University of Texas at El Paso

DigitalCommons@UTEP
Open Access Theses & Dissertations

2018-01-01

Shape Detection Of Structural Electronics Using

Machine Vision System Embeded Into Scara Robot
Varaha Venkat Vyas Kumar Naidu
University of Texas at El Paso, vyas.binu@gmail.com

Follow this and additional works at: https://digitalcommons.utep.edu/open_etd

Part of the Industrial Engineering Commons

Recommended Citation
Naidu, Varaha Venkat Vyas Kumar, "Shape Detection Of Structural Electronics Using Machine Vision System Embeded Into Scara
Robot" (2018). Open Access Theses & Dissertations. 1496.
https://digitalcommons.utep.edu/open_etd/1496

This is brought to you for free and open access by DigitalCommons@UTEP. It has been accepted for inclusion in Open Access Theses & Dissertations
by an authorized administrator of DigitalCommons@UTEP. For more information, please contact lweber@utep.edu.
 SHAPE DETECTION OF STRUCTURAL ELECTRONICS USING MACHINE

VISION SYSTEM EMBEDED INTO SCARA ROBOT

VARAHA VENKAT VYAS KUMAR NAIDU

Master’s Program in Manufacturing Engineering

APPROVED:

Tzu-Liang (Bill) Tseng, Ph.D., Chair

Amit J. Lopes, Ph.D. Co-Chair

Yirong Lin, Ph.D.

Charles Ambler, Ph.D.

Dean of the Graduate School
 Copyright ©
By
Varaha Venkat Vyas Kumar Naidu
2018

iv
SHAPE DETECTION OF STRUCTURAL ELECTRONICS USING MACHINE

VISION SYSTEM EMBEDED INTO SCARA ROBOT

by

VARAHA VENKAT VYAS KUMAR NAIDU, B.Tech

THESIS

Presented to the Faculty of the Graduate School of

The University of Texas at El Paso
In Partial Fulfillment
of the Requirements
for the Degree of

Master of Science

Department of Industrial and Systems Engineering

THE UNIVERSITY OF TEXAS AT EL PASO
May 2018

iv
 Acknowledgements

I would like to express my deepest gratitude to my thesis committee chair Dr. Tseng Bill

for the opportunity of being a part of his research team, and for his continuous assistance and

support at each level of my master’s journey. I sincerely thank my thesis committee Co-chair Dr.

Amit Lopes for all his support and guidance; for sharing his dedication, knowledge, and passion

for his work that highly motivated me to give my best effort in this presented research. I especially

would like to thank my committee member Dr. Amit lopes for his time and providing feedback. I

would like to acknowledge Sheakar Charith Boppana, for teaching me how to operate the robot.

My heartfelt appreciation to the faculty and staff of Industrial, Manufacturing and Systems

Engineering Department for making my stay extremely enjoyable.

Finally, I must thank my parents for their blessings, love and constant encouragement

throughout my master’s program. Also, I would like to thank all my friends and family members

who supported and motivated me all the time.

iv
 Abstract

Additive manufacturing processes combined with enhanced technologies for electronics

production enables a highly flexible manufacturing of personalized 3D electronic devices. Due to

the technical progress within the last years, these technologies made their way from prototyping

towards manufacturing. The growing request for products manufactured in batch size, making use

of a robot to incorporate the electronic components necessary to make functional 3-Dimensional-

Structural Electronics will extend the mass production possibilities. There is significant interest in

automation of a process in which parameters are influential in the occurrence of assembly defects

and which adjustments should be made to those variables to reduce assembly defects. The present

study aimed at investigating the accuracy of the electronic parts mounted on the 3D printed board,

and up to which extent these factors are significant for the accurate shape detection process. The

full factorial experiment was conducted to illustrate and explain the findings in the parameter

optimization. Finally, optimal settings for the robot are proposed for more repeatedly and

accurately.

v
 Table of Contents

Acknowledgements……………………………………………………………………….. iv
Abstract……………………………………………………………………………............ v
Table of Contents………………………………………………………………................. vi
List of Tables.…………………………………………………………………………….. viii
List of Figures…………………………………………………………………….............. ix
Chapter 1:
1. Introduction…………………………………………………………………………….. 1
1.1 Robots in Industry………………………………………..…………................ 1
1.2 Present Industrial Robotic Control and its Development…………………….. 5
1.3 Rapid Prototyping and 3D printing……………………………….................... 6
1.3.1 3D Printing Architecture:…………………………………………... 9
1.3.2 G-Code…………………………………………….………………... 9
1.4 Measuring the 3-Dimensional Objects Using Vision Optical Method….......... 10
1.4.1 Photogrammetry……..……………………………………………... 12
Chapter 2:
2. Theoretical Background……..…………………………………………………………. 18
2.1 Printing of Structural Electronics Using 3D Printer Process…………………. 18
2.2 Experiment Set-up………………………………………………………......... 19
2.2.1 Machine Vision System (Cognex)………………………………….. 20
Chapter 3:
3. Research Methodology………………………………………………………………… 21
3.1 General Review………………………………………………………………. 21
3.2 Design of Experiments (DOE)………………………………………………... 21
3.2.1 Factorial Design Advantages………………………………….......... 22
3.2.2 Experimental Model Based on Four-Factor………………............... 23
3.3 Design of Experiments (DOE) Methodology………………………………… 24
3.3.1 Planning Phase………………………………………........................ 24
3.3.2 Designing Phase…………………………………….…….……….... 26
3.3.3 Experiment Conducting Phase………………………….................... 26

vi
Chapter 4:
4. Data Analysis and Obtained Results…………………………………………………… 33
4.1 Analysis of Variance (ANOVA)……………………………………………... 33
4.1.1 Interaction Effect……………………………...…………….......... 34
4.1.2 Main Effects……………………………………………................. 37
4.2 Factorial Plotting……………………………………………………………… 38
Chapter 5:
5. Conclusions and Recommendations ……………..…………………………................. 40
5.1 Conclusion…………………………………………………………................. 40
5.2 Recommendations…………………………………………………….............. 41
References………………………………………………………………………………… 42
Appendix I- Yamaha SCARA YK180X Specifications………………………………….. 45
Appendix II- Vacuum Cups Used in the Experiment………..…………………………… 46
Appendix III Program for the Robot………………………………………..................... 47
Vita……..…………………………………………………………..................................... 50

vii
 List of Tables

Table 1.1: Original Values of DH Robot Parameters……………………………….......... 3

Table 1.2: Several Possible Features of Analyzed Products…………………………........ 15
Table 1.3: Industrial Vision System Classification………………………………………. 17
Table 3.1: Levels of Variables Used in the Experiment………………………………….. 26
Table 3.2: Design Layout with Accuracy Score………………………………………….. 27
Table 4.1: Analysis of Variance Table…………………………………………………… 33

viii
 List of Figures

Figure1.1: Schematic overview of the stereovision system and calibration procedure

experimental setup ……………………………………………………….………………. 4
Figure 1.2: The Process of the Robotic Arm …………………………………………….. 4
Figure 1.3: Model-Based Control Systems ……………………………………………… 6
Figure 1.4: 3D Printing Flow Char Diagram …………………………………………….. 8
Figure 1.5: 3D Printer Hardware Architecture …………………………………………... 9
Figure 1.6: Typical Vision Sensor System ………………………………………………. 11
Figure 1.7: 360 Degrees Vison Measuring Process ……………………………………… 12
Figure 1.8: Setup Layout for Conducting an Experiment ………………………………... 13
Figure 1.9: Depth Estimation with Camera ……………………………………………… 14
Figure 1.10: Predicted Orientation of the Pen …………………………………………… 14
Figure 1.11: Major Categories of Industrial Vision Applications.……..………………… 15
Figure 1.12: Major DoFs of Industrial Vision Systems ……………………………..…… 15
Figure 2.1: 3D Printed Circuit Board (Source: W.M. Keck center) ……………………... 18
Figure 2.2: Internal Structure Achieving the High-Rigidity and High-Accuracy ……….. 19
Figure 2.3: Insight Explorer Machine Vision System …………………………………… 20
Figure 3.1 Process Variable of Pick and Place the Assembly of Electronic Components . 25
Figure 4.1: Normal Probability Plot of Residuals for Accuracy Score…………………... 35
Figure 4.2: Box-Cox Plot of Accuracy Score…………………………………………...... 36
Figure 4.3: Normal Probability Plot of the Transformed Response……………………… 37
Figure 4.4: Main effects Plot for Transformed Accuracy Score…………………………. 38
Figure 4.5: Interaction Plot for Trans-Accuracy Score…………………………………... 39
Figure 5.1: Contour Plots of Accuracy Score…………………………………………….. 41

ix
 Chapter 1: Introduction

1.1 Robots in Industry

The use of robots in industries has revolutionized the industrial production and manufacturing

system since late 90’s. Robots were used to perform simple tasks such as pick and place objects from

one place to another with a point to point teaching process. By that time, they did not have any type of

sensors or visual analysis equipment to analyze the further steps (neither external sensors).

