CARACTERIZATION OF STRIP VIBRATION AT THE WIPING NOZZLES

Transcription

1 CARACTERIZATION OF STRIP VIBRATION AT THE WIPING NOZZLES GAIGNARD Sylvie, DUBOIS Michel R&D ARCELOR Wallonnie 1 Rue SOMPRE 44 IVOZ RAMET Belgium Tel (32) (4) Fax (32) (4) Michel.Dubois@arcelor.com ABSTRACT All strip movement in front of wiping nozzles leads variations of the zinc coating weight. To improve the zinc uniformity and thane reduce the overweight which must be done to warranty the minimum thickness specified by the customer, the strip movement at the wiping nozzles should be reduced. However before any improvement can be done, the first thing to start with is quantification of the amplitude and frequencies involved in relation with process parameters as strip size, pot hardware adustment, power of the cooling tower A portable measuring system has been elaborate to measure the strip vibration oat the nozzles on industrial line. It consist in 3 inductive sensors: 2 are localised close to the strip edge and the last in the centre of the strip. This system has shown reproducible results provided the right calibration procedure is used. The device is very easy to install on line and safe to use. The results have been compared to those obtained by lling and show the similar trend for the effect of strip tension and size. Typically a peak to peak amplitude is from 2 to 3mm which represents 2% to 3% of the nozzle to strip distance (typically versus 1). It then predicted a coating weight variation of: 2 to 3 µm for a 1µm coating. The frequency analysis shows that whereas all the frequency spectrum is below 1 Hz, the submerged equipment is not the only responsible of the vibration: the higher the cooling power in the tower the higher the amplitude. The expected beneficial effect of strip tension not clear for the range of 2 to 2.kg/mm².. BACKGROUND Mastering the coating weight uniformity is a key issue for coated product ands especially galvanized steel. Non uniform coating in fact implies that excess thickness has to be realized to be sure the minimum value specified by the customers are warranted on any point of the surface. On a theoretical base, the aimed value by the operators should be the minimum single point value added of 3 standard deviation of the coating weight distribution. The standard deviation is mainly related to any variation of the nozzle to strip distance during wiping. The typical overcaoting cost is estimated at.2e/t/excess gram /m². Taking into account that any saving on one side induce the same saving on the other side, the total stake for an automotive line is estimated at 1kE par gram (T with 1 gram saving on each side) The variation of the nozzle to strip distance in a well known and may be a maor problem of the hot dip galvanizing process which affects the coating uniformity. The variations can be:

2 either in the transverse direction : in that case they can usually be corrected by adustment of the nozzle skewness or the positioning of the submerged roll to correct the strip cross bow or related to time as due to strip vibration: poor rotation of the pot hardware and/or vibration induced by the air impingement in the cooling tower are the main responsible. If the strip crossbow has been studied by some researchers and then quite well described in literature, the strip vibration is conversely a undocumented topic. Many process people can and have talked about it but values for amplitude and frequencies observed in various process widows are never reported. The only reference known are papers of LTV and Cockerill Sambre at the Galvanizers meeting (REF) The first proposed a simple equipment to use in line for characterization and the later showed the effect of the cooling tower on the measured amplitudes at the nozzles. If the peak to peak amplitude of the vibration is as described in the previous reference in the range of 3 to 4mm at the nozzles distant from the strip of about 1mm, the use of wiping ls predict a coating weight variation of about 3 to 4µm on a 1µm coating, typical thickness required by the automotive industry. In addition to a poor Zn yield value, the galvanized surface can be affected by areas of different brightness showing the vibration pattern. The stake related to an improvement of the coating weight uniformity is very high keeping in mind that every gram/m² saved will decrease coating cost by.2e/t or 1E/year /line for a 1 gram Zn saving on each side. OBJECTIVES Driven by the high stake related to the coating non uniformity, work has been conducted to better understand and quantify the strip vibration at the nozzle. This is indeed only a first step before any improvement on the pot process can be done. It is in fact a first priority to quantify the measured amplitude on different lines independently of the pot equipment installed and to identify the main frequency peaks which are a signature of the device inducting the vibration. The final obective is to reach a peak to peak amplitude lower than.mm, a quite ambitious value for a 4m strand from the sink roll to the top Roll. The coating weight can then be expected to vary by less than 4g/m² in the longitudinal direction. 3 key obectives were followed in this work : definition and construction of a system easy to implement and use in different line of the Arcelor Group to measure the strip vibration at least on 3 point across the width, identification of the right procedure and data processing to obtain absolute amplitude value, comparable from one measuring campaign to another and from one line to others, Characterisation of some process parameters on the frequency spectrum and amplitude, firstly the cooling power in the tower, the strip size and the strip tension. Because the measurement is done at 3 point on the width, in addition to the identification of the frequency peaks specific data processing was conducted to e able to identify if the vibration is Pure string, Twisting and Flapping (Fig 1) time 1 Pass Line time 2 String Mode Twisting Flapping Fig 1 Vibration Mode EXPERIMENTAL DEVICE Measuring the strip vibration reliably in line with an accuracy obective in the range of.1-.2mm is not an easy task due to the hostile environment: hot ambiance, liquid Zn with high brightness, variation of thickness and strip, cost of line stop in case of problem on the devices.

