Automating the task of tracking the delivery and receipt of fabricated pipe spools in industrial projects

Jongchul Songa, Carl T. Haasa, , , Carlos Caldasa, Esin Ergenb and Burcu Akincib

aDepartment of Civil Engineering, University of Texas at Austin, Austin, TX 78712, USA

bDepartment of Civil and Environmental Engrg., Carnegie Mellon University, Pittsburgh, PA 15213, USA

Accepted 2 March 2005. 

Available online 1 June 2005.

Abstract

Piping in industrial projects is a critical and costly process as thousands of unique pipe spools go through design, fabrication, interim processing, delivery, storage, installation and inspection. Current methods for tracking pipe spools through this long supply chain are subject to many problems. This paper evaluates the use of RFID technology as a possible solution to some of these problems through automation of the current tracking process. The technical feasibility of RFID applications is analyzed based on field tests. A model of current tracking process is presented to identify potential economic benefits from using RFID technology in automated tracking.

Keywords: Automated data collection; Radio frequency identification; Tracking; Pipe spool

Article Outline

1. Introduction

2. Related efforts and enabling technologies to automate tracking

3. Overview of field tests

4. Description and results of Phase I field tests

5. Phase II field tests

5.1. RFID system used and the testing procedure

5.2. Technical performance metrics and results of field tests

5.3. Factors influencing read rate

6. Potential benefits from using RFID in pipe spool tracking

6.1. Inefficiencies associated with the current tracking process

6.2. Potential benefits

6.2.1. Reduced time in identifying and locating pipe spools

6.2.2. More accurate and timely information on material availability and for craft work planning

6.2.3. Reduced time in searching for misplaced pipes and potential improvements on the pipe fitting schedule

7. Conclusions

Acknowledgements

References

1. Introduction

Of the elements that comprise a constructed industrial facility, construction materials and installed equipment may account for 50–60% of the total cost of a construction project [13]. Materials for a construction project can be classified into three categories: off-the-shelf, long-lead bulks, and engineered items [8]. The different categories of materials vary in cost, supply lead time, and interchangeability. Generally, engineered items are available at higher costs in smaller quantities and with more unique properties, thus requiring longer lead time and more front-end planning.

Among engineered materials, pipe spools are of particular interest to industrial projects as piping has been recognized as a critical and costly process [22]. Industrial process facilities often involve hundreds or thousands of fabricated sections of pipe spools, many of which are unique in material (e.g., cast iron), shape, finish, and other properties including final installation location on site. In a typical size industrial project with total installed cost ranging from US$200 to $300 million, there may be as many as 10,000 pieces of pipe spools [21].

Many industrial projects are executed on fast track, due to the pressing need to bring products to market fast. Given this characteristic, some industrial projects may take the opportunity to fabricate pipe spools off-site while prerequisite work is occurring on site. Several process models based on this scenario have been studied by Tommelein using the discrete event simulation approach implemented in Stroboscope [22]. In fact, piping has seen significant increase in the use of prefabrication and preassembly over the preceding 20 years [6]. However, piping in fast-track projects still poses potential uncertainty in deliveries and in completing prerequisite site work, leading to “mis-matches that foul up scheduled work sequences” [22].

Under this uncertainty, the constructors' materials managers may choose to rely on large buffers of pipe spools in an effort to secure flexibility in workable backlogs for pipe fitting crews so that they have “at least 60 percent of all pipe on site when 20 percent of the pipe had been installed” [9]. Interviews conducted as part of this study indicated that this situation is still the norm. Such large buffers of pipe spools are accumulated in a constructor's laydown yard from deliveries received 5 to 6 months prior to scheduled installation, and received pipe spools dwell in the laydown yard until pipe fitting crews file requisitions. This practice in industrial piping is comparable to the case with precast components which are often stored in the plant's storage areas, possibly as long as 6 months, until shipping to the erector [2].

When pipe fitting crews make a requisition for certain pipes, the constructor's laydown yard personnel will locate and identify the pipe spools and issue and/or stage them at the crew's work area. In some cases, they may not be able to locate pipe spools in their laydown yard within a reasonable time and have to search for the “misplaced” pipe spools. While such misplaced pipe spools may represent about 2% of all pipes for a single project [21], the constructor's search for misplaced pipe spools often requires collaboration and coordination with other project entities. If specific pipe spools cannot be located within the laydown yard, it is likely that they are in other premises, for instance, in the fabricator's storage areas. Thus, successful recovery of misplaced pipe spools would require extensive search effort across the entire supply chain.

Field materials management was identified by a recent construction technology needs assessment as one of the areas with the greatest potential for improvement and the greatest positive development impact on engineering construction work processes [23]. The fundamental function of field materials management is to assure availability of materials when they are needed for installation and make such availability information readily accessible for crew level work planning. Thus, underlying this functionality of field materials management is the capability to track materials accurately and in a timely manner. This paper is concerned with the ways in which the use of current RFID technology can be extended to tracking uniquely identified pipe spools during their delivery and receipt. Joint deployment with complementary automated data collection technologies is considered as well.

