CONTENTS. 1. Abstract Introduction Statement of the problem Significance of the problem Statement of purpose

CONTENTS 1. Abstract…………………………………………………………………………………6 2. Introduction…………………………………………………………………………….8 3. Statement of the problem……………………………………………………………...
Author: Guest
3 downloads 0 Views 2MB Size
CONTENTS 1. Abstract…………………………………………………………………………………6 2. Introduction…………………………………………………………………………….8 3. Statement of the problem……………………………………………………………..13 4. Significance of the problem…………………………………………………………...14 5. Statement of purpose………………………………………………………………….15 6. Objectives………………………………………………………………………………16 7. Definitions……………………………………………………………………………...17 8. Literature………………………………………………………………………………18 8.1 Introduction………………………………………………………………………..18 8.2 Machine Vision for identifying attributes………………………………………..19 8.3 Radio Frequency Identification…………………………………………………...20 8.4 RFID for tracking and tracing……………………………………………………21 8.5 RFID in manufacturing……………………………………………………………22 8.6 Integration of vision systems and RFID………………………………………….23 8.7 Understanding the optimization of RFID systems……………………………….24

9. Limitations and Delimitations…………………………………………………………25 9.1 Limitations………………………………………………………………………….25 9.2 Delimitations………………………………………………………………………..25 1

10. Methods…………………………………………………………………………………26 10.1 Research components……………………………………………………………..26 10.2 Equipment…………………………………………………………………………26 10.3 Methods……………………………………………………………………………27

11. Data Collection, Analysis and Interpretation…………………………………………30 12. Conclusion………………………………………………………………………………49 13. References………………………………………………………………………………51 14. Appendix I………………………………………………………………………………54 15. Appendix II……………………………………………………………………………...57 16. Appendix III…………………………………………………………………………….59

2

LIST OF TABLES

Table 1: Factors and its levels………………………………………………………………… 32 Table 2: Factors and its levels …………………………………………………………………36

Table 3: Average number of tag reads……………………………………………………….. 36 Table 4: ANOVA Table ………………………………………………………………………..37

Table 5: Average RFID Readings for factorial experiments……………………………….. 38 Table 6: Full factorial design ANOVA table …………………………………………………43 Table 7: Fractional factorial design ANOVA table ………………………………………….45 Table 8: Summary of Full Factorial and Fractional factorial Design ………………………46 Table 9: Summary of General Factorial Design ……………………………………………..50

3

LIST OF FIGURES Fig. 1: Position and placement of tags ………………………………………………………...11

Fig. 2: Mobile Mark Antennas; Dorner conveyor belt; Imaging Source Camera………… 26

Fig. 3: Block diagram of the Experimental Setup…………………………………………… 28 Fig. 4: The Experimental Setup ……………………………………………………………….29 Fig. 5: Side view of conveyor belt experimental setup ……………………………………….30 Fig. 6: Alien squiggle ALL9440-02 ……………………………………………………………31 Fig. 7: Avery Dennison – AD814 ……………………………………………………………...31 Fig. 8: Software identifying all attributes …………………………………………………….33 Fig. 9: Software identifying errors ……………………………………………………………34 Fig. 10: Surface plot to analyze the tag reads at various speeds and distances …………….37 Fig. 11: Factorial plot of the read rate for Full Factorial design ……………………………40 Fig. 12: Full Factorial Design (Main Effects) ………………………………………………...41 Fig. 13: Full Factorial Probability Plot ……………………………………………………….42 Fig. 14: RFID readings for stationary and moving belt ……………………………………..48 Fig. 15: Fractional Factorial Design (Main Effects) …………………………………………54 4

Fig. 16: Fractional Factorial Design (Main and Interaction Effects) ………………………54 Fig. 17: Fractional Factorial Design Factorial Plot ………………………………………….55 Fig. 18: Fractional Factorial Probability Plot ………………………………………………..55

Fig. 19: Fractional Factorial design Significant and Insignificant factors……………….....56 Fig. 20: Full Factorial Design (Main and Interaction Effects) ………………………………57 Fig. 21: Full Factorial design Significant and Insignificant factors ………………………...58 Fig. 22: Successful detection in test mode …………………………………………………….59 Fig. 23: No pin No OK tag No RFID tag in test mode ……………………………………….60 Fig. 24: No RFID tag No OK tag in test mode ………………………………………………..61 Fig. 25: Check Mode Success ………………………………………………………………….62 Fig. 26: Check Mode Fail ……………………………………………………………………...63 Fig. 27: Code to check for attributes ………………………………………………………….64

5

1. Abstract Machine-vision programs are used as a tool in the chain of events that will lead to making a final decision or used to make a final decision based on pre-determined attributes. A typical example would be an application involving explicitly defined single-variable pass-fail criteria where the task is to check several critical quality characteristics of an item and reject it if any of the characteristics is out of compliance. RFID systems that have interrogators (readers) are placed on manufacturing lines and conveyor lines to control and verify inventory accuracy. The read-write capability of tags on the item could be programmed to indicate the status and this feature could be used to write real-time information such as operation summary, data on critical quality characteristics and final decision on pass-fail information on the chip of the tag. The tag will then become a permanent record of the item and its history along the supply chain and information retrieval by other parties becomes possible. The problem with successful RFID implementation is the lack of application specific optimization. In a manufacturing environment the variation in product, speed of conveyor, tag position, distance of tag from antenna cannot be generalized. To gain optimum performance from an RFID system these factors should be optimized. The aim of this study is to investigate the relationship between speed of the conveyor belt, type, orientation and placement of the tag. Full factorial and fractional factorial designs are considered to determine the effect of critical factors and the interaction between them based on tag readability. Recommendations are provided for the combinations of the potential factors and their levels in order to obtain the best possible readability. This research describes a system that has been developed to integrate the vision

6

assisted fault diagnosis to continuously monitor items and record critical characteristics associated with it that can be retrieved, on a RFID tag.

