Summary

This research focused on the development of a diffusible hydrogen sensor for steel welds. The sensor is based on the chemochromic reaction of WO₃ with hydrogen. The sensor is inexpensive, produces results quickly and has been shown to correlate with hydrogen concentrations derived using the standard gas chromatography method. Further research on the sensor should lead to the development of a field model for complete nondestructive testing with respect to hydrogen assisted cracking, in particular at pipelines.

Messung des diffusiblen Wasserstoffs in Schweißverbindungen mit faseroptischen Sensoren. Im vorliegenden Beitrag werden die Forschungsarbeiten zur Entwicklung eines Sensors zur Bestimmung des diffusiblen Wasserstoffs in Staldschweißungen präsentiert. Der neue Sensor basiert auf der chemochromischen Reaktion des Wolframoxyds WO₃ mit Wasserstoff. Es handelt sich um einen preisgünstigen Sensor, der schnell Ergebnisse liefert. Es hat sich gezeigt, Class die damit ermittelten Wass erstoffkonzentrati onen mit denen mit Hilfe von Gaschromatografie bestimmten Werten korrelieren. In zukünftigen Forschungsarbeiten soll ein Sensor zur feldmässigen Anwendung entwickelt werden, um zu einer umfassenden zerstörungsfreien Prüfung hinsichtlich einer möglichen wasserstoffbeeinflussten Rissbildung insbesondere an Pipelines zu gelangen.

Background

The problem of hydrogen damage in welded structures has been studied extensively. When welding operations are performed, moisture from the atmosphere can enter the high-temperature welding arc and form hydrogen which can then enter the welding spot. Some of this hydrogen is captured or trapped by microstructural features in the metal, and is known as residual hydrogen. The diffusible or mobile fraction is of more interest because it is primarily responsible for hydrogen damage. Diffusible hydrogen either escapes from the metal or it migrates to flaws or microcracks in the weld and actually accelerates the cracking process.

The current methods for diffusible hydrogen analysis require up to 72 hours to complete, and are thus time consuming and expensive. These tests are per formed on laboratory specimens using welding parameters that must then be duplicated in the field. The goal of the research presented in this contribution was the develop of' a hydrogen sensor that can be attached to the actual weld in question. The sensor measures the rate of hydrogen evolution from the weld deposit, which is then used to back- calculate the initial hydrogen concentration. The results are thus obtained in minutes as opposed to days, considerably reducing the analysis time.

The sensor is based on the chemochromic reaction of certain transition metal oxides [1] with hydrogen in air. The chemochromic material used for these sensors was a thin film of tungsten oxide or WO₃. The basic principle of the sensor is the application of thin films to the end of a fiber optic cable. The color change of the film in the presence of hydrogen is defected by reflectance spectroscopy.

The reaction of tungsten oxide and hydrogen in air is represented by the equations

Pd + xH₂ + x/4O₂ + WO₃
---> H_xWO₃ + x/2H₂O + Pd        (eq. 1)
hv + W⁵⁺(A) + W⁶⁺(B)
---> W⁶⁺(A) + W⁵⁺(B)               (eq. 2)

Fig 1: Hydrogen calibration data for fiber optic hydrogen sensor

The species formed is a type of ion-insertion compound known as a hydrogen tungsten bronze. The reaction is catalyzed by palladium or platinum [2] which adsorbs hydrogen on the surface and converts it to its atomic form. Electrons are localized on W⁵⁺ sites and optical absorption occurs by electrons undergoing W(V)-V(VI) intervalence transfer. The transferred electrons can be considered as small polarons, or electrons trapped in distorted lattice sites [3]. These electrons are, thought to be responsible for the chemochromic effect. This compound absorbs light in the red region of the spectrum and thus appears dark blue.

Sensors utilizing the reaction of WO₃ have previously been devised for detecting hydrogen [4]. In particular, fiber of the sensors for hydrogen leak detection in air [5-7] and plate-like sensors for hydrogen producing bioorganisms [8] have been developed. The same concept was applied to detection of hydrogen evolving from the weld metal of high strength steel samples.