2. Related efforts and enabling technologies to automate tracking

Automated tracking of materials, equipment and labor on site has become technically more viable with recent advances in automated data collection (ADC) technologies. Among ADC technologies, Global Positioning System (GPS) technology has drawn much attention from researchers endeavoring to find effective ways to automatically track the location of construction labor and equipment on site [14], [16] and [18].

Navon and Goldschmidt showed in [14] that workers' locations can be automatically collected by GPS and converted into labor inputs based on a building project model developed by Sacks et al. [18]. Since the regulatory measure to degrade civilian GPS signals, known as the selective availability, was discontinued, state-of-the-art GPS can now provide positions of centimeter accuracy under non-stationary situations [15]. Undoubtedly, the present GPS can be used to precisely track the location of craft workers and machines over a great range of geographic and geometric scales, but the technology is still expensive for dense deployment to automate tracking individual material items. Tagging hundreds of individual pipe spools with GPS receivers for tracking purposes would be prohibitively expensive, and still other means for identification would be required.

Other enabling technologies for automated tracking include Radio Frequency Identification (RFID). Like barcodes, RFID is an ADC technology for identifying, locating, or tracking objects or assets and people, but presents several advantages over barcoding in that its operation does not require physical contact, line-of-sight, or clean environments devoid of noise, contaminants, glare and dirt. Current RFID systems consist of transponders or tags, interrogators or readers, and a host computer. Attached to host objects or people to be identified or tracked, an RFID tag carries data about the host, such as identification and item specific information or instructions, on its internal memory. A reader is a fixed or mobile device that reads and may write data to the tag through RF wireless communication when tags come within its read range (varying from one inch to 100 ft or more) and passes the data to the host computer for particular application needs.

RFID tags vary in many specifications, such as power source, carrier frequency, read range and rates, data storage capacity, memory type, size, operational life, and cost. Since their power source dictates other characteristics directly or indirectly, RFID tags are primarily classified as passive, semi-passive, or active, depending on the manner in which they derive operating power to run the digital logic on the chip and transmit the stored data to the reader [19]. With an independent power supply, active RFID tags allow greater communication range, higher carrier frequency, higher data transmissions rates, better noise immunity, and larger data storage capacity than passive tags, and are typically read/write. The trade-off is a finite lifetime (optimally, 5 or more years), and greater size and higher cost compared with semi-passive and passive tags.

While RFID technology had already seen significant beneficial applications in manufacturing, retailing, and transport and logistics industries, its potential applications in the construction industry have only begun to be explored. Jaselskis et al. [10] discussed conceptual RFID system designs to track material delivery vehicles, material-handling equipment, and the material itself in general, and for a particular case with tracking concrete delivery vehicles. Peyret and Tasky [17] applied this concept of tracking delivery vehicles using RFID technology to the case with plant mixed asphalt for quality control purposes. Production data related to a batch of asphalt were automatically collected on an RFID tag mounted on the asphalt hauling truck and transferred to the asphalt paver on site, while the position of a particular batch of asphalt being laid was provided by the GPS receiver mounted on the paver.

Jaselskis and El-Misalami [11] also reported on several pilot tests conducted to explore the application of passive (opposed to active) RFID technology in the receiving process of palletized pipe hangers and pipe supports at job site laydown yards. The pilot tests demonstrated the usefulness of the technology in receiving the unique engineered materials, but technical difficulties were encountered in that the RFID handheld reader had to be within a few inches of a tag for proper reading. Schell [20] reported a pilot test at an oil refinery plant that suggested the effectiveness of RFID technology in the pressure relief valve tracking process involving data entry of maintenance records.

For identification and tracking of construction components on site, Furlani and Pfeffer [4] developed a prototype system Comp-TRAK, based on an overall system architecture proposed by Furlani and Stone [5]. In the Comp-TRAK system, the identification and tracking of the tagged structural steel component is accomplished with different technologies, i.e., bar coding or RFID, coupled with 3D fanning laser systems. Most recently the use of RFID technology was considered for tracking construction components through a supply chain. Akinci et al. [2] and Ergen et al. [3] proposed the use of RFID technology in tracking precast concrete pieces and storing information associated with them through a supply chain.

In summary, related efforts to automate tracking construction resources have been focused on tracking the location of craft workers and equipment on site using GPS, and on identifying unique materials received at the job site using RFID. Some combination of the two technologies may prevail in the long run. On the other hand, tracking such unique materials as pipe spools using RFID was identified as one of the potential applications in the construction industry [11]. However, the ability to effectively and simultaneously read active RFID tags installed in pipe spools from longer distances (feet as opposed to inches), with minimal human efforts, and in moving platforms (flatbed trucks) under realistic shipping conditions, has yet to be studied. This ability would eliminate the need to read RFID tags individually from shorter distances using handheld readers and hence minimize associated efforts, such as knee bending [11]. This paper discusses how the use of current RFID technology can be extended to track uniquely identified pipe spools, and delineates potential benefits that are possible with the use of the technology beyond material shipping and receiving. Based on the findings from the field tests conducted, the technical feasibility of RFID technology in automating the tracking of pipe spools is discussed. A model of the current tracking process is then presented, and potential benefits from the use of the technology in the process are described.

