This site uses cookies in order to improve your user experience and to provide content tailored specifically to your interests. Detailed information on the use of cookies on this website is provided in our Privacy Policy. You can also manage your preferences there.

By using this website, you consent to the use of cookies.

Learn more
OK
Internet of Things

“Data is the new gold” can be heard everywhere. Outside the production world, no one would disagree, however, can this be also true for all the production lines all over the world?

Millions of Process-Data are being generated during production cycles on the OT (Operation Technology) level by sensors, controllers, cameras, contactors etc. Most of them are being used in the sequence program of the PLC (programmable logic controller) but these data have much more potential as just the input of an if/else function.

Why not use these Process-Data for your prediction? For example, as a prediction for the next maintenance cycle of a machine, or for determining the quality of the produced item. As we know, predictions based on Process-Data are much more efficient to ensure for example 100% quality than to organize expensive testing facilities.

I would like to show you now, how Process-Data can be used to guarantee 100% quality. For this purpose, I am taking you on a journey to our Siemens factory in Amberg, where more than 16 million SIMATIC products (Controller, HMI etc.) are being produced. 75% of the value chain here is automated, making the factory one of Siemens leading manufacturing sites for digitalization around the world. Our customers demand 100% quality for the SIMATIC products. Consequently, PCB (Printed Circuit Board) of the SIMATIC controller is tested by X-Ray(s) during the manufacturing process to ensure absolute excellence of the solder contact. But the problem with X-ray tester is that it requires long throughput time and hence it is the factories bottleneck. An additional X-Ray machine would result in capital investment of €500,000 and must be integrated into the manufacturing environment. This is why we decided to face the challenge of relieving the X-Ray bottleneck by developing an algorithm to predict the likelihood of manufacturing defects and thereby to increase the throughput in the production. How was that possible?

1. Big Data become Smart Data. First, we wanted to gain transparency – Plenty of data are being generated at the soldering OT environment which is also called Big Data. 40 different data sets such as XY-Offset, soldering temperature and volume of solder paste, were defined for the quality prediction by Siemens Data Analysts. In some cases, we didn’t even have the data yet to reach our goal. This meant we needed first to get the transparency by installing the right sensors and automation with TIA (Totally Integrated Automation), in order to get the data very easily in a standardized way available, and then to structure and prepared the 40 different records to converted into Smart Data.

2. Algorithm becomes Artificial Intelligence (AI): The next step was, collecting data and training the algorithm – we had to gather Process-Data from the physical equipment on the OT level together with the X-Ray results of produced pieces in order to train the AI algorithm. In our factory in Amberg, we collected Data from the solder paste printer, Solder pate Inspection, Pick & Place and Oven & AOI (Automatically optical inspection) and transferred it through the Siemens TIA Portfolio (controller, Edge Device etc.) into our Cloud System MindSphere. There, the AI algorithm was trained based on much more powerful CPU (central processing unit) and more RAM (random access memory) than those of a local PC. As the algorithm was trained, the model was put in an Edge application, which took care of the data acquisition and the data preprocessing and then guaranteed the quality prediction for the PCB production.

3. AI Algorithm is ready for prediction. The final step was uploading the AI algorithm to the Siemens Industrial Edge Management System. In this step, the Edge Application was deployed on the local Edge Device from the production, in order to analyze the data locally, without having to transfer the data to the Siemens Cloud MindSphere anymore. As otherwise the Re-training was needed. The important production data stayed inside the production critical level. No interaction with the cloud was needed at that point, therefore, the prediction could have been done in real-time. The result of the prediction: “if an additional testing via X-Ray is necessary or not” was fed into the manufacturing execution system (SIMATIC IT) so it could then decide whether to send the PCB to the X-Ray machine or not. If the X-Ray was skipped, the costly bottleneck was avoided. We call this local interaction with the process “closed loop analytics”: meaning analyzing the data and using the results to optimize the underlying process in real time.

In Amberg, we currently work on optimizing the machine learning workflows in automation from the data acquisition, data preprocessing and AI training into Model deployment. These improvements will give us the opportunity to continuously train the algorithm in the Cloud (MindSphere) with different production lines all over the world. It will help us create an even more powerful algorithm. In my opinion, if this closed loop is running, the full bloom of Artificial Intelligence will be shown based on Process-Data. It will change the way machines are being developed, as machines efficiency will be automatically increased without the need of writing a new program or updating it manually.

OT meets IT. In our factory in Amberg, there are plenty of these real- examples, of how Siemens challenges the integration of OT and IT by using new technologies such as Edge Computing, Cloud Computing and AI.

How much potential lies dormant in your factory? The challenge as I see it, is to understand the process, get the transparency of all available data and finally see the potential value of the data. In our case, we showed, that we can reduce test efforts for the X-Ray machine in up to 30% in the first step. We continuously train our algorithm (AI) to reduce permanently test efforts until zero X-Ray tests are needed. Furthermore, we showed that the costly bottleneck in manufacturing can be avoided and the capital investment could be reduced by €500,000.

For us, Process-Data is the new gold in the production process – it will pay off if you dig for it!

Comments

2 comments