Hello, my name is Stephane Gouttebroze, and I will present you this document entitled Zinc Coating Excellence: Leveraging JMP for Accurate Pressure Control and Thermodynamic Modeling. This is the content of my presentation. First, a short introduction on Clecim. After, I will speak about the background and experimental setup. After, I will make a short jump demonstration, and I will finish with some conclusions. Clecim, who we are? We are a French integrator of equipment for the global steel industry. We are located in France, near Lyon, in Montbrison.

The company is 100 years old, and we are about 200 employees in this company. We have a design office, workshop, and on-site installation and commissioning. The main activities of Clecim are the supply of individual machines and also of complete production lines like pickling lines, annealing, galvanizing, painting, and so on. Clecim is also able to provide service like spare parts or maintenance.

We also have a consulting department which can bring expertise and process audits to our customers. The examples of installation are the following. For example, rolling mills, hot or cold, leveling lines, surface inspection systems for strip quality, and laser welders, and many more.

On the bottom of this slide, you can see a galvanizing line example dedicated to automotive market. The length of this line from here to here is about 500 meters, and the approximate height is 35 meters. In the following slides, we will focus on this area, which is the zinc pot area.

Concerning the background and the experimental setup, we are in the zinc pot area, and here you have liquid zinc bath. Here you have to imagine a strip going out from a furnace here, going into the zinc bath and going out after being coated by zinc in this liquid zinc. The strip is continuously in motion. It's a continuous production. You can imagine that once the strip is going out from the liquid zinc, there is a coated zinc layer on its surface.

For quality and productivity and corrosion resistance, we need to be sure, or we need to ensure that there is a constant zinc thickness and a homogeneous zinc thickness along strip width and strip length. This is ensured by a specific device, which we call air knife. It's this design here. It's in fact just like an air dryer. It is blowing air onto the strip surface in order to remove excessive zinc.

The objective of the study were to conduct experimental measurements. We also want to model the impact pressure with respect to various parameters and also to compare with finite element modeling. At the end, to adjust the finite element model in order to be in accordance with what we have measured. Here you have the results of the modelization with finite element, and we will now switch to the test rig.

The test rig is, you can see here, the air-blowing system in functioning. Here, you have the air knife device. It's the width of this device represented by the Y axis, and the altitude is represented with the Z axis. On this test rig, we added a small part here, which is a metal sheet, instrumented with a pressure transducer in order to measure the air impact pressure on this metal sheet. This will be useful for representing the pressure of the air knife on a fictive strip.

This shows you the experimental measurements. We have one measurement of the blower speed, which will be varying and which will change the pressure inside the air knife. We have also three position X, Y, and Z, can can vary for the measurements. We measure the impact pressure here with the pressure transducer and the air temperature. With all these measurements, we can build a data table, and we will see after this table.

The workflow of the study is the following. We start here with the measurements. With these measurements, we choose X and Y position, which are fixed, and we are varying Z position in order to measure the impact pressure. We see that the curve will have this shape. With this shape, we decide to model as a Gaussian curve with the tool in JMP called Gaussian fitting curve. We will have now three parameters representing these curves, A, B, and C, which are the peak value, the critical point, and the growth rate.

These three parameters will be incorporated in our data table, and we will make some measurement correction. For example, we will add a Z offset in order to recenter all the curves because our experimental device can drift a little in Y axis. After that, we have chosen to model the peak value and the growth rate as a function of process parameters. We will make a new table with predicted formula in order to be able to predict pressure in all the 3D space in X, Y, and Z, and also in motor speed and air temperature.

The objective is to obtain this contour plot. This contour plot here is representing predicted pressure with respect to Z value and X value with air temperature fixed and the blower motor speed fixed. We have to remain this curve, and I will show you now how to obtain this contour plot.

Now, I have to switch to JMP. I will open the data table. The data table is here. You have here 13 measurements with the same date and hour, which are representing a set of measurements for a given motor speed, X, Y. We are varying Z value from minus three to plus three, and we measure impact pressure, and air temperature. I can show you all the measurement points here. You can see that we have chosen to make X values like this, Y and Z values. We don't have an optimized representation. In the future, we would like to do a design of experiment in order to have a better representation of this 3D space.