7

2. Introduction Recent years have seen a rapid development in the evolution of radio frequency identification, better known as RFID. Although this technology for automated identification has been available for several decades, advances in several integrating technologies have facilitated continuous improvement. An increase in the pace of standardization and optimization techniques has helped the widespread adoption of this technology (Cassett, 2004). Much of the initial research and was focused on the implementation of RFID to increase product visibility in a supply chain. Efforts have been made to push the use of the technology further into other sectors (Cassett, 2004). One of the sectors where RFID is currently finding application is in the manufacturing sector which involves converting a raw material to a usable product and is a process which adds value to the unit being manufactured. It involves moving components from one location to another for machining, inspection or packaging. The manufactured products which reach the customer must be ensured that it is defect free. Hence the manufacture of products requires careful quality control to ensure compliance with product specifications (Godfrey, 2005). It is a common practice today to check for quality and specifications of the product prior to packaging and an inspection sheet is attached. However, this lacks the history of the component throughout its manufacturing steps. In spite of the process to check for defects before packaging is successful there are customers who still receive defective units. Thus improving the process involves change from quality control to quality assurance which checks for quality of the product at every step of manufacturing and documented in real time (Godfrey, 2005). This involves the use of a machine vision system which can help analyze the attributes like size, shape and presence or absence of a label on the unit being manufactured (Lin, 2006), (Rao, 1996), (Ravn 1994). The

8

advantages increase many fold if the vision technology integrated with RFID (Brusey, 2009), (Godfrey, 2005), (Li, 2008), (Lin, 2006), (Rao, 1996). Factory automation utilizing a machine vision system in such tasks can bring many benefits. Machine vision systems can perform repetitive tasks faster and more accurately, with greater consistency over time than humans. They can reduce labor costs, increase production yields, and eliminate costly errors associated with incomplete or incorrect assembly. They can help automatically identify and correct manufacturing problems on-line by forming part of the factory control network. The net result is greater productivity and improved customer satisfaction through the consistent delivery of quality products. RFID has great advantages than any other existing tracking technology like the bar code system because it can monitor inventory more effectively and provide real time information remotely with minimal staffing requirement. The other advantages are that the RFID tags are resistant to heat, dirt, acids and other solvents. The tags are also capable of being read from a long distance which makes RFID more suitable in many manufacturing applications [Shepard, 2005]. RFID tags are of two types – active and passive. In active systems, the tag contains a small power source that enables it to control communication with the reader continuously whereas a passive tag has no power source but it derives its power for its operation from the reader‟s radio communication signal. Due to cost limitations passive tags are generally preferred, the only limitation being the range of tag readability (Hunt, 2007). The process would then involve RFID tagging of every unit being manufactured and the data written on to the tag. To accomplish this task the tag must be on the unit at the beginning of the manufacturing process, in the absence of which an error message will be generated. Once the tag

9

is on the unit, subsequent processing and quality attributes that are critical to quality (CTQ) are identified and stored in the tag. This facilitates accept or reject criteria at each stage of the process thus preventing defective units being sent to the next stage of the process. The units that are accepted have embedded documentation in the tag which helps when any recall, traceability or tracking is needed (Symonds, 2010), (Recall Checklist, 2001), (EAN Approach) (Shanmugam, 2009). A recall is a request to return a batch of units to the manufacturer due to safety issues, tracking means following a product along the supply chain stage by stage and tracing means getting information about a specific product and is usually performed at the end of the supply chain. To accomplish this, a fully optimized passive RFID system is a required. Passive RFID is being used in a variety of applications which demand its optimization, but limited research is performed in this area (Hoong, 2007), (Huang, Wang, Chen, 2008), (Singh, Deupser, et.al, 2005). One such application is manufacturing. Although the use of RFID systems in manufacturing applications are gaining popularity, the variables such as linear distance of the tag from the antenna, tag placement, tag orientation and type of antenna being used at different conveyor speeds are not fully investigated or understood. The identification of the appropriate factors and their levels is the main problem in RFID readability (Aryal, 2009). The factors taken into consideration in this study are linear distance between the tag and antenna, speed of the conveyor, type of tag used, orientation of the tag and placement of the tag. The data is analyzed using Design of Experiments (DOE) to understand the effect of variables on RFID tag readability in manufacturing industry where the use of conveyor belts is common. A cardboard box 6 in.  6 in. 6 in. is placed on a conveyor belt and the height of the antenna from the ground is

adjusted in such a way that it is in-line with the center of the box which is moving on the conveyor belt. The conveyor is operated at two different speeds to simulate a real time 10

manufacturing environment. The other factors investigated are, the horizontal distance from the tag to the center of the antenna, tag type from two manufacturers, orientation of the tags, and the placement of tag (Fig. 1).

A:Horizontal tag orientation, in front of antenna B:Vertical tag orientation, in front of antenna C:Horizontal tag orientation, parallel to the movement of conveyor D:Vertical tag orientation, perpendicular to the movement of conveyor

Fig. 1. Position and placement of tags

In this research 2k factorial design is used to analyze the data to identify the main and interaction effects of the factors that were investigated (Montgomery, 2009). This approach is used to draw conclusions from the data gathered from the experiments. For comparison purposes and to create baseline information, additional data was gathered by keeping the tagged box stationary in front 11

of antenna for the travel times that corresponds to different speeds. Data is analyzed using Statistical Analysis Software (SAS) (SAS Institute, 2004). This analysis helps identify the factors has no significant effect on the tag readability.

12

3. Statement of the problem Manufacturing industries usually face the problem of generating waste in the form of scrap or re-work during manufacturing and also tracking the unit along the supply chain. The receiving customer may also be interested in knowing the history of the unit that is being received. Most of the manufacturing companies implement Lean methodologies to reduce waste. However, it might be impossible to detect imperfections if it undergoes several operations during the manufacturing process. If a part has a defect in the initial stages of manufacturing, it may go undetected until the final checking or at the testing stage. This wastes time and resources on machining and additional remedial steps that may be necessary to correct the defective unit. Moreover, if a defective product reaches a customer then tracing the history of the item becomes virtually impossible. Hence integration of Machine Vision with RFID system would help to address both these issues simultaneously. Research has been performed on integrating tagging systems with vision systems (Godfrey, J.W., 2005) and (Brusey, J., 2009) but not much published research is available on the application of the integrated technology in manufacturing industry. Thus the aim of the research is to integrate the RFID and Machine Vision technologies to provide a solution which can minimize or eliminate the defect and tracking related issues during and after manufacturing. The design used in this research is quantitative, experimental with extensive literature search, learning vision technology and the use of statistical techniques to analyze the experimental data.

