Tool Management in Construction: An Initial Comparison of Accuracy and Scan Times for RFID and Bar Code Systems

Introduction

Research Problem

When categorized from the perspective of how tools are identified, there are basically three types of tool management systems: manual, bar-code labels, and radio frequency identification (RFID) tags.  The manual system requires a person to visually identify an individual tool.  This is usually accomplished by observing a unique number that has been somehow secured to each tool.  A bar-code system relies on a label affixed to the tool and a hand-held bar-code reader.   A RFID system relies on an attached RFID tag and a tag reader that can be programmed to automatically sense the presence of RFID tags within its range of operation.  The purpose of this study is to measure and compare the accuracy of identification and the time it takes to identify tools for bar-code and RFID systems.

This study is significant because it compares the newer technology, RFID, to the older technology, bar code.  Overall, the components that are substantially different between these two types of tool management systems are 1. the data source (labels or tags), 2. the actual data acquisition (readers) and 3. the ability of RFID systems to be programmed to periodically scan the environment for the presence of RFID tags.  This study will only test to see if there is a significant difference between the scan times and scan accuracy of these two types of systems.  This is important because any measured differences in these two parameters could influence the relative economic value of the systems to a construction business.

Variables

The dependent variable is the System Type. This variable has two possible conditions, RFID or Bar Code.  The first independent variable is Scan Time measured in seconds. This is the elapsed time required to scan a set of subject tools.  The next variable is Accuracy, the percent of tools correctly identified.

The last variable is Location and is related to the shelf on which the tool is placed within the tool storage box (commonly called a gang box).  In this experiment it has three possible conditions: Top, Middle and Bottom.  This last variable was included to see if the tool location within a gang box would have any effect on the ability of the RFID scanner to find and identify an individual tool. There was some concern that the metal gang box itself or the distribution and quantity of metal tools within the box might interfere with the radio signal used by the RFID system to retrieve data from the RF Tags.  This variable only pertained to the RFID system.

Hypotheses

There is no difference between scan times for the two systems.

There is no difference in accuracy between the two systems.

There is no Location effect on RFID system accuracy.

Literature Review

In January, 2004 an RFID research project was conducted in Houston, Texas involving Fluor and FIATECH (Collins, RFID Journal). The study objective was to determine whether RFID could help automate the shipment and delivery of pipe spools from fabrication plants to construction sites.

In this study between twenty and thirty pipe spools were loaded on a truck. RFID tags were then attached to the pipe using lanyards. There were two types of tag readers used in the experiments: a hand-held reader and a data portal. The data portal was built so that a fully loaded truck could drive through it simulating the condition that might be found at a fabrication plant from which pipe would be shipped to a construction project.

The study found that in every case the hand reader successfully read the tags 100 per cent of the time. The researchers also said that after some experimentation using the portal, they came up with a method that resulted in 100 percent accurate readings. The successful technique that evolved required stopping the truck while it was in the portal. The truck stopped only for a moment and then started. Some of the tags were not properly identified on a few of the trials where they did not stop the truck.

The researchers concluded that both the hand-held reader and the portal reader accurately identified the pipe spools. They also thought that the portal reader was most appropriate for this application.

Researchers also said that the method of attaching the tag to the pipe spools was critical to accurate material identification. Mounting the tag inside the pipe or on a bracket welded to the underside of the pipe did not work well. The most accurate readings occurred when the tag was attached to the pipe with a lanyard thus allowing it to swing freely.

This pipe-spool study suggests the location of the RFID tag could have an influence on the ability to correctly identify the item. The researchers speculated that this was due to the fact that the objects to be identified were metallic and this could interfere with the radio waves used to communicate information between the RF tag and the RF tag reader. Therefore RFID tag attachment to tools and the tool location within a metal gang box could affect the ability of a RFID tool-management system to correctly identify tools.

Method

Sample Description

The population of interest for this study is the tool inventory at the tool management facility of Zachry Construction in San Antonio, Texas.  This population is further limited by tool size to those that might be placed in a gang box for use at the construction site.

A gang box (a large tool box for a work crew) was loaded with a representative set of tools. Nine of these tools were randomly selected as subject tools and had RFID tags attached.  A second set of nine similar tools was placed on a counter. A bar code reader was set up to scan these tools.  Results from this study can only generalized to the Zachry tool management facility and its Houndware tool management system.