Experimental

The sensors consisted of polymer optical fibers coated with thin films of WO₃ and palladium. The films were deposited by vacuum evaporation. The average thickness of the WO₃ films was 500 nm. The thickness of the palladium films varied from 3 to 30 nm. The change in optical properties was measured using an infrared (IR) reflectometer equipped with a laser photodiode emitting at 850 nm. The reflectance decreased as the hydrogen concentration increased due to the cater change in the WO₃ film, and was measured in units of reflected power (µW)

Fig 2: Design for prototype diffusible hydrogen weld sensor

Calibration tests were performed to determine the linearity of the sensor's response to known quantities of hydrogen. A sensor was placed in a Plexiglas reaction chamber and exposed to hydrogen introduced through a calibrated leak valve. The sensor was allowed to react for fifteen minutes, a reading was taken, then additional gas was added. Initial calibration tests were made with up to six data points. The results are presented in Figure 1. The correlation coefficient for the displayed curve is 0.9994. 'The linear range, is estimated to be from 200 to > 1000 ppm.

The, response of the sensor to hydrogen diffusing from a weld was then investigated. A sensor housing was designed for attachment to the base plate of a welded sample as shown in Figure 2. The housing contained a short section of fiber optic cable coated on one end with WO₃ and palladium. The housing was threaded to accept a standard optical cable connector. The inside of the housing was scaled and the sensor was coupled to another length of fiber optic cable using optical matching gel, which excluded air and allowed fight to pass into the section with only a small loss in power. The bottom of the housing was fitted with a small o-ring to form a seal with the metal surface. Pressure was applied to the sensor using a Plexiglas plate to hold it in place.

Initial experiments were, conducted on AST M A36 steel specimens that were gas metal arc welded (GMAW) with 0.1 vol.-% H₂ in Argon, then quenched in ice water and stored in liquid nitrogen (LN₂) until analysis. The sensor was placed directly adjacent to the weld deposit. Using this practice it was found that the amounts of hydrogen diffusing from the base plate were too small for a detectable response.

The sensor was therefore modified to sample. from the curved surface of the weld deposit using a rubber gasket adapter as shown in Figure 3. Data was collected using a General Purpose Instrumentation Bus (GPIB) interface connected to a laptop computer; measurements were made every minute. As shown in Figure 4, the amount of hydrogen diffusing from the weld deposit was more than adequate for detection. I he same specimen was allowed to diffuse at room temperature for extended periods of time, then re-analyzed every hour. These experiments were, repeated up to an interval of five hours after welding. The hydrogen from the weld deposit was still detectable by the sensor after a period of five hours.

Fig 3: Design for prototype diffusible hydrogen weld sensor with soft rubber gasket

Fig 4: Hydrogen response data as a function of time after quenching gas metal arc welded ASTM A 36 steel (0. 1 % H2/Ar shielding gas) using prototype diffusible hydrogen sensor with soft rubber gasket

Results and Discussion

For generating quantitative data, the inside volume of the sensor was calculated and used to correlate the sensor response from the weld specimen to that

from a calibration rising a gas mixture. The response of the sensor was in the shape of a sigmiodal curve. It was there fore decided to correlate the slope of these curves with the, theoretical flux from a diffusion equation. If the sensor response was proportional to this flux the initial concentration could be obtained.

To correlate the sensor response, to the GC results, calculations were performed using a diffusion equation based on the error function. For these calculations the simplifying assumption was made that the diffusion occurs from a semiinfinite plant, sheet with a constant surface concentration of zero. For this case the solution for the error function equation takes the form



[9] where M_t is the amount of diffusing substance, D is The diffusion coefficient for the substance in a particular medium, C, is the initial concentration, and t is time.

Calculations were made for five initial concentrations and adjusted for the average amount of weld metal per sample and the amount of surface area sampled. The weld metal geometry was assumed to be cylindrical, and the total surface area was taken as half the area of rite cylinder plus the full area of both ends. Figure 5 shows the results of the theoretical calculations. Initial calculations were made with a diffusion coefficient of 10⁵ cm²/sec [10].