13

4. Significance of the problem Many key tasks in the manufacture of products, including inspection, orientation, identification, and assembly, require the use of visual techniques. Human vision and response, however, can be slow and tend to be error-prone either due to boredom or fatigue. Replacing human inspection with machine vision can go far in automating manufacturing operation, but implementers need to carefully match machine vision options with application requirements. Moreover different tasks may require different vision attributes. Inspection requires an ability to examine objects in detail and evaluate the image to make pass/fail decisions. Assembly, on the other hand, requires the ability to scan an image to locate reference marks (called fiducially) and then use those marks to determine placement and orientation of parts. The results from this research can help a manufacturing organization becoming a lean organization by detecting defects at the point of origin during manufacturing using the vision system. This helps to save time, money and other resources by preventing subsequent manufacturing on a defective product. The RFID tags help provide information about the manufactured product with information like place of manufacture and details of testing which can help tracing the product with ease in case of defect. The equation developed by the factorial design helps to solve optimization problems of implementing RFID in a manufacturing environment.

14

5. Statement of Purpose The purpose of this experimental research is to reduce costs by identifying defects at every stage of a manufacturing process and eliminating manual tracking and tracing related issues within and beyond the organization boundaries by integrating Machine Vision and RFID systems.

15

6. Objectives To identify the problems involved in the process of product tracing and tracking and to update information from vision system onto the tag.

To identify the problems involved during manufacturing like checking specifications before the product is sent for subsequent manufacturing, thereby reducing costs from manufacturing on a defective product.

To provide an integrated solution to solve multiple issues mentioned above.

Develop a regression model for predicting the RFID read rate

16

7. Definitions Antenna: The physical object that transmits RF waves and receives RF waves returning from tags. Defect: A product which does not comply with specifications. Electronic Product Code (EPC): The numbering scheme for storing data on an RFID tag. The EPC number is the equivalent to an eventual replacement for UPC barcodes for most general supply chain applications. Gen 2 (Generation 2): The next step of evolution of RFID tags. They work better than generation 1 tags, and work more reliably. Machine Vision: A processing system which includes camera, a signal processor which is connected to a middleware and is capable of capturing images and processing them as well. Optical Character Recognition (OCR): An algorithm which is used to identify characters optically.

17

8. Literature Review 8.1 Introduction In today‟s world there are many competent technologies which are available to check for the presence of defects in a manufactured object. The most often used method is the Laser technology. But there are no mechanisms available which can feed the pass-fail information about the product on to the item history automatically. Therefore, the item or the product history is not retrievable at any stage of the process or after the product reaches the customer. This process also makes it difficult to trace back the product to its manufacturer. Hence the process of identifying defects using various cost effective technologies is studied. The research this discusses about the use of sophisticated technology to reduce defects and cost, track and trace units with ease. This section would involve the following process of identifying the advantages of the chosen technology and how the accomplishment of the integration can help the manufacturing industries. The optimization of the tracking system is also studied because of the availability of only standardized process and not optimized processes.

18

8.2 Machine Vision for identifying attributes Laser technology is one of the most efficient ways of compliance checking, the disadvantage being that the laser is not capable of identifying color differences and shades. Hence, the use of vision technology is one of the sophisticated methods of process control. Today the vision systems are capable of identifying, color, shape, texture and depth which are far superior to its predecessors (Rao, 1994). Today‟s manufacturing environment is completely based on quality improvement, cost reduction, increased volume, increasing complexity of parts and shorter cycle times for manufacturing (Rao, 1994). Rao (1994) in the research discusses about the improvement of the image resolution by understanding the use of color, depth and texture and their integration. The research also discusses the use of gray scale images for most of the applications and that algorithms have been developed for enhancing the gray scale images using non-conventional imaging modes. The advantage of using a vision system is to reduce manual interception to the manufacturing process. The vision system uses less than a second to detect defects in any unit irrespective of the number of attributes it measures in each process which can take a few minutes for detecting manually. The challenges of the deployment of a vision system include lack of standardization and lack of appropriate images when identifying defects (Rao, 1994). Another recent publication includes the use of vision system to identify serial number on the beer keg (Lees, Campbell, Keir, 2007). The research identifies the existing solution for identification like Radio Frequency Identification (RFID) and Bar codes but are not cost efficient. Hence, the machine vision is used to identify the characters on the beer keg. The beer keg has a serial number attached to it which is identified using machine vision (Lees, et.al, 2007). The Optical Character Recognition (OCR) capability of vision systems is explored to 19

identify the serial numbers on the beer kegs. This methodology helped the organization achieve a recognition rate of 97%. The challenges faced by the organization is classified in to two types, pre-defined due to physical properties of the kegs and some defined in consideration of the operational requirements within the context of the a production line (Lees, et.al, 2007). Lees et.al (1994) identifies the key characteristics and constraints to be measured. The illumination is one the most important criteria taken into consideration and the capture area is defined. This research takes the machine vision application to new dimensions by identifying two methodologies i.e. Template matching and neural networks to understand the better of the two methods. The Neural networks provide a better overall recognition rate of 92%. A technique similar to the neural network methodology is embedded in the software to identify and detect the characters. Hence, machine vision is a comprehensive tool which can help identify defects in a manufacturing environment. 8.3 Radio Frequency Identification Radio Frequency Identification (RFID) is defined as a set of technology that provides network delivered information services dependent on physical object identification captured by radio waves. Applications of RFID have been promoted in several fields and have been used in areas of transportation, distribution, and manufacturing, processing and packaging (RFID study Group, 2006; Singh, Deupser, Olsen and Singh, 2007). A typical RFID system consists of four main components, a tag (transponder), a reader (interrogator), an antenna and middleware (computer). When the tag is in the electromagnetic field of RFID antenna the tag‟s presence is detected and the relevant information stored in the tag is reflected back to the antenna to be processed by the reader. The reader is interfaced to the middleware using a network cable which displays the number of reads on a screen. This entire process can be completed with no human intervention 20

