7 Automatic Liquid

Article pubs.acs.org/ac

Laser-Induced Breakdown Spectroscopy for 24/7 Automatic Liquid Slag Analysis at a Steel Works Volker Sturm,*,† Rüdiger Fleige,† Martinus de Kanter,† Richard Leitner,‡ Karl Pilz,‡ Daniel Fischer,‡ Gerhard Hubmer,‡ and Reinhard Noll† †

Fraunhofer-Institut für Lasertechnik ILT, Steinbachstrasse 15, 52074 Aachen, Germany voestalpine Stahl GmbH, voestalpine Strasse 3, 4020 Linz, Austria

‡

ABSTRACT: Laser-induced breakdown spectroscopy (LIBS) is applied for the inline analysis of liquid slag at a steel works. The slag in the ladle of a slag transporter is measured at a distance of several meters during a short stop of the transporter. The slag surface with temperatures from ≈600 to ≈1400 °C consists of liquid slag and solidified slag parts. Automatic measurements at varying filling levels of the ladle are realized, and the duration amounts to 2 min including data transmission to the host computer. Analytical results of the major components such as CaO, Fe, SiO2, MgO, Mn, and Al2O3 are compared with reference values from the steel works laboratory for solid pressed slag samples as well as for samples from the liquid slag. Stable 24/7 operation during the first three-month test run was achieved.

A

These setups use argon-flushed sample stands with mechanical contact to the sample, i.e., the mechanical working distance (MWD) is zero. The optical working distances (OWD) range from 0.15 to 0.25 m. Remote analysis of slag is performed at a laboratory furnace at temperatures up to 850 °C.29 For liquid slag analysis directly at the plant, the challenges are the harsh environmental conditions due to heat and dust as well as the varying sample distances due to the cragged surface or due to bubbling up of the melt. Therefore, large distances (OWD) of several meters are preferable to reduce splashes, fume condensation, and heat exposure of the optics as well as for an easier control of the sample distance. Mechanical contact to the sample should be avoided (no sample stand, MWD > 0). In addition, robustness and complete automation of processes are decisive for routine industrial use.

t the steel works of voestalpine Stahl GmbH, Linz, the Linz−Donawitz process (LD or basic oxygen process) is used for crude steel production. After steel tapping, approximately 15−20 tons of liquid LD (converter) slag is filled into ladles. Transporters carry the ladles to slag pits where the slag is poured out for cooling and further process ing. The slag temperature in the ladle is in the range of ≈600 to ≈1400 °C and sometimes even more. The amount of slag at this plant is about 600 × 103 tons per year. Because the slag composition varies from ladle to ladle, the analysis of each ladle is of interest for the further use of the slag, such as for bituminous road construction and others.1,2 Standard X-ray fluorescence (XRF) analysis3 requires sampling, sample transport, and preparative actions prior to analysis. For a fast analysis of each ladle, this is considered as being too labor-intensive and time-consuming. Furthermore, a complete automation is hard to implement. In contrast, laserinduced breakdown spectroscopy (LIBS) allows measurement directly in the process at a distance of several meters and can be automated more easily while avoiding mechanical sampling, sample transport, and mechanical sample preparation. Several aspects and applications of LIBS are described in the literature.4−10 The LIBS analysis of liquid steel,11−15 liquid aluminum alloys,16−18 and liquid glass19−23 are noted because of needs similar to that for liquid slag analysis. Recently, industrial applications of LIBS with 24/7 operation were published for sorting and analyzing materials (coal, minerals) on moving conveyor belts.10,24 Analysis of steel slags is reported for solid samples which are prepared by crushing, milling, homogenizing, and pressing to pellets prior to LIBS analysis,25−27 as well as measured directly at a steel works site next to a vacuum degasser unit.28 © 2014 American Chemical Society