3. Overview of field tests

In response to the compelling opportunity presented in recent construction industry research and advances in RFID technology, the FIATECH (Fully Integrated and Automated Technology) Smart Chips project, in conjunction with Shaw Fabricators and Fluor Corporation, undertook the field tests of current RFID technology. The primary objective of the tests was to determine the current technical feasibility of using RFID technology to automatically identify fabricated pipe spools and collect other information about them, such as purchase order number, in a laydown yard and through a shipping portal as part of realistic transport environments.

The field tests were conducted in two phases that span from September 2003 to March 2004, to allow a staged assessment of RFID capability in field construction applications. Phase I was intended to document technical issues and learning related to the envisioned applications of RFID technology. Based on the findings of Phase I, Phase II focused on determining the reliability of RFID technology to some statistical significance to automatically identify individual pipe spools as they pass through portal gates in typical transportation conditions.

There have been many technical limitations that prevented RFID technology from working effectively in the construction field environment in the past, though related problems have been solved for automated vehicle identification applications in transportation [12]. Technical issues addressed in these field tests that were not directly investigated in the previous research studies reported in Section 2 included: (1) the RF signal read ranges, which typically need to be longer than the current common commercial RFID applications in the manufacturing and retail industry, (2) metal interference with radio signals, which has been a problem in many common RFID applications, (3) the density of tags in a congested area, (4) the position of RFID tags relative to spool pieces and to the readers, and (5) the amount of information that can be stored on and read from the tags. In order to best assess the capabilities of RFID technology in addressing these issues, recent commercially available active (as opposed to passive) RFID systems were used during the field tests.

4. Description and results of Phase I field tests

The Phase I trials were conducted using two different types of RFID systems, equipped with handheld and fixed readers. The handheld system was used in determining the ability to read signals at long distances and around metal in manual receiving and inventory application at laydown yards. The handheld reader included a RFID reader in PC card format and an antenna that were inserted in a handheld PC and could be carried around a laydown yard or a flatbed trailer (see Fig. 1). The handheld RFID reader works with two types of active tags that operate at the same frequency but differ in memory capacity and read/write range (Fig. 2).

Fig. 1. Handheld reader unit.

View Within Article

Fig. 2. Short range tag (left) and long range tag (right).

View Within Article

For the field trials using the handheld system, fabrication shop workers placed fabricated pipes in a laydown yard and on a flatbed trailer, as they would normally do prior to shipping. Next, RFID tags were attached to 12 individual pipe spools in a variety of sizes and shapes using plastic tie wraps or double sided mounting tape. Most of the tags were positioned under large pieces, or on very congested pallets, where the reader would not be in direct line of sight and/or tag RF signals could be more difficult to reach the reader during the tests. Finally, the handheld reader was carried around about 2 to 3 ft above the pipe laid down in the yard or loaded on the trailer to collect unique ID.

The results of field trials using the handheld system indicated that current active RFID technology could function well in a congested, highly metallic environment to improve efficiency in receiving and inventory storage applications, where relatively long read range is desirable. The only difficulties in reading tags in the trials seemed to develop when either: (1) tags were fully surrounded by solid metal (e.g., placed more than an inch or two inside of a spool, or shielded completely by multiple layers of spools creating a “Faraday cage”), especially with the reader's RF power lowered, (2) or tags were placed in full contact with a surface such as flat metal plate, concrete beam, and the ground. Detailed test logs can be found in [1] that record each trial with different tag placement under varying levels of RF power.

Confirming that read distances and metal interferences could be addressed, a fixed reader system was installed on a portal structure through which a flatbed trailer could be driven, simulating a typical pipe spool transport and receiving operations (see Fig. 3). Fig. 4 shows the fixed reader to be mounted on the portal and the tag to be attached to a pipe spool. Description of both RFID systems used in Phase I is summarized in Table 1.

Fig. 3. Portal structure with four fixed readers installed.

View Within Article

Fig. 4. RFID reader and tag for the fixed system.

View Within Article

Table 1.

RFID systems used in Phase I

Full-size table

View Within Article

For the field tests with the fixed reader system (or “portal” system), 20 RFID tags were attached to fabricated pipe spools after being inspected for quality control, and were loaded on a flatbed trailer to be driven under the portal equipped with four readers (Fig. 3). In addition to the unique ID number of each tag, data, such as piece marked number, spool number, sketch number, and purchase order number for each pipe spool, had been written to the tags. The tests were conducted under presumed shipping en route to a construction site, with varying conditions; (1) the density of tags on the trailer, (2) the amount of tag data to be captured—ID only versus ID with additional data stated earlier, (3) the movement of trailer under the portal—pass through or stop-and-go at different speeds, and (4) the number of readers activated—all of the four, those two on top or side, or only one on top center of the portal. The fixed RFID system was tested in 25 truck passes under the portal gate, including 10 passes involving reading identification and other data associated with individual pipe spools.