hence RFID system can be truly automatic. Hence this is the most preferred methodology of tracking and tracing objects within and beyond the organization‟s borders. 8.4 RFID for tracking and tracing The EAN Approach (n.d.) discusses the product identification and traceability using the bar coding technology. The document provides information on the procedures for identifying and tracing product, their specifications, their status and knowledge of exact content of products that have been delivered. The bar coding methods of identification and traceability include coding requirement, machine readable language, encode data, quality control, apply markings, decode data, transmit data and use the data. The document also discusses about a fresh produce supply chain tracking and traceability model which discusses the methodology of tracking using block, batch, product and packer ID and tracing using invoice, delivery slip, consignment note delivery advice and record of receipt. Certain businesses call for a high level of traceability to track high value products (Symonds, Parry 2008). In the paper on tracing high value products using RFID in a health care environment, the use of RFID is being considered to the bar coding technology. But the use of RFID in a health care environment would require a lot of considerations like standards, cheaper alternatives and regulations. Symonds and Parry (2008) discuss the supply chain strategies used by the health care facility and the methodology to be used to identify and collect data. The time considerations about how RFID can perform tasks quicker than the bar code technology and cites Hou and Huang (2006) who measured that bar codes take 33 minutes to identify 1000 items whereas RFID takes 1 minute and 40 seconds. The benefits of RFID based on line of sight and information reliability is studied. Shanmugam (2009) combines RFID with Product Life cycle Management to solve multiple business issues. The product life cycle issues are divided into several categories of which product recall, tracking and tracing issues and 21

early identification of product failures as some. Understanding the PLM and integrating with RFID would provide cost benefits of 600 billion pounds for European food and drink industry estimates a CIAA report. A number of areas where difficulties are found is first identified and classified. Right part at the right place at the right time, Service and maintenance, Asset management are some of the areas where RFID implementation can make a lot of difference (Shanmugam, 2009). The integrating RFID and PLM in a manufacturing environment ensure reduction of rework, ability to track and trace objects, complete automation of the process. Brusey and McFarlane (2009) discuss the use of RFID for object tracking and tracing by pattern recognition techniques to identify aggregated objects. 8.5 RFID in manufacturing The current scenario would require the use of RFID in a manufacturing environment. Li, Li, Gao and Fan (2008) discuss the intelligent manufacturing and RFID. Intelligent manufacturing technology refers to simulating the human experts manufacturing activities using computers in a completely automated environment. Integration of RFID and a concurrent manufacturing model can reduce the response time to customer inputs (Li et.al, 2008). Li (2008) discusses a method integrating a dynamic database and RFID to help customer make changes midway through a manufacturing process. Brewer and Sloan (1999), in the paper, Intelligent tracking in manufacturing exemplify the use of RFID in a manufacturing environment. Dynamic scheduling is an important new innovation in manufacturing and supply chain management which can provide real-time information. The paper also understands the other tracking techniques like Global Positioning System (GPS) and other wireless technologies. A pilot test at an electronic manufacturing unit was conducted to trace and track a particular type or lot of printed circuit boards. The problems faced include making the employees understand that the use of RFID can 22

help reduce workload and increase the efficiency of the process. The tasks that go into designing the system include developing the software, installing the hardware and software the system and specific knowledge in each area. 8.6 Integration of vision systems and RFID Godfrey and Bonney (2005), in their patent Compliance Tracking Method provide a methodology for automatically tracking compliance in a manufacturing environment which involves selecting a test unit, defining attributes to be tested on the unit, on which manufacturing operations are performed, writing the pass-fail information on a memory and the same steps performed on successive manufacturing operations. The main aim of the method is to shift the process from quality control to quality assurance by making the entire compliance tracking system automated by the use of tagging techniques. RFID systems have been used to track the movement of objects to which RFID tags have been attached. In this invention the RFID tag is affixed to an object which is to be tracked. The RFID system is used as a trigger mechanism to a vision system which determines whether the object is moving towards or away from a particular position. Another similar invention is the integration of video surveillance system with the RFID tracking system. The calibration of RFID tracking system is enhanced by the use of information provided by the video surveillance system. Moreover the calibration of the vision system is enhanced by the information of the RFID system. RFID systems are calibrated by placing RFID tags at visually apparent locations to determine appropriate correction factors for use in subsequent RFID locations. This invention provides the advantages of RFID tracking system and the video surveillance system to overcome the disadvantages of either systems or both (Lin 2006).

23

8.7 Understanding the optimization of RFID systems RFID system implementation always requires that the system be optimized to every application. For optimizing the RFID system experiments need to be performed to identify the placement, location, orientation of tag antenna and other factors which interact with the system. The importance of optimization is explained by Hoong (2007) and Ammu, Mapa and Jayatissa (2009). Ammu (2009) identifies the effect of distance between tag and antenna and performs design of experiments to analyze the resulting data. The main measurement is the number of reads for every experiment and its repetitions. The difference between the levels of each factor is also analyzed using statistical analysis software. The best placement and location of the tag is identified using the factorial approach. This approach gives an application specific optimization. Similarly Hoong (2007) performs paired t-test and design of experiments to identify the optimum placement of the tag.

24

9. Limitations and Delimitations 9.1Limitations 

The position of the tag must be in a place where no machining should take place.



The temperature of the manufacturing environment must be within the operating temperature range of the tag.



The lighting around the unit being manufactured should be adjusted for the machine vision unit, which might not be possible in some situations.

9.2 Delimitations 

The current scope involves only the manufacturing industry.



There are many techniques for identification but the most feasible would be the use of RFID taking into consideration the type of environment where the unit is being manufactured.

25

10.

Methods

10.1 Research components The research follows a quantitative design that focuses on problems associated with manufacturing and encoding critical quality characteristics in a manner that could be useful to the end customer. This will include various steps in a manufacturing process, where dimensions or other attributes of the product being manufactured are needed by customers. The research would have components like data collection, data analysis, optimum tag location, camera and antenna placement and other equipment consideration. A simulation generated will show the process flow with the unit moving from one manufacturing activity to another. 10.2 Equipment

(i)

(ii)

(iii)

Fig. 2. (i) Mobile Mark Antennas (ii) Dorner conveyor belt (iii) Imaging Source Camera The RFID reader used for performing the experiments was manufactured by Motorola Technologies. Reader configuration was Electronic Product Code (EPC) class I Gen2, model XR 450. The objective of the EPC is to provide unique identification of physical objects. This is used to address and access information about individual objects from the computer network, similar to

