Utility-Locating Technology Development Using Multisensor Platforms (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

90 C h a p t e r 4 Conclusions The purpose of this research was to bring together and develop creditable nondestructive geophysical techniques on a land- based, towed platform capable of detecting and locating under ground utilities under all geologic conditions. The R01B research project resulted in the development and improve- ment of advanced technologies (detection sensors, analysis software, and work procedures). These technologies contrib- ute to a multisensor approach that offers subsurface utility engineers and geophysical service providers the best chance to completely and accurately detect, locate, and characterize sub- surface utilities at any location across the United States and the world. However, even with improved technology, the nec- essary resources and technologies would not likely be deployed to a site without a cost-benefit analysis that considers potential project delays, safety issues, and cost overruns that could occur if utilities are not effectively identified and located. Thus, tech- nological advances in locating and characterizing utilities must be accompanied by complementary improvements in manage- ment and procedures to allow this technology to be used effec- tively. In fact, the management and funding of efforts to locate utilities, the required training, and the prohibitive cost of imple- menting effective operations are as much factors in prevent- ing the effective use of advanced technologies in the field as the technology’s limitations. Recommended System Deployment Strategy Based on the in-service testing and experience outside the R01B program, the recommended deployment of multi- sensor geophysical systems is as follows: 1. Perform SUE Level D. 2. Perform SUE Level C. 3. Perform SUE Level B, using a. Pipe and cable locators; b. 2-D GPR, as appropriate; c. 3-D GPR, as appropriate over areas of more complex utility networks or where unknowns are expected; d. TEM, as appropriate—but cover nearly every site as completely as possible due to low cost, fast coverage, ability to detect “hard to tone” utilities such as cast iron and ductile iron; e. Final interpretation of all system data to produce a combined map of all targets; and f. Chosen sites for test holes. 4. Perform test holes as in SUE Level A at selected sites: a. Use test hole data to refine depth parameters for GPR, if GPR was performed; b. Refine GPR depth mapping based on test hole data; and c. Produce final mapping. 5. If desired by DOT or engineer, use all data produced to construct 3-D modeling of utilities. Multichannel GPR and SPADE Software Several GPR technologies are available that range in their abil- ity to detect targets. A common technology used for SUE appli- cations today is single-channel GPR. These systems generally use a grid of data collection transects spaced at some interval greater than the detection footprint of the GPR antenna; as a result of the relatively coarse spacing of these transects, very little information about the subsurface is generally obtained using this technique. By contrast, the multichannel 3-D GPR system (TerraVision) used by UIT contains 14 antennas similar in capability to those commonly employed in single-channel systems. These antennas are spaced approximately 4 in. apart, resulting in much higher data density than can be reasonably achieved with a single-channel system (this would require a single-channel operator to perform individual transects at a spacing of 4 in. to obtain equivalent data density). By collecting adjacent swaths of multichannel data, it is possible to obtain coverage over 100% of accessible areas of a project site. These data can generally be collected more rapidly than single-channel data, reducing field time and project costs while also increasing Conclusions and Suggested Research

91 data density. SPADE is the sophisticated software package designed to assist data analysts in the visualization and inter- pretation of these GPR data sets. The R01B project resulted in several enhancements, which were applied to SPADE and offer data analysts proven methods to improve the efficiency and accuracy of geophysical analysis. The segmentation algorithms and threshold methods developed through this R01B research offer a radically differ- ent way of analyzing GPR data. The potential improvements from a more intuitive selection process are large—it is thought that presenting data as a set of forms and surfaces will allow rapid identification of features and rejection of noise. The algorithms that were tested are efficient at identifying the majority of features present in the data. A number of image processing algorithms can be applied to remove ringing artifacts. A good summary of these can be found in the article, Removal of Ringing Noise in GPR Data by Signal Processing (Kim et al. 2007). However, in UIT’s TerraVision data the ringing