Combining AI and IoT in manufacturing - our BOSCH practitioners approach unlocking golden opportunities

Shaping the transformation: the use of artificial intelligence within the Internet of Things

In the IoT, we connect intelligent products and services, generating added value for our customers. By adding AI, we create a closed value creation cycle and focus even more on users. The data resulting from the use of intelligent, connected products and the interaction between people and machines and machines themselves are the key factor in this context. By linking IoT with AI and machine learning, we can draw the right conclusions from huge quantities of data and react to these data during product engineering in seconds. We learn from the data and can thus improve our products and services on an ongoing basis.

AIoT makes it possible to develop new products more quickly. At the same time, we can optimize customers’ products during their lifetime for example by using over-the-air updates or we can add new functions.

By using data as a basis for optimizing and personalizing our products, the need for privacy and AIoT security grows. The greater people’s trust in the AIoT, the greater their acceptance.

AI in manufacturing combines big data with production experience

The key factor for smart factories and Industry 4.0 is the data that arise from the use of intelligent and networked products, the interaction between humans and machines and machines with one another. At Bosch we have about 240 plants around the world in which numerous networked production systems are in use. These generate an enormous amount of data that require the use of efficient data processing methods. The Bosch manufacturing and logistics platform helps to access and structure this data. By using AI methods, we gain useful and - compared to human analyzes - objective knowledge very quickly. At the same time, we have accumulated a considerable amount of knowledge and models for many manufacturing processes over the past few decades. Our ability to combine these two elements - data and knowledge of production processes - in what are known as hybrid models - is a great advantage when it comes to implementing AI solutions for manufacturing and logistics.

Develop, connect, improve: the AIoT cycle

The so-called AIoT cycle – a value creation cycle comprising four phases – shows the benefits of linking AI and the IoT:

1. Value creation

Connected products provide data. Bosch uses these data during research and development to improve applications and revise or supplement functions. At the same time, we can improve the security and reliability of our products on an ongoing basis and adapt them to meet the individual needs of customers. Making AI in AIoT products secure, robust and explainable is a key issue for us. Research projects such as machine learning testing and AI safety help to achieve this goal.

2. Products in interaction with customers

We deliver connected products and services to our customers. When these products are used, they generate data, which we use in the following phases of the cycle to improve products and applications. AIoT products ensure greater security for users. Research projects such as embedded AI based siren detection show this.

3. Data processing

The data which are produced when connected products are used are the basis for this phase of the AIoT cycle. They are collected and stored in a structured manner. With technologies such as self-sovereign identities (SSI) and trustworthy computing, we ensure that users can keep control and maintain sovereignty over their data at all times and that these data are always protected.

4. AI algorithms

In this phase, we process the data using AI algorithms and machine learning and gain new findings on this basis. The visual analytics research project shows how this process leads to greater security for users: in autonomous vehicles, AI is used for image recognition. In rarely occurring situations where several unusual conditions converge, so-called corner cases, the AI in the image recognition system points out weaknesses. For example when a red traffic light is hard to see from a certain angle in inclement weather. Visual analytics helps to detect blind spots and to automatically supplement the existing data set with the help of a second AI. As a result, shortcomings of the first AI are remedied and overall system accuracy can be increased.

A variety of new possibilities

The use of the huge amounts of data analyzed in real time with the help of AI will make it possible to manufacture products faster and with higher quality, bringing artificial intelligence production and logistics to the next level. Models based on deep learning are able to detect and avoid even those errors that humans or classical systems have difficulty perceiving. But predictive analytics will also be able to optimize production in the future in other ways:

• AI will enable companies to more strongly individualize their products without much additional process-related effort.
• AI will help robotics find its way into areas where it was previously inconceivable. With the help of AI, the learning effort is reduced, the robot's vision is improved and thus transfer learning is made possible in the first place.

Speed and quality for smart factories

Analyzing data from manufacturing control systems has been a very time-consuming process. With the help of AI, we can not only analyze the data faster, but also send the insights drawn from it directly to dashboards that inform operators when something is wrong. In the future, these insights can also be used to give the machine's control system new instructions, such as adjusting parameters for example. In the end, this might lead to AI-based closed-loop production systems, which are self-regulating or self-optimizing. A list of possible root causes for problem resolution is provided right along with it.

In other words, the use of AI allows to speed up and to optimize production processes overall. It helps improving quality by greatly reducing scrap. In addition, with the help of AI, the utilization of machines and equipment can be planned more efficiently. Most importantly, customer satisfaction increases in the end.

AI in manufacturing the Bosch Way

MAS - Manufacturing Analysis Solutions

In this area, Bosch is developing scalable AI and analytics solutions to detect anomalies and malfunctions in the manufacturing process at an early stage and determine the root causes. These solutions process millions of data points from various sources.

AOI - Automated Optical Inspection

Behind this are advanced deep learning approaches designed to improve optical inspection processes for goods produced by Bosch.

Robotic assembly

Here, we are developing novel perception and control algorithms that enable skilled/efficient processing of parts and tools. The goal is to be able to automate even highly complex assembly processes.

Process Optimization

In this field we focus on the development of algorithms that automatically adapt machines to the respective production and processing requirements.