Event Driven Data Analytics –
How having fun gave rise to the Industrial IoT Showcase
When our hack team from Accenture came to the Bosch Connected Experience 2017 (#BCX17), a major objective was to learn more about the Bosch IoT ecosystem. 48 hours later we had developed a new i4.0 showcase for our new Industrial IoT Innovation Center.
Design thinking and streamlined data analysis
In addition to taking part as “hack enablers” with two coaches from FJORD, my colleague Ralf Schumacher signed up a team to participate in the Manufacturing Hack Challenge. While Anna Ciechanowicz and Sonja Meriläinen explained the basic principles of design thinking to interested participants, Alexander Dieterle, Sven Mentl and I developed a project called “EDDA – Event Driven Data Analytics”. Would we be able, in the short time that the hackathon takes place, to develop a showcase for our new Industrial IoT Innovation Center in Garching near Munich with the equipment from Bosch? We were eager to see the result.
The objectives of the EDDA hack project.
Our project: Event Driven Data Analytics (EDDA)
Since it was important to us to test a typical end-to-end setup and the overall technology pipeline, we chose a typical problem that frequently shows up when designing an i4.0 use case: On the one hand, as much data as possible should be analyzed in the cloud in order to obtain comprehensive information about the optimization of processes, availability and quality. On the other hand, owners and operators are quite hesitant to share their data or to pay large sums of money for the infrastructure and cloud for the transfer and storage of each individual data point in an industrial setup. Our proposed solution: Instead of “blowing” all available data into the cloud, we wanted to make the data stream event-driven. In normal operation, we limit ourselves to the most necessary sensor data. In special situations, for instance in cases of certain process events or altered ambient conditions, we transmit more sensor data, for example to warn about an impending failure in a timely manner.
Our hack project solution.
The design of our infrastructure
The events are defined by the crossing of previously defined threshold values for the sensor data generated in our case by a Bosch XDK sensor. Since this includes multiple sensors on a small footprint, it is perfectly suitable for a demo case. The other building blocks of our tool chain are the IoT Gateway from Bosch Rexroth, the Production Performance Manager PPM from Bosch Software Innovations as well as GE Predix and its user interface Predix UI.
The data flows between the XDK sensor, IoT Gateway, PPM and Predix.
Where and how is the data filtered?
An important thing to consider when it comes to data selection is where the decision to extend the data stream should be made. In our structure, the filtering can be either close to the sensor in the IoT Gateway or downstream in the PPM. We opted for preprocessing in the IoT Gateway because the criteria for decision making when it comes to the forwarding of data can be set very quickly based on simple rules. In the respective application, the rules in terms of edge computing are ideally applied at the latest when additional costs are incurred or network bottlenecks occur.
The XDK sensor from Bosch can be connected to the IoT Gateway via Wi-Fi and delivers status and production data for the EDDA project.
Building a demo in 48 hours
To program the XDK sensor, we used the Eclipse environment for C in addition to the supplied workbench. A processing app for the IoT Gateway can be programmed using Java OSGi. We immediately had process data visualization using PPM. The hack coaches from GE Digital Team who traveled from Paris supported us in setting up the cloud tool Predix. As a result, we were able to successfully visualize the event-driven increases and decreases in the data stream as well as finish our demo setup within 48 hours. The end-to-end technology stack is certainly still not ideal, however the feasibility check was completed without a shadow of a doubt.
The IoT Gateway prefilters the gathered sensor data according to the rules. The data stream to the downstream analysis system widens when thresholds are exceeded.
The showcase in Garching
The EDDA Showcase developed at the #BCX17 has now been set up at the IIoT Innovation Center in Garching, where it will also be presented, and includes the XDK sensor and IoT Gateway. In addition, the possibility of embedding the setup at the Smart Product Division Center is also being seriously considered. For instance a product that is equipped with a sensor could run through the process chain, thereby collecting data, and then, exclusively in the case of irregularities, deliver data to be analyzed. EDDA is essentially very suitable for environments that tend to be repetitive. For example, all neuralgic points for analyzing machine health can be queried while retrofitting several similar production machines and then evaluated using machine-learning methods without the need to transfer and store large amounts of production data.
At the new Industrial IoT (IIoT) Innovation Center in Garching near Munich, Accenture showcases its work such as EDDA, which was developed at the Bosch Connected Experience.
A good starting point for further use cases
We were able to gain clear insights into the components from Bosch from a developer perspective at the BCX within a relatively short time and we have already been able to apply them. The proof of concept was a success and the demo case that was developed gives our Innovation Center a further idea that we can share with our visitors. We thought the event itself was great. We met with many highly motivated students, start-ups and Bosch employees from across the company to whom we would have liked to talk longer. In the end, we all were amazed at how much can be achieved and learned in such a short time. I am looking forward to experimenting further with Predix, PPM, XDK and the IoT Gateway and to taking advantage of one or two of these technologies for an actual customer project.
The Accenture hack team at the BCX17 (from left to right): Alexander Dieterle, Ralf Schumacher, Falk-Moritz Schaefer and Sven Mentl