Summer School | Artificial Intelligence in Industrial Applications

Information and Communication Technology (ICT) plays an important role in modern production environment. As examined in the first module, the generation of an added value for production relies on the collection and integration of data from the field. The second module of the summer schools therefore focuses on how to enable smart devices in factory environments to exchange information in a seamless way. After a general introduction into the basic technologies that enable information exchange in computer networks, examples for protocols and data exchange standards are given.

Industrial environments are increasingly characterized by distributed devices and information sources. These devices are subject to constant change, rearrangement and modification. Thus, in order to create an infrastructure capable of sustaining the interoperability in such dynamic environments, open standards such as the internet have to be established in the factory environment. One solution to this requirement is summarized with the Industrial Internet of Things (IIoT) term that describes a pattern that enables the formation of dynamic networks in modern industrial environments. The IIoT is based on the Internet of Things (IoT) paradigm that has the major goal of enabling interoperability between all devices that have access to the internet. The IIoT has the goal to establish such holistic information exchange capabilities between all devices in a factory by means of open, lightweight and easy-to-use standards. This module focuses on how to enable IIoT for existing production systems.

In this module, the use of machine learning and other Data Science methods will be prepared by focusing on the programming and database methods that are required to implement an Artificial Intelligence workflow. One of the most popular programming language for applying Data Science is Python. That is why in this module the Python programming language is introduced including basics about libraries that facilitate the querying of data, the preprocessing of this data and subsequently the preparation of data for the application of machine learning methods and techniques.

Additionally, the module will cover database programming, i.e. database querying methods that have to be used to initially extract information from existing databases to apply machine learning. The basic methods covered in this module serve as the fundament for all later steps involving advanced analytics through machine learning.

Machine learning is generally separated into two general training methods, of whom one is the data-driven approach, i.e. supervised and unsupervised learning, and the other is trial-and-error learning or so-called reinforcement learning. In reinforcement learning an agent learns by interacting with an environment, for example represented by a robot and its surrounding. The results of such learning process are organized by something like a human brain, i.e. an artificial neural networks that sort of remembers situations, which the learning agent already “experienced”. Based on such learning the system can learn independently and reach some sort of state that enables the agent to reason without the help of a human, thus developing an artificial intelligence.

In this module, the instructors will introduce the fundamentals and basic methods of artificial neural networks and how they are applied for deep learning. The participants will learn how to train these models and generate a working agent.