Fri, 17 January, 2025
A three-year long, game-changing project has recently concluded, successfully delivering a new solution that mitigates the risk of failure associated with parts that have been produced using laser powder bed fusion (LPBF), a type of metal additive manufacturing (AM), while also ensuring consistency and quality.
LPBF is a promising technology for manufacturing complicated metal parts that uses a laser to melt metal powder into a solid, and produce three-dimensional objects of almost any shape or design. However, components produced by LPBF can be prone to defects such as porosity, cracking and residual stresses, arising from factors such as laser power, scan speed and powder bed temperature.
The iAM-3DPO project: Intelligent advanced additive manufacturing enabling dynamic process efficiency aimed to tackle these challenges head-on by providing control over the parameters of LPBF in combination with real-time process monitoring. Spearheading this technically demanding research and development (R&D) activity were consortium members the Joining 4.0 Innovation centre (J4IC) – a strategic partnership between Lancaster University and TWI, Materialise, NQuiringMinds, ThinkLaser and TRUMPF Laser UK.
Working collaboratively, the partners created an advanced, novel, integrated, laser-based system for LPBF parts production, capable of achieving zero distortion and defects during the 2D build process, and combining real-time analysis of optical and thermal monitoring to enable the detection of anomalies and process optimisation. The solution leverages partner Materialise’ Materialise Control Platform (MCP) technology to enable total beam path control, linked to TRUMPF Laser UK’s unique, beam shape control technology for single mode beams, to optimise the factors associated with build quality. Cloud-based data sharing, artificial intelligence (AI) and closed-loop technology have also been integrated.
J4IC’s contribution to iAM-3DPO spanned:
- Creating a neural network algorithm incorporating ‘Long Short-Term Memory’ (LSTM) layers to process temporal dependencies, and a custom-made class structure for data handling and management
- Building a quality control system with a quality-aware loss function that interfaces with infrared camera feed data, utilising a sliding window pre-processing pipeline and maintaining time-series relationships
- Designing the system architecture which features power control policy generation, and an extraction algorithm that reduces the dimensionality of the training state space while preserving essential information, and is also modular
- Establishing a testing framework with data augmentation capabilities, and model saving / loading functionality to support system deployment and maintenance
Speaking about the industrial impact of iAM-3DPO, Dr Darren Williams, Director of J4IC, said “The novel LPBF system that was achieved by the project partners will help to both meet the increasing demand for production of components with no defects or distortion, and accelerate the take-up of LPBF in the metal AM market. Prior to iAM-3DPO, the LPBF process put limits on component size, and the use of multi-materials with varied physical and chemical properties meant managing multiple phases to obtain the final metal product. Thanks to the consortium partners’ R&D, manufacturers will now be able to take advantage of a new technology that allows right-first-time production, and in turn contributes to greater efficiencies and cost savings.”
iAM-3DPO received funding from Innovate UK via a EUREKA R&D:SMART Advanced Manufacturing grant, reference 76940.