can lead to saturation, at which point image data are irretrievably lost. Sagentia has recommended that UIT address the causes of ringing in the sensor hardware. Sagentia has produced an improved method for quickly assessing the migration parameters for GPR data and an improved version of the 3-D migration algorithm. The devel- opments also include a method for performing operations on a chosen subset of GPR data that allows for faster computer calculations for heuristic parameter selection. Those param- eters can then be applied to the whole data set just one time. A new 3-D migration algorithm fixed some older shortcomings and enhanced the processing speed and accuracy of the migra- tion calculation. Figure 4.1 illustrates this newly developed functionality. These changes are already being implemented in SPADE and are proving to be a big help in improving both the speed and quality of GPR data interpretations. High-Frequency Seismic Imaging This component of the R01B research focused on proof-of- concept ideas. The soil seismic properties investigated under a SHRP 2 subcontract (OES/Psi-G Task II.A, Soil Seismic Properties and Testing) showed S-wave and P-wave velocities to be in reasonable agreement with data presented previously (Lew and Campbell 1985). That work was based on soil prop- erties evaluations performed at a large number of California field sites. The SHRP 2 soil properties tests were designed to yield S-wave and P-wave velocity and attenuation profiles at three regional field sites: Manteno, Illinois; Houston, Texas; and Manassas, Virginia. These results show the S-wave veloci- ties to be in the range 200–300 ft/s at 1-ft depth, increasing to the range 800–1,400 ft/s at 12-ft depth (Figure 3.28). The P-wave velocities at these sites were in the range 450–700 ft/s at 1-ft depth, increasing to the range 1,700–3,600 ft/s at 12-ft depth (Figure 3.30). The velocity gradients were found to be accurately described by fractional power-law regression curves with depth exponents of 0.521 for S-waves and 0.586 for P-waves. The corresponding quality factor (Q) profiles at these three sites were found to be relatively constant at equal mean values of about QS = QP = 20, with variations in the range QS = 10–27 and QP = 12–28. The attenuation rates cor- responding to these Q values are als = 2.73–1.01 dB per wavelength and alP = 2.27–0.975 dB per wavelength, respec- tively, with approximately equal mean values of 1.36 dB per wavelength. These summary velocity and attenuation values and ranges are generalized results derived from recorded field data con- taining noise and spurious resonance effects in the measure- ment system. For some of the test sites, the measurements were physically limited to shallower depths than planned. Therefore, the indicated values are only approximate; and in some cases, they are only inferential with respect to the deeper depths. While the quantitative accuracy of these results for each specific test site cannot be determined, the similarities in S-wave and P-wave velocity profiles and their depth trends at the five independent borehole test locations tend to validate the analyses and interpretations. For guidance in any further use of these field results, an empirical accuracy bound of ±10–15% is assigned to the results presented in this report. The soil seismic properties at each regional test site were used to predict the SH-wave reflection system pipe detection capability for pipe sizes of 3-in. diameter and larger. The seis- mic system operating characteristics were specified to be existing state-of-the-art capabilities with an effective system dynamic range of 120 dB. The system performance model was formulated in combination with the soil seismic proper- ties derived from the regional field tests. The absolute receiver detection threshold was determined to be approximately -138 dB below 1 g peak acceleration source radiation, giving the recorded signals a reasonable signal-to-noise ratio at the system threshold signal recording level. The two-way attenu- ation loss in the regional soils dominated the model-derived detection depth limits. Table 4.1 summarizes the maximum detection depths for pipes in the range 3-in. to 10-in. diameter using a two-octave seismic vibrator sweep frequency range of 400–1,600 Hz. These results are based on accurate alignment of the incident SH-wave polarization parallel to the pipe axis. The results show that, for the soil conditions at four of the five soil test boreholes, 3-in.-diameter pipes are only detectable at depths of about 6 ft below the ground surface. The soil at Houston Borehole D2 is the exception, with 3-in.-diameter pipes detectable at depths about twice as deep as at the other sites. These detection performance results are qualified by the accuracy of the derived soil seismic properties and the fact that the threshold detection intercepts are at or near the rec- ognizable resolution of the reflected seismic waves (Table 4.1).