They have replaced humans in monotonous, repetitive, heavy, and hazardous operations locations.

Nowadays, these industrial Robots have been divided into three types based on operations, they are:

1) Material Handling Robot

2) Process Operation Robot

3) Assembly Robot

There are five fundamental laws of robotics that robots are always expected to follow, however,

the following three laws are the most important:

1. “A robot may not injure a human being, or, through inaction, allow a human being to come

to harm.”

2. “A robot must obey the orders given by human beings except where such orders would

conflict with the First Law.”

3. “A robot must protect its own existence as long as such protection does not conflict with

the First or Second Laws” [1]

1
 The first three laws of robotics have been complemented by two more laws by Stig Moberg

from ABB Robotics that have been presented in the movement of robot and also taken into

consideration. These two laws are:

4. A robot must follow the trajectory specified by its master, as long as it does not conflict

with the first three laws.

5. A robot must follow the velocity and acceleration specified by its master, as long as

nothing stands in its way and it does not conflict with the other laws. (Translated from

Swedish by the author [9, p.8]).

Robots have been very successful in the Industrial Manufacturing field in the area of

production, welding and material handling applications, to widening the range of quality control

and machining process, for Robots are capable of handling heavy loads and sensitive operations [2].

By 1985, nearly 16,000 industrial robots were present in the United States of America. Where

several benefits related to the manufacturers are observed:

1. They have reduced the labor cost.

2. Due to accurate operations, they have increased the productivity.

3. They were capable to operate in dangerous and difficult conditions.

4. They have improved the quality of the product.

5. Due to accurate operations, they have reduced the material waste during production.

6. Due to easy operations, they have increased the flexibility of manufacturing operations.

7. Industrial robots have very low capital cost.

8. Robots are easy to control and to fasten in the inventory [3].

2
The Vision Analysis is a process of controlling and manipulating its environment by observing,

which includes the following two processes behind its operation:

 Look and Move

 Visual Analysis

This Visual Analysis method, is precisely calibrated with the camera and the manipulator,

and it follows the Jacobin matrix, which involves the image displacement of camera motion [4].

The not contact method which is also known as stereovision system, attaches cameras to the robotic

arm to measure and calibrate the points within its operational space. For example, the system

consists of 6 design of freedom (DOF) KUKA KR 6R900 robotic arm with a repeatability of ±0.03

mm and 6kg payload. DH robot parameters are shown below in table 1:

Table 1.1
Original values of DH robot parameters [7].
d (mm) Ɵ (deg) a(mm) a(deg)

Link 1 400 ᶿ1 25 90

Link 2 0 ᶿ2 455 0

Link 3 0 ᶿ3 35 90

Link 4 420 ᶿ4 0 -90

Link 5 0 ᶿ5 0 90

Link 6 80 ᶿ6 0 180

The robotic arm has an interchanging option tool, which allows a flexible automation tool

exchange. The vision system, as shown in the below figure, consists of CCD cameras and a robotic

tool as center point. One pixel is roughly mapped to 0.05mm. The camera system allows high

3
definition image and accurate measurements of sphere center [7]. Figure 1.1 present an example

Vision system and the calibration procedure.

Fig. 1.1
(a) Schematic overview of the stereovision system; (b) Calibration procedure experimental setup [6].

Time based visual mapping and moving of the robot into a desired location, is possible with

the camera feedback loop system. The camera provides the desired location on the path and by using

a manipulator sensor the robotic arm will be controlled. This type of model can be applied to

different type of systems like eye-in-hand system, part-in-hand system, and mobile robot system.

The following diagram explains how the process of the robotic arm and the vision sensor work.

Fig.1.2, The process of the robotic arm [8].

4
Once the process of the camera is completed, the data is sent to the manipulator sensor. During that

time, several image processing problems can occur, as it follows:

1) The visual sensing and processing of the data can be too much slower than the
mechanical operation of the robot.
2) There can be serious lag between the object and the sensor while visually

understanding the object.

3) There is a possibility of motion lag due to slow updating in the visual feedback. This

causes a bigger motion to the other updates, making a delay when the mechanical

system finishes the motion [8].

The development of industrial robots is going to go further and further by using advance

technology and performance control in order to reduce the manufacturing cost. Most robots are

reliable and efficient to maintain and operate them safely. Presently, there are certain auto

manufacturing industries which use fully automation programs while assembling powertrains,

painting, courting, and metal stamping. Looking into the future, we see more about Artificial

Intelligence (A.I.) robots are used in industries during the final assembly process. As industrial

robots are able to handle complex assembly operations, they are more likely to be seen in coming

decades. Nowadays, new robotic technology and flexible automation are required to operate

complex operations like assembly and product geometries. In the next few years, the biggest

challenge of robots will be to obtain feasible solution for complex operations.

1.2 Present Industrial Robotic Control and its Development

There is a tendency of development of industrial robots timely. The development of new

products originates new production concepts brought into implementation in automotive industry.

The recent development in multi robot control and different solution are presented in many robot
5
development companies. The main reason to introduce multi robot control into the industry is to

reduce the cost of production. The multi control robots requires high performance control concepts

and efficient to program the robot [9]. The following image (1.3) shows an example of the multi

control robots is a production process.

Figure 1.3, Multi robot control requires high performance model-based

control concepts and efficient methods for robot programming [8].

1.3 Rapid Prototyping and 3D Printing

Rapid prototyping is a group of techniques which are utilized to produce quick fabrication

of different types of parts using CAD programs in computers. RP, offers promising powerful

technology that has the potential to revolutionize the prototyping industry. This process involves

various ideal features, ideas, concepts, functions, and performance. This process of production is

also called as 3d printing or layer-based manufacturing. The history of RP comes from 1960es by

Prof Herbert Voelcker, who has devised the tool of mathematical and the three-dimensional objects,

resulting in mathematical and algorithmic theories for solid molding and fabrication. In 1987 Carl

Deckard, a researcher from University of Texas, developed layers of manufacturing and printed a

model using 3d printers. He utilized laser fusion metal powder in solid prototypes single layer at a
6
time. The first patent of apparatus of producing 3d printed objects using stereo lithography was

awarded to Charles Hull, whom many have believed him as the father of RP industry.

Steps involved during RP prototyping models

1. Imaging the model using MRI or CT scan.

2. Acquisition of DIACOM files.

3. Conversion of DIACOM Into STL files.

4. Evaluation of the design.

5. Surgical Planning and superimposition if desired.

6. Additive Manufacturing and creation of model.

7. Validation of the model.

The above steps were involved while getting MRI or CT scan from patients. CT scan is

preferred to be high sliced caliber than slice thickness is of 1-2mm. Most of the MRI and CT scan

software’s gives output in form of digital imaging popularly known as DIACOM image format [10].

Reducing the time produced during creation of the new prototypes is the key to produce new faster

products. Based on this fact, many companies utilize RP techniques. One of the major disadvantage

about RP process is that they use only paper, polymers, and waxes [11, 12]. Metallic porotypes can

be manufactured in two different steps. The 1st step is to design the component in the CAD

application and send that file to 3D printer or rapid Prototyping equipment [13].

3D printing process refers to a process of converting 3D CAD design to 3-dimensional

physical object. The above figure describes the end-to-end flow chat with functional entities and

they are explained

7
 Fig. 1.4, 3D Printing Flow Char Diagram [9].

In the above figure (1.4), the diagram is representing a step by step 3D printing process. In the

cyber domain stage the application layer will be processed as a (stl) file that later on will be

developed into a compiler layer through the instructions that have been sent to the printer. Next, in

the physical domain, the (stl) file is transformed in the form of a (G-core) file, to finally be

transformed into a hardware layer. The following steps, describe this process in more detail.

