The modular and transparent approach used in the software suite developed for the ML4P (Machine Learning for Production) project could reduce the barrier to entry for small companies, said the developers at Fraunhofer in Germany.
The suite, which aims to make industrial manufacturing more efficient by using ML, is reportedly very flexible and can be applied to existing production processes.
“Modern machines fitted with extensive sensors supply an increasing amount of data, resulting in huge optimisation potential for production by means of data analysis using artificial intelligence (AI) and ML,” the researchers said. A consortium of several Fraunhofer institutes sought to harness that untapped potential for the benefit of industry.
During the four-year project, the consortium combined process models and ML software tools, aiming to make production faster and more energy- and resource-efficient. The suite can analyse machine data to discover hidden trends and use these to optimise the manufacturing process, for example. It can also ‘learn’ over time to improve production on an ongoing basis.
“We don’t just pop up with a completed software solution for a company – instead, we guide them through the process model, taking a methodical, step-by-step approach,” said Christian Frey, ML4P project manager.
The first step, the team said, is to analyse the current state of the production process. The experts then identify potential areas for optimisation, set targets, and develop a concept for implementing ML4P. They then examine whether the concept can actually be implemented with the available machinery and data, and how it lines up with the company’s objectives.
The following step involves transforming process data from the machinery into a ‘comprehensive, digital information model’, requiring expert knowledge from engineers about all steps of production. That is integrated into an ‘ML4P pipeline’, which learns a process model from machine data. Implementation and test operations follow, and eventually the process model is deployed and daily production begins.
The software suite includes generic tools for typical tasks, such as monitoring a machine’s operating state. These are compatible with several industrial communication interfaces, such as OPC UA (Open Platform Communications Unified Architecture). Wherever possible, developers tried to avoid using proprietary software protocols, relying instead on established standards and programming interfaces.
Once put into operation, each module can be customised to continuously update the process model and highlight potential for further optimisation. The Fraunhofer researchers said both new and old machines can be integrated, even those that are 30-40-years-old.
“It is not so much about the machines, but whether it can provide suitable data, for example if it is equipped with dedicated sensors,” said deputy project manager Lars Wessels.
Smaller companies can also apply ML4P, the developers said, even if they only want to optimise specific parts of a manufacturing process.
“Many companies are still sceptical about the use of AI or ML because they have not yet recognized the enormous potential that ML offers for production. However, the modular platform from Fraunhofer provides transparency, flexibility and scalability, thus reducing the barrier to entry,” said Frey.
The team tested the integrated concept across various applications, including hot sheet metal forming, production of membrane filters and glass bending.
Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.
Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.