3 If many systems could be used in a first approach, the laser type had to be reected due to its cost and the environment. The final choice was an inductive type from TELEMECANIQUE based on the very successful results obtained by LTV. The system is very cheap, easy to install, and the signal easy to process. The principle consist in the measurement of the reluctance of the magnetic circuit (the sensor, air and the strip) due to the gap variation between the sensor and the strip. The only drawback is the possible dependence of the signal on the magnetic properties of the sheet (Fe+Zn) as well as drift due to temperature variations. Preliminary studies at the lab as well as some uncertainties about the possible variations of the magnetic properties of the strip at 4 C drive us to the conclusion that it is compulsory to conduct an in line and systematic calibration procedure ust before each measuring campaign. Ideally, it is recommended to ensure that the sensors have no drift during the run (for example due to temperature). This is verified by re conducting the calibration procedure. To minimize the drift in response or failure due to overheating during operation, the sensor has to be set in a cooled housing which uses compressed air for simplification reasons. The device is then only designed for experimental purpose and not for continuous use. Because the vibration, as described in Fig1, is as important as the frequency spectrum, at least 3 sensors have to be use at the same time and at different locations on the width. They are supplied by DC courant and in series with a multi channel converter to 4-2mA output and have a filter to protect the data logger. This one gives 3 variable voltages with the distance of each sensor to the strip which are then processed by a software written to compute directly the FFT of the signal. This procedure allows to have quickly frequency results and verify the record quality. The 3 signals are then filed for post processing and detail analysis. The mechanical fitting of each device on the beam supporting the nozzles is done by magnets. This method is very easy and well working provided stainless steel is not used for the beams. Of course the beam must be very stiff in order to not induce artefacts in the measurements. Each sensors is cooled by air which is supplied by a small plastic pipe. FIG2 : Sensors on the nozzle beam on the back side Fig 2 show the implementation of the system. If 6 units seems to be present it is due to the reflection of the 3 sensors in the bright liquid Zn. It will also be noted that the sensors are not connected together which means that a cross bow profile cannot be identified with accuracy with the equipment as such The inline procedure consist in first positioning the 3 sensors at about 3-4mm from the strip. With that distance range, the sensors gave an almost linear response for the total strip displacement expected (typically 2 to mm). If that distance is quite reasonable to avoid touching by the strip, it is nevertheless still quite close in case the sensor would be used continuously, including the problems of heat buckles, poor welds, unexpected pass line shift. Because the distance cannot be measured with reliability due to the strip movement, the correct positioning is checked by verification of the measured voltage which must stay in the right range, typically 4 to 6 volt. Sampling is done at 4Hz during 2.6sec which correspond to 124 points. The sampling rate of 4Hz was selected because in previous studies had shown that no frequencies over 1Hz exist. This is quite normal due to the very long strand in the tower.. The number of points helps in the FFT computation. Whereas a higher number of point would improve the accuracy in the frequency peaks, the in line experience showed that increasing the recording time increase also changes in the line conditions: new coil, line speed change