However, a best fit to the experimental data was obtained with an apparent diffusion coefficient of 7.5 x 10⁵ cm²/ sec. For each curve the slope was calculated at a time interval of 2.25 hours and plotted as a function of initial concentration.

The data gathered with the gasket adapter produced a correlation between the actual specimens and the theoretical values that was unsatisfactory. It was assumed that the surface area actually sampled was greater than that estimated due to a poor gasket seal. A new sensor adapter was therefore designed as shown in Figure 6. The. original sensor head was again used. The new adapter was threaded at both ends. One end accepted the sensor head and the other a proboscis-like tip that could be inserted into a hole drilled in the. weld metal.

Fig 5: Theoretical hydrogen evolution curves using diffusion equation lot semi infinite plane sheet (based on erf(x) five different initial diffusible hydrogen concentrations; Deff - 7.5 x 105 cm2/sec)

Fig 6: Proboscis adapter design for prototype diffusible hydrogen sensor)

Figure 7 shows some representative sensor data for GMA welded HSLA 100 steel samples. A total of nineteen samples were analyzed. The samples were welded in duplicate with varying levels of hydrogen in the shielding gas. The duplicate samples were analyzed for diffusible hydrogen using the standard gas chromatography method [AWS A4.3/93] and results were obtained in ml/100 g weld metal. The samples analyzed with the sensor were allowed to remain at room temperature for two hours prior to analysis. The slope of each sensor response curve was then calculated from inflection point to inflection point, and yielded results in nW/min. The results from the curves were then converted to µL/min using the calibration data.

Fig 7: Experimental hydrogen response curves for prototype diffusible hydrogen sensor (gas metal arc welded HSLA 100 steel)

Fig 8: Comparison of experimental and theoretical data for fiber optic diffusible hydrogen sensor (gas metal arc welded HSLA 100 steel, Deff - 7.5 x 105 cm2/sec, r - 0.970)

Fig 9: Comparison of experimental and theoretical data for fiber optic diffusible hydrogen sensor (gas metal arc welded HSLA 100 steel, Deff - 7.5 x 105 cm2/sec, r - 0.989)

Figure 8 shows the results from all nineteen analyses. The correlation for the sample curve is 0.970. A standard statistical analysis could not be performed since there was no mean value; the sample curve was therefore analyzed and a residual standard deviation of 0.00865 µl/min was obtained, as shown by the error bars. Based on this result certain data points were omitted and the data was re-plotted.

Figure 9 shows the data with seven outlying points omitted. The correlation is now 0.989. As seen by the two curves, the fit to the theoretical data is excellent. Based on these results it was decided that the sensor could indeed be used to generate quantitative data for welded steel.

The use of this sensor would provide numerous advantages. First, the, actual welded structure can be analyzed as opposed to a laboratory specimen. The sensor allows data acquisition in less that an hour, making the test much faster than the present standard methods. The equipment is portable, inexpensive, and can be designed for use by welding personnel. Finally, the. use of an optical signal will eliminate interference caused by magnetic fields produced by any nearby welding operations.

Other possible applications for the sensor include any other structures where hydrogen damage is considered a problem, such as galvanically protected structures, pipelines, cannon barrels or high pressure vessels.

Current work on hydrogen sensors at NREL is focusing on improved materials for better performance. The materials cur really being investigated show faster response times and decreased susceptibility to degradation. Specific improvements for the weld sensor will involve a redesign to allow complete nondestructive testing. This will undoubtedly include a method for internal calibration that will eliminate the need for volume measurements. More experiments will need to be performed to insure compliance with the standard methods, as well as generating correlations for different types of steels.

Conclusions

A fiber optic transition metal oxide based diffusible hydrogen sensor has been developed. The sensor generates results quickly and allows analysis of an actual welded structure as opposed to a laboratory specimen.

The sensor has been calibrated to yield quantitative results for HSLA 100 steel in ml/100 g weld metal as a function of hydrogen diffusion rates from the Sample surface. An effective diffusion coefficient of 7.5 x 10^-5 cm²/sec provides a good fit to measured data.