Instrumentation

Description

The gang box was fitted with an active RF tag reader.  The reader was attached to electrical power and the computer network.  A bar code reader was setup to scan the second set of nine similar tools that were placed on a counter.

Each observation consisted of scanning all nine tools in either the RF Tag set or the Bar Code set. One observer recorded the elapsed time of each observation. An observer also recorded the number of tools correctly identified by location within the gang box.   A scanner, RFID or Bar Code, was connected to a Houndware tool management system during data acquisition.  A total of 47 observations were made: nine for the Bar Code set and thirty-eight for the RF Tag set.

Procedure

RF tag tool set.

Bar code set.

 Results

Scan Times

Plot

Descriptive Statistics

t-Test

System Average Scan Time per Tool

The average scan-time for correctly identifying a tool with an RF Tag was 1.66 seconds per tool. The average scan-time for correctly identifying a Bar Coded tool was 4.92 seconds per tool.

RF Tag Tool Identification

There were 38 observation attempts of 9 subject tools. Out of these 342 attempts there were only seven instances when the system failed to correctly report all tools.  This yields a raw success rate of 98 per cent.

Of the nine subject tools, only two failed to be correctly identified. One was a beveling torch and the other a 5-ton hydraulic ram. The system failed to correctly scan the beveling torch 6 of 38 attempts. The system failed to correctly scan the hydraulic ram 1 time out of 38 attempts.  The beveling torch was located on the middle shelf and the ram on the top shelf of the gang box.

Discussion

Scan Time

The plot gives a strong visual impression that the average scan-time for the RF Tag system is much lower than for the Bar Code system.  Since the ranges never overlap on even a single observation, it also suggests that this difference will be significant even though the sample size for the Bar Code system is small (only nine observations). The descriptive statistics offer no evidence contradicting the plot. The t-test strongly confirms that the average time to scan nine tools for the RF Tag system (14.6 seconds) is significantly less than the average time to scan nine tools for the Bar Code system (44.2 seconds).  On a per tool basis, the RF Tag system required an average of 1.66 seconds to correctly identify each tool. The Bar Code system required 4.92 seconds. This means it takes three times as long for the Bar Code system to scan a tool as it does the RF Tag system. Consequently, the authors reject hypothesis one that states there is no difference between scan times for the two systems and conclude that the average RFID scan time is less than the average Bar Code scan time.

RF Tag Tool Identification

The evidence suggests that the Bar Code system is marginally more accurate.  However there are mitigating circumstances that may call this result into question.  One factor is that under the existing experimental setup, the only way to determine if a tool is properly identified in a scan sequence is through the reporting feature of the tool-management software itself. Another factor is that the RF Tags in the experiment have internal settings that determine their reporting period. The reporting period is the time interval between successive reports to the tag reader.

Since the researchers wanted the system to work as fast as possible, they set the RF Tags to the lowest value available, 16 seconds. This meant that every 16 seconds each tag would send a signal identifying itself to the reader. Under normal field operations a user would not want this value set so low.

The software vendors theorized that under these conditions, the few times the system failed to properly identify a tool were due to longer data base update time requirements rather than RF Tag identification problems. Since under normal operating conditions this would not occur, they believe the accuracy of the RF Tag system would be at or near 100 per cent.  The authors find that the study is inconclusive regarding the accuracy difference between the two systems and that there is any location effect on RFID accuracy. 

Study Implications

The evidence in this study suggests that there may be some time advantage to using RF Tags rather than Bar Codes for tool management. The advantage of bar codes is that the labels are cheap and can be attached to tools without concern for safety issues. Additionally, bar code labels do not interfere with operation of the tool. One disadvantage is that they are easily damaged or removed.

RF Tags are much more expensive although they can be used over and over if not damaged. On some tools it may be difficult to attach RF Tags due to tool shape, use or size. The active tags used in this experiment are about as big as a domino, much larger than a bar code label.

RF Tag durability and attachment effectiveness are other concerns. In the end, however, RF Tag technology holds the promise of increased productivity of tool-room personnel, reduced capital investment in tools and generally better information regarding tools in support of managerial decisions.

Future studies may want to more thoroughly investigate the relative accuracy of these two systems.  They may also want to do detailed studies regarding the effect of tool shape and tool location on system accuracy.

References

Collins, Jonathan; Case Builds for RFID in Construction, RFID Journal, January, 2004.