If I plot now the pressure with Z, you can recognize this Gaussian curve shape. All curves are not aligned. It's because of the phenomenon I explained previously with the drift in the test rig. The objective now will be to model all these curves as a function of these parameters. In order to do that, I use the Fit curve platform in JMP. I choose to fit-curve pressure as a function. Sorry, as a curve like this. Do you recognize the three parameters previously introduced? With this, I do the sitting curve with respect to all set points of measurements.

Now, with this, I can incorporate, sorry, I can extract the values of the parameters in a new table like this. You see that with this measurement, the same date and hour, we have the three parameters and the three values. I have to modify a little to split this table in order to be able to incorporate it in the original one. I use this split function. Now a row is representing the critical point, the growth rate and the peak value for each test point.

Once I have this, I can integrate this table to the original one. This is done by doing this. Now you see the original table here with these columns added. These four columns are just the calculation of a fit curve platform. This one is a formula I created. It's just to recenter the Z value with respect to the critical point. If I use this value now, and I can plot the pressure with respect to this value, and you see that now all the curves are aligned to zero.

With this curve, now I am interested in knowing or in finding which parameters or how the parameters are varying the peak value, for example. This is what I can show you here. If I change the motor speed, for example, how it is changing the peak value, or what is the influence of the X value or the air temperature. In order to obtain these two profiler, the same for the growth rate. In order to obtain these two profiler, I have to use a new platform, another platform. It's the fit model platform. You can find it here. I choose to keep only three parameters with, it's a linear model. For me, it's a good model because we have a good R squared.

We can improve a little this model because of here the shape, the pattern in the residual values seems to be optimized, but we choose to keep like this because it will be a simple formula. I can show you the formula. The prediction expression. You see it's linear and it's very simple. We decided to keep this model. With this model, I can add the prediction formula for the peak value. Here you can see it's added with the formula you have here, you recognize it. I will do the same for the growth rate. Same conclusions. We have a very good R squared, only the same parameters, linear also, and the same conclusion on the residual, we can maybe improve a little, but it's not the point here. We will keep like this.

Now I have two new columns, prediction formula for peak value and for growth rate. I can add, of course, the last one, the last column, which will be the prediction formula for the pressure here. If I show you the formula, you recognize the Gaussian curve formula with the prediction formula for growth rate and the prediction formula for peak value.

Now, we have a prediction formula for the pressure and as a function of all these parameters plus the air temperature. Now, the idea is to construct a new table containing these parameters with values inside the bounds of the measurements, but with, for example, an optimized repartition. I will show you how to do this. First, I will just open. I prepared the four tables. You recognize here the motor speed, the X value, the Z value, and the air temperature. What I have to do now is to combine these four tables into just one with all combination possible between the four parameters.

For that, I use the platform table, concatenate table, and I choose to concatenate, for example, sorry, excuse me, it's not concatenate, it's join. We choose to join table. We choose to join the motor speed with, for example, D, and instead of using by matching column, I have to change and to choose Cartesian join. You can see here, all combination between these two parameters I computed by JMP. It creates me a new table. If I use it again with Z and again with air temperature, you will have a complete data table with these four parameters.

I have done previously, so I will open it. Here you have my four parameters with all combination possible leads to about 3,000 rows. With this table, I can add my formula. Here we have the four columns. Here we have the prediction formula calculated previously and the prediction pressure the same as previously showed, added in these columns. The advantage of doing this is that I can change the repartition or the precision in Z, X, air temperature or motor speed.

With this table, I can now make the contour plot I show I showed you. Here, you can recognize it, and I can animate it with respect to motor speed. Now, you see that we have obtained what we would like to have as shown here.

I would like just to give you some words in conclusions. The conclusions are that we have here on the left part, a finite element model results imported in JMP. On this center part, it's the experimental results visualized with JMP. In comparison, we can plot these two curves, which are cuts of this contour plot with X constant value. Here you see a very good matching between experimental and 3D finite element. Whereas here we have some difference. In this case, we would like to improve our finite element model by changing some assumptions or some modeling parameters.

In conclusions, we can say that JMP helps Clecim to explore data intuitively, analyze and model measurements in a simple and effective manner. We validated our experimental setup. We can easily compare a finite element model and experimental results. We can also extrapolate our measurements, validate our assumptions. We can present the results in an attractive manner and communicate and convince our customers of our performance and study capabilities. This is the conclusion of my speech, and I would like to thank you for your attentions. If you have questions, please do not hesitate to contact me by email or by LinkedIn or by phone. Thank you.