Acknowledgement

The authors would like to thank the National Renewable Energy Laboratory, Golden, CO, for the use of their facilities. The authors also acknowledge and appreciate the support of the United States Army Research Office.

References

1. Granolvist, C.G.: Handbook of Inorganic Electrochromic Materials, Elsevier Publish-ing Co., New York, 1995, p. 81
2. Khoobiar, S.: Particle to Particle -Migration of Hydrogen Atoms on Platinum-Alumina Catalysts From particle to Neighboring Particles. J. Phys. Chem. 68, 1964, pp. 411-412
3. Lampert, C.M.: Electrochromic Materials and Devices for Energy Efficient Windows. Solar Energy Mat. II 11, 1984, pp. 1-27
4. Ito, K.M.; Ohgami, T.: Hydrogen Detection Based on Coloration of Anodic Tungsten Oxide Film. Appl. Phys. Lett. 60 (8), 1992, pp. 938-940.
5. Benson, D.K.; Tracy, C.E.; Bechinger, C.: Design and Development of a Low Cost Fiber Optic Hydrogen Detector. Proceedings of the DOE Hydrogen Program Meeting Vol. 11, 1996, Miami, FL, pp. 605-624.
6. Benson, D.K.; Tracy, C.E.; Hishameh, G.A.; Ciszek, P.A.; Lee, S.; Haberman, D.P.: LowCost Fiber-Optic Hydrogen Gas Detector Using Guided-Wave Surface Plasmon Resonance in Chemochromic I Thin Films. Proceedings of the SPIE International Symposium on Industrial and Environmental Monitors and Biosensors, 1998, Boston, MA, pp. 185-202.
7. Benson, D.K.; Tracy, C.E.; Bechinger, C.: Fiber Optic Device for Sensing the Presence of a Gas. U. S. Patent 5,708, 1998, 735.
8. Seibert, M.; Flynn, T.; Benson, D.; Tracy, E.; Ghirardi, M.: Development of Selection and Screening Procedures for Rapid Identification of H₂, Producing Algal Mutants With Increased O₂, Tolerance. Proceedings of the 1997 Biohydrogen Meeting, 1997, Waikoloa, HI, pp. 227-234.
9. Crank, J.: The Mathematics of Diffusion.. Oxford Science Publishing Co.. London, UK, 1992, p. 32.
10. Völkl, J.: Alefeld G.: Diffusion of Hydrogen in Metals. Hydrogen in Metals I, Völkl, J. and Alefeld G. ed., Springer Verlag, New York, 1978, pp. 321 348.

The authors of this contribution

Rodney D. Smith II received his B.S. in Chemistry from Colorado State University in 1986 and his Ph.D. in Applied Chemistry from Colorado School of Mines in 1999. His research at CSM focused on advanced methods for hydrogen in high strength steels. He is currently employed by DCH Technology in Valencia, California, and is working on new technologies for hydrogen detection and safety.

David K. Benson (retired) was the principal scientist and team leader in the Center for Basic Sciences at the National Renewable Energy Laboratory in Golden, Colorado until 1999. He received his B.S. and M.S. in Physics at the University of Missouri.

J. Roland Puts, holds a Ph.D. in Physics from the University of Denver. He joined NREL in 1978 and is currently a Senior Scientist and Opto-Electronics team Leader, engaged in research on the surface modification of materials and the development of photoelectrochromic and other optical devices.

David L. Olson is the J. H. Moore Distinguished Professor of Physical Metallurgy at the Colorado School of Nimes, Professor Olson received his Ph.D. in Materials Science from Cornell University. fie was honored as a fellow by ASM International and the American Welding Society. He performs research on topics associated with welding metallurgy, reactive metals and microstructural evolution.

Dr. Thomas R. Wildeman has been teaching and doing research at the Colorado School of Mines for over 30 years. He is currently a Professor of Chemistry and Geochemistry. fie has 35 years of research experience in trace element and environmental chemistry and in methods of chemical analysis.

© NDT.net - info@ndt.net

|Top|