The results of field tests indicated that it is technically feasible to use commercially available active RFID technology in automating the tracking of the shipping and receiving of fabricated pipe spools beyond simple identification, in typical transport conditions. In the field trials, ID and other information about pipe spools were captured from more tags when the trailer stopped for a short time under the portal, allowing the readers more time, if on the order of a few seconds, to read data. When reading ID and other data about pipe spools was attempted in a situation where the trailer stopped under the portal, using only one reader on the top center of the portal resulted in more tags to be unread than using multiple readers. However, when reading ID only was attempted, the number of readers did not make any difference, provided that the trailer stopped under the portal and then proceeded slowly through it. For more information on the field tests, refer to [1].

5. Phase II field tests

Phase I had targeted the investigation of many technical issues related to applications of current RFID technology in shipping and receiving the deliveries of pipe spools, and indicated that further trials would be promising. Phase II was pursued to determine the reliability of the technology in such an application that would enable automated identification of individual pipe spools as they pass through a portal, or “portal” application, to some statistical significance. Phase II field trials were conducted using a fixed reader system with the same types of tags as in the handheld system of Phase I (Fig. 2).

5.1. RFID system used and the testing procedure

For the fixed reader system to be functional, the reader was connected up to four antennas via a cable on one end, and on the other end to a host computer running software via Ethernet cable (Fig. 5). According to the vendor Identec Solutions (www.identecsolutions.com), the fixed reader can transmit/receive data at distances of 6 m (20 ft) from a short range tag or up to 100 m (300 ft) from a long range tag.

Fig. 5. Fixed RFID reader used in Phase II.

View Within Article

The general procedure for the field trials is depicted in Fig. 6. Steps 1 through 5 comprise the set up process of the overall testing procedure, and Step 6 starts the testing process for technical performance of the technology by determining the values of several parameters that form a particular set of field conditions. The test parameters, which are expected to impact the technical performance, can be divided into two categories, according to how easy it is to change their values during the field trials, as shown in Table 2.

Fig. 6. General test procedure in Phase II.

View Within Article

Table 2.

Categories of test parameters in Phase II

Full-size table

View Within Article

Parameters fall in the Static category if changing their values meaningfully would require much time and effort to affect the fabricator's tight delivery schedule. As such, each parameter in the Static category was set to a uniform value over 1 or 2 days of field tests, or “test bed.” Though limited, the relative tag positions have been changed by moving tags around spools or by reversing the traveling direction of the trailer through the portal. On the other hand, parameters under the Dynamic category are those that were given different values within a test bed, but were expected to have some constant value across a group of truck passes (or “trips”) in the same test bed. For instance, one trip in a test bed might include the same type and total number of tags as any other trips in the same test bed, but may be characterized by a different travel speed than some trips in the same test bed. Different categories of test parameters may be thought of as different levels of control that govern the technical performance of the technology.

As the truck, loaded with tagged pipes, approached and left the portal at a predetermined traveling speed (Step 9), the reader was activated and deactivated, and the read data were saved and exported to Excel to determine the number of different tags that were read in each pass (Step 10). Finally, the number of different tags read in the pass was noted along with the set of field conditions (Step 11), and the subsequent passes started with Step 6. If some tags were not read in the previous pass, Step 6 involved selectively changing some of the previous parameter values to increase the number of different tags to be read. Following the procedure described above, 4 days of field tests were completed over a month period and a total of 70 truck passes were made, as shown in Table 3.

Table 3.

Overview of Phase II field tests

Full-size table

a Indicates long range tags.
b Short range tags.

View Within Article

5.2. Technical performance metrics and results of field tests

In determining the technical feasibility of RFID technology for the portal application, the read rate is used as a metric to assess the ability to automatically identify pipe spools as the shipment departs or arrives through the portal. The read rate measures in percentage how many different tags of the total loaded are read in each pass. Since in a single pass, tags could be read more than once via any one of the active antennas, the duplicate read ratio is also defined to quantify how many times a particular tag is read in each pass at a cost of energy and redundancy. Table 4 shows summary values of the metrics resulting from each test bed which can be thought of as a sample with size being the number of truck passes. The median read rate is the ‘middle’ read rate, so exactly a half of the passes in the test bed resulted in the read rate greater than the median read rate.

Table 4.

Summary of read rates and duplicate read ratio

Full-size table

View Within Article

Test bed 3 yielded a lower mean read rate, but a higher median read rate, than other test beds, as can be seen in Fig. 7. This is due to several extreme cases (outliers) that Test bed 3 ensued as the read rates observed in Test bed 3 had the most skewed distribution. This skewed distribution may be explained by the fact that Test bed 3 underwent highly dynamic test conditions which arose from parameter values being more actively changed. Test bed 3 also resulted in the largest mean duplicate read ratio, close to seven. This means that if read at all, a single tag was read on average seven times in each pass. This high duplicate read ratio is due to the reader's RF power set to the maximum sensitivity to RF signals transmitted from tags. This increased the reader's burden to handle seven times more but essentially redundant data received from tags. This may have contributed to the low mean read rate by creating a transmission environment in which a relatively weak, shielded signal from a particular spool would be “drowned out” by the other signals.