26

the internet protocol (IP) address allows the computers to identify, organize and communicate with one another. The antenna used for data collection was Mobilemark Communications circularly polarized 915MHz, PN7-915 series (Fig 2 (i). The tags used are Class 1 Generation 2 manufactured by Alien with model Squiggle 9640-02. These tags have a user programmable space of 512 bits which can help write information from the vision system on to the tag. Gen2 tags are the second generation of tags used for better performance. The conveyor belt used for this study is manufactured by Dorner 7400 series (Fig 2 (ii)). The camera is monochrome manufactured by The Imaging Source with model number DMK 21BUC03 (Fig 2 (iii)). A Netgear gigabit Ethernet switch is used to interface the RFID and the Machine Vision equipment. 10.3 Methods The methods of integrating RFID with vision technology include data collection, data analysis and simulation of the data obtained as well as the process involved. A two way communication between the RFID equipment and the Camera system has to be established. This enables data and control information to travel within the integrated system (Fig. 3 and Fig. 4). Once the data communicates between the reader and the camera processor, further experiments are to be performed to understand the optimal experimental setup. The placement of the antenna and the reader and the placement of the tags on the unit being manufactured are of primary importance. The other factors needed to be taken into consideration are the distance between the camera and the unit being manufactured, lighting involved for obtaining images, speed of the conveyor on which the unit will move to be read by the RFID reader.

27

Fig 3. Block diagram of the Experimental Setup

28

Fig 4.The Experimental Setup Experiments were conducted by varying the factors like speed of conveyor, distance between the unit with tag and camera and the distance between the unit and the RFID reader and antennas. A simulation of the layout will be generated using simulation techniques to better understand the process. Various levels of machining are done on the unit before it is sent out for packaging. The RFID with Machine Vision system is placed between each of the levels of machining. At each level the unit is checked for attributes by the Machine Vision system and the pass/fail information in binary form is sent out to the RFID system to be written on the tag. If a „fail‟ is written on the tag, a signal is sent to a programmable logic controller circuit to reject the product to prevent further machining of a defective unit.

29

11.

Data Collection, Analysis and Interpretation

Data was collected in the laboratory with minimal interference from external factors such a metal objects and signals from other RF sources. This was confirmed using Airview™ Spectrum Analyzer which detects the presence of any stray signals between 895-935 MHz ranges in the laboratory. Any potential sources of interference are isolated using static shields. The data was collected at different placement of the tag on the package, orientation of the tag, speed of the conveyor, distance between tag and antenna and different tag type. To ensure the reliable effect estimates of each factor the distance between the low (-) and high (+) level is chosen as appropriate as possible. The distance between the antenna and the tag is given by linear distance (LD) at 2 inches and 12 inches as shown in Fig. 5.

Fig. 5. Side view of conveyor belt experimental setup The position of the antenna with respect to the package remains constant in the entire experiment. The speed taken into consideration is 10ft/min and 50 ft/min. The two types of tags used are Alien Squiggle ALL 9440-02 (Fig. 6) and Avery Dennison AD 814 (Fig.7) which were positioned vertically and horizontally. 30

Fig. 6. Alien squiggle ALL9440-02

Fig. 7. Avery Dennison – AD814 The other factor is the placement of the tag on the package, in the middle or on the top. In all the experiments the factors are given very high priority to make sure that every combination of experiment is performed as per factorial design. A factorial design is a type of statistical method to determine effect of various factors in different levels which includes all possible combinations for all factors and its levels (Montgomery, 2009). There are five factors each at two levels and hence the experiment is 25 factorial experiments. It helps in analyzing the data set and also the interaction between the factors and its levels. The experiment is performed with the factors and their levels as listed in the Table I.

31

Table 1. Factors and its levels Factors

Designation

Low(-)

High(+)

A

2

12

B

10

50

Avery Dennison

Alien Squiggle

Horizontal

Vertical

On package facing

On top of the

antenna

package

Distance from tag to antenna (in.) Speed of conveyor belt (ft/min.)

TagType C

Tag orientation D Tag Placement

E

In this research a test unit is subjected to several simulated activities on its surface with different attributes like presence and absence of tag, size, color and position of the critical characteristic are incorporated as shown in Fig. 1. The unit used in this research is a cardboard box (6in x 6in. x 6in.) where the presence of the RFID tag is identified. Presence of three light colored pins and a dark pin are segregated; the positioning and size of a circular cut out and an OK label are recorded by the vision software. The camera used provides images in gray scale; hence it is important to understand the contrast between colors. Although the software used with the camera is programmed to identify specific colors, in this research only the gray scale was used to contrast between colors in identifying the attributes.

32

Determining each attribute would involve the following steps in the configuration mode of the vision software, (1) identifying the presence of tag, pins and position and location of circular cut out and OK label (2) creating a window around the area where the presence of the attribute is assumed to be found (3) define the attribute to be measured with specifications (4) test run the code and (5) debug the code if any errors are found. The vision system displays a success message if all the attributes are present as shown in Fig 8.

Fig 8.Software identifying all attributes

33

Fig 9.Software identifying errors An error message is generated if more than one or absence of an RFID tag is detected and fails to identify other attributes in the defined window. The RFID tag is identified irrespective of the position and placement provided the tag is present within the defined window. The system identifies the three light colored pins and generates an error message if less than or more than three pins are identified. The colored pin is not considered as an error indicating that the vision system can differentiate between shades. The center of the circular cut out is to be placed at a distance of six centimeters from the center of the unit. The camera provides numerical values in terms of pixels, so the equivalent of six centimeters of linear distance is measured in terms of pixels and the nominal value with tolerance is fed into the code. The code for each process is executed at their respective stages of the process or multiple attributes can be measured at each process step. The attributes are measured and checked using the check mode in the vision 34

software. The pass-fail criteria are shown on the image by two shades. The darker shade (black and red) in Fig. 9 indicates that the product failed to meet the required specifications and a lighter shade (green) in Fig. 8 indicated that the unit can proceed to the next stage of manufacturing process. Other possibilities of errors and the software in Run mode can be found in Appendix III. The window where the product attribute is defined is highlighted with the shapes. The same information can be obtained as one bit information which is stored in a register. The register is an option given in the software where data can either be input to the code or the output from the vision system can be extracted and saved as a file or interfaced to other system. In this research the data is extracted from the register and is interfaced to the RFID reader by saving it as a text file. A zero (0) indicates that the product failed to meet specification and a one (1) means the product is eligible for the next level of operation. At the final level OK label is affixed which is also detected using the character recognition capability of the camera. Errors are introduced manually to simulate a real-time manufacturing environment and the results gathered. Once the vision system generates the pass-fail information the entire focus shifts to the RFID system. The first point of contact of the data in the RFID system is the reader which receives the information through the TCP Ethernet cable. The data is processed by customized RFID software to convert the one bit data into a 24 bit tag ID field, for example, the data received is either zero or one (0/1) which is then converted to a 24 digit tag ID like 00000000000000000000000(0/1). For simplicity purposes the data will have the 1st digit as zero or one (0/1) from the vision system and the rest of the digits as zero. The antenna receives the data from the reader which wirelessly writes the information on the tag. The RFID system has fast response time and hence the data which is written can be read immediately by the same system confirming the data transfer. 35

