In our 15th iteration of the Technology Circle (20.04.24), we invited nine speakers to learn what it takes to implement AI in manufacturing companies.
In this recap of the presentations, you’ll discover five reasons why manufacturing companies need to implement AI as soon as possible and learn how to do it for your own company — in just six steps.
What is AI?
Misunderstandings of and the resulting mistrust in AI are the biggest blockers keeping manufacturing companies from implementing AI into their own processes.
AI is a black box for most. We need to demystify it. Understanding what AI is and how it works helps you and your team accept and embrace it.
Every manufacturing company should use AI.
We’ve seen AI again and again drive down costs, increase efficiency and improve sustainability across the board in the manufacturing industry.
AI is made up of six key components
Machine learning: Training algorithms on large datasets lets the AI identify patterns and make predictions. This type of data-based training is the basis for any AI system.
Natural Language Processing (NLP): Enabling machines to understand and interact with human language facilitates tasks such as automated customer service, sentiment analysis and language translation.
Computer vision: Using sensors and cameras to let AI interpret and process visual data is crucial for applications like quality control in manufacturing, facial recognition and autonomous vehicles.
Deep learning: A subset of machine learning that uses neural networks with many layers to model complex patterns in data. It’s especially powerful for tasks like image and speech recognition.
Neural networks: These computing systems are inspired by the human brain. They consist of interconnected nodes (like neurons) that process information in layers. Neural networks are fundamental to deep learning and are used for recognizing patterns, classifying data and making predictions.
Cognitive computing: Systems can mimic human thought processes in complex situations. Self-learning systems use data mining, pattern recognition and natural language processing to simulate human reasoning and decision-making.
Learn more important terms used in industry 4.0
Many new terms are emerging with industry 4.0 (the digitalization of industrial production).
Stay on top of them with our updated glossary!
Five reasons your manufacturings needs AI
Predictive maintenance
By analyzing data from various sensors, AI can predict potential failures before they occur. This proactive maintenance approach minimizes downtime, reduces repair costs and extends the life of equipment.
Energy efficiency
AI systems analyze energy consumption patterns and optimize the use of energy. This results in significant cost savings and improves the sustainability of manufacturing operations.
Quality control
With computer vision, AI can detect defects and anomalies in real time during production. This ensures high product quality and reduces waste.
Data-driven decision making
AI systems analyze large datasets to support rational decision-making. This helps manufacturers optimize production processes, manage daily operations and innovate with new products and services.
New business opportunities
Client-side AI can become part of your services portfolio. For example, AI-assisted setup or remote technical support can ease the burden on your support staff.
Liebherr shows how it’s done
Stéphan Kohler, Adjoint au Directeur général at Liebherr Colmar, presents an example of Liebherr’s AI journey.
Liebherr identified machine maintenance as the area with the biggest potential for improvement through AI implementation.
Together with our AI assistance program, AI4SME, two goals were defined:
- Improve the availability of production tools
- Minimize downtime for repair and maintenance
The project started off simple to ease into the complexity of AI. The first use case was monitoring a hydraulic power unit dedicated to the automated tool-changing system, a subsystem of a big machining center. This selection was based on the pain point analysis: The mechanical troubleshooting is mostly generated during the tool-changing sequence.
After connecting the machine to the network, the next step was to direct the data flow to the external domain of Predict. To fulfill IT security standards, Liebherr had to configure a secured data channel. In-house experts first needed to learn to visualize the data. Using Grafana software, Liebherr developed a dashboard combining machine movements and temperature.
Visualizing this vast amount of data is only possible by focusing on one machine component at a time. Here, the focus is on the spindle, for example. Using Predict’s visualization environment, a larger set of parameters can be simultaneously represented. This can help analyze the interaction between several variables.
When a deviation on the machine’s spindle is automatically detected, Liebherr can stop drilling, avoid rupture and save repair time.
This successful integration of AI showcased the practical benefits of AI in enhancing operational efficiency and productivity in manufacturing. In the future, Liebherr plans to integrate AI more widely in their machine park.
How you can get started with AI in your manufacturing
Step 1: Make sure you’re AI-ready
Skepticism, mostly driven by a lack of knowledge, is the #1 blocker to AI adoption. You have to bring your workforce fully on board with AI before your project starts.
Another critical point is generating the right data and processing it fast enough. Your IT infrastructure is the most important thing to implement AI, besides the people ultimately using it.
Step 2: Choose a pilot project
Even if your entire manufacturing is AI-ready, start small.
A small-scale pilot project will help you learn practically and allow you to solve unforeseen problems without much stress.
You’ll refine your approach and measure the impact before full-scale deployment.
Choose your pilot project according to its potential impact.
This isn’t a gut decision.
Use your data, both historical and real-time, to pinpoint bottlenecks and set measurable objectives.
This could include reducing downtime due to machine maintenance, improving product quality through quality control or optimizing energy usage, to name just a few examples.