 Application: The Model is created in the form of stereo lithography (stl) and Computer-

aided design (CAD).

 Compiler: The compiler is program to process the STL file and to generate tool path

file, which is called G-CODE file in many cases. The G-CODE has the instructions to

direct the printing process.

 Firmware: The firmware on the 3D printer interprets the g-code and transmits the

required control signals to the hardware.

 Hardware: The physical units on the 3D printer like Motors, heaters and the cooling

fan are being controlled according to the inductions given in G-code.

8
1.3.1 3D Printer Architecture:

FDM (Fusion Deposition Modeling) is the most commonly used 3D printer [14]. The FDM

3D printer process is based on forming structures by the material that is basically put under the last

parts of the object to be built, and that is developed through forming layers. Once a layer is

completely finished, then the following layer can start its process. This pattern will continue until

the product and its supporting part are finished.

The figure below (1.5), represents the hardware architecture of the FDM 3D printer. The

motors shown in the image, are represented with X, Y, and Z which control the movement of the

nozzle. The Extrusion motor monitors the extrusion of the material. The heater melts the material

and lays it down on the printed platform layer by layer. The cooling fans are made to work to prevent

the excluder from overheating.

1.3.2 G-Code:

The G-code (see image 1.5), is also known as RS-274 and is also known as the set of

numerical control instructions for the 3d printer to process [15].

Fig. 1.5, 3D Printer Hardware Architecture [9].

9
1.4 Measuring the 3-Dimensional Objects Using Vision Optical Method:

In the present industry, accurate measuring of 3D dimensional objects is needed to precisely

determine the quality and dimensions of the product. In various applications, 3D measurements

require intelligent robots to identify the shapes. This type of analysis includes measurements of die,

stamping panel geometry, stress/strain, and vibrations measurements. In this process the vision

analysis is converted into 3D shape measurement of objects under the inspection process. The

automation on inspection and recognition issues can be converted into 3D shape measurement

object under inspection. For example, the dents in the metal body or paint can be easily identified

by the vision sensor. 3D shape measurement has been recently commercialized and widely used in

different industries. For some scale of depth or shape, micrometer or nanometer measurement can

be reached if con focal or microscope is used during measurement, however, it has been a

challenging to measure 0.5mm depth measurement but frequent usage of more cameras and camera

position are required to obtain several shapes from which the final lag shape has been patched.

Various optical techniques have recently developed to measure the 3D shapes. Machine

vision provides innovation solution towards industrial automation [18]. A plethora of industry

activities have various benefits form the application of industrial vision technology in the

production process. These activities include manufacturing delicate electronics component [19],

machine parts, and integrated circuits [20]. Typical industrial vison system computer is employed

to acquire the object image data. Usually, images are acquired by one or more cameras which are

placed above the scene under proper inspection. These cameras of the vision system are regularly

fixed, and in most of the cases, industrial vision system cameras are used to measure the objects

which are fixed in its position. The following image (1.6) explains the setup of the experiment. It

10
shows the main processor that connects to the camera and identifies the object to examine any flaws

regarding dimensions.

Fig. 1.6, Typical Vision Sensor System [10].

The types of Processes that can be performed with the set up are as follows:

 Control over manufacturing process pick and place, printed circuits, painting circuits.

 Propagate the other external devices for further processing.

 Identifying the different type of defects and taking action for the reported items and

correcting or replacing with the correct item from the production line.

The basic requirement to develop and design the industrial machine vision system depends

on the application domain that is related to the task to accomplish, environment, speed, etc. There

are no existing industrial visions system that have ability to handle all the applications. Only one

requirement of particular application domain is specified, then appropriate decision has been taken

while design and development.

11
1.4.1 Photogrammetry:

Complex Photogrammetry employs stereo techniques to measure 3D shape objects. Also,

others methods such as shading and scaling can be used. Photogrammetry uses some bright markers

such as retroreflective painted dots on surface of the object.

1.4.2 General Approach to Measure 360-degree Shape of the Object:

In global cordinate system is a setup and local cordinates are measured during registration.

The imaging sensor (camera) is placed aproximate position to accuratly map the 3D object

succusfully in a single view, and abslute phase value at each object point is calculated. Image 1.7,

is shwoing the proces of how photogrammetry works. The camera is mesuring the dimensions of

the objects below to compare data to givenh specifications.

Fig. 1.7, 360 Degrees Vison Measuring Process [11].

These phase values and geometric optic model of mesurment system determine the 3D

coordinates of the object point. There are three ways that are commonly used to measure 360

degrees of the desired object shape by a rotation method, the camera system transmits the imaging

system with multiple camera system. Image 1.8 presents the set up layout of the measurement of

360 degree. On top, the sensor of cameara will optically track location and orientation of the object
12
 Fig 1.8, Setup Layout for Conducting an Experiment [17].

The final measurements of the 3D coordinates are compared with the CAD data in the

computer using various methods, in which the different comparison techniques are used and square

fits are often used [17]. Nowadays modern robots can be hand-calibrated by teaching them or they

can be even scripted to carry out operations easily. However, autonomous grasping of the object

and handling them still is unknown and challenging. To grasp the previously known object or trying

to obtain 3d model of object, various types of approaches can be applied. It has been so difficult to

obtain the 3D reconstructions of the object during first time vision as it’s hard to identify the textures

and even where stereopsis works best, since it helps in reconstructing the visible portions of the

object.

The next figure (1.9) presents 2 type of vision. In the first image(A), first time vision, as the

object appearance is transparent and light reflective, the texture can be barely determined, what

limits the 3D reconstruction. Image (B), in the gray scale, allows to measure the depth of the object,

however, the stereovision will not restore this measurement thus, a laser scanner should be used in

order to determine the depth. Figure 1.10, presents a case where a robot tries to analyze the

dimensions of an object such as geometrical characteristics in 2D and 3D using a visual sensor. This

13
object, opposite to the previous example, is not a transparent object but an opaque one. The two

different lines, dotted and solid, represent the true and the predicted orientations of the object.

(A) (B)

Fig. 1.9, Image (A) is the texture less, transparent and reflective object. Image (B) shows the depth
of the object in the gray scale. The stereo vision failed to return the depth estimate. Specialized sensors
like laser scanners are utilized to estimate the object shape [12].

Fig. 1.10. Laser scanner measuring non-reflective object. Predicted orientation at the grasping point
for pen. The dotted line represents the true orientation and the solid. Line represents the predicted
orientation [12].

Through the use of a visual sensor, or a laser scanner, different features of the inspected

product can be determined. Characteristics such as shape, surface flaws, measurements, presence

of external objects, and quality issues can be detected. Table 1.2, explains these characteristics in a

more detail.

14
 Table 1.2
Several possible features of analyzed products [10].

Dimensional Dimensions, shape, positioning, orientation, alignment,

roundness, corners.

Structural Assembly (holes, slots, rivets, screws, clamps)

Foreign Objects (Dust, bur, swarm)

Surface Pits, scratches, cracks, wear, finish, roughness, texture

Seams-folds-laps, continuity.

Operational Incompatibility of operation to standard and specifications.

Fig. 1.11, Major Categories of Industrial Vision Applications [10].

Most of the applications of industrial vision are focused on inspection of a single

characteristic. The remaining characteristics are considered as degrees of freedom since these

applications indicates the flexibility of the vision system. A popular real time application is

inspected and classified of shouldering joints on printed circuit boards (pcb). Figure 1.11, shows a

graphical example of the single characteristic inspection.

Fig. 1.12, Major DoFs of Industrial Vision Systems [10].

15
 Industrial vision applications have been classified based on different features whose

measurements does not affect the inspection process. Some of the most common degrees of freedom

(see figure 1.12), meanwhile the following table (1.3) illustrates the relationship between Degrees

of freedom and the quality inspection techniques development and use in different applications [21].