Fig. 7. Box-whisker plot of read rates.

View Within Article

Nonetheless, in general and as expected, a higher duplicate read ratio due to higher RF power is associated with a larger number of different tags read per pass for all three test beds. This can be seen in Fig. 8 where duplicate read ratio and read rate is plotted for each pass (the slopes of the fitted line are all positive). Yet, this apparent association may not be generalized to say “the more RF power leads to the larger number of different tags to be read and hence to a higher read rate.” Indeed, Test bed 3, with RF power set to maximum, resulted in a lower read rate on average than the other test beds. In a sense, it created a higher “risk” of costly read misses.

Fig. 8. Pairwise scatter plot of duplicate read ratio and read rate per pass.

View Within Article

Another interesting observation from Fig. 8 is that the slope of the fitted line for Test bed 2 is steeper than that of Test beds 1 and 3. If slopes for each test bed are thought of as the strength of the relationship between duplicate read ratio and read rate, the steeper slope means the read rates in Test bed 2 were more sensitive to the reader's RF power. In fact, only Test bed 2 included tags with a shorter read range, while the other test beds were dedicated to longer read range tags. Thus, with a small change to the RF power, short range tags would end up with a rather large gain or loss in the read rate. Since the signal strength decays as the square of the distance, this makes sense from the perspective of basic physics [7].

5.3. Factors influencing read rate

The variability of the read rates in Test bed 3 has lent itself to further analysis, but the sample that arose from Test bed 3 may not be taken as random since the read rate of one pass is dependent on the outcome of the previous pass to some degree. This dependency stemmed from the unconscious effort to increase the read rate by selectively changing some of the previous parameter values. To alleviate this dependency structure in the sample, 38 passes from Test bed 3 were re-organized into groups that have similar values for some test parameters, as shown in Table 5 (several passes do not fall into any one of the groups). Moving from Groups I to V, the mean read rates tend to increase while the variance of read rates is decreasing (Fig. 9). This observation suggests that under the set of field conditions (table column values) classifying passes into Group IV or V, the technology under consideration is most likely to achieve 100% reading of 56 tags every time the load of pipe spools is shipped/received through the portal.

Table 5.

Test bed 3 decomposed into groups of passes

Full-size table

View Within Article

Fig. 9. Mean and variance of read rates for groups of passes in Test bed 3.

View Within Article

Whether the field test data supports the assertion that the differences of the mean read rates between groups of passes are statistically significant is answered by means of statistical hypothesis tests on the differences in mean read rates between groups; (1) Groups I and IV, (2) Groups II and IV, and (3) Groups II and III (Table 5). Noting that the number of passes is not a test parameter but rather represents a sample size for each group, these pairs of groups considerably differ in only one test parameter: reader activation timing, travel speed, or the number of active antennas. Thus, comparing each pair of groups allows determining the significance of the effect that a particular test parameter may have on the read rate. Statistical hypothesis testing is usually based on some test statistic (e.g., t or F statistic) to define a rejection region from the sample space where the null hypothesis (H0) is rejected, and hence the alternative hypothesis (H1) is accepted. A statistic used in our hypothesis tests is described below.

Let pi denote the probability that each tag will be read (i.e., the read rate) during the passes of Group i, and each tag is assumed to have the same pi in every pass under the field conditions characterizing Group i. Then the number of tags of the total 56 that are read in each pass of Group i, Yi has a binomial distribution with parameters 56 and unknown pi, or Yi  Bin[56, pi]. Further, let ni denote a sample size of Group i (i.e., the number of trials or passes, e.g., n1 = n4 = 9), and Yij be the number of tags read in jth pass of Group i. Then observed values of Yi1, Yi2, … , Yini would have yielded a random sample of size ni since (1) they arise from the identical binomial distribution and (2) every Yij is independent of one another provided that the underlying dependency between successive trials has been addressed by the reconstruction of the overall sample. Without loss of randomness, Yij can then be added up to represent the total number of tags read in passes of Group i, and approximated to a normal random variable Zi:

(1)

ΣYijBin[56ni,pi]≈ZiN[56nipi,56nipi(1−pi)].

Manipulating Zi gives us another normal random variable Zi/(56ni) ≈ N[pi, pi(1 − pi)/(56ni)], which represents the read rate for Group i (note that its mean is pi). The statistic to be used in the hypothesis tests involves standardizing Zi/(56ni) − Zj/(56nj) the difference between two normal random variables that represent the read rate of each group.