92 Table 4.1. Maximum Detection Depths for Pipes in the Range of 3-in. to 10-in. Diameter Regional Test Site Maximum Pipe Detection Depth (ft) for SDR  120.4 dB 3 in. Dia. 4 in. Dia. 5 in. Dia. 6 in. Dia. 8 in. Dia. 10 in. Dia. Manteno B2 6.5 8.2 10.2 11.5 14.6 17.1 Houston D2 11.7 16.6 >17 >17 >17 >17 Houston E1 6.3 8.4 10.5 12.7 >17 >17 Houston F2 6.0 7.9 10.3 13.0 >17 >17 Manassas H1 6.0 7.4 8.8 10.2 13.4 16.0 Figure 4.1. Previewing migration with “hyperbola overlay” filter.

93 In practice, the realizable detection performance will gen- erally be less than that indicated in Table 4.1. For, example, the slopes of the curves in Figures 3.32 through 3.34 for pipes smaller than about 6 in. in diameter are high enough that, when a 10-dB detection margin is allowed, the maximum detection depths decrease by about 0.5–1 ft, depending on the pipe diameter. On this basis, 3-in.-diameter pipes are detectable at depths of about 5.5 ft, and pipes 6 in. in diame- ter or larger can be expected to be detected at depths down to about 10–12 ft in the tested soils and in soils similar to those at the three regional field sites. Seismic Modeling Software The achieved objectives of the work at Louisiana Tech Uni- versity related to the R01B research work were (a) to develop finite difference time-domain (FDTD) software capable of simulating propagation of acoustic and elastic waves in a real- istic soil, (b) to study the properties of acoustic and elastic wave propagation through a soil and, (c) to analyze the physi- cal consequences of the obtained results. On the basis of the obtained results, Louisiana Tech was supposed to look into the possibility of developing a virtual testing laboratory for the simulation of acoustic and elastic methods of detecting buried pipes and conduits. Limited resources and time did not allow for full develop- ment of the FDTD software. Although LTU developed, tested, and applied several in-house computer models in different situations, the software chosen for simulating the acoustic and elastic wave propagation in soils was the open source Wave Propagation Program (WPP) from the Lawrence Livermore National Laboratory (Center for Applied Scientific Comput- ing 2011). The software is offered “as is,” and users are free to modify it. The WPP software had options for variable wave speeds, attenuation coefficients, and soil densities throughout the computational domain. While the WPP code was used in large-scale seismic cases, it had never been used at the scale of applications of the R01B project. Very detailed studies of the WPP were performed at LTU, including the performance of the code in the high-frequency range, up to the limits allowed by the capabilities of the Louisiana Optical Network Initiative (LONI) system of supercomputers. It was found that the soft- ware is adequate for the scope of the R01B project. Studies of the properties of acoustic and elastic wave propa- gation through a soil and characteristics of the reflections from the buried pipes were performed using both experimen- tally obtained data and theoretically possible waves. It was found that there were no limits in applications of the WPP assuming a realistic situation. The physical consequences of the obtained results were analyzed, and some of the results are presented in this report. LTU is in a position to fully develop a virtual testing labo- ratory for the simulation of the acoustic and elastic methods for the detection of buried pipes and conduits. Such a virtual lab would be very beneficial in designing the devices for the detection of buried objects by reducing the development cost. While not completely finished, some work was also put into novel methods for signal processing. It was shown that the method would have capabilities of real-time signal analysis. TDEMI Technology The time-domain electromagnetic induction (TDEMI) sys- tem prototype has been developed to operate advanced trans- mit (Tx) and receive (Rx) coil pairs at variable time gate intervals on a towed, nonmetallic instrument platform. The Tx and Rx parameters are fully programmable, which is use- ful in optimizing the detection and location of underground metallic utilities. The system was designed after SAIC’s cur- rent time-domain EMI array, which was developed collabora- tively with the U.S. Naval Research Laboratory. During this research the TDEMI array was adapted for data processing and analysis procedures to support data import and manipulation within Geosoft’s Oasis Montaj geophysical software package. Target classification and discrimination of underground utility anomaly sources remain important objectives of the proposed research. UIT has fully assembled the system and conducted a series of system bench tests, as well as demonstrated the TDEMI prototype’s functionality during in-service testing. This testing has indicated that the TDEMI prototype offers an