4 To have reliable results, 2 consecutive measurements are done in stable processing conditions. It means especially on the same coil, with the same line speed and without changing the submerged roll set up which will change the cross bow. DATA PROCESSING & INTERPRETATION The data processing consist in a quick in line verification of the results by the use of a software implemented in the data logger. To be able to interpret the results, the recorded files are reprocessed by a Mathcad software which does the followings: Conversion Volts to mm based on the calibration curve of the day Conversion of the distance measured at the sensors to distance at the nozzles. This is done based simple proportionality based on the distance ratio Last Touch Roll to Senor /Last Touch roll to nozzle That ratio can be different from one campaign to another for example du to a disfference in the wiping height. Computation of the RMS, peak to peak values for each sensors FFT calculation of each signal Drawing of few graphs. Some corresponds to the distance variation with time (FIG 3) but others are especially dedicated to the identification of the vibration. If a phase analysis seems the best tool to identify that vibration, it was found to be not practical because too sensitive to the exact value of the frequency peaks which can indeed be slightly different on both edges. That s why it aws preferred to do contor plot which give immediately a visual evaluation (Fig 4) It will be noted that the data processing can give valuable results only if the centre position of the strip has not changed during a trial as it is the case on Fig 3 Fig 3 : pass line shift during a record (Distance (mm) versus Time (sec) MAPTno Fig 4 Contour plot Color f(amplitude X=sec total Y= width If the FFT is a well efficient practice and the RMS value quite easy to understand physically, the interpretation of then results are sometimes more complex. The frequency analysis is much more difficult to interpret for the following reasons. The frequency of the string which is that of the strip in the tower is low and can be computed by : Freq = SpecificTension MaterialDensity The frequency in galvanizing is than typically in the range of.4 to.6 Hz. To compute with an accuracy of.1hz by FFT a high number of point and then a long recording time is needed and the risk of changes in the process increases. It will be noted that the higher the tension, the higher the frequency. It is than an easy mean to identify the source of a peak. However, because the frequency varies with the square root of tension, shift in the

5 peak location can only be observed is the tension variations are high enough. Typically, at least 1% variation is need to expect to see something. Moreover, the vibration frequencies are not only due to the string. Flapping and twisting can also exist and it is only by analysing the phase between edge and centre that identification of the true behaviour of the strip can be obtained. The frequencies induced by the rotation of the submerged are computed based on the line speed and the roll diameter. They range typically in.4 to.9hz for the sink roll, very close to the range of the free string, and in 2 to 4Hz for the stabilizers. The peaks induce by the pot hardware can be more or less identify but changing the line speed as the peaks will move. The frequencies induced by the resonance of the roll rigs is another story. They are usually unknown, never computed. If that last approach could be done, cares must be taken to include the effect of liquid Zn which change the Mass value in the classic resonance formula. ϖ = Elastic _ Cons tan t Mass In practice, multiple s use to be observed. More over, sometimes the higher s have more amplitude than the lower. It becomes than very difficult for example to identify a 3.1Hz: is it a 4 th of free strand, the 3 rd of sink roll of the 1 st of a stabilizer? b PS PS freq,, 6, 8, 1 2 Fig : Identification of the 3.1Hz peaks Fig express the problem. In the FFT we observe a peak at 1.2, 2and 3.1Hz. The vertical lines corresponds to the 6, 8 and 1th of the string. The 1.2 Hz could be 2 of the string, 2Hz peak may be due to stabilizer rolls provided it fits with the roll diameter, and the 3.1 Hz may be a multiple of the string, or a first of a stabilizer roll or a resonance frequency. In order to identify the origins of the peaks and than the action to undertake on the line to improve the strip stability, the following procedures are recommended when possible. Line speed increase: in that case, the peaks due to submerged rolls will move. The variation of at least 2-2% is needed to see clearly the effect. However, the distinction between the 1 st free string and the 1 st due to sink roll is usually never easy. Variation of strip tension: with such an action, the total amplitude at the nozzle will change and the peaks will move in frequency. However the shift is never very high because the free vibration varies as the square root of tension. More than 2% variation of tension is require. Moreover, that action can modify the excitation due to the pot hardware because the friction conditions in the bearing is changed.