16
 Table 1.3
Industrial Vision Systems Classification [10].
Quality Inspection Application Field DOFs

Surface Mini resistor painting Resistor orientation

Aluminum sheet casting Sheet width
Railroad line inspection Illumination/rail-foot-head position
Oil seals Illumination
Chicken meat defects Illumination/skin
Water surface inspections Distortion/scale/orientation/position
Surface approximations Illumination
Granite surface inspection Texture
Directional texture Illumination/rotation of direction
Surface roughness Orientation
Surface defects Pose
Textile seam defects Translation/rotation
Internal wood defects Wood density
Wood veneer surface Scale/intensity
Surface corrosion Shape

Dimensional Machined part inspection Scale/translation/orientation

Solder joints inspection Orientation
Solder joints inspection Orientation/position
Solder joints inspection Position/orientation/size
External screw threads Thread position
Banknotes inspection Position
Image segmentation Shape/texture
Object classification Scale/orientation
Object classification Shape
Circular parts Peripheral occlusion
Packaging Position/orientation
Line segment measurement Orientation/scale
Fruit harvesting Maturity/illumination/occlusion
Packaging Shape/size
Packaging Illumination/size
Automotive industry Size/shape/pose

Structural Object assembly Orientation (limited)

Railroads part inspection Orientation/shape
Railroads part inspection Illumination/shape/texture
Automotive industry Illumination/position
Automotive industry Illumination/position/shape/size
PCB inspection illumination

Operational Laser butt joint welding welding path shape/gap size/beam position
Wrist watch quality hands shape/size/orientation/distortion

17
 Chapter 2: Theoretical Background

2.1 Printing of Structural Electronics Using 3D Printer Process

3D printing which is also known as additive manufacturing, in this process of production

we design the objects using cad software. This process came into existence since 1980s [22]. Since

the last few decades, there has been research and development in a variety of additive manufacturing

process and materials because of the ability to print board range structures with accurate and cheaper

price. With the help of 3D, the printing process can fabricate parts from scratch in very few hours,

as It enables the power to create virtual object into 3D physical object.

Researchers and developers were creating different ways to utilize 3D printing or additive

manufacturing techniques to 3D printed electronic components, due to various advantages of 3D

printing that includes easy duplicating of product and low cost. Commonly, STL format file is used

before giving this design file to the 3D printer. The image below shows a 3D electronic component

made in W.M. Keck Center for 3D innovation in The University of Texas at El Paso.

Fig. 2.1, 3D electronic circuit board (W.M.Keck center)

18
2.2 Experiment Set-up

The consideration of 3D printing was selected randomly to conduct the experiment. The

Yamaha Yx180 equipment is a robotic arm used to perform the experimenal operations such as

picking and placing the the elecronic components from the platform to the designed location, in

order to test the performance accuracy. The following set of images introduces this example:

Fig. 2.2, Above, Internal structure of Robotic Arm. Below, Robotic Arm set
over the platform [13].

19
2.2.1 Machine Vision System (Cognex)

Machine Vision is a techonology and method that provides an image-based inspection

technique. Through this vision system process, automatic inspection, control, and guidance are

monitered. The machine vision system is used as well to inspect the production process and to

mesure the accuracy. In the vision system, any image can be set as an standard image. For this

research, four models of images are representing the electonic components and are standard images.

When the experiment runs multiple times, the accuracy of identifying the object dimensions will

increase to obtain the accurate results. The next image ilustrates the user intriface of the Insight

Vision system.

Fig. 2.3, Insight Explorer Machine Vision System [14].

20
 Chapter 3 Research Methodology

3.1 General Review

In this chapter, the methodology proposed to evaluate the process variables that are

embedded into structural electronics will be discussed. The robotic arm is going to be evaluated and

operated at different parameters.

3.2 Design of Experiments (DOE)

Design of Experiments (DOE) refers to the planning process of an experiment so that

appropriate data can be obtained and evaluated to acquire a conclusion. To solve the experimental

errors, statistical analysis is the appropriately developed. Many industrial experiments involve the

gathering of the appropriate data from two or more number of cases studies. To carry out different

experiments and to study the effect of factors at different levels, many experiments need to be

carried out. Factorial experiments depend on the following reasons:

1) The main moto of this experiment is to determine the quickest effect on each of the

number of factors over specified range.

2) To analyze the interaction among the different factors.

3) To carry out different practical experiments to obtain results over different

conditions.

Factorial design has been used to solve problems due to the versatility that it provides to

study the different individual effects, its levels accurate the process. Factorial has another

advantage, it has the possibility of changing factors with giving up the randomization. The reason

of choosing certain design depends on the number of factors that has been analyzed. It also depends

21
on a number of different levels of each specific factor of a given number of experiments which are

been carried out.

3.2.1 Factorial Design Advantages

Factorial design is normally more efficient than one factorial at a time of experiment to

determine the couple or more responses. A factorial design normally is efficient even if there were

interactions between the factors of experiment. The factorial Design gives freedom to estimate at

various levels of other factors and to obtain yield conclusion which are valid over a range of

experimental conditions. Factorial design is necessary for each complete trail or replication of levels

of factors, these are investigated without loss of randomization, and they are highly efficient for

industrial experiments.

A factor is an independent variable which is used in research of an experiment. An effect of

factors is defined as a change in response produced by different level of factors. This type of effect

is also known as main effect. An interaction is said to be happening in the experiments if the

difference between the levels of single factor is not the same at all levels of the other factors. An

interaction is said not be occurring if the response of the factor A at all levels of factor B is not

dependent on factor B. The basic principles of randomization and replication that were followed by

experiments which were done by Keyboards. Randomization is the process in which the order to

runs the experiment is randomly determined. It tends to have an average effect of extraneous factor

in an experiment that is not under specific control. Replication number has been observed or sample

size treatment. Independent values of experiments are usually studied through experiments.

Mathematical representation model for four-factorial representation can be written as explained in

the next section.

22
3.2.2 Experimental Model Based on Four-Factor

The Four-Factorial Experiments Involves four different types of factors, also known as

treatments. The normal model has ‘a’ level of factor A, ‘b’ level of factor B, ‘c’ levels of C, ‘d’

level of factor D, and n replicates. The observation can be described as a statistical model as shown

below:

Yijklm=U+Ai+Bj+ABij+Ck+ACik+BCjk+ABCijk+Dl+ADil+BDjl+ABDidl+CDkl+

ACDikl+BCDjkl+ABCDijkl+Eijklm

U = overall mean effect

Ai; = effect of ith level of factor A

Bj = effect of jth level of factor B

ABij = interaction between factors A and B

Ck = effect of kth level of factor C

ACik = interaction between factors A and C

BCjk = interaction between factors B and C

ABCijk = interaction between factors A, B and C

Dl= interaction between factors A and D

BDj; = interaction between factors B and D

ABDjj; = interaction between factors A, B, and D

CDkl = interaction between factors C and D

ACDikl = interaction between factors A, C, and D

BCDjkl = interaction between factors B, C, and D

ABCDijkl = the four factor (ABCD) interaction

23
Eijkim = random error component

In these experimental models, various experiments were carried over the ultraviolet coating

machine. Using this experiment, four-factors were analyzed. In the upcoming section the

experimentation is explained in more detail. The experiments were performed using Analysis of

Variance (ANOVA) for each component with a design software. Plots probability and residuals

were also analyzed.

3.3 Design of Experiments (DOE) Methodology

The Methodology of DOE was basically divided into four phases [23], as shown below:

A. Planning phase

B. Designing phase

C. Experimental phase

D. Analyzing phase

3.3.1 Planning Phase

The successive procedure of outlining the Montgomery followed during the planning phase

of the experiment is made of problem recognition, Planning phase is made up of problem detection,

selection of response, selection of process variables, classifying different types of variables and

analyzing the levels of process variables.

(a) Problem recognition: A clear statement can be made for better understanding of what

should be done. Finding pick and place process spread of 3D printed electronics

components using SCARA robot and finding optimal setting to reduce the process

speed, which might lead to poor compatibility.

24
 (b) Selection of responses: The selection of appropriate response for experiment is critical

to the success of any industrial design experiment. For detection of dimensions and

quality of the object accuracy is important. So, the responsive variables were later

analyzed through assembly process. In this experiment we used Cognex vision sensor

to measure the accuracy as explained in previous chapter. Few trail runs were done on

the vision system to see the robustness to the environmental changes.

(c) Selection of Process variables: To identify possible design parameters, engineering

knowledge is used. As a result of the planning, the constant variables were held and

nuisance factors were determined. The control variable was differentiated during the

experiments and determined based on the working knowledge of the robot. The below

illustration shows the variables consideration during the experiment.