improve- ment to detection of small metallic utilities compared with current TDEMI digital geophysical mapping systems. This is a direct result of the early time gate measurements that can be achieved with the prototype system. Several other aspects of the system, however, require further investigation and improve- ment; these include system ruggedization and the consolidation of hardware components, system integration with laser-based positioning systems, a more detailed assessment of bistatic measurement(s) capabilities for target characterization, and problems associated with the time step variations. Suggested research System Deployment Improvements All of the geophysical systems experience difficulties when working over terrain that is other than flat. First, there are issues with working on uneven surfaces, including over curbs; these sometimes contain metal—such as pipe and cable locators—that can confuse or provide spurious signals to electromagnetic systems. On irregular terrain, GPR systems sometimes experience apparent noise due to air gaps beneath

94 the antennas. Sometimes obstacles, such as vegetation or engineered structures, exist above the utilities being mapped. On steep slopes, all of the systems produce signatures that must be mapped perpendicular to the surface, not directly (vertically) beneath the spot on the surface where the signa- ture is indicated. The rough surface issues must be handled in the field, with careful use of the tools by experienced operators; and for the most part, they are straightforward. Obstacles simply provide challenges that must be considered in the project planning; in the case of vegetation, the site can be mowed, which is often done. The steep slope issue can be handled in two ways. First, if a digital terrain model is available, those data can be used to recompute depths to targets—or accurate surface locations in the case of pipe and cable locators. The second way to han- dle slopes is to include an inclinometer or other method of determining slope on an inch-by-inch basis as data are col- lected. This solution applies to cart-based systems; it has not been deployed for regular use by anyone as yet but is an area of future development. Currently, consultants and contrac- tors handle these issues on a case-by-case basis, using solu- tions appropriate for the situation. An additional issue is the need to provide utility elevations, not just depths. The geophysical systems that provide depth information need to be augmented with a method of measur- ing current surface elevation so that elevations of utilities can be computed. UIT currently does this by obtaining a digital elevation model from the project engineer’s surveyor. When the geophysical cart-based systems use high-quality RTK or laser robotic survey devices, good quality surface elevations are available, but only on those areas surveyed by the geo- physical cart. In the future, lidar scans of the surface will be used to obtain elevations over the whole site. That is another area in which development is needed. Multichannel GPR and TDEMI Technology UIT’s work experience and the results obtained from the in- service testing indicate that the use of both GPR and TDEMI technologies offers a more complete assessment of the sub- surface condition. The horizontal locations of subsurface utility targets can be determined with relative ease using these methods. However, depth information on these targets is a more challenging aspect of the geophysical investigation. With some study of the local soils for calibration, the depth to the pipe can be measured to a tolerance of a few inches with the GPR data. Since the TDEMI system’s response has amplitude affects that depend on the distance of the sensor from the pipe, it would be valuable to have a technique for find- ing the depth to the pipe independently. Further study of GPR signals may also yield information about the pipe itself (e.g., the corrosion state of rebar in bridge decks is estimated on the basis of variation in the amplitude response of GPR signals). Aside from depth estimation capability, GPR and TDEMI technologies may provide a realistic approach for pipeline assessment from the ground surface. For TDEMI sensors, pipeline detection should give a very consistent instrument response if materials, depth to the pipe, soil type, and pipe condition are the same along the mapping zone. When any of those conditions change at any section of the mapping zone, the geophysical anomalous signature is expected to change as well. As mentioned, the TDEMI prototype developed through this research was built on a TDEMI system designed for find- ing UXO targets. Theoretical analysis indicates that a system set up differently—to focus on measuring the response that occurs from interaction of the signals with the surface of the pipe—could be used to more accurately examine the pipe’s