Sustained Testing of Yamari Aluminum Sensors in Galvanizing Production

Transcription

1 101 Sustained Testing of Yamari Aluminum Sensors in Galvanizing Production by G.R. Adams, N-Y.Tang, N.Qiang Cominco Ltd. Product Technology Centre Mississauga, Canada Presented at the 89th Galvanizers Association Meeting San Francisco, CA October 12-15, 1997

2 102 Sustained Testing of Yamari Aluminum Sensors in Galvanizing Production G. Adams, N-Y.Tang, N.Qiang Background With the ever-increasing pressure from customers to improve quality and consistency and to show that the coating process is under control, the modern galvanizer is looking more closely at zinc bath management practices. While focusing on quality, he is also forced to look at productivity and production costs at the same time to remain competitive in today s business world. The importance of aluminum in the zinc bath is well known to all galvanizers, and the need to control aluminum levels in the bath to optimal narrow ranges is becoming more and more important. It is known that poor management of bath aluminum can lead to coating quality problems such as poor formability, poor weldability, non-uniform appearance, and other surface blemishes from dross drag out1,2. Excessive dross formation is often associated with poor control of aluminum in the bath, and dross removal can be expensive. It also leads to inefficient use of zinc and aluminum, and increased wear of pot hardware. CGL bath management technology has seen some impressive advances in recent years, including a better understanding of the Zn-Fe-Al metallurgical system and how it affects both the zinc coating and what is happening in the bath, including dross formation. Effective aluminum is being used more often and methods of accurately determining effective aluminum have been developed. Methods of bath sampling and analysis are being revisited for improved accuracy of Fe and Al measurements in the bath. And now, with the introduction of aluminum sensors, it is possible to measure effective pot aluminum directly, immediately and continuously during galvanizing production. The Yamari aluminum sensor was invented by Professor Yamaguchi for measuring aluminum dissolved in molten zinc and it was licenced to Yamari Industries for production. It has been used in Japan by Nippon Steel for bath aluminum control. Cominco recently became the exclusive distributor for the sensors and now markets them around the world. The sensors continue to be manufactured in Japan by Yamari. The Yamari Al sensor is a disposable sensor, based on a semi-solid composite mixture of molten salts in a submerged probe as shown in Figure 1. Acting as an electrochemical concentration cell, it generates a small voltage which is dependent on the aluminum in the molten zinc as well as the local temperature. Cominco developed a user-friendly computer system and other instrumentation to go with the Yamari sensor to complete the package. Details of these developments have all been reported elsewhere and will not be repeated here. As part of the Cominco development work, it was necessary to operate the Yamari sensors and

3 103 software systems on a sustained basis on a galvanizing line. Engineers at a North American steel mill expressed interest in evaluating the Yamari Al sensor for CGL baths and co-operated with Cominco in a sustained series of tests on one of their lines. This paper describes the results of those tests. The line provided a very demanding test of the sensor system for a number of reasons. This line is a Zenzimir type, producing approximately 250,000 tons of galvanized product per year. It has a very small 70-ton pot, which is very congested with hardware, leaving little undisturbed room for sensors to be installed. The line alternates frequently between two products having distinctly different bath aluminum requirements, nominally 0.06 % for one product and 0.16% for the other. The large and frequent swings in bath aluminum provided an ideal opportunity to test the performance of the sensor systems; but at the same time, it was recognised that this would be a very demanding test, because of the nature of the line and production schedules. A transverse catwalk is located above the sinker roll on the line, and this is where the two sensors were installed for the tests. They were attached to the handrail, in front of the snout, at both ends of the catwalk, approximately 3 in. from the kettle wall. This location is shown in Figure 2. The stainless steel sensor holder design is shown in Figure 3. It provides a method of fastening the disposable sensors in place, including the electrical connection. It also allows for vertical adjustments to set immersion depth, and includes a thermocouple, complete with its own stainless steel sheath. The thermocouple tip was located within 2 in of the sensor tip, to provide the accurate local temperature required for aluminum determination from the sensor emf output. Wires from the sensor and thermocouple were carefully shielded and routed to the line operator control booth near the pot. Here, they were connected to a computer for calculating aluminum from the sensor and thermocouple outputs, displaying temperature and aluminum, and for storing the data for later analysis. A typical screen from this computer program is shown in Figure 4. Some features seen in this figure are the digital displays of time, bath temperature, and bath aluminum, along with 2 instant graphs on the bottom half of the screen, for temperatures on the left and aluminum readings on the right. The large digital displays can be used to show temperatures or aluminum for both sensors or any combination of 2 outputs, such as aluminum only for both sensors. When called for, as shown in Figure 5, the whole screen can be used to display all of the data saved since the current sensors were installed. For the trials reported here, aluminum readings were collected every 2 seconds and this data was automatically saved every 2 minutes to the hard drive disk on the computer.