■

MATERIALS AND METHODS The mass fractions (in m.-% or 10−2 g/g) of the major components of the LD slags are typically within the following ranges: CaO 35−51, SiO2 9−21, Fe 12−32, MgO 4−8, Mn 2−7, Al2O3 1−5. “Fe” and “Mn” stand for the mixture of the different iron and manganese oxides. The liquid slag in the ladle (diameter at the top, ≈3 m) exhibits at the surface a partly solidified and cragged crust, the free surface of the liquid slag, or a spatial mixture of solidified parts and melt. The filling height varies from ladle to ladle by more than 1 m. It is evident that the precondition for a representative measurement at the Received: June 18, 2014 Accepted: August 27, 2014 Published: September 9, 2014 9687

dx.doi.org/10.1021/ac5022425 | Anal. Chem. 2014, 86, 9687−9692

Analytical Chemistry

Article

Figure 1. (a) Slag transporter with ladle during analysis underneath the measuring cabin. Inside the cabin the LIBS analyzer, sensors/control devices, and supply units are installed. (b) View inside the measuring cabin with the optical module and the measuring probe mounted at motorized x- and z-axes. (c) Schematic diagram of the LIBS analyzer setup. The distance from A to B is approximately 3.7 m.

surface is that the slag composition is sufficiently homogeneous within the volume of the ladle; see below. As seen in Figure 1a, the transporter with the slag ladle stops on its way from the steel tapping to the slag pits underneath a measuring cabin in which the LIBS analyzer is installed. The LIBS analyzer consists of an optical module with a nitrogenpurged measuring probe (steel tube, length ≈3.5 m) at its bottom side. The optical module is mounted at a motorized x- and z-axis to move the module with the measuring probe in the vertical (z) and horizontal (x) direction as illustrated in Figure 1b,c. The measuring probe is fed through an elongated opening in the cabin floor and the heat shielding to access the slag in the ladle underneath. The LIBS plasma is generated by a Q-switched Nd:YAG laser beam (wavelength 1064 nm, pulse energy ≈180 mJ) which is focused through the measuring probe onto the slag surface at an OWD of ≈3.7 m. The MWD between the lower end of the probe and the slag surface is controlled to ≈0.15 m. For LIBS, the laser-induced plasma light is detected in reverse direction through the measuring probe, and subsequently it is imaged onto the entrance of a 12 m long optical fiber to be guided to an echelle spectrometer (ESA4000, LLA, Berlin, Germany, resolution λ/Δλ = 40 000). The optical module contains all optical components as well as the laser head, a laser distance meter, and an infrared pyrometer for the slag temperature measurement. The z-position of the optical module is automatically controlled according to the laser distance meter measuring the distance to the slag surface. Thus, the distance between the end of the probe and the slag surface is held close to the set-point of 0.15 m throughout the LIBS measurement. The sequence of the LIBS analysis is as follows. The automatic measurement is released by the driver after positioning the transporter. The motorized z-axis moves the optical module with the measuring probe toward the slag surface until the set-point distance is reached. Then the optical module with the measuring probe is moved in the x-direction for spatial averaging. During the x-travel, the measuring probe

Figure 2. Camera view into the slag ladle during the LIBS measurement. In the example shown, a solidified crust has formed at the slag surface. The measuring probe and the laser-induced plasma plume are moving across the slag surface.

follows approximately the x−z-surface contour due to the controlled distance as shown in Figures 1c and 2. Simultaneously, the laser pulses with a repetition rate of 50 Hz are triggered in a burst mode to generate the LIBS plasma. The burst parameters can be adjusted, herein 20 bursts of 40 laser pulses each are used. After each laser pulse, the LIBS plasma light is detected with a delay time and an integration time gate of 4 and 10 μs, respectively. The spectra of a burst are accumulated within 0.8 s on the CCD chip of the spectrometer and read-out as a single spectrum. Subsequently, these spectra are evaluated using univariate calibration of the mean of the spectral line ratios (ratio of peak area of analyte line and reference line, e.g., Si(I) 390.55/ Ca(I) 397.37, wavelengths in nm) with the mass fraction ratios of reference samples in the case of solid slags or a reference analysis in the case of liquid slags; see below. The mass fraction ratios (e.g., SiO2/CaO) are the analyte mass fraction to the mass 9688



Article

fraction of the major component CaO (except for the basicity CaO/SiO2). For analysis, the mass fraction ratios are calculated from the measured spectral line ratios by the calibration curves (linear or quadratic fit). At the end of the x-measurement run, the optical module is lifted up and the analytical results are transferred to the steel works host computer. The slag transporter continues the drive to a defined slag pit where the liquid slag is dumped according to the LIBS results. The measurement takes 2 min from the release of the measurement until the results are transmitted.