Design of experiments is performed to determine the optimal distance between the antenna and tag and the speed of the conveyor. Here, a four level two factor general factorial design is performed to analyze the data. The factors and its levels are shown in Table 2. The average of the number of reads from the experiment can be seen in Table 3. Table 2 Factors and its levels Factors\Level

Level 1 (-1)

Level 2 (-0.3333)

Level 3 (0.3333)

Level 4(1)

Speed (ft/min)

10

30

50

70

Distance (in.)

6

12

18

24

Table 3 Average number of tag reads Speed(ft./min)\Distance (in.)

6

12

18

24

46.

53.

57.

61.

8

2

6

2

14.

18.

10

20. 19

30 6

8

2

10.

11.

12.

2

2

6

7.2

8.2

8.8

8.8

50

6.6

70

36

Table 4 ANOVA table Source

DF

SS

MS

F

Pr>F

Distance

1

436.8

436.8

5.1

0.0274

Speed

1

21815.3

21815.3

252.4

0.0001

Model

2

22252.1

11126.1

128.7

0.0001

Error

77

6655.45

86.4

211.2

0.0001

(Lack of fit)

13

6503.85

500.3

(Pure Error)

64

151.6

2.4

Total

79

28907.6

Fig 10 Surface plot to analyze the tag reads at various speeds and distances

37

The data is analyzed using Statistical Analysis Software SAS™ (Sweeney, 2005) and the ANOVA table is generated as shown in Table 4. This table helps to analyze whether the factors and error is significant or not. In this case, it is seen that the distance is having a p-value of 0.0274 which is greater than the critical value of 0.005, hence not significant, whereas the other factor, speed is significant. In this experiment the distance between the antenna and the tag does not significantly make a difference in the number of tag reads. The surface plot in Fig. 10 shows that the trend is similar for any distance but the number of reads increase at 10 ft/min and the number of reads fall significantly at the other speeds. The purpose of the present study is to determine the effect of critical factors and the interaction amongst them based on tag readability. The read rate produced by the various combinations of the variables is summarized in Table 5. The combinations are listed in the Yates‟s order (Montgomery, 2009). Table 5 Average RFID Readings for factorial experiments Run Combination

Factors A B C D E

Average reads

1

(1)

-

-

-

-

-

48.4

2

A

+

-

-

-

-

20.8

3

B

-

+

-

-

-

8.6

4

AB

+

+

-

-

-

4.2

5

C

-

-

+

-

-

53.2

6

AC

+

-

+

-

-

55.6

7

BC

-

+

+

-

-

10.0

38

8

ABC

+

+

+

-

-

11.0

9

D

-

-

-

+

-

46.8

10

AD

+

-

-

+

-

43.6

11

BD

-

+

-

+

-

8.0

12

ABD

+

+

-

+

-

8.6

13

CD

-

-

+

+

-

70.8

14

ACD

+

-

+

+

-

70.2

15

BCD

-

+

+

+

-

13.2

16

ABCD

+

+

+

+

-

13.6

17

E

-

-

-

-

+

26.0

18

AE

+

-

-

-

+

44.2

19

BE

-

+

-

-

+

5.4

20

ABE

+

+

-

-

+

8.4

21

CE

-

-

+

-

+

54.6

22

ACE

+

-

+

-

+

60.8

23

BCE

-

+

+

-

+

10.4

24

ABCE

+

+

+

-

+

11.4

25

DE

-

-

-

+ +

18.8

26

ADE

+

-

-

+ +

36.4

27

BDE

-

+

-

+ +

3.2

28

ABDE

+

+

-

+ +

7.0

29

CDE

-

-

+

+ +

57.4

30

ACDE

+

-

+

+ +

68.6

39

31

BCDE

-

+

+

+ +

10.0

32

ABCDE

+

+

+

+ +

12.6

The factorial plot of the subject data is displayed in Fig. 11 below. In the figure, -1 represents the low level and 1 represents the high level. From this we can infer that when factor B (speed) is in the negative level the read rate becomes high.

Fig. 11. Factorial plot of the read rate for Full Factorial design

40

Fig 12. Full Factorial Design (Main Effects) Fig. 12 is a Main effects plot which is a diagrammatic representation of the most significant and the least significant factors. From this we can see that there is a significant drop in the number of reads when factor B (Speed) is changed from low level (10 ft/min) to the high level (50 ft/min).

41

Fig 13. Full Factorial Probability Plot The probability plot shows that most of the data lies around the normal value indicating that the data is normal. From the above figure we can also identify possible outliers which lie far from the center. The plots for the Fractional factorial design are available in Appendix I. Interaction plot shows the interaction between all the variables which is available for both Fractional and Full factorial designs in Appendix I and II. The complete analysis of variance (ANOVA) for full factorial 25 design is given in Table 6. Notice that second and higher order interactions are considered as error.