Temperature (24-25 degrees)

Motion type (curvilinear)

Control
Held Constant Nuisance
Variables
Variables Variables

Vacuum Pressure Operation Bias

Velocity Process
Friction
End Effectors Variables
Work Envelope

Fig. 3.1, Process Variable of Pick and Place the Assembly of Electronic Components [15].
Determining the levels of process variables: There are two qualitative (type of vacuum end

effectors, Work envelope) and two quantitative (vacuum pressure, velocity) variables in this

experiment.
25
 All the four variables were studied at three levels as the decision was made to use three-

level factorial designs.

Table 3.1
Levels of Variables Used in the Experiment.
Variable name Units Level 1 Level 2 Level 3
A Vacuum Pressure Psi 50 70 90
B Velocity Mm/s 9 45 180
C Type of Vacuum Gripper Xp-B5 XP-B8 XP-B10
D Position and shape of the part
PAT1 PAT2 PAT3

3.3.2 Designing Phase

The factorial design is considered based on the variables and number of levels. Selection of

sample size is one of the most important aspects of any experimental design. Three replicates of the

experiment are conducted and blocking is used to reduce the variability caused by noise factors or

nuisance factors. The operator takes the results over the experiment to remove operator to operator

variability.

3.3.3 Experimental Conducting Phase

Several considerations are recognized after performing the experimental trials:

 Ensuring the experiment is not affected by external sources of noise like humidity

and temperature.

 Record the observed response variables on the prepared data sheet.

 Attention was paid to variables which should be held constant so that they won’t

inflate the error in the experimental results.

Randomized Experimental runs are conducted based on the run order generated by the Design

Expert software. The run order for the experiment can be seen in Appendix IV.