■

RESULTS AND DISCUSSION Measurement of Pressed Slag Powder Samples. The LIBS analyzer allows measurement of solid samples inside the cabin when the optical module with the measuring probe is lifted to an upper z-position (approximately the position shown in Figure 1b). This enables the automatic monitoring of the analyzer with a set of monitor samples in routine use. In addition, for experimental use it allows to measure successively a large set of solid reference samples in order to examine the analytical performance, see Figure 3. Pressed pellets from slag powder have been collected during a long period from the plant in order to achieve a wide calibration range for the elemental composition. The samples are prepared and analyzed by XRF in the steel works laboratory to provide the reference analysis. The linear fit of the XRF reference vs the measured LIBS values yields the ideal identity line y = x with a coefficient of determination R2 of 0.972 for the basicity, and 0.93 for the Fe/CaO ratio, respectively; see Figure 3a,b. This is assessed as an outstanding correlation, taking into account the 3.7 m distance (OWD) of the LIBS measurement. All available 94 samples are included in the graphs exhibiting no outliers. In addition, the rootmean-square error of prediction, RMSEP, is calculated, giving a measure of the average prediction error; see for example refs 15, 28, and 30. In Figure 3c, the mass fractions are calculated from the measured elemental ratios by using the 100% method.15,31 Measurement of Liquid Slag. For liquid slag, the calibration is carried out with ≈50 slag ladles which are dumped in separated partitions of the slag pits. After cooling, slag samples are collected from different spots of each partition and mixed to get a representative sample for each ladle. These samples are analyzed at the steel works laboratory with standard XRF analysis methods, and the results are used as reference values for each ladle. The corresponding graphs to Figure 3 are shown in Figure 4. The fitting lines are still approximately the identity lines, but as expected the coefficient of determination is lower than for the more precisely defined case of solid samples. This is not surprising, taking into account the demanding measuring conditions as well as the complicated gathering of reference values for several tons of material. In addition, the available range of slag compositions (calibration range) was narrower than the one applied for the solid sample set. Table 1 summarizes the RMSEP values for solid samples in column d. The values are comparable or better than literature values reported for solid slag samples (column b), and solidified glass samples (column c) despite the much larger OWD distances. The RMSEP values for liquid slag (column e) are greater, at maximum, by a factor of 2, compared to the values for solid slag samples (column d). This justifies the assumption of sufficient homogeneity within the ladle and the measurement at the surface within the significance range of the sampling of the solidified slag partitions.

Figure 3. Analysis of pressed slag powder samples (see picture at the top, diameter ≈35 mm). (a−c) Comparison of the reference XRF values with the LIBS measurement for the ratios (a) CaO/SiO2, i.e., basicity, and (b) Fe/CaO as well as for (c) the mass fractions calculated from the elemental ratios measured by LIBS. The LIBS emission line ratios are Ca(I) 397.37/Si(I) 390.55 (a) and Fe(I) 382.12/Ca(II) 396.85 (b). R2 coefficient of determination, RMSEP root-mean-square error of prediction. 9689



Article

Figure 5. Repeated measurements of the basicity of two monitor samples showing the stability of the LIBS measurement (solid lines are linear fitting curves). The LIBS emission line ratio for the basicity is Ca(I) 397.37/Si(I) 390.55. At the beginning, narrower intervals are used compared with those used later.

Figure 4. Analysis of liquid slag in ladles as shown in Figure 2. The graphs correspond to those of Figure 3. The LIBS emission line ratios are Ca(II) 373.69/Si(I) 251.92 (a) and Fe(I) 372.26/Ca(I) 422.67 (b). For the reference values for each slag ladle, see the text. The calibration range is restricted due to the process-intrinsic variability given by the existing production conditions. Figure 6. Basicity (a) and Fe oxide content (b) of liquid slag during the test run. Each circle marks a slag ladle, and the (blue) solid lines are the moving average of 30 slag ladle measurements. The LIBS emission line ratios are (a) Ca(II) 373.69/Si(I) 251.92 and (b) Fe(I) 372.26/Ca(I) 422.67 together with the 100% method.15,31

Three-Month Test Run in Plant. A test run over a threemonth period was performed in 24/7 automatic routine operation at slag ladles of transporters during normal steel works production; see Figure 1a. To monitor the stability of the LIBS analyzer during the test run, in addition two solid monitor samples (glass monitor samples, no LD slag) are measured automatically in defined intervals between the liquid slag measurements as described above. The temporal behavior exhibits no significant drift; see Figure 5. The stability in terms

of relative standard deviation of the data shown in Figure 5 amounts to 2.1% for sample 1 and 1.7% for sample 2. The major part of the standard deviation is due to small parameter 9690