42

Table 6 Full factorial design ANOVA table Source

DF

SS

MS

F-value

P-value

A

1

162.0

162.0

7.33

0.0075

B

1

62133.8

62133.8

2812.76

0.0001

C

1

9378.9

9378.9

424.58

0.0001

D

1

486.5

486.5

22.02

0.0001

E

1

412.8

412.8

18.69

0.0001

AB

1

41.0

41.0

1.86

0.1752

AC

1

41.0

41.0

1.86

0.1752

AD

1

166.0

166.0

7.51

0.0068

AE

1

1410.2

1410.2

63.84

0.0001

BC

1

4378.6

4378.6

198.22

0.0001

BD

1

278.3

278.3

12.60

0.0005

BE

1

178.5

178.5

8.08

0.0051

CD

1

288.9

288.9

13.08

0.0004

CE

1

120.7

120.7

5.46

0.0207

DE

1

770.0

770.0

34.86

0.0001

Error

144

3181.7

(LOF)

16

2888.9

(PE)

128

292.8

Total

159

83428.9

22.09

43

The above ANOVA table clearly shows the significant factors. Any P-value which is below the 0.05 level is significant. The only two in the above table which are not significant is the interaction between Factors A and B and Factors A and C. The regression model for predicting the read rate is given by

yˆ  28.806  1.006 x1  19.706 x 2  7.656 x3  1.744 x 4  1.606 x 5  0.506 x1 x 2  0.506 x1 x3  1.019 x1 x 4  2.968 x1 x5  5.231x 2 x3  1.319 x 2 x 4  1.056 x 2 x5  1.343x3 x 4  0.0869 x3 x5  2.194 x 4 x5 Where the coded variables xi , i  1,2,3,4 ,5 represent the factors A, B, C, D and E respectively and xi x j , i  j represent the interactions. Residuals are obtained as the difference between observed and the predicted read values and then the model adequacy is checked using these residuals. Since the number of factors considered in the subject study is large it is more likely that the system may be primarily driven by some of the main effects and low order interactions. In order to investigate this, half fractional factorial design is used. The complete ANOVA table from the study is given in Table 7.

44

Table 7 Fractional factorial design ANOVA table Source

DF

SS

MS

F-value

P-value

A

1

31.25

31.25

14.49

0.0003

B

1

27306.05

27306.05

12663.68

0.0001

C

1

5281.25

5281.25

2449.28

0.0001

D

1

756.45

756.45

350.82

0.0001

E

1

92.45

92.45

42.88

0.0001

AB

1

0.05

0.05

0.02

0.8794

AC

1

296.45

296.45

137.48

0.0001

AD

1

6.05

6.05

2.81

0.0988

AE

1

328.05

328.05

152.14

0.0001

BC

1

2714.45

2714.45

1258.88

0.0001

BD

1

806.45

806.45

374.01

0.0001

BE

1

4.05

4.05

1.88

0.1753

CD

1

54.45

54.45

25.25

0.0001

CE

1

6.05

6.05

2.81

0.0988

DE

1

530.45

530.45

246.01

0.0001

Error

64

138.00

2.15

Total

79

38351.9

The regression model for predicting the read rate using the half fractional factorial design is given by

45

ˆ  27.975  0.625 x1  18.475 x 2  8.125 x3 y  3.075 x 4  1.075 x5  0.025 x1 x 2  1.925 x1 x3  0.275 x1 x 4  2.025 x1 x5  5.825 x 2 x3  3.175 x 2 x 4  0.225 x 2 x5  0.825 x3 x 4  0.275 x3 x5  2.575 x 4 x5 Where the coded variables xi , i  1,2,3,4 ,5 represent the factors A, B, C, D and E respectively and xi x j , i  j represent the interactions. Residuals are computed as the difference between observed and predicted read values and then the model adequacy is checked using these residuals. Table 8 below summarizes the results of using full factorial and fractional factorial design. Table 8 Summary of Full Factorial and Fractional factorial Design

Factorial Design

Fractional Factorial Design

Mean

28.80625

27.975

R-Square

96.19%

99.64%

Adj. R-Square

95.79%

99.56%

RMSE

4.700584

1.4684

CV

16.3179

5.2490

To determine the better design of the two, the root mean square error values (RMSE) and the adjusted R-squared values for full factorial and fractional factorial designs are compared. A design with less RMSE value and higher R-square value is the most suitable among the two

46

designs. Minimizing the error is the objective of any experimental design. The RMSE value was improved from 4.70 to 1.46 from full factorial to fractional factorial designs. The adjusted coefficient of determination, adjusted R-square, which refers to the proportion of variation in y accounted for by independent variables was improved from 95.79% for a full factorial design to 99.56% to fractional factorial design. These values indicate that the model best used for reading is the fractional factorial design. Baseline experiments are performed keeping the box stationary in front of the antenna to represent the speed of the conveyor. Instead of the conveyor speed, the corresponding time taken by the box to cover the distance (5 feet) is taken. Fig. 14 compares the average readings at different treatment combinations with respect to the motion of the conveyor belt. It is clear that the average read rate are consistently low when the conveyor belt is in motion.

47

RFID Readings: conveyor belt Motion VS. Stationary 80

Motion Stationary

A verage number of reads

70 60 50 40 30 20 10 0 3

6

9

12

15 18 Runs

21

24

27

30

Fig. 14. RFID readings for stationary and moving belt This graph helps to analyze the loss of efficiency when the tag is stationary and when the tag is moving along the conveyor. Each of the 32 experiments performed in both setups is run with 5 replications having total of 160 runs for analytical purposes.

48

12.

Conclusions

The attribute pass-fail information on the tag can be read by any other passive RFID readers capable of reading tags at UHF (895-925MHz) range. This confirms the universal use of this technology. Thus the product information can be retrieved at any point of time along the supply chain. Durability test using RFID tags suggest that RFID tags can last longer provided the tag is under the operating limits. Hence the tagged unit is capable of providing relevant information like manufacturing location, product information and testing attribute pass-fail information whenever it is read. Thus, the tagging of the unit using RFID serves two purposes (1) storing pass-fail information to prevent defects from moving to the next stage of manufacturing or to the customer and (2) tracing back a unit to its manufacturer and the process step becomes easy if the unit is considered faulty. The images shown in the Appendix III shows that the project is successful with successful detection of defects and subsequently writing the information on the tags. The code used for detecting the defects is given in Appendix III. Tests were performed under various lighting environments and the images provided were obtained from a particular light setup. The camera must be reprogrammed based on the type of lighting and the type of attributes it has to identify in any particular setup. Table 9 gives the Root Mean Square Error Value (RMSE) and the Adjusted R-Square value of the test experiment performed with 4 different distances and 4 different speeds. The RMSE value is 9.29 and the adjusted R-squared value which indicates the adjusted coefficient of determination is 76.38%.