The resulting standard normal variable denoted by Zij can be used to test the null hypothesis H0: pi = pj (no difference in mean read rates between Groups i and j) against H1: pi < pj. The hypothesis tests result in H0 being rejected at a significance level α if the observed sample value of the test statistic Zij is smaller than a critical value Zα. Given α, the critical value is determined such that Pr(Z ≤ Zα)=Φ(Zα) = α, where Φ is the cumulative distribution function of a standard normal variable Z. Table 6 shows that there is a statistically significant difference in mean read rates between Groups I and IV (α = 0.01). This result suggests that the number of active antennas, which is the single factor notably different between the groups, has a significant impact on the technical performance of the portal application measured by the read rate. Similarly, the results of hypothesis testing for other groups are presented in Table 7 and Table 8, suggesting that traveling speeds of the truck also have a significant impact on the read rate. However, the duration from reader activation to deactivation does not have such a significant effect on the read rate. Note the present α is 0.05, and if it were set to 0.10, the test would result in the rejection of H0.

Table 6.

Hypothesis test on the difference of the mean read rates between Groups I and IV

Full-size table

View Within Article

Table 7.

Hypothesis test results for groups II and IV

Full-size table

View Within Article

Table 8.

Hypothesis test results for groups II and III

Full-size table

View Within Article

Based on the data obtained from field tests, statistically significant factors that can affect the technical performance of RFID technology in the portal application were found. Specifically, using four active antennas and driving the trailer at a speed 1–2 mph will allow automatically identifying all 56 tags precisely. Nonetheless, it should be noted that accountability of our statistical inference is challenged by the small sample size ni. Recall that our construction of the test statistic was based on approximation of the binomial random variable (total number of different tags read during passes of each group) to the normal random variable. The rationale behind this approximation is that exact probabilities concerning a discrete random variable are difficult to compute.

6. Potential benefits from using RFID in pipe spool tracking

The field tests indicate that current RFID technology can be used to automatically identify unique pipe spools effectively not only as they are stored at laydown yards, but also as they are shipped and received through portal gates. This result suggests that current active RFID can address the technical difficulties with the read range, the need for individual manual tag reading, and metal interference that were encountered during the previous pilot tests on material receiving at laydown yards [11]. Thus, the applicability of the technology is not limited to material receiving but can be extended to tracking of uniquely identified materials when portal systems are deployed along the supply chain.

To put this extended use of the technology into perspective, problems and inefficiencies in a model of the current pipe spool tracking process are described, and activities that RFID technology might support and improve are identified. Potential benefits from the use of RFID in tracking pipe spools are also described.

6.1. Inefficiencies associated with the current tracking process

A sample model of the overall tracking process from fabrication to job site receipt is given in Fig. 10, with three parties involved: a fabricator, a painter as a typical intermediary processor, and a constructor. This model reflects the process that was observed during the field tests and that was identified during interviews with managers and supervisors. There may be some variations, but a typical pipe spool supply chain uses processes similar to the ones described here.

Fig. 10. Sample process of tracking pipe spools and inefficiencies.

The sample process exemplifies the iterative cycle in which pipe spools are manually identified, located, shipped/received, and stored at multiple times along the supply chain. In this modeled process, potential problems and inefficiencies have been identified and grouped into three categories:

▪ Time-consuming identification and finding the location of pipe spools, as is prerequisite to shipping and receiving (marked with shaded boxes in Fig. 10). In steps 5, 13 and 25, there is a potential problem of identifying spools with those that look similar and hardly differ in 18 alpha-numeric ID numbers on the metal tags. This problem can be compounded if spools are to be located in a crowded or large laydown area. The similar problem affects steps 10 and 20, which involve verifying (‘kick and count’) the receipt of the spool pieces against the packing list.

▪ Error-prone data recording and transcription (shown as thick outlined boxes). Time-consuming verification of the receipt in steps 10 and 20 is also prone to error. When pipe spools are identified, located and verified, the packing list is produced or updated manually prior to shipping to the downstream party, as in steps 6, 14–16, and 26–28. The manual recording and transcription of shipping and inventory information are prone to error.

▪ Delayed information on shipping, receiving, and inventory (as in boxes with dotted outline). The error-prone recording and transcription of shipping, receiving and inventory information is also subject to delay. In particular, verification of the receipt in step 20 will not immediately inform managers which spools are now in receipt. Materials managers can know about that only after the packing list marked with the storage location of verified spools (step 21) is keyed into the materials control system (step 22). As a consequence of error-prone and delayed information, some of requisite spools may not be able to be located within a reasonable time, and have to be searched over an entire laydown yard and in certain cases in other parties' premises. The misplaced pipes may also have some impact on pipe fitting schedule, depending on the requisition lead time since pipe fitting is a critical path activity on a typical industrial project.

6.2. Potential benefits

The observed process of tracking pipe spools presents the problems and inefficiencies in manually identifying and locating pipe spools and collecting data on shipping, receiving, and inventory. A discussion on how this manual tracking process may be improved with handheld and portal RFID systems deployed at laydown yards and portal gates along the supply chain is included below.

