Industry 4.0
Turbocharging artificial intelligence in production

Artificial intelligence (AI) offers a lot of potential for Industrie 4.0, but there are technological barriers that make it more difficult to implement. Researchers at Fraunhofer IKS are developing a framework to support and optimize the data and AI life cycle. This will significantly increase the scope of value creation provided by AI.

mask Cars on the production line

Industry 4.0 is a term for continuing digitalization in production — machines and processes are intelligently networked, meaning that more and more data can be generated. Artificial intelligence (AI) has the potential to generate information from these data to improve production and services. Possible application scenarios include predictive maintenance, optimization and automation of processes and quality control. However, it is not yet possible to make full use of this potential in Industry 4.0 because there are multiple technological barriers limiting the generation and processing of information.

The first barrier is the multi-vendor landscape in today’s production sites, with machines made by different manufacturers, from different generations of technology, and with different — and often proprietary — communication interfaces and protocols. This heterogeneity makes uniform data access impossible. Instead, there are many isolated technology-specific solutions that require knowledge of a particular domain.

Incomplete datasets create problems

The second barrier is the lack of support for the data scientist, who has no knowledge of the specific domain and therefore requires support when obtaining real-time or historical data. There is also the issue of incompatible, inconsistent and incomplete datasets and missing metadata. As a result, data processing is often laborious, lengthy and manual and requires a lot of coordination.

The third barrier is inflexible AI operation. AI applications are often run rigidly in the cloud or on a local server. This means that the applications are unable to make optimum use of the available resources. Furthermore, updates to the AI application are needed — to respond appropriately to changes at the production site or in processes — for which there is currently no general solution.

KI-Hub Bayern: Presentation of the research project »REMORA«

In the context of the KI-Hub Bayern, Fraunhofer IKS hosts the first networking event: On Wednesday, May 25, 2022 at 4 pm, the author of this blog post will present the research project»REMORA«.

Find out more here.

The solution: a framework for the data and AI life cycle

In order to overcome these problems, researchers at the Fraunhofer Institute for Cognitive Systems IKS are working on the project “REMORA — Multi-Stage Automated Continuous Delivery for AI-based Software & Services Development in Industry 4.0” to develop an open, interoperable and technology-neutral framework to support and optimize the data and AI life cycle. The aim is to ensure an automated, continuous and dynamic process flow consisting of

  • data acquisition
  • data aggregation
  • data preparation
  • AI development
  • AI training
  • AI integration
  • AI operation
  • Data analysis
  • AI monitoring
  • AI update

Specifically, the framework is intended to achieve the following objectives:

  • support for the data scientist,
  • automated and flexible AI integration and
  • automation of AI processes.

It begins with the development of an interface for the data scientist to support the AI development process. This interface makes it possible to retrieve data in a simple and uniform way without the need to consider technology-specific aspects such as communication interfaces and protocols. The interface then deals internally with mapping the data onto the technologies and with the necessary data transformations. The interface also provides an overview of the topology and the metadata and an interface for training and operating an AI model. This interface can be operated not only by a data scientist, but also by a layperson together with an AutoML framework, for example.


Conference Paper, 2021

Nicola Franco, Hoai My Van, Marc Dreiser, Gereon Weiß: Towards a Self-Adaptive Architecture for Federated Learning of Industrial Automation Systems Pfeil nach rechts

Managing AI applications

An application management component then enables automated and flexible AI integration — from the component level to the cloud — based on the required resources and optimization targets. In addition, the AI application manager works with the data interface to ensure that the AI applications are networked to secure the flow of data.

Finally, an AI management component aims to enable the automation of AI processes, i.e. the automated retraining and redeployment of an AI model, to ensure continuous improvement of the data analysis. For example, when replacing machinery, new training data could be collected automatically to train a new AI model. Operations could also be automated in response to the data analysis (e.g. cooling an overheated system) or to increase the efficiency of real-time AI analysis (e.g. adjusting the sampling rate).

Simple AI integration for Industrie 4.0

For more information on the REMORA project, please visit our website.

Find more information Pfeil nach rechts

What this means for the production of the future is that, with this framework, the potential that AI holds for Industrie 4.0 can be better utilized — through simplified, technology-neutral data access, support for AI development, flexible and automated AI integration and updates — thereby increasing the efficiency of AI operation.

This project is funded by the Bavarian Ministry of Economic Affairs, Regional Development and Energy as part of the project "REMORA - Multi-Stage Automated Continuous Delivery for AI-based Software & Services Development in Industry 4.0" and supported by Bayern Innovativ - Bayerische Gesellschaft für Innovation und Wissenstransfer mbH.

Read next

A software toolkit for flexible and robust architectures

Florian Wörter
Florian Wörter
Cognitive systems / Fraunhofer IKS
Cognitive systems