Adapted from TRB Paper No. 00-0558, "An impulse radar tomography imaging system for NDE of timber structures," by Hernandez and Duwadi. The original paper was prepared for the 79th Annual Meeting of the Transportation Research Board on Jan. 9-13, 2000.

There are about 40,000 timber bridges in the United States, and another 40,000 steel stringer bridges with timber decks. More than 40 percent of the timber bridges are considered structurally deficient or functionally obsolete.1 These conditions limit bridge utility and, if they are not adequately monitored and maintained, could pose a threat to the safety of the bridge users.

In 1971, the National Bridge Inspection Standards came into being for bridges constructed of all materials. In light of this, all states have some form of bridge inventory and inspection program, the record of which are annually sent to the Federal Highway Administration (FHWA) to update the National Bridge Inventory. Conducting a detailed inspection for any type of bridge is time-consuming. For major structures, the amount of time may be justifiable, and the inspections are fairly accurate if the appropriate tools and techniques are used. However, many timber bridges are not major structures, and the available tools and techniques are not sophisticated. In addition, many inspectors do not have the required background in wood; therefore, in general, many of the timber bridges tend to get subjective ratings, which can lead to inaccurate assessment of the condition of these bridges.

The state of the practice in terms of methods and techniques available for inspection of any bridge is largely

dependent on the material and the design of a bridge. The most common inspection tool, however, for any structure type is still "visual qualitative" inspection. Other tools and tech-niques commonly used for the inspection of timber bridges include: probing, pick test, pilodyn, sounding, drilling, coring, and moisture meters¹ However, most of these techniques are destructive or semi-destructive and are only good for detecting wood deterioration close to the surface (except for coring).

For these reasons, most of the current research in the inspection of timber structures focuses on the development of nondestructive evaluation (NDE) techniques. Ground-penetrating radar (GPR) is one of several technologies currently under investigation for NDE of timber and other structures. In particular, the micropower impulse radar (MIR) technology developed at the Lawrence Livermore National Laboratory (LLNL) shows good potential for this application due to its small size and low power consumption, and MIR can easily be packaged into a portable (hand-held), lightweight, battery-operated, relatively inexpensive tool for actual use in the field. (See figure 1.) Furthermore, its imaging capability is expected to accurately show the extent and location of problem areas and to produce data that can be inter-preted more easily than conventional GPR data.

System Overview

The current MIR imaging system prototype consists of a radar module that houses the radar electronics, a rechargeable battery, horn antenna, portable computer equipped with a data acquisition PCMCIA card, and a cable kit for con-necting the radar to the computer. (See figure 2.) The radar electronics consists of off-the-shelf components surface-mounted on a 7.5 centimeter x 11.5 centimeter printed circuit board. The transmitter section of the radar has a patented circuit, which generates ultra-wideband pulses with an approximate bandwidth of four gigahertz (GHz) (billion cycles per second) at a rate of five megahertz (MHz) (million cycles per second) and an average power of about 100 micro watts. The receiver section consist of a swept-range gated integrating receiver with very low power con-sumption. A single horn antenna is used for both transmitting and receiving the radar pulses. The antenna has a very broad response -- about 12 Ghz -- and a beamwidth of about 90 degrees. The portable computer is used for data acquisition, real-time graphics, and data processing.

Data Acquisition

The radar data consists of a collection of waveforms representing time-delayed reflections of the transmitted pulses from subsurface dielectric interfaces. The data is collected by scanning the radar unit over the object surface and triggering the data acquisition system at uniform inter-vals so that the final waveforms lie on a square grid. This implies that some type of position encoding mechanism is needed to control the overall data acquisition. A computer-controlled mechanical system was used to conduct the experiments presented in this article. (See figure 3.) Work is underway to add a RF- (radio frequency) based position encoding system to the current prototype to keep the system portable and battery-operated.

The radar data is acquired and stored in the computer's internal hard drive via the data acquisition card. Custom software under the Linux operating system is used to display the raw data in real-time as individual waveforms or as a two-dimensional color waterfall display. (See fig-ure 4.) Additional software is available for further data processing and image formation.

Image Reconstruction

(a)

(b)

The purpose of the image-reconstruction algorithm is to map the radar data into a high-resolution spatial image of the scattering interfaces in the object. The image-formation algorithm currently being used is based on a multifrequency diffraction tomography algorithm.^2,3 Unlike conventional diffraction tomography methods in which the tomographic nature of the data is obtained from the rotation of the object, a synthetic aperture pulse-echo radar method is employed. In this method, the tomographic information is obtained from the wideband pulse that provides multiple illuminating wavelengths. Each wavelength in the received wavefield represents a different portion of the spectrum of the object distribution. A frequency domain plane-to-plane backward propagation method is used to coherently superimpose the spectral components from each wavelength and focus the individual wavelengths back to the source. The underlying propagation model is flexible enough to account for multiple layers of different materials.⁴

Experimental Results

Initial tests of the MIR imaging prototype have been performed on a 42-centimeter by 22-centimeter by 122-centimeter Douglas Fir glulam beam sample that was fabricated for this study with a variety of known size voids. Some of the holes were filled with sawdust for identifying changes in wood density. An illus-tration of the beam sample showing the fabricated defects on each lamination and a photo of the actual beam are shown in figures 5 and 6.