26
Table 3.2
Design Layout with Accuracy Score
Std. Block Run Vacuum Velocity Type of Position Accuracy
Order Pressure Vacuum & shape score
Gripper of part
1 Black 1 46 50 9 5 PAT1 92.9
2 Black 1 18 70 9 5 PAT1 100
3 Black 1 58 90 9 5 PAT1 92.4
4 Black 1 20 50 45 5 PAT1 85.4
5 Black 1 78 70 45 5 PAT1 95.8
6 Black 1 76 90 45 5 PAT1 65.9
7 Black 1 72 50 180 5 PAT1 91.4
8 Black 1 13 70 180 5 PAT1 83.2
9 Black 1 8 90 180 5 PAT1 82.4
10 Black 1 31 50 9 8 PAT1 100
11 Black 1 70 70 9 8 PAT1 93.6
12 Black 1 68 90 9 8 PAT1 88.4
13 Black 1 52 50 45 8 PAT1 88.6
14 Black 1 5 70 45 8 PAT1 95.7
15 Black 1 65 90 45 8 PAT1 76.1
16 Black 1 54 50 180 8 PAT1 98.9
17 Black 1 6 70 180 8 PAT1 100
18 Black 1 32 90 180 8 PAT1 97.6
19 Black 1 45 50 9 10 PAT1 87.5
20 Black 1 48 70 9 10 PAT1 89.5
21 Black 1 74 90 9 10 PAT1 86.5
22 Black 1 34 50 45 10 PAT1 74
23 Black 1 24 70 45 10 PAT1 100
24 Black 1 11 90 45 10 PAT1 97.8
25 Black 1 40 50 180 10 PAT1 100
26 Black 1 57 70 180 10 PAT1 100
27 Black 1 17 90 180 10 PAT1 100
28 Black 1 39 50 9 5 PAT2 81.1
29 Black 1 44 70 9 5 PAT2 97
30 Black 1 27 90 9 5 PAT2 68.8
31 Black 1 2 50 45 5 PAT2 100
32 Black 1 77 70 45 5 PAT2 86.2
33 Black 1 21 90 45 5 PAT2 75
34 Black 1 7 50 180 5 PAT2 100
35 Black 1 61 70 180 5 PAT2 60.5
36 Black 1 51 90 180 5 PAT2 86.5
37 Black 1 63 50 9 8 PAT2 100
38 Black 1 37 70 9 8 PAT2 100
39 Black 1 50 90 9 8 PAT2 99.7
40 Black 1 33 50 45 8 PAT2 88.7
27
41 Black 1 60 70 45 8 PAT2 100
42 Black 1 41 90 45 8 PAT2 98
43 Black 1 9 50 180 8 PAT2 97.3
44 Black 1 38 70 180 8 PAT2 100
45 Black 1 28 90 180 8 PAT2 100
46 Black 1 23 50 9 10 PAT2 100
47 Black 1 79 70 9 10 PAT2 100
48 Black 1 22 90 9 10 PAT2 100
49 Black 1 42 50 45 10 PAT2 96.9
50 Black 1 19 70 45 10 PAT2 100
51 Black 1 53 90 45 10 PAT2 100
52 Black 1 15 50 180 10 PAT2 100
53 Black 1 10 70 180 10 PAT2 100
54 Black 1 75 90 180 10 PAT2 100
55 Black 1 4 50 9 5 PAT3 98.2
56 Black 1 81 70 9 5 PAT3 100
57 Black 1 71 90 9 5 PAT3 100
58 Black 1 55 50 45 5 PAT3 96.2
59 Black 1 66 70 45 5 PAT3 84.5
60 Black 1 16 90 45 5 PAT3 98.3
61 Black 1 1 50 180 5 PAT3 100
62 Black 1 69 70 180 5 PAT3 100
63 Black 1 25 90 180 5 PAT3 99.9
64 Black 1 56 50 9 8 PAT3 97.7
65 Black 1 47 70 9 8 PAT3 89.6
66 Black 1 36 90 9 8 PAT3 94.7
67 Black 1 35 50 45 8 PAT3 100
68 Black 1 30 70 45 8 PAT3 98.5
69 Black 1 43 90 45 8 PAT3 85.8
70 Black 1 29 50 180 8 PAT3 0
71 Black 1 14 70 180 8 PAT3 95.8
72 Black 1 59 90 180 8 PAT3 100
73 Black 1 49 50 9 10 PAT3 98
74 Black 1 73 70 9 10 PAT3 92.7
75 Black 1 80 90 9 10 PAT3 82.2
76 Black 1 26 50 45 10 PAT3 95.4
77 Black 1 67 70 45 10 PAT3 96
78 Black 1 3 90 45 10 PAT3 95.5
79 Black 1 64 50 180 10 PAT3 0
80 Black 1 62 70 180 10 PAT3 100
81 Black 1 12 90 180 10 PAT3 99.5
82 Black 2 144 50 9 5 PAT1 71.6
83 Black 2 141 70 9 5 PAT1 100
84 Black 2 128 90 9 5 PAT1 81.9
85 Black 2 82 50 45 5 PAT1 90.5
28
 86 Black 2 105 70 45 5 PAT1 71.6
87 Black 2 126 90 45 5 PAT1 71.6
88 Black 2 98 50 180 5 PAT1 92.4
89 Black 2 115 70 180 5 PAT1 95.7
90 Black 2 99 90 180 5 PAT1 90.2
91 Black 2 153 50 9 8 PAT1 92.1
92 Black 2 158 70 9 8 PAT1 91.7
93 Black 2 107 90 9 8 PAT1 89.4
94 Black 2 152 50 45 8 PAT1 73.4
95 Black 2 139 70 45 8 PAT1 99
96 Black 2 135 90 45 8 PAT1 99.7
97 Black 2 117 50 180 8 PAT1 100
98 Black 2 148 70 180 8 PAT1 100
99 Black 2 136 90 180 8 PAT1 98.9
100 Black 2 162 50 9 10 PAT1 93.5
101 Black 2 129 70 9 10 PAT1 82.4
102 Black 2 143 90 9 10 PAT1 100
103 Black 2 86 50 45 10 PAT1 97.9
104 Black 2 84 70 45 10 PAT1 92.2
105 Black 2 110 90 45 10 PAT1 100
106 Black 2 113 50 180 10 PAT1 100
107 Black 2 160 70 180 10 PAT1 99.7
108 Black 2 147 90 180 10 PAT1 100
109 Black 2 146 50 9 5 PAT2 91.4
110 Black 2 122 70 9 5 PAT2 76.1
111 Black 2 119 90 9 5 PAT2 61.8
112 Black 2 95 50 45 5 PAT2 85.6
113 Black 2 106 70 45 5 PAT2 100
114 Black 2 159 90 45 5 PAT2 83
115 Black 2 88 50 180 5 PAT2 100
116 Black 2 155 70 180 5 PAT2 99.5
117 Black 2 138 90 180 5 PAT2 66.7
118 Black 2 133 50 9 8 PAT2 99
119 Black 2 157 70 9 8 PAT2 100
120 Black 2 90 90 9 8 PAT2 92.4
121 Black 2 125 50 45 8 PAT2 66.2
122 Black 2 123 70 45 8 PAT2 100
123 Black 2 150 90 45 8 PAT2 100
124 Black 2 83 50 180 8 PAT2 100
125 Black 2 101 70 180 8 PAT2 100
126 Black 2 111 90 180 8 PAT2 100
127 Black 2 112 50 9 10 PAT2 100
128 Black 2 161 70 9 10 PAT2 100
129 Black 2 103 90 9 10 PAT2 100
130 Black 2 130 50 45 10 PAT2 100
29
131 Black 2 100 70 45 10 PAT2 100
132 Black 2 114 90 45 10 PAT2 100
133 Black 2 96 50 180 10 PAT2 65.1
134 Black 2 151 70 180 10 PAT2 97
135 Black 2 127 90 180 10 PAT2 100
136 Black 2 120 50 9 5 PAT3 71.6
137 Black 2 85 70 9 5 PAT3 100
138 Black 2 137 90 9 5 PAT3 100
139 Black 2 89 50 45 5 PAT3 99.5
140 Black 2 124 70 45 5 PAT3 97
141 Black 2 97 90 45 5 PAT3 99.6
142 Black 2 93 50 180 5 PAT3 98.2
143 Black 2 91 70 180 5 PAT3 100
144 Black 2 116 90 180 5 PAT3 97.1
145 Black 2 121 50 9 8 PAT3 99.5
146 Black 2 140 70 9 8 PAT3 91.8
147 Black 2 109 90 9 8 PAT3 90.2
148 Black 2 131 50 45 8 PAT3 89.2
149 Black 2 104 70 45 8 PAT3 97.2
150 Black 2 87 90 45 8 PAT3 98.7
151 Black 2 92 50 180 8 PAT3 0
152 Black 2 102 70 180 8 PAT3 99.8
153 Black 2 108 90 180 8 PAT3 98.3
154 Black 2 94 50 9 10 PAT3 94.1
155 Black 2 145 70 9 10 PAT3 98.4
156 Black 2 134 90 9 10 PAT3 100
157 Black 2 142 50 45 10 PAT3 94.9
158 Black 2 118 70 45 10 PAT3 97.4
159 Black 2 149 90 45 10 PAT3 97.4
160 Black 2 154 50 180 10 PAT3 0
161 Black 2 156 70 180 10 PAT3 99.5
162 Black 2 132 90 180 10 PAT3 100
163 Block 3 233 50 9 5 PAT1 76.4
164 Block 3 189 70 9 5 PAT1 78.8
165 Block 3 198 90 9 5 PAT1 92.2
166 Block 3 193 50 45 5 PAT1 93
167 Block 3 205 70 45 5 PAT1 71.5
168 Block 3 204 90 45 5 PAT1 74.5
169 Block 3 168 50 180 5 PAT1 74.1
170 Block 3 200 70 180 5 PAT1 97.7
171 Block 3 229 90 180 5 PAT1 72.2
172 Block 3 163 50 9 8 PAT1 91.3
173 Block 3 177 70 9 8 PAT1 96.3
174 Block 3 197 90 9 8 PAT1 93.5
175 Block 3 191 50 45 8 PAT1 100
30
176 Block 3 236 70 45 8 PAT1 100
177 Block 3 199 90 45 8 PAT1 81.8
178 Block 3 231 50 180 8 PAT1 97.9
179 Block 3 206 70 180 8 PAT1 99.8
180 Block 3 212 90 180 8 PAT1 73.9
181 Block 3 215 50 9 10 PAT1 91.6
182 Block 3 226 70 9 10 PAT1 89.4
183 Block 3 203 90 9 10 PAT1 88
184 Block 3 210 50 45 10 PAT1 100
185 Block 3 196 70 45 10 PAT1 91.3
186 Block 3 180 90 45 10 PAT1 100
187 Block 3 185 50 180 10 PAT1 91.1
188 Block 3 228 70 180 10 PAT1 100
189 Block 3 173 90 180 10 PAT1 100
190 Block 3 224 50 9 5 PAT2 100
191 Block 3 217 70 9 5 PAT2 93.3
192 Block 3 240 90 9 5 PAT2 78.4
193 Block 3 176 50 45 5 PAT2 93
194 Block 3 222 70 45 5 PAT2 0
195 Block 3 194 90 45 5 PAT2 61.8
196 Block 3 195 50 180 5 PAT2 74.1
197 Block 3 178 70 180 5 PAT2 100
198 Block 3 230 90 180 5 PAT2 84.3
199 Block 3 174 50 9 8 PAT2 99.7
200 Block 3 172 70 9 8 PAT2 100
201 Block 3 208 90 9 8 PAT2 100
202 Block 3 238 50 45 8 PAT2 99.9
203 Block 3 223 70 45 8 PAT2 78.4
204 Block 3 218 90 45 8 PAT2 99
205 Block 3 165 50 180 8 PAT2 99.9
206 Block 3 221 70 180 8 PAT2 99.7
207 Block 3 182 90 180 8 PAT2 99.7
208 Block 3 170 50 9 10 PAT2 100
209 Block 3 187 70 9 10 PAT2 100
210 Block 3 225 90 9 10 PAT2 100
211 Block 3 171 50 45 10 PAT2 100
212 Block 3 164 70 45 10 PAT2 100
213 Block 3 175 90 45 10 PAT2 94.6
214 Block 3 235 50 180 10 PAT2 0
215 Block 3 211 70 180 10 PAT2 100
216 Block 3 242 90 180 10 PAT2 100
217 Block 3 192 50 9 5 PAT3 96.1
218 Block 3 190 70 9 5 PAT3 100
219 Block 3 169 90 9 5 PAT3 92.5
220 Block 3 181 50 45 5 PAT3 95.7
31
 221 Block 3 186 70 45 5 PAT3 89.1
222 Block 3 227 90 45 5 PAT3 98.3
223 Block 3 220 50 180 5 PAT3 99.8
224 Block 3 202 70 180 5 PAT3 99.8
225 Block 3 209 90 180 5 PAT3 88.7
226 Block 3 232 50 9 8 PAT3 96.5
227 Block 3 166 70 9 8 PAT3 96.1
228 Block 3 201 90 9 8 PAT3 94.8
229 Block 3 207 50 45 8 PAT3 98.7
230 Block 3 243 70 45 8 PAT3 91.5
231 Block 3 237 90 45 8 PAT3 96.9
232 Block 3 219 50 180 8 PAT3 99.9
233 Block 3 179 70 180 8 PAT3 99.2
234 Block 3 184 90 180 8 PAT3 95.7
235 Block 3 213 50 9 10 PAT3 87.9
236 Block 3 216 70 9 10 PAT3 92.1
237 Block 3 239 90 9 10 PAT3 99.1
238 Block 3 167 50 45 10 PAT3 96.7
239 Block 3 188 70 45 10 PAT3 91.8
240 Block 3 234 90 45 10 PAT3 95.8
241 Block 3 183 50 180 10 PAT3 0
242 Block 3 214 70 180 10 PAT3 100
243 Block 3 241 90 180 10 PAT3 93.9

In the table above, the main concentration is on the vacuum pressure and on the type of vacuum

gripper. Based on that, a high accuracy score has been obtained as observed in some of the presented

values in the results column. As shown, the velocity is not affecting the accuracy final scores. Since

however the accuracy may vary, it is only determined by the mentioned vacuum pressure and the

type of gripper. Analyzing phase and results are discussed in the following chapter.

32
 Chapter 4 : Data Analysis and Obtained Results

The the costom factorial desigin is analyzed using Minitab Software. The normal

plot pabability for standartizied resudilies from the resudials from the data is normally distributed.

Due to the process has many values which are close to the nutral limit, the data is skewed.

4.1 Analysis of Variance (ANOVA)

The Analysis of Variance (ANOVA), is a technique that gathers statistical models and their

variation to evaluate the differences between several groups. In order to develop a significance

test, the variance (a variable) is separated in several factors referable to distinct origins of variation.