4 104 Procedures With 2 sensors installed and the computer program running in the control room, the system was operated continuously for 3 shifts per day and 7 days per week for 14 weeks. During this period, there were 12 bath transitions to allow for 6 runs of one product and 7 runs of the other. Engineers from both companies monitored the progress closely, and replaced spent sensors when they failed. Data on aluminum and temperature readings from the 2 sensors were down loaded daily from the computer for charting and analysis. Throughout the run, bath samples were collected every 4 hours during steady production, and every 2 hours during transitions between products. These were analysed by ICP and AAS methods. These assay results for aluminum and iron were calibrated using the Cominco DEAL software to determine effective aluminum. These results were then superimposed on the aluminum charts from the Yamari sensors to check the accuracy of the sensors. Results Accuracty Figure 6 shows a typical chart from the trial, with sensor output and bath assay results over a 14-day period, which included a transition between products. Although somewhat difficult to see, there are 2, sensor output curves in this chart, agreeing very well with each other. Charts like this were produced for the entire run and included over 500 bath assays. It can be seen that the sensor outputs were very close to the effective aluminum results determined from bath assays. Occasionally there were short periods when data spiking occurred, and this was shown to be due to bath turbulence near the sensors due to drossing and bubbling, a technique used occasionally to prevent freezing of zinc in regions of the bath surface. The two sensors gave results that varied by less than %A1 in normal operation. When the outputs varied by more than this, it was a sign that one of them was failing and would soon have to be replaced. The sensor at one location always gave higher readings than the other, indicating that there were some small aluminum gradients in the pot and repeatability between sensors was thus better than this 0.005% figure. No attempt was made to make quantitative statements about sensor accuracy from this study, since the issue of accuracy and precision of analytical methods is very complex and even has an ASTM spec to cover how it must be done4. For example, it would require tests to be repeated in 6 laboratories and it is necessary to take into account the errors associated with

5 105 AAS and ICP analysis, where human errors in sampling and assay lab work are involved as well as variations in assaying instrument performance. However, some statistical analysis was carried out on the results, comparing sensor results with effective aluminum values based on AAS and ICP assays. The results are shown in Tables 1 and 2 respectively. Since sensor results are effectively continuous, an attempt was made to find and use only the instantaneous sensor output at the time that each bath sample was taken in this analysis, and after calibrating to determine effective Al using the DEAL program, the results were compared as shown in the tables. It can be seen from these tables that statistical distributions for aluminum determined in the 3 different ways, were very similar. Comparing Table 1 with Table 2, it is also clear that the more bath samples analysed, the wider the dynamic range (as indicated by the minimum and maximum Al). The Al sensor could give the best dynamic range among these three methods because it reports bath Al concentrations continuously with time. Table 1 Al concentration statistics as measured by Al sensors and AAS plus DEAL. Analysis Average Al Standard Minimum Al Maximum Al method (wt%) Deviation (wt%) (wt%) Product 1 Al sensor (23 Samples) AAS+DEAL Product 2 Al sensor Samples) AAS+DEAL I Table 2 Al concentration statistics as measured by Al sensors and ICP plus DEAL. Analysis Average Al Standard Minimum Al Maximum Al method (iv-t%) Deviation (wt%) (wt%) Product 1 Al sensor (92 Samples) ICP+DEAL Product 2 Al sensor (213 Samples) ICP+DEAL

