of the IEEE International
Conference on Robotics and Automation in Nagoya, Japan, May 21-27, 1995.
95. R.P. Berretty et al., “On Fence Design and the Complexity of Push Plans for
Orienting Parts,” Proceedings of the 13th ACM Symposium on Computational
Geometry in Nice, France, June, 1997, 21-29.
96. M. Brokowski, M.A. Peskin and K. Goldberg, “Optimal Curved Fences for
part alignment on Belt,” Proceedings of the IEEE International Conference on
Robotics and Automation in Atlanta, Georgia, USA, May, 1993, 467-473.
97. R.P. Berretty et al., “Computing fence designs for orienting parts,”
Computational Geometry, Theory and Applications 10, no. 4 (1998): 249-262.
98. M. Mani and R.D. Wilso, “A Programmable Orienting System for Flat Parts,”
Proceedings of the 13th North American Manufacturing Research Conference
in Berkley, California, USA, May, 1985.
153
100. M.T. Mason, “Mechanics and Planning of Manipulator Pushing
Operations,” International Journal of Robots Research 5, no. 3 (1986): 53-71.
101. L.E.N. Lim et al., “Flexible vibratory bowl feeding using modular
orienting devices,” Journal of the Institution of Engineers 33, no. 4 (1993):
25-29.
102. M.S. Branicky, G.C. Causy and R.D. Quinn, “Toward a Science of
Flexible Feeding,” Proceedings of the IEEE International Conference on
Advanced Intelligent Mechatronics in Atlanta, USA, September, 1999.
106. A.S. Rao and K.Y. Goldberg, “Shape from Diameter: Recognizing
Polygonal Parts with a Parallel jaw Gripper,” International Journal of
Robotics Research 13, no.1 (1994): 16-37.
110. A.H. Redford, E.K. Lo and P.J. Killeen, “Parts presentation to multi-arm
assembly robots,” Annals of the CIRP 32, 1983.
111. J.L. Goodrich and G.P. Maul, “Programmable Parts Feeders,” Industrial
Engineering Magazine 15, no. 5 (1983): 28-33.
154
112. M. Kosanchick and G.P. Maul, “Programmable Feeding Systems: The
Right Step for Flexible Systems,” Synergy Conference Proceedings in
Chicago, Illinois, USA, November13-15, 1984. 188-189.
116. K.R. Snyder and C.S. Elliot, “Barriers to Factory Automation: What are
They, And How Can They Be Surmounted,” Industrial Engineering 20,
(1988): 44-46.
117. R.K. Baird, “Vision Keeps Bowl Feeder Parts In Line,” Sensor Review
Magazine 11, no. 3 (1991): 29-31.
119. G.P. Maul and J.L. Goodrich, “Programmable Parts Feeders,” Proceedings
of 15th CIRP International Seminar on Manufacturing Systems in Amherst,
Massachusetts, USA, June 20-22, 1983.
122. G.P. Maul and C. Ou-Yang, “Predicting the Cycle Time for a Sequence of
Parts in a Sensor-Based Vibratory Bowl Feeder,” International Journal of
Production Research 12, (1987): 1705-1714.
155
125. N.I. Jaksic, Programmable Air-Jet Tooling For Vibratory Bowl Feeder
Systems, Doctoral Dissertation, The Ohio State University, 2000.
127. G. Maul and N. Jaksic, “Active Air Jet Tooling for Vibratory Bowl
Feeders,” Active Air Jet Tooling for Vibratory Bowl Feeders,” Proceedings of
the Second International Conference on Engineering Design and Automation
in Maui, Hawaii, USA, August, 1998.
129. N. Jaksic and G. Maul, “Orienting Parts in a Plane Normal to the Track
and the Bowl Wall of a Vibratory Bowl Feeder with Active Air-jet Tooling,”
Proceedings of the Fifth International Conference on Engineering Design
and Automation in Las Vegas, Nevada, USA, August 5-8, 2001, 97-102.
130. P.K. Agarwal, A.D. Collins and J.L. Harer, “Minimal trap design,”
Proceedings of the 2001 IEEE International Conference on Robotics and
Automation (ICRA) in Seoul, Korea, May 21-26, 2001.
131. P.S. Tan et al., “Using the Modular Approach for the Design of Orienting
Devices in Vibratory Bowls,” Proceedings of International Conference on
Computer Integrated Manufacturing at GINTIC Institute of CIM, NTU,
Singapore, September 6-10, 1993, 431-439.
133. A.J. Cronshaw, “A Practical Vision System for use with Bowl Feeders,”
Proceedings of the 1st International Conference on Assembly Automation in
Brighton, UK, 1980.
134. T. Suzuki and M. Kohno, “The Flexible Parts feeder which helps a robot
Assemble Automatically,” Assembly Automation 1, no. 2 (1981).
156
135. K.G. Swift and R.J. Dewhurst, “Versatile Automatic Handling Systems
Based on a Laser Electro-Optic Component Recognition Technique,”
Proceedings of the 19th International Seminar on Manufacturing Systems,
CIRP in University Park, Pennsylvania, June 1-2, 1987.
138. N.D. Edmondson and A.H. Redford, “Flexible Parts Feeding for Flexible
Assembly,” International Journal of Production Research 39, no. 11 (2001),
2279-2294.
140. J.W. Hill and A.J. Sword, “Programmable Part Presenter Based on
Computer Vision and Controlled Tumbling,” Proceedings of the 10th
International Symposium on Industrial Robots in Milan, Italy, May, 1980,
129-140.
141. J.W. Hill, “Programmable bowl feeder design based on computer vision, “
Assembly Automation1, 1 (1990).
145. R.D. Quinn et al., “Design of an Agile Manufacturing Work Cell for Light
Mechanical Applications,” Proceedings of the IEEE International Conference
on Robotics and Automation in Minneapolis, Minnesota, USA, April, 1986.
157
146. H. Hitakawa, “Advanced parts orientation system has wide application,”
Assembly Automation 8, no. 3 (1988): 147-150.
157. W.A Morcos. and M.F. Massoud, “On the Design of Oscillating
Conveyors,” Journal of Engineering for Industry, Transactions of the ASME
91, no. 2 (1969): 353-356.
158
158. A.E. Quaid, “A Miniature Mobile Parts Feeder: Operating Principles and
Simulation Results,” Proceedings of the IEEE International Conference on
Robotics & Automation in Detroit, Michigan, USA, May, 1999, 2221-2226.
159. l. Han, W.L. Xu, and S.K. Tso, “Decoupled Vibratory Bowl Feeder,”
Proceedings of the 5th International Conference on Mechatronics and
Machine Vision in Practice in Nanjing, China, 10-12th September 1998, 481-
486.
167. Automation Devices Inc, 7050 West Ridge Road, Fairview, PA 16415-
2099, http://www.autodev.com/
159
171. W.J. Stronge, Impact Mechanics (New York: Cambridge University Press,
2000)
174. J.R. Bauler, An Analysis of the Effect or Orientation on the Feed Rate in a
Vibratory Bowl Feeder, Master’s Thesis, The Ohio State University, 1998.
160