Using artificial intelligence methods in additive manufacturing

Robust production in line with requirements from single-unit production

Additive manufacturing processes, which originated in the 1980s, are capable of manufacturing complex, personalized products in small quantities down to single-unit batches in an economical and sustainable manner.

Plastic-based processes use photocrosslinkable or thermoplastic materials to produce components layer by layer. This eliminates the need for tooling.

Additive manufacturing has now reached a level of maturity that allows more and more industrial applications to be successfully realized. The corresponding materials, systems and peripheral processes are available. However, the manufacturing processes are complex and still pose a challenge for industrial use in connection with the production of individualized products and single-unit production. The methods and procedures for process optimization and quality assurance known from mass production are often impractical or uneconomical. Artificial intelligence methods can be used here to make additive manufacturing processes more robust and their use more economical.

Benefits and challenges

The past 20 years have seen significant advances in performance in the field of artificial intelligence (AI) and machine learning. AI is also an established and accepted technology in the industrial environment thanks to cost-effective IT infrastructure and powerful software packages. It is therefore also available for various applications within additive manufacturing - here are just a few examples:

• Identification of relevant process parameters for specific component properties
• Automated parameter optimization: Component and material-specific optimization to reduce costs and waste.
• Prediction of component properties based on the condition of the raw materials and process data
• Inline control of relevant sub-processes (e.g. material discharge, thermal control, cross-linking processes,...)
• Optimization of user interaction
• Quality assurance methods for individualized components, single-unit and decentralized production

However, not all fields of application and potentials of AI methods for additive manufacturing are currently known. It needs to be clarified which methods are best suited to the respective case, how training data is recorded and labeled, which sensor technology is suitable for recording data, whether data can be generated using simulation and whether the trained AI system is ultimately suitable for the task.

Experts from the fields of Artificial Intelligence and Machine Vision and additive manufacturing are therefore working together at Fraunhofer IPA to offer companies a comprehensive range of services for the use of AI methods in additive manufacturing. Together with our customers, we are working to exploit the potential arising from the combination of these two future technologies.