6 106 Sensor Life During the first 2 weeks of the run, there were serious problems with mechanical damage to the sensors, causing them to be removed and replaced before reaching the normal end of life. The damage was from fractures and short circuits caused by contact with drossing tools and splashes of zinc during drossing. Remember that this pot was very small and congested, allowing very little room for operators to avoid the areas of the sensors during skimming. In some cases, the sensors had to be removed from the zinc prematurely, because of sink rig changes associated with product changes and normal submerged hardware maintenance. At this point in the run, protective stainless steel cages were installed on both sensors. These were designed to protect the sensors from contact with drossing tools. Improvements were also made to shielding of the wires from molten metal splashes. It was recognised that a longterm production installation would be hard-wired differently to prevent this type of damage. After these changes, the average sensor life increased to 3.4 days, and most failures were due to the natural expiration of the sensor, instead of from mechanical damage. This natural failure of the sensors would be difficult to determine without having the second sensor to compare with. The normal failure causes sensor output to decay or to increase gradually, and this can be confused with actual pot aluminium changes in some cases. DISCUSS ion & Conclusions Test Severity This test was demanding due to the small pot size, frequent product changes, including large swings in bath aluminum and temperature, and the disturbances caused by pot hardware changes. Sensor Accuracv Comparison of outputs from the 2 sensors with each other and with over 500 bath assays, showed excellent accuracy. Sensor Life Premature failures occurred early in the trial due to mechanical damage from pot drossing in the congested work space of this very small pot. After protective cages were installed, the average sensor life over 32 sensors was 3.4 days. It was apparent that industrial ruggedness was good. Detection of sensor failure would be difficult without 2 sensors operating at the same time. Gradual changes in sensor output in the early stages of natural failure can be confused with normal bath aluminum variations. Cost-Benefit The sensors are accurate enough for production control, and conventional bath sampling and assays can potentially be eliminated. Direct and immediate readings of aluminum at all times reduces operator uncertainty and the use of sensors has the potential to

7 107 improve product quality and reduce off-grade production. If 2 sensors are required at all times, the cost can be significant. In the case of the line used in these trials, for example, the cost was calculated to be 40 cents per ton of product. Of course, this cost depends directly on line speed and production throughput, as well as product gauge, etc. The line used in this study was not as new and fast as most lines in use today, and so this 40-cents cost figure is not expected to be typical. If a single sensor could be used, the cost would become more acceptable in general. This and other potential cost reduction alternatives need to be developed and one approach is showing promise in a research-project as a means of detecting sensor failure. These factors are less significant when sensors are used as a tool for understanding the operation of a line and for monitoring only during more critical times, such as product transitions and production problem solving. Future trends Yamari aluminum sensors are becoming an important bath management tool, promising to be the most modern, most informative and easiest method of controlling bath aluminum during galvanizing production. With completion of current efforts by Cominco and Yamari to detect sensor failure automatically and to reduce sensor costs both directly and indirectly, it is expected that this will become the predominant method of measuring aluminum in continuous galvanizing production. A logical step in the future will be to automate the zinc bath additions, based on continuous signals from aluminum sensors. Acknowledgement The authors would like to thank STELCO Inc. for their support in carrying out this study. References 1. E.W. Fossen, J.P. Landriault and M.G. Lamb, Aluminum Control on Stelco s Z- Line, in Proc., Galvatech 95, Chicago, II, USA, ISS, 1995, V. Jagannathan, The Effect of Steel-Bath Interfacial reactions on the Surface Quality of Galvanized and galvannealed Steel, in Proc., 37 Mechanical Working and Steel Processing, Hamilton, Canada, Oct.22-25, N. Qiang, N,-Y. Tang and G.R. Adams, Applications of Al Sensors in Continuous Galvanizing, in Proc TMS meeting, Orlando, USA, February 1997, E691-92, Standard practice for conducting an interlaboratory study to determine the precision of a test method, Annual Book of ASTM Standards, Vol , 1992.

8 108! Shrinking (black tube scaling) lectrolvte Notch snap for I tube Figure 1 : Yamari aluminum sensor construction. Walkway \ Al Sensor or (East) (West) Zinc Loading Area Figure Sensor 2: locations and pot layout.

9 109 Al sensor Figure 3 : Sensor holder construction. Figure 4 : Typical computer screen displaying aluminum and temperature.

10 110 Figure 5 :! I I : : I II... : : : I I0 Date (September, October) Figure 6 : Typical chart showing Yamari sensor results together with bath assay results.