6.2.1. Reduced time in identifying and locating pipe spools

The most direct, but not necessarily the most substantial, benefits are expected in verifying the receipt of pipe spools. In the current manual approach, the average kick and count time per load of one hundred spools ranges from 4 to 6 h, as per the perception of industry practitioners and as reported in [11]. Though RFID portal systems can automatically identify pipe spools with near 100% reliability as they arrive through portal gates, the industry practitioners felt that the current kick and count step would not be completely eliminated. Instead of comparing every pipe spool to the packing list, ‘kicking and checking’ will suffice that determines if the number of spools unloaded agrees with that given by RFID systems.

In addition to efficient identification, RFID technology can also help to locate pipe spools along the supply chain. The portal application of RFID technology can tell who received and shipped which pipe spools upon their arrivals and departures through a particular portal gate, thus indicating whether the pipe spools are still within the premises or not. When pipe spools within the premises need to be located for shipping to the immediate supply downstream, the handheld reader can be carried around the premises, and when entering into the read range of the spools to be located, it will indicate the proximity by triggering beeping sound and/or flickering LED on RFID tags. Integration with GPS technology may also be advantageous for this purpose.

6.2.2. More accurate and timely information on material availability and for craft work planning

Other immediate benefits from the use of RFID technology in tracking pipe spools are a function of its capability to automatically collect data on shipping, receiving, and storing inventory, more accurately and in a more timely fashion. The ID number of pipe spools would not need to be transcribed as many times as in the sample manual approach; from a metal tag to packing lists and to a computerized materials control system, to name a few. Since pipe spools are automatically identified as they are shipped and received through portals, packing lists will be rapidly and precisely generated or updated. Thus, for instance, the constructor will be able to flag pipe spools as available as soon as they arrive, without the need to wait until warehouse personnel go through the time-consuming verification and storage processes. As a result, information that certain pipe spools are in receipt can potentially be delivered 1 or 2 days earlier, allowing early start of crew level work planning at a construction site.

As the pipe spools are finally issued to the crews, inventory at the constructor's laydown yard will be updated quickly so warehouse personnel will not have to look for the spools that are no longer at their laydown yards. More accurate and timely information about shipping, receiving, and inventory will not only streamline the tracking process and improve efficiency, but also prevent spools from potentially being “misplaced.”

6.2.3. Reduced time in searching for misplaced pipes and potential improvements on the pipe fitting schedule

If pipe spools are tracked accurately and expeditiously, it not only helps decrease the probability of spools being misplaced, but also reduces search time and re-fabrication of misplaced spools. According to interviews with materials management personnel, 2% of all pipes for a single project get misplaced with the current tracking process, and the constructor's search for a single misplaced spool can take up to 24 h on average. Since the initial search in the constructor's laydown yard can sometimes be unsuccessful, the pipe fabricator may also need to join the constructor searching for the misplaced spool (in its own yard), spending one third as much as the constructor's search time [21].

The potential risk of unsuccessful initial searches may be the unwarranted by-product of massive inventory (“buffers”) of pipe spools that materials managers have built in an attempt to secure flexibility in workable backlogs for pipe fitting crews. Periodically, the search effort turns out to be unsuccessful so that the lost spools must be reproduced by the fabricator after the initial delay due to search. In addition to requiring an extensive search effort and potential re-fabrication, misplaced spools may cause delays or disruptions in pipe fitting schedules, depending on the requisition lead time since pipe fitting is a critical path activity on a typical industrial project.

7. Conclusions

In response to the compelling need to track uniquely identified materials through the supply chain, field tests of current RFID technology were conducted to determine its technical feasibility for automatically identifying and tracking individual pipe spools in laydown yards and under shipping portals. The field tests indicated that the technology could function effectively in the construction field environment involving large metal objects and requiring relatively long read range. It was also shown to some statistical significance that commercially available active RFID technology can automatically identify pipe spools with 100% accuracy and precision if they are driven at a speed less than 2 mph through portal gates equipped with four antennas.

Potential benefits from the use of RFID technology in automated pipe spool tracking may include (1) reduced time in identifying and locating pipe spools upon receipt and prior to shipping, (2) more accurate and timely information on shipping, receiving, and inventory, (3) reduced misplaced pipes and search time, and increased reliability of pipe fitting schedule. However, most of the potential benefits will be realized when the use of RFID technology is extended through construction and other stages of the project life cycle. This suggests that new applications should be developed so that they can leverage portal and/or handheld systems in other project stages. One example of such applications is locating tagged spools beyond the proximity level and tracking their location on a construction site thus providing the backbone of automated piping work progress tracking.

Acknowledgements

The authors greatly appreciate the opportunity provided by Charles Wood to participate in the field tests part of the FIATECH Smart Chips project. The support provided by the technology vendors and other participants is also deeply appreciated; John Wadephul and Don Dart, Fluor Corporation; Trent Roundtree, Shaw Fabricators; Barry Allen, IDENTEC SOLUTIONS; and Rick Pollack, Phase IV Engineering.

