Hanover At the Hanover Fair in the past few days, artificial intelligence (AI) was one of the topics that moved the around 4,000 exhibitors and 130,000 visitors the most. The technology offers industry potential to make its processes more efficient, faster and safer or to develop new business models.
The possible uses are diverse. “In the factory and in industrial production, one quickly thinks of predictive maintenance,” explains Patrick Glauner, Professor of Artificial Intelligence at the Deggendorf University of Applied Sciences. “But this is just where it starts – we can do so much more with AI than just predictive maintenance.”
Word has gotten around in the industry, says Glauner. “More and more deep learning is being used. It works very well – and there is great progress in inspection there.”
Deep learning is a category of AI that uses artificial neural networks for data analysis and works very reliably, for example, in image recognition.
From the predictive maintenance and repair of machines to quality control and programming with voice commands: the Handelsblatt shows some examples of how AI is used in industry.
Predictive Maintenance
Fixing the machine before it breaks down: That’s the idea behind predictive maintenance. Such a system is offered by the start-up Ai-Omatic, for example, which was at the Hanover Fair for the first time. It presented predictive maintenance software designed to reduce unplanned machine downtime.
There were around 4000 exhibitors and 130,000 visitors at the fair.
(Photo: dpa)
The company’s maintenance assistant evaluates the operating data of a system using deep learning and uses this to create a statistical model for the investigation of future processes. On this basis, the program can monitor the sensor data from machines live and point out anomalies. Users can also monitor the condition of their machines via a dashboard.
Particularly difficult in times of crisis: distinguishing downtimes or imminent faults from anomalies and thus avoiding false alarms. Therefore, the Ai-Omatic system relates current sensor values to the values of other days and thus calculates the probability of an emergency – or a false alarm.
Factory quality control
A task that is error-prone because it is uniform is quality testing in the factory: workers on the assembly line have to examine numerous products closely to discover damage and irregularities.
This activity can be automated: SAP has developed a function called Visual Inspection for its factory software, which automatically analyzes camera images and issues warnings in the event of irregularities.
Smart Press Shop, a joint venture between Porsche and Schuler AG, uses it to examine body parts – and thus relieves the workload of the factory employees. The press shop employees train the algorithm themselves: During the configuration phase, they enter their evaluation into the system. After just a few hours, the database is often large enough.
Industrial companies should be able to reduce the error rate with such processes. The result: fewer rejects, fewer complaints. The SAP system is also able to transfer the production data to a digital product file, which makes it easier to take targeted action in the event of recalls, for example.
Accessibility features for robots
With large language models and chatbots like ChatGPT, a new generation of artificial intelligence is emerging, which may one day also be used in factories – at least once the liability issues have been clarified.
Robots were also presented at the Hanover Fair.
(Photo: Bloomberg)
Hewlett Packard Enterprise (HPE) and the Heidelberg start-up Aleph Alpha presented a possible application scenario at the trade fair: an intelligent assistant that factory staff can use to communicate in natural language and with images.
This means that questions about the installation, maintenance and operational safety of the robot can be answered quickly. “Because the factory staff does not have to read manuals that sometimes have several thousand pages, processes are accelerated,” explains HPE Germany boss Marc Fischer. “And in an emergency, the AI assistant can guide the employee to stop production processes.”
>> Read here: New technology from Aleph Alpha: Artificial intelligence should stop talking nonsense
According to Fischer, the AI-based application also makes a significant contribution to occupational safety. In addition, the AI assistant always has the appropriate sources from the manuals ready for his answers.
Aleph Alpha has been developing artificial intelligence on an HPE supercomputer since 2019, which competes with the AI projects of US companies such as OpenAI and Google.
operation and programming
Siemens and Microsoft also showed the potential of large language models at the Hanover Fair. The two groups are upgrading the Teamcenter software for what is known as Product Lifecycle Management (PLM) with AI.
Among other things, Microsoft showed the potential of large language models at the Hanover Fair.
(Photo: Reuters)
The Dax group is working on an app that service technicians or production employees can use to report problems with quality or design via voice message – the software evaluates them and automatically sends a report to the appropriate teams, whether in design, research or production.
The new language technology could also prove useful for programming: systems such as Github Copilot and Amazon Code Whisperer convert user input into program code. They are therefore also suitable for programming machines and robots. It is conceivable that the code for so-called programmable logic controllers, such as those sold by Siemens, could be generated by voice command.
>> Read here: With these innovations, companies can save energy and CO2 in the future
In this way, development teams could significantly reduce their time and the possibility of errors by entering commands in natural language and letting the computer “code” it. This would also enable maintenance teams to identify errors and generate solutions more quickly.
simulation
According to Glauner, another approach to AI-based applications could become popular among industrial companies in the future: simulations of production processes and behavior. “Simulations take a long time, are very expensive and also energy-intensive.
With the help of machine learning from previous simulations, an AI can often predict the simulation result within milliseconds for new cases,” says Glauner.
Only if the AI is not sure does it have to be simulated again. “That takes days again – or sometimes several weeks, but the AI then learns even better from it.”
More: When a tablet replaces the service technician: This is how AI is revolutionizing the industry