The ANOVA test will yield a “statistical test” according if the means (group of variable) of the

tested groups are alike generalizing them to several groups. ANOVA test is basically a comparing

test of variable of groups for statistical purposes. Table 4.1 offers an example of an analysis of an

ANOVA test. The listed models in the table are values of the robotic arm that were used as input

for the test.

Table 4.1
Analysis of Variance Table. Adapted from Minitab.
DF Adj SS Adj MS
Model 32 34076384323 1064887010
Linear 8 8103073255 1012884157
Vacuum Pressure 2 2239080204 1119540102
Velocity 2 146693827 73346914
Type of Vacuum gripper 2 4966570645 2483285323
Position & Shape of part 2 743020400 371510200
2-Way Interactions 24 25961754765 1081739782
Vacuum Pressure*Velocity 4 4606268499 1151567125
Vacuum Pressure*Type of Vacuum gripper 4 4866562363 1216640591
Vacuum Pressure*Position & Shape of part 4 2665894581 666473645
Velocity*Type of Vacuum gripper 4 2062432382 515608095
Velocity*Position & Shape of part 4 2907184048 726796012
Type of Vacuum gripper*Position & Shape of part 4 8867895045 2216973761
Error 210 54050558358 257383611
Lack-of-Fit 48 21150843846 440642580
Pure Error 162 32899714512 203084657
Total 242 88126942681

33
Source F-Value P-Value
Model 4.14 0.000
Linear 3.94 0.000
Vacuum Pressure 4.35 0.014
Velocity 0.28 0.752
Type of Vacuum gripper 9.65 0.000
Position & Shape of the part 1.44 0.238
2-Way Interactions 4.20 0.000
Vacuum Pressure*Velocity 4.47 0.002
Vacuum Pressure*Type of Vacuum gripper 4.73 0.001
Vacuum Pressure*Position & Shape of the part 2.59 0.038
Velocity*Type of Vacuum gripper 2.00 0.095
Velocity*Position & Shape of the part 2.82 0.026
Type of Vacuum gripper*Position & Shape of the part 8.61 0.000
Error
Lack-of-Fit 2.17 0.00

4.1.1 Interaction Effect

The model presents six interaction effects to firstly be evaluated. The P-value of 0.097 for

the velocity by gripper type interaction is not less 0.05. Therefore, there is no significant interaction

effect. Which means, there is no evidence that the velocity of the robot on the accuracy of the

assembly depends on the type of the gripper. As well, interaction between 1) vacuum pressure and

velocity 2) vacuum pressure and gripper type 3) velocity and position and shape of part 4) gripper

type and position and shape of part 5) vacuum pressure and position and shape of part. The

following graph assesses the equal distribution of the data sets. The resulting points should form a

pattern along the straight line if a normal distribution was to be achieved.

34
 Fig. 4.1 Normal Probability for Resudial Accurtely. Acquired from Minitab Software.

The accurate score has been transformed using Box-cox Transformation process. From the

above box-cox plots the transformation is said to be considered as conflict interval for λ does not

include 1.

A Box Cox transformation is a method to turn a non-normal dependent variables into a

normal shape. For lots of statistical techniques normality is very important. For non-normal data

applying a Box-Cox means that a wider number of tests can be run. The assumption of normality

often leads to simple tests, mathematically tractable, and powerful compared to tests that do not

make the normality assumption. Unfortunately, many real data sets are not accurately normal.

However, an appropriate transformation of a data set can often yield a data set that does follow

approximately a normal distribution. This increases the applicability and usefulness of statistical

techniques based on the normality assumption. Figure 4.2, presents a graph with acquired data

through a Box-Cox Plot.

35
 Fig. 4.2, Accurate score for box-cox. Acquired from Minitab Software. The aim of the Box-Cox
transformations is to ensure the usual. Assumptions for Linear Model hold. That is, clearly not all
data could be power-transformed to normal

The A normal probability plot is extremely useful for testing normality assumptions. It’s

more precise than a histogram, which can’t pick up subtle deviations, and doesn’t suffer from too

much or too little power, as do tests of normality.

36
 Fig. 4.3, Normal probability of the transformation response. Acquired from Minitab Software.

4.1.2 Main Effects

The model contains two significant main effects (vacuum pressure and gripper type) are

evaluated in the presence of a significant interaction. The p-value of vacuum pressure and gripper

type effect on accuracy score which are less than 0.05. Therefore, you can conclude that there is

significant effect. The p-values of velocity and position and dimensional shape of the part are not

less than 0.05. There is no evidence that accuracy score depends on the velocity and position and

shape of the part.

37
4.2 Factorial Plotting

The main effects plot is most useful where several categorical variables are present. It helps

to compare the changes in the level means to see which categorical variables influences the

response the most. The main effects were present when there is mean of the response varies at

different levels than at another level [11]. Variables with two levels, the mean is highest at one

level of the variable than another level. This main effect is caused because of this difference. The

main effect is only interpretable if the interaction which are not significant.

Fig. 4.4, Main Plot for Transformation Accuracy. Acquired from Minitab Software.

Minitab creates the main effects plot by plotting the fitted mean for each value of a variable

in the model. Minitab can plot data means for variable that are not in the model. A line connects

the points for each variable. Looking at the line we can determine whether or not the main effect

is present for the variable. When the line is horizontal (parallel to the x-axis), then there is no main

effect present [6]. Each level of the variables will affect the response in the same way, and the

response mean is the same across all level. When the limit is not horizontal (parallel to the x-axis),
38
then there is a main effect present. Different levels of categorical variables affect the response

differently. The greatest the difference in the vertical position of the potted points, the greater the

magnitude of the main effect. To determine if a different is statistically significant, check the p-

value of the term in the analysis of variance table. By comparing the slopes of the lines, you can

compare the relative magnitude of the effects. Minitab also draws a reference line at the overall

mean. Factorial plots do not use the data in the worksheet for the fitted means, instead, Minitab

estimates the fitted mean based on a stored model. Factorial plots are accurately only if the model

represents the true relationship [11].

Fig. 4.5, Interaction plots for Trans -Accuracy score. Acquired from Minitab Software.

39
 Chapter 5: Conclusions and Recommendations

5.1 Conclusion

The following conclusions can be drawn from the experiment:

1) In the factorial experiments that are conducted for the assembly process, the results show

that vacuum pressure and type of vacuum gripper plays a very important role. In addition

to that, the interaction of 1) vacuum pressure and velocity 2) vacuum pressure and gripper

type 3) velocity and position and shape of part 4) gripper type, position, and shape of the

part also play an equally important role. This is supported by the ANOVA table as well.

2) Although velocity plays an important role in assembling the electronic components

surprisingly, the analysis does not prove it. From the residual plot, it was evident that

there is still room for improvement of the model by adding additional variables.

3) Optimum Settings of the robot used for the experiment are found out by using response

surface design. At the optimal settings counter-plots of accuracy score are prepared to

analyze the interaction

The following image (5.1) presents the interface of the contour of accuracy scores that were

mention above

40
 Fig. 5.1, Contour Plots of Accuracy Score. Acquired from Minitab Software.

5.2 Recommendations

A better gripping arm has to be designed by calculating the adhesion forces acting on the micro

parts. An optical sensor will help automatic pick and orientation of the micro parts. The experiment

only investigated the accuracy of the pick and place of micro objects. For a better understanding

of the process, other quality characteristics like repeatability have to be studied.

41
 References

[1] R. K. Miller, [CITATION] Robots in industry, Chapter 1, 1980.

[2] G. Granosik, The OminiTrwad Serpentine Robot for Industrial Inspection and

Surveillance (Pg. 139), Proceedings of the 1999 IEEE/RSJ International

Conference of Intelligent Robots and Systems.

[3] S-H. Han, W.H. Seo, K.S. Yoon, M-H. Lee, Real-Time Control of an Industrial

Robot Using Image-Based Visual Servoing (Pg. 1762), Proceedings of the 1999

IEEE/RSJ International Conference of Intelligent Robots and Systems.

[4] J. Wallénv, Division of Automatic Control (pg. 4), The history of the industrial

Robot 2008.

[5] NyTeknik. Stopp för Slängar, NyTeknik Automation (5), Swedish Weekly

Technical Magazine, in Swedish, October 2007.