References

[1] B. Akinci, E. Ergen, C. Haas, C. Caldas, J. Song, C.R. Wood and J. Wadephul, Field Trials of RFID Technology for Tracking Fabricated Pipe, FIATECH Smart Chips Report, FIATECH, Austin, TX (2004).

[2] B. Akinci, M. Patton and E. Ergen, Utilizing radio frequency identification on precast concrete components—supplier's perspective, Proc. ISARC '02, IAARC, Washington, DC (2002), pp. 381–386.

[3] E. Ergen, B. Akinci and R. Sacks, Formalization and automation of effective tracking and locating of precast components in a storage yard, Proc. 9th EuropIA International Conference, EIA9: E-Activities and Intelligent Support in Design and the Built Environment, Europia, Istanbul (2003), pp. 31–37.

[4] K.M. Furlani and L.E. Pfeffer, Automated tracking of structural steel members at the construction site, Proc. ISARC '00, IAARC, Taipei (2000), pp. 1201–1206.

[5] K.M. Furlani and W.C. Stone, Architecture for discrete construction component tracking, Proc. ISARC '99, IAARC, Madrid (1999), pp. 289–294.

[6] C.T. Haas, J.T. O'Connor, R.T. Tucker, J.A. Eickmann and W.R. Fagerlund, Prefabrication and Preassembly Trends and Effects on the Construction Workforce, Report No. 14, Center for Construction Industry Studies, Austin, TX (2000).

[7] D. Halliday, R. Resnick and J. Walker, Electromagnetic waves, Fundamentals of Physics, John Wiley & Sons, Inc., New York (1997), pp. 841–862.

[8] D.W. Halpin, A.L. Escalona and P.M. Szmurlo, Work Packaging for Project Control, Source Doc. 28, Construction Industry Institute, Austin, TX (1987).

[9] G.A. Howell and H.G. Ballard, Managing Uncertainty in the Piping Function, Research Report 47-13, Constr. Industry Institute, Austin, TX (1996).

[10] E.J. Jaselskis, M.R. Anderson, C.T. Jahren, Y. Rodriguez and S. Njos, Radio-frequency identification applications in construction industry, J. Constr. Eng. Manag., ASCE 121 (1995) (2), pp. 189–196. Full Text via CrossRef

[11] E.J. Jaselskis and T. El-Misalami, Implementing radio frequency identification in the construction process, J. Constr. Eng. Manag., ASCE 129 (2003) (6), pp. 680–688. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (54)

[12] J.A. Khoury, C.T. Haas, H. Mahmassani, H. Logman and T. Rioux, Performance comparison of automatic vehicle identification and inductive loop traffic detectors for incident detection, J. Transp. Eng., ASCE 129 (2003) (6), pp. 600–607. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (2)

[13] D.U. Kini, Materials management: the key to successful project management, J. Manage. Eng., ASCE 15 (1999) (1), pp. 30–34. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (9)

[14] R. Navon and E. Goldschmidt, Can labor inputs be measured and controlled automatically?, J. Constr. Eng. Manag., ASCE 129 (2003) (4), pp. 437–445. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (17)

[15] R. Navon and E. Goldschmidt, Monitoring labor inputs: automated-data-collection model and enabling technologies, Elsevier Autom. Constr. 12 (2002), pp. 185–199.

[16] R. Navon, E. Goldschmidt and Y. Shpatnisky, A concept proving prototype of automated earthmoving control, Elsevier Autom. Constr. 13 (2004), pp. 225–239. Article | PDF (605 K) | View Record in Scopus | Cited By in Scopus (8)

[17] F. Peyret and R. Tasky, Asphalt quality parameters traceability using electronic tags and GPS, Proc. ISARC '02, IAARC, Washington, DC (2002), pp. 155–160.

[18] R. Sacks, R. Navon and E. Goldschmidt, Building project model support for automated labor monitoring, J. Comput. Civ. Eng., ASCE 17 (2003) (1), pp. 19–27. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (17)

[19] S. Sarma, D.W. Engels, On the Future of RFID Tags and Protocols, Technical Report (Auto-ID Center, Cambridge, MA, 2003).

[20] D. Schell, Oil refinery uses RFID to eliminate rework and reduce data entry by 70% (Business Solutions, 2001) http://www.businesssolutionsmag.com.

[21] J. Song, C. Haas, C. Caldas, E. Ergen, B. Akinci, C.R. Wood and J. Wadephul, Field Trials of RFID Technology for Tracking Fabricated Pipe—Phase II, FIATECH Smart Chips Project Report, FIATECH, Austin, TX (2004).

[22] I.D. Tommelein, Pull-driven scheduling for pipe-spool installation: simulation of a lean construction technique, J. Constr. Eng. Manag., ASCE 124 (1998) (4), pp. 279–288. View Record in Scopus | Cited By in Scopus (55)

[23] M.C. Vorster and G. Lucko, Construction Technology Needs Assessment Update, Research Report 173-11, Constr. Industry Institute, Austin, TX (2002).