condition. One suggestion would be to begin research with a newer, more programmable TDEMI that may offer clues to subsurface pipe conditions. Seismic Modeling Software The ultimate goal is the modeling of multisensor platforms consisting of the electromagnetic technique and the acoustic technique. The project team hopes that acoustic code devel- oped during the R01B project will be used along with electro- magnetic code to study the efficacy of multisensor platforms. The data collected individually from both codes may provide target signatures for carrying out the development of data fusion algorithms. More development of the postprocessing program will occur in future WPP scenarios. Presently, the postprocessing program is hard-coded to deal with specific names of files and would have to be recompiled to be adapted to the output of future WPP scenarios. If the postprocessing program needs to be shared with other users, or if it begins to be applied to many different WPP scenarios, the program could be adapted to use a configuration file to specify tasks and program options. An easy implementation of this feature is possible via boost pro- gram options, a library in the Boost C++ libraries. The code in the postprocessing program is largely pro- cedural and localized to a few functions. Large portions of the procedural code could be unified into more generic sub- routines, and following from that, the code would not be so localized to a small handful of functions. Documentation would be easier to maintain and create if the code were reduced. Doxygen, an automatic documentation generation program, is used for what little documentation presently exists. The pres- ent documentation is no better than the weak documentation in WPP’s MATLAB postprocessing scripts. As the demand and/ or opportunity arises, these deficiencies should be dealt with.

95 SPADE Workflows Three potential workflow options for UIT have been assessed for efficacy and time-efficiency: depth-slice analysis, time- slice analysis, and migration analysis. Further controlled evaluation of these methods could lead to the optimal data analysis procedures used for subsurface utility investigations. The narrative below provides a brief explanation of these methods. Depth-Slice Analysis This is the workflow currently used by UIT. An operator scrolls through vertical depth slices, picking the peaks of hyperbolae. Picks are then joined into features, with the assis- tance of time slices to provide a plan view of the data. Migration Analysis Migration analysis is a reorganization of the workflow. An operator uses a small sample of data to identify the correct z-factor. Then the entire data set is migrated, and horizontal time slices of the migrated image are assessed for features. Areas of uncertainty can be checked using depth-slice analy- sis. Once a feature is found, the operator’s mouse click initi- ates a tracing algorithm which attempts to locate the entire feature. The operator refines and improves the machine’s effort of feature tracing. Time-Slice Analysis A workflow oriented around horizontal time slices, but without the need for migration, this option strikes a balance between the two workflows above. Having merged several swaths to provide a broad data set, an operator scans time slices, focus- ing on subsections of the whole image. Identified features are picked and the picks joined into features. Areas of uncertainty can be checked using depth-slice analysis. These three workflows will need to be thoroughly assessed against a set of criteria: • Preprocessing time. This includes the time to choose the correct migration z-factor and the time to migrate the data set. • Processing time. This is the speed of the workflow in pro- cessing features in a data set. • Frequency of false positives. The score assigned is a relative assessment, not a probability or frequency of occurrence. • Frequency of false negatives. The score assigned is a relative assessment, not a probability or frequency of occurrence. • Clarity of features. Despite overlap with the two previous classifications, this criterion is explicitly included to charac- terize the visibility of features against the noise back- ground and the ease of identification of a whole feature—its form (linear, point-like) and directionality. • Facility of process. The final criterion assesses ease of use and the associated drain on the operator. The initial subjective assessment suggests that geophysical service providers may be able to improve their workflow with a transition to analysis based on time slices, potentially incor- porating migration transformations. UIT currently operates a quality-checking method which involves one operator reas- sessing an entire data set. This provides an ideal environment for thorough testing of alternative workflows—a new work- flow can be trialed and assessed against the old workflow dur- ing the quality process. The project team also suggests that a “polarized view” toggle be introduced to SPADE for the manipulation of TerraVision data. While there is no strong need to process images for migration with separate polarizations, being able to toggle unmigrated images between the three states (both polarizations as viewed currently, A only, or B only) may clar- ify features exhibiting polarization effects. High-Frequency Seismic Imaging A focus of this research was to demonstrate the SH-wave seis- mic system concept. Practical soil seismic properties investi- gated and reported on here impose limitations on detecting 3-in.