In the end, the AI project should demonstrate a positive return on investment (ROI), showing value potential to convince stakeholders to invest in AI on a larger scale.
Step 3: Choose your AI methodology
AI methodology is the systematic approach to developing, training and implementing AI models. It encompasses the selection of algorithms, data preparation techniques, model training processes and validation methods.
There are many different AI-driven solutions for the same manufacturing problems.
You need to find the AI techniques and frameworks that align with your specific goals and manufacturing environment.
Step 4: Get your data ready
Even if you’re generating data, you might not be collecting, storing and processing it AI-appropriatly. The entire concept of AI hinges on data, so you need to properly collect, store and process it from the getgo.
Collecting data
Start by identifying all potential data sources within your manufacturing process that you need for this specific project.
This can include sensors on machinery, production logs, quality control records and maintenance logs.
Ensure that data is being collected consistently and accurately. Implement IoT devices and sensors to gather real-time data, which is crucial for dynamic and responsive AI models.
Storing data
Use a scalable data storage solution, such as cloud storage or data lakes, that can handle large volumes of data. Ensure your storage solution supports easy retrieval and processing of data.
Processing data
Processing data involves cleaning, transforming and organizing it for AI use. Clean your data to remove any inaccuracies, duplicates or irrelevant information.
Use data normalization techniques to ensure consistency across data sets. Transform your data into formats that are compatible with AI algorithms. This may involve aggregating data, creating new features or labeling data for supervised learning.
Use data processing tools and frameworks to automate and streamline these tasks, ensuring your data is ready for AI model training and analysis.
Step 5: Develop and deploy your AI solution
Development and deployment depend entirely on your project and your AI methodology.
If all virtual tests succeed, you can connect your AI models to your manufacturing systems and run a live test.
Step 6: Keep optimizing
Don’t be afraid of roadblocks and setbacks.
It’s important to monitor the performance of your AI systems continuously. Use dashboards and visualization tools to track key metrics.
Be prepared to iteratively improve your AI models based on feedback and changing conditions. Regular updates and retraining of models may be necessary to maintain their effectiveness.
The shortcut: Collaborate with experts!
Implementing AI in manufacturing can be complex and time-consuming, and collaborating with AI experts can streamline this process.
Stéphan Kohler, Adjoint au Directeur général of Liebherr France said about the above-described project with AI4SME, “This project showed us the power of expert collaboration. We had to recognize our limits and open our doors.”
Leverage these organizations’ knowledge and resources to implement AI more effectively, gaining a competitive edge in the manufacturing industry:
EDIH Grand Est
Olivier Horent leads the EDIH Grand Est initiative, providing a comprehensive one-stop shop for industrial SMEs looking to adopt AI and digital technologies.
The EDIH Grand Est offers diagnostics, test-before-invest services and support for finding financing. Their goal is to enhance the competitiveness of SMEs by facilitating access to advanced digital tools and expertise.
DIZ Digitales Innovationszentrum GmbH
Gennadi Schermann manages the EDIH-AICS Karlsruhe, focusing on applied AI and cybersecurity.
DIZ Digitales Innovationszentrum GmbH helps SMEs and public sector organizations implement AI by providing readiness checks, workshops and connecting them with AI experts. Their mission is to bring AI into practical use, ensuring businesses can leverage these technologies to improve their operations.
HE-Arc
Hatem Ghorbel leads the Data Analytics Group at HE-Arc, which spearheads the Swiss AI Center for SMEs.
This center offers prototyping projects, customized training and a sandbox environment for testing AI applications. HE-Arc supports SMEs by making advanced AI techniques accessible and practical, helping them accelerate their digital transformation.
AI4SME
Our own collaborative initiative, AI4SME, is designed to help small and medium-sized enterprises discover and implement AI solutions.
We provide expert support, financial backing for proof-of-concept projects and foster cooperation between SMEs, startups, universities, and research institutions. By offering a structured approach to AI adoption, AI4SME aims to demystify AI and promote its practical applications in various industries.
Download the free AI4SME guide
Our AI4SME guide contains all the information you need to develop and implement AI in your manufacturing SME:
Its learnings are based on a 3-year period where we helped develop 7 POCs. You’ll learn about:
- Strategic planning
- Collaboration
- Data and workload management
- Project management
- Organizational and people development
- Expectation management
You may also be interested in
Nouscom secures 67.5 million euros in financing
Celebrating six months of innovation: i4Challenge accelerator New Ideas 2022/2023
How open innovation in healthtech hubs is fueling the rise of digital healthcare
Paradigm shift in pain management
BOOM Summit in Basel accelerates health technology
Investors stump up 3.7 million Swiss francs in Onena Medicines
”Are you interested in innovation and networking? In our Trinational Industry 4.0 Technology Circles we bring together researchers, developers, entrepreneurs and other stakeholders for cross-disciplinary exchange.
Sébastien MeunierDirector Industrial Transformation