Article

Table 1. Root Mean Square Error of Prediction (RMSEP) for Different Analytes in Comparison to Other Published Resultsa solid slagb,26

solidified glassc,23

solid slagd

liquid slage

analyte

range

RMSEP

range

RMSEP

range

RMSEP

range

RMSEP

CaO SiO2 Fe MgO Mn Al2O3 mean

43−64 4−20 6−34 0.7−11 2−6 0.5−2

1.05 0.84 0.85 0.48 0.24 0.21 0.61

1−32 33−63 1−16 0.4−25 − 4−32

0.88 2.1 1.4 1.1 − 1.4 1.4

31−55 6−26 4−30 4−14 3−8 0.8−5

1.32 0.74 1.1 0.72 0.39 0.21 0.75

36−47 11−21 14−29 4−8 3−6 1−5

1.68 1.4 2.1 0.66 0.39 0.35 1.1

The range is the minimum to maximum of mass fractions used for the calibration. bLaboratory, OWD 0.2 m, 20 °C, sample stand, Ar. cLaboratory, OWD 0.25 m, 20 °C, air. dThis study, OWD 3.7 m, 20 °C, measuring probe, N2. eThis study, slag ladles at plant, OWD 3.7 m, ≈600−1400 °C, measuring probe, N2. a

(3) Bennet, H.; Oliver, G. XRF Analysis of Ceramics, Minerals and Allied Materials; Wiley: New York, 1992. (4) Miziolek, A. W.; V. Palleschi, V.; Schechter, I. Laser-induced breakdown spectroscopy; Cambridge University Press: Cambridge, 2006. (5) Cremers, D. A.; Radziemski, L. J. Handbook of Laser-Induced Breakdown Spectroscopy; Wiley: Chichester, 2006. (6) Hahn, D. W.; Omenetto, N. Appl. Spectrosc. 2010, 64, 335A− 366A. (7) Hahn, D. W.; Omenetto, N. Appl. Spectrosc. 2012, 66, 347−419. (8) Fortes, F. J.; Moros, J.; Lucena, P.; Cabalín, L. M.; Laserna, J. J. Anal. Chem. 2013, 85, 640−669. (9) Harmon, R. S.; Russo, R. E.; Hark, R. R. Spectrochim. Acta, Part B 2013, 87, 11−26. (10) Noll, R.; Fricke-Begemann, C.; Brunk, M.; Connemann, S.; Meinhardt, C.; Scharun, M.; Sturm, V.; Makowe, J.; Gehlen, C. Spectrochim. Acta, Part B 2014, 93, 41−51. (11) Jowitt, R.; Abell, I. D.; Jones, J. G. S. Liquid Steel Analysis; Technical Steel Research, Final Report 7210-GD/802, EUR 11905 EN, Commission of the European Communities, Luxembourg, 1988. (12) Aragon, C.; Aguilera, J.; Campos, J. Appl. Spectrosc. 1993, 47, 606−608. (13) Sturm, V.; Peter, L.; Noll, R.; Viirret, J.; Hakala, R.; Ernenputsch, L.; Mavrommatis, K.; Gudenau, H. W.; Koke, P.; Overkamp, B. Elemental analysis of liquid steel by means of laser technology. In Materials Technology and Testing, International Meeting on Chemical Engineering, Environmental Protection and Biotechnology, ACHEMA 2000, Frankfurt am Main, Germany, May 22−27, 2000; pp 9−11. (14) Peter, L.; Sturm, V.; Noll, R. Appl. Opt. 2003, 42, 6199−6204. (15) Hubmer, G.; Kitzberger, R.; Mörwald, K. Anal. Bioanal. Chem. 2006, 385, 219−224. (16) Sabsabi, M.; Cielo, P. Appl. Spectrosc. 1995, 49, 499−507. (17) Rai, A. K.; Yueh, F. Y.; Singh, J. P. Appl. Opt. 2003, 42, 2078− 2084. (18) Weisberg, A.; Craparo, J.; De Saro, R.; Romero, C.; Pawluczyk, R.; Whitehouse, A. I. Presented at Applied Industrial Optics: Spectroscopy, Imaging and Metrology, Tucson, AZ, June 7−8, 2010; conference paper AMC4, ISBN: 978-1-55752-892-6. (19) Yun, J.; Klenze, R.; Kim, J. Appl. Spectrosc. 2002, 56, 437−448. (20) Yun, J.; Klenze, R.; Kim, J. Appl. Spectrosc. 2002, 56, 852−858. (21) Matiaske, A. M.; Gornushkin, I. B.; Panne, U. Anal. Bioanal. Chem. 2012, 402, 2597−2606. (22) Panne, U.; Neuhauser, R. E.; Haisch, C.; Fink, H.; Niessner, R. Appl. Spectrosc. 2002, 56, 375−380. (23) Laville, S.; Sabsabi, M.; Doucet, F. R. Spectrochim. Acta, Part B 2007, 62, 1557−1566. (24) Barishnikov, A.; Gaft, M.; Isaenko, G.; Mansurova, N.; Tikhonov, D. Presented at 6th International Congress on the Science and Technology of Ironmaking − ICSTI, Rio de Janeiro, RJ, Brazil, Oct 14−18, 2012; pp 2577−2582. (25) Doujak, G.; Mertens, R.; Ramb, W.; Flock, J.; Geyer, J.; Luengen, S. Stahl Eisen 2001, 121, 53−58 (in German).