[6] M. Švaco, B. Šekoranja, F. Šuligoj, B. Jerbić, Calibration of an Industrial Robot

Using a Stereo Vision System, (Pg. 460-461), ELSEVIER, 24th Daam.

International Symposium of Intelligent Manufacturing and Automation 2013.

[7] Y. Liu, A.W. Hoover, I.D. Walker, A Timing Model for Vision-Based Control of

Industrial Robot Manipulators, (pg. 891-892), IEEE Tansactions on Robotics,

Vol . 20, No. 5, October 2005.

[8] T. Broga˚rdh, Present and Future Robot Control Development— An industrial

Perspective (pg. 69-70), ELSEVIER, May 2007.

[9] J. Ajay, CH. Song, A.S. Rathore, CH. Zhou, W. Xu,” 3D Gates: An Introduction

–Level Energy Analysis and Optimization of 3D Printers. (Page 420)

42
[10] E.N. Malamas, E.G.M. Petrakis, M. Zervakis, L. Petit, J-D. Legat. “A Survey on

Industrial Vision System, Applications and Tools”

[11] Willrich Precion Instruments, URL https://willrich.com/product/starrett-galileo-

ez300-manual-vision-systems/, last access 5/6/2018.

[12] A. Saxena, J. Driemeyer, A.Y. Ng “Robotic Grasping of Novel Objects Using

Vision” (Page 158,171)

[13] Yamaha Industrial Robots, URL https://global.yamaha-

motor.com/business/robot/lineup/ykxg/tiny/index.html/, Last Access on 5/7/2018

[14] User Interface of Machine Vision System, URL

https://www.youtube.com/watch?v=Wy076DRFe7s/, last accessed on 5/7/2018

[15] V. Bagaria, D. Rasalkar, S.J. Bagaria, J. Ilyas, Medical Applications of Rapid

Prototyping - A New Horizon (Pg. 1, 2), Intechopen.

[16] J.P. Kruth , “Material Increases Manufacturing by Rapid Prototyping

Techniques”, CIRP Annals, Vol. 42 No. 2, Katholieke Universiteit, Leuven, 1991.

[17] K. Jensen, “State-of-the-Art of Different Available and Coming RP-Systems”,

Proceedings of the 2nd Scandinavian Rapid Prototyping Conference, Exhibition

and Course, Danish Technological Institute, Aarhus, 1993.

[18] M. Greulich, M. Greul, T. Pintat (pg. 20), Fast, Functional Prototypes Via

Multiphase Jet Solidification, Emerald Insight, Rapid Prototyping Journal, Vol. 1,

Number 1 – 1995.

[19] 3D Hubs. 3D Printing Industry Trends Q3-2016. https://www.3dhubs.com/trends.

Accessed: 12/27/2017.

43
[20] J. Ajay, Ch. Song, A. S. Rathore, Ch. Zhou, W. Xu, 3D Gates: An Instruction-

Level Energy Analysis and Optimization of 3D Printers.

[21] F. Chen, G.M. Brown, M. Song, Overview of Three-dimensional Shape

Measurement Using Optical Methods.

[22] E.R. Davies, Automated Visual Inspection, Machine Vision, Second Edition,

Chapter 19, pp. 471–502. Academic Press, New York, 1998

[23] J.L.C. Sanz, D. Petkovic, Machine Vision Algorithm for Automated Inspection of

Thin-Film Disk Heads, (pg. 830–848), IEEE Transactions on PAMI 10, 1988.

[18] H. Li, J.C. Lin, Using Fuzzy Logic to Detect Dimple Defects of Polished

Wafer Surfaces, (pg.1530–1543), IEEE Transactions on Industry Applications 30

(1994).

[24] E.N. Malamasa, E.G.M. Petrakisa, M. Zervakisa, L. Petitb, J-D Legat, A Survey

on Industrial Vision Systems, Applications and Tools, ELSEVIER, 2002.

[25] A. Goswami, A. Quaid, M. Peshkin, Complete Parameter Identification of a

Robot from Partial Pose Information, Proceedings of the 1993 IEEE International

Conference on Robotics and Automation (pg. 168-173), 1993.

[26] K. Young, C.G. Pickin, Accuracy Assessment of the Modern Industrial Robot,

Industrial Robot: An International Journal, 27(6), 427-436.

doi:10.1108/01439910010378851.

[27] A. Mehrez, M. Hu, O.F. Offodile, Multivariate Economic Analysis of Robot

Performance Repeatability and Accuracy´, Journal of Manufacturing Systems,

15(4) (pg. 215-226), 1996.

44
Appendix I- Yamaha SCARA YK180X Specifications

45
Appendix II- Vacuum Cups Used in the Experiment

46
 Appendix III- Program for the Robot

ACCEL 90 'SET ACELERATION FOR THE ENTIRE PROGRAM

DECEL 70 'SET DEC FOR THE ENTIRE PROGRAM

SPEED 60 'SET MAX SPEED THE ROBOT WILL RUN
MOVE P,P0,Z=0 'GO HOME EVERY TIME WE START CYCLE
DO2(2,1,0)=0 'TURN OFF ALL OUTPUTS
' TEST
'***INPUT AND OUTPUT DEFINITION***
' INPUT 20, 21 AND 22 ARE PROGRAM SELECTION
' INPUT 23 IS START CYCLE
' INPUT 24 IS PRESSURE OK
' INPUT 36 AND 37 ARE FOR COLOR CAMERA
' OUTPUT 20 IS FOR DONE
' OUTPUT 21 IS FOR VACUUM ON
' OUTPUT 22 IS FOR LASER ON
' OUTPUT 27 IS FOR COLOR CAMERA
*************************PROGRAM**********************************
*MAIN: 'MAIN PROGRAM
DO2(0) =1 'ROBOT READY AND IN AUTO
IF DI (23) = 0 THEN GOTO *MAIN 'WAIT FOR SET CYCLE
'THE COMMAND BCD IS INPUTS DI20, DI21 AND DI22
'COMMAND BCD 0 = RUN PICK_N_PLACE PROGRAM
IF DI2(2,1,0) = 0 THEN GOSUB *PICK_N_PLACE
GOTO *MAIN
'****************************SUBRUTINES********************************
*TURN_VACC_ON:
DELAY 100 'DELAY 100 MS
DO2(1) = 1 'TURN VACCUM ON
DELAY 200 'WAIT 200 MS
RETURN
*TURN_VACC_OFF:
DELAY 100 'DELAY 100 MS
47
DO2(1)=0 'TURN VACCUM OFF
DELAY 200 'WAIT 200 MS
RETURN
*LASER_ON:

DO2(2)=1 'TURN LASER ON

DELAY 200 'WAIT 200 MS
RETURN
*LASER_OFF:
DO2(2)=0 'TURN LASER ON
DELAY 200 'WAIT 200 MS
RETURN

'****************************PICK_N_PLACE*************************

*PICK_N_PLACE:
MOVE P, P1000, Z=0
DO2(1) = 1
MOVE P, P1001, Z=0
DO2(1)=0
DELAY 200
MOVE P,P1002, Z=0
DO2(1)=1
MOVE P,P1003, Z=0
DO2(1)=0
MOVE P,P1004, Z=0
DO2(1)=1

MOVE P,P1005, Z=0

DO2(1)=0
MOVE P,P1006, Z=0

48
DO2(1)=1
MOVE P,P1007, Z=0
DO2(1)=0
RET

49
 Vita

Vyas Kumar Naidu was born on the 6th of July in 1993 in Visakhapatnam, Andhra Pradesh,

India. The eldest son of Shankar Rao Naidu (Father) and Aruna Devi (mother). He graduated from

the JNTU University, in Kakinada, in the fall of 2014, with a bachelor's degree in Mechanical

Engineering. During the summer of 2014-2016, he worked as a trainee Engineer in Vizag Steel

Plant. By fall of 2016, he was accepted in the graduate school of the University of Texas at El

Paso, where he began a master's degree in Manufacturing Engineering. While pursuing his master's

degree, he worked as a Research Assistant in the IMSE Department, at UTEP. He also worked as

an Event staff for Whelan El Paso, Sodexo food services and the University Ticket Center at the

University campus.

Contact Information: vnaidu@miners.utep.edu

This thesis/dissertation was typed by Vyas Kumar.

50