Automating Control Charts with JMP Live

Hello. I'm Ruben Lenaerts. I will tell you a little bit about how we used JMP Live at JSR Micro to automate the control charts that we have for our finished products. At JSR Micro in Belgium, we produce around hundred different products, and all of these products we have to control. On top of that, we also have to control all the raw materials that we use to make these products.

One product has between 20–400 different parameters, so it means that daily, we have to really go over a lot of control charts and really have to follow up a lot of materials. To do that, we are using JMP Live, and I will show you here how we did it. Most of our data is already stored in the SAP 4/HANA database. On top of that, we also have data in the OSI PI data historian.

Our production team, they use OSI PI directly. They follow up continuously the tanks and all the production parameters directly from the control charts in OSI PI. But then for the finished products, then we have all the QC measurements that we store in our SAP 4/HANA database and which implies we can continuously have access to the latest data in these databases. We also have a complete control over the reports that we want to make and how we want to process that data.

JSR, it's a company that was established in 1957 in Japan. It is a Japanese multinational with around 8,000 people that work in that company. They're mostly working in organic chemistry. They have a division for making polymers, a division in the life sciences, and then the division that I work in, that's the semiconductor division.

More about the semiconductor business that we work in. We make photoresist. If you have silicon wafer or silicon carbide wafer, for example, and you want to make the patterns in it, then you first put a photoresist on top of it. That photoresist that you can expose to make the patterns that you want to make the device that you want to make.

In NV, in Belgium, we produce KrF and EUV technology. If you are in the EUV scale, it means you make patterns that are below 14 nanometers. A human hair is 100,000 nanometers, so it really shows that you are at a very small scale. If you're in that small scale, then a couple of things are important. Like, you cannot have particles in your photoresist. You cannot have a lot of metals, because metals, they will make the structures that you make behave not as they should behave, which can cause a defect. On top of that, you also have to control the functional parameters, which is film thickness of your resist and sensitivity. After exposure, what kind of latitudes, what kind of structure they will give. To control all these parameters, we use JMP Live.

Now how we do that? I will show you that now. First, we need direct data access to the database where the finished products are in. For that, everything is in SAP, so we can use the ODBC driver, which you can directly use within JMP with the query builder, for example, here. You can also script it using the query builder in JSR codes, or you can script it with a Python code. It doesn't matter which way you use. With all these three methods, you can access the latest data of the SAP 4/HANA database.

On top of that, also have the CoA data of the raw materials and the finished products in SAP, the internal control limits, and the customer limits. All of these data, you can access with the ODBC driver in JMP. Then to make the SPC Charts. All customers have a certain way of looking at the SPC of the products that we make. We want to look at the same way at our data as our customers do, that we are on the same page as our customer.

How they do that is they calculate the control charts without the last batch because it is the last batch that you want to evaluate. Then you make the SPC of the last 20 batches only because the older batches are less representative of how the process is going at the moment. On top of that, they can have extra SPC rules, for example, looking at rule 2 or rule 3. All of these items you can configure completely during the scripting in JMP.

If you have the data, and you did the calculations, then you can make the reports. How we did that is that you make a batch report of every batch that you produce, and on that, you have one tab, so one sheet in the report, where you have all the control charts of the parameters that are out of spec or out of control or have a violation of rule 2 and rule 3. You have a very fast overview of the items that are wrong with this batch.

On top of that, you also have two other tabs, one with all the control charts of the last 20 batches and one with all the control charts of all batches that you have produced for this product. If something is wrong, then you can immediately look at the control charts of the last 20 batches and follow the batches to see the long term trends. On top of that, we also generate a report on the shared drive, which can be added to the data packages by the production team.

Then I will show you how it looks like in JMP Live. This is an example of how the SPC report looks like in JMP Live. This is, of course, a fake material and it has fake parameters. But you can see that for this product, for example, you have Parameter AA that is out of control but in spec. This one is also out of control. Then you have your first overview of what is out of control. This parameter, for example, is out of spec.

If you then want to know more about your parameters, for example, Parameter AA is linked to one other parameter, then you can go to the other tab with all the parameters and see how a related parameter, for example, AC behaved. This report then is updated every 5 minutes if new data is available. If no new data is available, it stays like it is. In one report, you have all the data available for this material. These reports we then make for all the materials that we produce.

To conclude how we set up the automation of our control charts for finished products with JMP Live. The first thing that is important is that you have to organize your data. You have to organize your data in a clean and structured database. In our example, it was SAP 4/HANA. It's also good if you implement some data validation on that. For us, all the data that is entered gets double-checked that all the data is really correct. Otherwise, your analysis that you will do can have some errors in that. Then you can also script the report making. It's making use of the specific analyst that your customer or your company uses, so you have complete freedom of that thanks to JSR. Then you can make it available to all users in your company.

I want to acknowledge all the people that helped in this project. Sorry. Mostly the people in my company and also someone from the value chain that helped us with installing the ODBC driver on SAP. I thank you for your attention.