variations of instruments, environmental conditions, and abrasion of the monitor samples. The liquid slag measurements show the varying slag composition from ladle to ladle during the test run in Figure 6 for the basicity (a) and for the Fe oxide content (b). No adjustment or interventions such as cleaning optics of the LIBS analyzer were required during this period.

■

CONCLUSIONS Despite much larger distances (OWD) in this application, the root-mean-square error of prediction (RMSEP) values for solid slags are comparable to other published results with LIBS using sample stands; see Table 1. The RMSEP values measured for liquid slags are greater, at maximum, by a factor of 2 only, compared to solid slag samples, although the reference sampling is quite different in both cases. This justifies the assumption of a sufficient homogeneous distribution in the ladle with several tons of slag. For the first time at a steel works, liquid slag is analyzed inline by an automated LIBS system within 2 min in the ladle of a slag transporter. Data transfer to the host computer is included as well as an automatic adaption to varying filling heights of the ladles. Stable 24/7 operation during the first three-month test run has been demonstrated successfully under these harsh environmental conditions. The presented application shows the versatile potential of LIBS for inline process analysis in industry.

■

AUTHOR INFORMATION

Corresponding Author

*Phone: (+49) 241 8906 0. Fax: (+49) 241 8906 121. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Many thanks to all who contributed in different ways to this work, especially to H. Kroiher, A. Pissenberger, L. Böske, M. Brankers, T. Eleftheriadis, M. Fohn, C. Heuer, and R. Prümmer. Furthermore, the authors express their thanks to the Fraunhofer Society for the support of the research activities which in parts are the basis of the work presented here.

■

REFERENCES

(1) Pilz, K.; Berndl, J.; Fritz, B.; Gangl, E.; Hubmer, G. Steelmaking slag measurement with LIBS. Global Slag Magazine; Nov 2007; pp 19−21. (2) Pilz, K. Berg- Hüttenmänn. Monatsh. 2012, 157, 250−257 (in German). 9691



Article

(26) Kraushaar, M.; Noll, R.; Schmitz, H.-U. Appl. Spectrosc. 2003, 57, 1282−1287. (27) Pedarnig, J. D.; Kolmhofer, P.; Huber, N.; Praher, B.; Heitz, J.; Rössler, R. Appl. Phys. A: Mater. Sci. Process. 2013, 112, 105−111. (28) Sturm, V.; Schmitz, H. U.; Reuter, T.; Fleige, R.; Noll, R. Spectrochim. Acta, Part B 2008, 63, 1167−1170. (29) Lopez-Moreno, C.; Palanco, S.; Laserna, J. J. Spectrochim. Acta, Part B 2005, 60, 1034−1039. (30) Esbensen, K. E. Multivariate Data Analysis, 5th ed.; Camo Process AS: Oslo, 2002; p 159. (31) Slickers, K. Die automatische Atom-Emissions-Spektralanalyse (in German), 2nd ed.; Brühlsche Universitätsdruckerei: Giessen, 1993; p 275.

9692


7 Automatic Liquid

Recommend Documents