Humans and machines team up in the factory of the future

“What makes this fab so special is the interplay of two technologies,” says Sarah, talking about artificial intelligence and the internet of things, whose combination results in the acronym AIoT. The Dresden fab is Bosch’s first AIoT factory. All its systems, sensors, and products generate data, which is collected in a central repository. The data quantities are enormous: every second, the plant generates the equivalent of 500 pages of text. That would amount to more than 42 million sheets of paper a day — big data indeed. This is where artificial intelligence comes in, as self-optimizing algorithms learn to make predictions based on the data. In this way, operations can be analyzed in real time without disrupting sensitive manufacturing processes.

Semiconductor manufacturing processes are complex. It all starts in the sub-fab, an area below the actual wafer fab fitted with equipment to provide gases, air, and electricity to the chamber above. Kilometers of pipes, cables, and hoses run through the sub-fab. Every pipe, cable, and hose has a digital counterpart, as does each valve and every other component throughout the plant. The entire building is mapped to a 3D model in exacting detail. This model is enhanced with other information such as consumption data to create a digital twin — that is, a virtual rendering of the plant and all its processes. “This enables us to simulate optimizations and structural alterations in the virtual realm without having to disrupt processes in the real world,” says Sarah.

Before Sarah can enter the heart of the fab, the cleanroom, she has to first slip into a special white coverall. Then she has to step into an air shower that blasts away every last speck of dust. Only then is she allowed to pass through the airlock to the chamber where the semiconductors are conveyed in wafer boxes. This area requires cleanroom class 1 conditions. The usual number of airborne particles, some 100,000 per cubic foot, is unacceptable here. A cubic foot of fab air — or around 28 liters — may contain no more than one particle weighing just half a microgram. To put this into perspective, that equates to a single cherry pit floating in Lake Constance, which holds about the same as 360 million bathtubs.

Even the most advanced machines require maintenance. However, these tasks can be optimized and executed just in time with predictive maintenance and AIoT. AI algorithms can pinpoint necessary repairs at an early stage based on the collected usage data. This work is not dictated by a set schedule but performed precisely when needed — in good time, before problems occur. A machine may even be maintained remotely using data glasses and augmented reality. For example, a specialist at an equipment manufacturer in Japan could perform maintenance work in Dresden without having to actually be there in person. The camera mounted on the data glasses transmits video images to Japan, allowing the expert there to guide the technician at Dresden through the maintenance task in real time.

The wafer fab will soon communicate via the new 5G mobile cell standard to accelerate these processes and data transfer between machines and computers. This is just one of several advances that feature prominently in the plans of the Dresden-based team of some 250 people. It also intends to use AI to develop and simulate entirely virtual prototypes before ramping them up for production. This way, Bosch will be able to set new quality standards, speed up the development process, and cut costs. What will it take to make that happen? State-of-the-art technologies and highly capable associates. Or as Sarah Fischbach puts it, “People and technology are one team in Dresden — and together, we’re stronger.”