-diameter pipes at depths greater than about 5.5–6.0 ft and 6-in.-diameter pipes at depths of about 10–12 ft. This detection performance is based on using state-of-the-art high-resolution SH-wave reflection technology adapted spe- cifically to underground pipe detection and mapping. Larger- diameter pipes are detectable at depths greater than 12 ft. On the premise that these predicted detection performance capa- bilities for state-of-the-art SH-wave reflection seismic tech- nology are potentially applicable to detecting and mapping underground utility pipes, the R01B project efforts may require certain technical revisions. The recommended technical revisions include the following: 1. Retain the SH-wave seismic operating concept and near- vertical illumination and reflection methodology using radiated SH-waves in the frequency range 400–1,600 Hz. 2. Proof test, troubleshoot, and suppress spurious resonances that were found to exist in the OES MicroVib SH-wave vibrator source. Retain this seismic source, with appropri- ate corrective adjustments, as a principal component of the

96 prototype seismic system. Note that, as defined in the license agreement between OES and UIT, no SHRP 2 R01B project work shall be performed to modify the proprietary MicroVib for the intended application. 3. Fabricate the recommended OES proprietary accelerom- eter sensor array as described in the project plan. Note that fabrication costs of the proprietary sensor array are sepa- rate from the SHRP 2 R01B project, as defined in the license agreement between OES and UIT. 4. Replace the automated trailer-based ground-scanning sys- tem design outlined in the project plan with a much sim- pler system configured only to test and evaluate SH-wave seismic reflections from underground pipes. In particular, instead of the previously planned prototype transport and ground scan platform, design a manually operated ground scan platform suitable for conducting controlled field tests of the prototype seismic system at sites having known underground pipe targets. 5. Modify the Dewetron Model 3201-8 data logging system used in soil properties testing to serve as the source excita- tion and data recording component of the prototype SH-wave seismic reflection system. The needed modifica- tions have already been identified as a direct result of using the Dewetron equipment in the soil seismic properties mea- surements. These modifications pertain to certain changes in the Model 3201-8 software to integrate several source excitation and signal recording functions for efficient seis- mic data logging under field operating conditions. The software modifications would need to be performed by Dewetron, Inc. The recommendations listed above represent a shift of the technical efforts from a directed prototype seismic system development and demonstration to a comprehensive applied research and development effort to refine and demonstrate the intended SH-wave reflection technology. Alternate Suggestions for Seismic Development The comments here are based on both the SHRP 2 R01B and R01C projects. Since UIT is the entity that joins the two proj- ects, the R01B principal investigator has made the final inter- pretations jointly and presents the following analysis and recommendations. It appears that the tasks proposed in the R01B and R01C projects of building both 2-D and 3-D imag- ing systems have proven to be too challenging for the early stages of this new area of technology. It is suggested that the aim be shifted to build a single 2½-D imaging system based on what we have learned from the testing thus far. What 2½-D means is that the final prototype will produce a single cross section (2-D) image by using a swath of sources and receivers (i.e., by making a partial 3-D measurement). It is believed that this approach is warranted and necessary to allow accountabil- ity for the complexities in scattering from targets and non- targets and to adequately characterize the velocity variations in the subsurface well enough to make an image. It is also believed that making 3-D images of the subsurface is beyond both the scope of this project and the currently available sci- entific knowledge of seismic data in the soils and depths of interest.