6 Fig 6 show a typical result and exercise when increasing the pot tension by 17%. The on a 127*.8 at 18m/min. Bothy records are done the same day in the same conditions. The Peak at.68 move to.81hz at 2.7kg/m². However its amplitude is decreased. That at 2.6 Hz remains at about the same location. It is suspected it is due to the stabilizer roll but it is not the case because it does not fit with the roll diameter which would give 3.48 Hz. It comes than then the peaks at 2 and 2.6Hz are not explained. May be is it resonance frequency of the rigs PS kg/mm² freq PS kg/mm² freq Fig 6 : Effect of Tension increase RESULTS Calibration procedure The calibration procedure is a maor key to obtain reproducible quantitative results. The principle consist in recording the sensor signal for the reference position (the position in which the sensor will be located for the measuring campagn), and after a known deplacement btoward the strip: typically and 1mm front and back. The average volt measured is computed and the results plots in a graph. A best fit formula is then computed and used during the data processing to convert the volt measured in function of time to distance in function of time. The exercise is done of course for the 3 sensors at the same time. It is usual to find different regression formula for each sensor. Typical results and formula found are shown on Fig 7. With such a procedure, the accuracy on the RMS value at the nozzle is estimated better then.1mm. P1 right side when looking to the pot P2 center P3 left side when looking to the pot FIG 7 Typical calibration curve

7 Results 3 Industrial lines from the Arcelor Group have been analyzed. Each of them have different design characteristics. Especially, the pot Hardware may be with 2 or 3 rolls and the tower height from 4 to 6m. Tower height (m) Cooling length (m) Nb of Pot Rolls Ligne A Ligne B Ligne C Table 1 All the run for a line have been conducted with the day that means with the same pot equipment. The process condition used for the campaigns are reported in Table 2 as well as the dimension of the roll the the concerned day. The Production in the 2 nd column is in Ton per hour for a 1m strip width. This value is interesting to know because the cooling power in the tower depend on it. Productivity * T/h/m Strip Size mm 2 Speed m/min. Rouleau de fond Stabilizer Roll Corrector roll Line A * Line B 127* Line C * * Table 2 Frequencies The frequency range observed on the 3 lines extend from to 1Hz and with the maority of peaks below Hz. Fig 8 shows typical FFT results. b PS PS b PS PS freq,, 6, 8, freq,, 6, 8, 1 2 Fig 8 Typical FFT Results The identification of the origins of the frequency peaks has never been easy. We have been able to observe the following things: Peaks due to the string with shift when the tension increases. However, there are cases in which it seems that higher s (3 to 6) are more important than the lower. Peaks to stabilizer and corrector rolls, moving with the line speed Peaks we attributed to resonance frequency of the hardware because to displacemlent was observed with line speed When the tension increases, some peaks are increased and others decreases. We suspected that the excitation was higher with higher tension in relation with an increase of friction in the bearing and then poorer rotation No general observation and trend could be drawn It seems that the best method to improve the understanding of the peaks is to note in a table the 3 or 4 frequencies which has the highest value for each process window.

8 Amplitude From the measurement done on the 3 lines, 3 main process variables have been tested : the strip tension, the cooling power in the tower and the imbrication of the stabilizer roll. Obviously when the cooling rate increases, the strip temperature at the top roll decreases All the results from the 3 lines are reported in RMS value at the nozzle (average Right middle left) in RMS(mm) 1,3 1,1,9,7 Width13 mm Width 18 mm Width16 mm C Line A, B & C A B 2, 2, 1, 1, RMS(mm), Cooling Power in % rotation, Fig 9:Effect of Cooling Power on Average Amplitude function of the percentage of the cooling power in the tower. used in the Fig 9. In the scale of line A & C is on the left, it is on the right for line B to make the graph more clear. It will be observed that the values are in the same range :.6 to 1.1mm for all 3 lines. It will be observed that the higher the cooling rate, the higher the amplitude. Some reader will maybe find the peak to peak amplitude very high as it is up to 4mm at the nozzle (4 times the RMS value) but that order of magnitude is consistent with what can be observed on video made on the line. It has to be kept in mind the amlitude discusse here is the average of the 3 sensors and it is usual that one location, either the center or edges exhibit much higher values a shown on Fig 1 The roll imbrication was alos observec as a key parameter in the amplitude. Fig 1 gives the results for line A for the conditions specified in Table 2.It is observed that increasing the roll intermesh increase substancieally the amplitude on the edges but not in the center. This is not fully understandable on a theoretical stand point and the observation could be related to a change in the process window in the furnace or in coil which if they have the same specification and run at 2 min interval are still physically different. This type of results show the complexity in the interpretation of the results as well as the clear identification of the process parameters. The effect of tension at the pot was found complex. As shown on Fig 11 & 12. as it can be seen on Fig 1 and 12, enter and edges are not always affected in the same way by a tension increase. This is presently RMS mm 4 3, 3 2, 2 1, 1, Motor side operator side Centre Fig 1 Effect of Stabilizer Roll Intermesh on Amplitude at nozzles Imbrication -4/-4 Imbrication +8/+8 explaine either by the fact that coils are different but also because when increasing the tension is increased, the stresses in then roll bearing increases and can deteriorate the friction behavior.