49

Table 9 Summary of General Factorial Design Mean

22.93

R-square

76.98%

Adj. R-square

76.38%

RMSE

9.29

CV

40.55

The research indicates that integrating vision and RFID can assist in continuous improvement of product quality, tracking and traceability along the supply chain. Factorial and fractional factorial designs can be used to analyze the tag detection in the presence of multiple factors. Factorial designs can also be used to test other possible factors at different levels in packaging industry where conveyor belts are commonly used. The ANOVA table for the full and fractional factorial designs shows that the main factors are significant. The interactions between distance and speed of conveyor, distance and tag type are not significant in full factorial design. On the other hand, fractional factorial design shows that the distance and speed of conveyor, distance and tag orientation are not significant. Higher coefficient of determination for fractional factorial design indicates that the fractional factorial design is more appropriate than the full factorial design. The model adequacy checked using the residuals indicate that the fractional factorial design performs better than the full factorial design with half the number of runs. The model developed here can be used to minimize failures, cost of implementation and predict outcomes with ease. It can be observed that in selecting an RFID system for a manufacturing application it is important to conduct design of experiments to identify the critical variables for optimum tag readability. 50

13.

References

Ammu, A., Mapa, L., Jayatissa, A.H., (2009). Effect of Factors on RFID Tag ReadabilityStatistical Analysis.IEEE International Conference on Electro/Information Technology. Aryal,G.,MapaL., Tsokos,C.,(2009), Read rate Analysis of RFID systems using mixed models, Neural, Parallel & Scientific Computations, 17, 339-349. Brusey, J., McFarlane D.C. (2009). Effective RFID-based object tracking for manufacturing. International Journal of Computer Integrated Manufacturing 22(7), 638-647 Godfrey, J.W. &Bonney, S.G. (2005).Compliance Tracking Method.US 6,839,604 B2. Washington D.C.: U.S. Patent and Trademark Office. Retrieved from http://www.uspto.gov Hoong, E.C.M. (2007). Application of Paired t-test and DOE methodologies on RFID Tag Placement Testing using Free Space Read Distance, IEEE International Conference on RFID. Huang,C.,Lo,L., Wang,W., Chen,H.,(December 2008). “A study for optimizing the reading rate of RFID tagged cartons in palletizing process,” IEEE International Conference on Industrial Engineering and Engineering Management,1138-1142 HuntV. D. (2007), RFID A Guide to Radio Frequency Identification, John Wiley Kator, Corinne (2008), RFID basics. Modern Materials Handling, Vol.63 no. 2, 38-40.

51

Lees, M., Campbell, D. and Keir (2007). A. Machine Vision for Beer Keg Asset Management, Mechatronics and Machine Vision in Practice, 125-137 Li, Z., Li, F., Gao, L., Fan, Y. (2008). Concurrent Intelligent Manufacturing based on RFID.Advanced Design and Manufacture to Gain a Competitive Edge. Lin, Y. (2006). Integrated RFID and Video Tracking System.US 6,998,987 B2. Washington D.C.: U.S. Patent and Trademark Office. Retrieved from http://www.uspto.gov MontgomeryD.C. (2009), Design and Analysis of Experiments, John Wiley Motorola XR450 Fixed RFID Reader Specifications available from the World Wide Web, Office of the Secretary, (1979).The Belmont report.The National Commission for the Protection of Human Subjects: Retrieved from http://www.hhs.gov/ohrp/humansubjects/guidance/belmont.htm#xbound Panel Mount Antennas, MobileMark Communication Antennas, Product identification and traceability (EAN Approach).Uniform code council Inc. and EAN international. Retrieved from http://elsmar.com/pdf_files/Product%20Identification%20and%20Traceability.doc Recall checklist, January 2001. Retrieved from http://www.cpsc.gov/businfo/recallcheck.pdf Rao, A.R. (1996). Future directions in Industrial Machine Vision: a case study of semiconductor manufacturing applications. Image and Vision Computing.

52

Ravn, O., Andersen, N.A., & Sorensen A.T. (1994).Auto-calibration in automation systems using vision.Lecture notes in Control and Information Sciences 200, 206-218 RFID study group at Pennsylvania State University (2006), Challenges in RFID Enabled Supply Chain Management, Quality Progress, 39(11) Shanmugam R. (2009). PLM-RFID combined Solutions to Solve New Business Issues. Tata Consultancy services. Symonds, J.A. & Parry, D. (January-March 2010). Using RFID to track and trace high value products: the case of city healthcare. Journal of cases on Information Technology Singh, J., Deupser, C., Olsen, E., and Singh, S.P. (2007). An Examination of the variables Affecting RFID Tag Readability in a Conveyor Belt Environment, Journal of Applied Packaging Research,2(2), 61-73 SAS Institute Inc. (2004). SAS Online Doc 9.1.3. Cary, NC-SAS Institute Inc. ShepardS. (2005), RFID Radio Frequency Identification, McGraw Hill Networking Professional SweeneyP.J. (2005), RFID for Dummies, John Wiley

53

Appendix I

Fig 15. Fractional Factorial Design (Main Effects)

Fig 16. Fractional Factorial Design (Main and Interaction Effects)

54

Fig 17. Fractional Factorial Design Factorial Plot

Fig 18. Fractional Factorial Probability Plot 55

Fig 19. Fractional Factorial design Significant and Insignificant factors

56

Appendix II

Fig 20. Full Factorial Design (Main and Interaction Effects)

57

Fig 21. Full Factorial design Significant and Insignificant factors

58

Appendix III

Fig 22.Successful detection in test mode Test mode is where the software is written to test for imperfections in the code used to identify the attributes. Here we can see that the code works perfectly detecting all the attributes successfully.

59

Fig 23.No pin No OK tag No RFID tag in test mode Test mode is where the software is written to test for imperfections in the code used to identify the attributes. Here the RFID tag, OK label, 1 pin is absent which is seen clearly by the red color in the figure above. The Circular cut-out is present hence successfully detected.

60

Fig 24.No RFID tag No OK tag in test mode Test mode is where the software is written to test for imperfections in the code used to identify the attributes. Here the RFID Tag and the OK Label is not present, the rest is successfully detected.

61

Fig 25.Check Mode Success The process is automated in the check mode which checks for the attributes repetitively. Here the attributes are detected successfully hence the Actual result is Success which can be seen in the right side of the figure under Process.

62

Fig 26.Check Mode Fail The process is automated in the check mode which checks for the attributes repetitively. Here all the attributes are not detected hence the Actual result is Error which can be seen in the right side of the figure under Process.

63

Fig. 27. Code to check for attributes This is the code which was used to first identify the image followed by providing pixel information to locate an attribute based on contrast information and location and also by providing the count to limit the number of attributes it detects.

64