To test the MIR imaging system, linear scans for each laminate were acquired. The top and bottom laminates were excluded for this test. The results are shown in figure 7. The two arc shape features in the first three images (laminates 2 through 4) correspond to the 5-centimeter and the 5-centimeter by 10-centimeter holes. There is some evidence in the fourth image (laminate 5) that the radar might be detecting the middle 2-inch hole, which was filled with sawdust. There is no obvious indication in the next image (laminate 6) of the radar detecting the two knots shown in figure 5 for this laminate. The last four images (laminates 7 through 10) show strong features corresponding to the large 10-centimeter by 61-centimeter hole. There is no obvious indication that the radar is detecting the 2.5-centimeter hole or the knots shown in figure 5 for these laminates.

The next set of images show the results of processing the individual linear scans with the image-reconstruction algorithm described previously. (See figure 8.) The results indicate some success in reconstructing the larger holes -- the 5-centimeter by 10-centimeter and the 10-centimeter by 61-centimeter. However, the smaller holes are not as easy to detect. The data also reveals some other clutter that is primarily due to pulse ringing from knots and other imperfections on the wood surface. (See figure 6.)

Finally, the next set of images show a couple of reconstructions for laminates 3 and 9 computed from a full three-dimensional data set -- for example, a hole sequence of linear scans similar to the ones discussed above. (See figure 9.) In this case, the image reconstruction algorithm makes use of the information in all three dimensions to reconstruct every point in the final image. As expected, there is more fidelity on these reconstructions compared to the previous results that were calculated from individual two-dimensional data sets only.

Conclusions

An MIR-based portable imager is being developed to aid bridge inspectors assess the condition of timber structures. An initial prototype has been developed and tested. The current results show that the radar is capable of detecting a variety of void-type defects on untreated Douglas Fir glulam beams. Currently, work is underway to finish the integration of an RF-based position encoding mechanism for the MIR imager. Further testing will be done once this next phase is completed. Further testing and optimization of the image-reconstruction algorithm is needed for this application. Post-processing algorithms may be needed to filter out unwanted clutter due to pulse ringing from surface knots and other imperfections.

References

1. Sheila Rimal Duwadi and Michael A. Ritter. Timber Bridge Research Program - Interim Report, Federal High-way Administration, McLean, Va., (unpublished).
2. Jeffrey E. Mast and Erik M. Johansson. Three-dimensional ground penetrating radar imaging using multi-frequency diffraction tomography, URCL JC-116340, Lawrence Livermore National Laboratory, 1998.
3. Jeffrey E. Mast. Automatic position calculating imaging radar with low-cost synthetic aperture sensor for imaging layered media, US Patent 5796363, 1998.
4. Jeffrey E. Mast. Microwave Pulse-Echo Radar Imaging for the Nondestructive Evaluation of Civil Structures, doctorial thesis, University of Illinois at Urbana-Champaign, 1993.
5. M. Born and E. Wolf, Principles of Optics, Pergamon Press, New York, 6th ed., 1980.

The authors acknowledge Kenny Lin, who performed all the experiments presented in this study; Jeffery Mast and Ming Liu, who developed the computer software; and Mark Vigars and Pat Welsh for building the radar prototype. They also thank APA-Engineered Wood Systems for donating the glulam beam sample and Calvert Company for fabricating the beam.

Jose E. Hernandez is a project engineer at the Lawrence Livermore National Laboratory. He has been at LLNL since 1984. His prior positions include image processing systems engineer, computer vision and pattern recognition data analyst, and signal and image processing engineer. His general research interests are radar imaging, computer vision, pattern recognition, signal and image processing, artificial intelligence, and software engineering. Hernandez has a bachelor's degree in electrical engineering from the University of Puerto Rico and a master's degree in electrical engineering from the Georgia Institute of Technology.

Sheila Rimal Duwadi is a research structural engineer in the Federal Highway Administration's Office of Infrastructure Research and Development. She manages research programs in the area of timber bridges, advanced wood composites, and horizontally curved steel bridges. She is currently on an assignment with the Office of Bridge Technology, working on the Historic Covered Bridge Preservation Program and serving on a team that is rewriting the Recording and Coding Guide for the Structure Inventory and Appraisal of the Nation's Bridges. Duwadi managed FHWA's Temporary Works Research Program, which produced guidelines and specifications for bridge falsework; the guidelines were adopted by the American Association of State Highway and Transportation Officials (AASHTO) in 1995. She is active in AASHTO and on committees of the American Society of Civil Engineers and the Transportation Research Board. She joined FHWA in 1984, and her career has included assignments in federal lands, federal aid, bridge design, and research and technology programs. She has a master's degree in civil engineering from Oregon State University, and she is a registered professional engineer in Virginia.