9 Fig 13 shows the FFT spectra of the case 12 with the shift in peaks and their increases in amplitude To investigate the effect of strip tension independently of the stabilizer roll effect (supposing perfect rotation,), numerical simulation have been done. The strip is led on its width and length in the tower RMS at Nozzles (mm) 1,9 1,7 1, 1,3 1,1,9,7 162 mm 184 mm, 1,6 1,7 1,8 1,9 2 2,1 2,2 2,3 2,4 Tension en kg RMS at Nozzles(mm),8,7,7,6,6, Center edge 3,3 kg/mm 2 2,24 kg/mm 2 3,3 kg/mm2, case Fig 11 Amplitude Versus tension Line C Average of 3 sensors Fig 12 Amplitude Versus tension Line A Each sensor separately b kg/ m m 2 : M ode 1 :.46 M ode 3 : 1.38 M ode 6 : freq,, 2, 3, 4 b PS PS kg/ m m 2 : M ode 1 :.4 M ode 3 : 1.6 M ode 6 : freq,, 2, 3, 4 Fig 13 Peak shift in frequency and amplitude with tension and the excitation source located at the height corresponding the cooling devices in the power. For a 1 kg/mm² Computed RMS (mm) Référence.2 1 kg/mm2.1 2 kg/mm² Location on width 2 kg/mm Kg/mm² Fig 14 : Computed Amplitude at Nozzles for same Excitation constant strip size and excitation, the vibration is computed at the nozzle for different tension levels. Typical results are at fig 14

10 It comes out that if it is exact that doubling the strit tension from 1 to 2 Kg/mm² decreases significantly the total amplitude, on increase of only 1% has an insignifiant effect no effect. The amplitude is not linearly dependant on the strip tension. Conclusions A study was conducted to quantify and analyze the vibration of running strips at the wiping nozzles. Therefore, an efficient, simple and reliable sensor was identifies and implemented on a transportable rig, easy to implement in line. A strict procedure has been identified and followed to generate comparable and reliable results. The estimated accuracy of RMS value of the amplitude at the nozzle is better than.1mm. 3 lines of the Arcelor Group have been characterized and it was possible to measure the frequency s and quantify the total amplitude in various cases. 3 sensors are used to record simultaneously edges and center. The effect of roll interemesh, strip tension and cooling power in the tower has been estimated The order of magnitude of the peak to peak amplitude for the 3 lines has been found between 2 and 4mm for strip width from 12 to 18mm and thickness between.6 &.77mm. The frequency spectrum is always below Hz but the identification of the origin of peaks appeared to be not easy and even risky. The variable which increases systematically the amplitude is the cooling power in the tower. For the strip tension, shift of frequency peaks are seen. However, regarding the amplitude, things are not as obvious as expected because it is suspected that the excitation source from the pot rolls varies with tension through its effect on the friction behavior in the bearings. Moreover, the same observations can not be done on center and edge of the strip. The stabilizer roll imbrication has also an effect on the amplitude and is that sense lower values use to induce lower vibrations. Here also is speculated that friction in the bearing is the responsible of the observed changes Future Work As it was quite difficult to identify the origin of the numerous peaks observed in the computed FFT, in the next trials, the strip vibration measurement will be coupled with the vibration measurement of the pot Hardware. This should improve significantly the identification of the frequency peaks. However, it will never be easy to identify the exact cause and effect of an observed vibration because of the high mechanical coupling between the strip and the submerged rolls. Finally, because measuring campaigns are always a sample of the normal production happenings, a device able to measure continuously the strip vibration will have to be implemented. Correlation between amplitudes and process variables will than be much easier, independently of the natural various existing from one coil to another. References