This efficiency is made possible by the methodology of Physics-Informed Neural Networks (PINNs), which the Springer review identifies as the most promising development direction for WAAM process control. PINNs combine classical neural networks with physical laws – specifically the partial differential equations for heat conduction, momentum, and mass conservation – and additionally integrate metallurgical models such as CCT diagrams and Maynier equations for microstructural development. Training such models occurs in two stages: in a first step, the network is pre-trained on large amounts of synthetic FEM data to internalize the fundamental physical relationships. In a second step, it is fine-tuned with real measurement data from the manufacturing process. The decisive advantage over purely data-driven models: PINNs deliver physically consistent results even with smaller training datasets – a critical advantage in an environment where experimental data is expensive and limited.
For industrial practice, this development represents a fundamental paradigm shift: instead of measuring residual stresses or component distortions after manufacturing and empirically adjusting the next job, the combination of FEM pre-simulation and AI-assisted real-time prediction enables proactive, anticipatory process control. Defects can be anticipated and prevented before they occur – a prerequisite for the certifiability of WAAM components in regulated industries such as energy generation or aerospace.
WAAM Meets Industry 4.0: Connected, Data-Driven, Future-Proof
Simulation and AI are powerful tools – but they only reach their full potential when production systems operate not as isolated islands, but as interconnected nodes within a digital manufacturing infrastructure. This step is demonstrated by a study published in November 2025 in Scientific Reports (Nature Portfolio) by RWTH Aachen University, involving the Welding and Joining Institute (ISF), Communication and Distributed Systems (COMSYS), the Institute of Automatic Control, and the Knowledge-Based Systems Group.
The research team analyzes WAAM for the first time systematically across three clearly defined process layers – workpiece, assembly, and product – presenting a framework that thinks far beyond the individual manufacturing cell.
The workpiece layer maps the immediate production environment. Here it is decided second by second whether a component meets specifications. In the Aachen experimental system, Gas Metal Arc Welding (GMAW) was used with voltages between 15 and 37 volts and currents between 190 and 410 amperes. A Proportional-Integral (PI) controller adjusted welding parameters in real time via a robot interface – with the goal of simultaneously ensuring the geometric accuracy of the deposited layers and minimizing welding fume emissions. The authors additionally discuss Model Predictive Control (MPC) as a future method for an even more capable closed control loop: MPC can plan several time steps ahead while simultaneously optimizing competing process objectives – such as deposition speed, geometric accuracy, and emission minimization.
The centerpiece of the quality assurance approach is not a complete digital twin, but the conceptually leaner digital shadow: while a digital twin maps the entire system in its full complexity, a digital shadow captures only those process parameters that are actually relevant for a specific task – such as predicting weld seam quality or monitoring welding fume emissions. This targeted information filter makes the approach computationally efficient and real-time capable. As a result, the Aachen team was able to reduce welding fume emissions by 12 to 40% through data-driven process optimization – a measurable gain for occupational safety, energy balance, and operating costs alike.
The assembly layer addresses the challenges that arise when multiple machines, sites, or companies need to communicate with one another. Interoperability between different stakeholders – machine manufacturers, operators, quality assurance, certification bodies – requires clear definitions of responsibilities, access rights, and data exchange protocols. The framework proposes standardized interfaces for this purpose, enabling controlled data exchange without dissolving proprietary system boundaries.
The product layer, finally, raises the question of the reliability and traceability of generated data across the entire product lifecycle. This is where techniques for the authentication and verifiability of process data come into play – central for industries where certification records must be provided without gaps. At the same time, the framework addresses data protection requirements through so-called privacy-preserving techniques, enabling meaningful data sharing between companies without disclosing sensitive production or design data.
A particularly forward-looking element of the Aachen framework is the connection to the Worldwide WAAM Library (WWL) – a data-sharing infrastructure developed within the Internet of Production initiative. The WWL enables digital shadows from multiple WAAM systems – across sites and operators – to be centrally stored, queried, and collectively evaluated. The more systems feed their process data into this shared pool, the more precise the quality models become, and the faster cross-site optimization is possible. The long-term goal: a global, learning network of WAAM systems that uses its collective experience to make each individual print job better than the last.
Last but not least, the connection of production systems to data networks also raises cybersecurity questions anew. With growing connectivity, the attack surface on manufacturing systems increases. The framework therefore integrates data security requirements from the outset – not as a retrofit solution, but as a structural design principle. For industrial users from safety-critical industries, this is an argument that goes beyond technological enthusiasm and addresses concrete compliance requirements.
Outlook: Wire as the Raw Material of the Next Manufacturing Generation
WAAM stands at a turning point: from a demanding specialist process to an industrially established serial process. The combination of high economic efficiency for large components, growing simulation accuracy, AI-assisted process monitoring, and increasing certification successes in sensitive industries opens markets that seemed hermetically sealed five years ago.
For the wire industry, this represents a fundamental repositioning of the classic welding wire value proposition: it is no longer merely a consumable – it becomes the feedstock for certified metal components of industrial scale. wire Düsseldorf offers precisely the platform where wire producers, machine manufacturers, research institutes, and end users can jointly shape the next steps of this development.
For more information on current technologies and innovations from the wire and tube industry as well as further industry and product information, please visit www.wire-tradefair.com and www.tube.de.
Further Information and Sources
- Ghazali et al. – A focused review on numerical computation in WAAM for HSLA steels (Springer, October 2025): link.springer.com
- Chen et al. – A comprehensive review of simulation approaches in WAAM (IOP Science, January 2025): iopscience.iop.org
- Mann et al. / RWTH Aachen – Connected, digitalized wire arc additive manufacturing (Nature/Scientific Reports, November 2025): 3dprintingindustry.com
- Recke / GEFERTEC – Optimization of series production with 3D printing and WAAM (All About Industries, October 2025): all-about-industries.com
- MX3D – Wire Arc Additive Manufacturing: A definition of this technology: mx3d.com
Infobox:
Wire Arc Additive Manufacturing (WAAM) – A manufacturing process in which metal wire is melted by an electric arc and deposited layer by layer to build up a component.
Directed Energy Deposition (DED) – An umbrella term for additive manufacturing processes in which material and energy source are simultaneously directed at a single point. WAAM is a subcategory of DED.
ISO/ASTM 52900:2021 – International standard for the classification and terminology of additive manufacturing processes.
Substrate – The base plate onto which the component is built up layer by layer.
Deposition Rate – The amount of material (in kg/h) that can be deposited per unit of time. In WAAM, this is typically 2–15 kg/h.
Near-Net-Shape – The manufactured blank already closely resembles the final geometry and requires only minimal post-processing.
Selective Laser Melting (SLM) – A powder-based additive manufacturing process in which metal powder is fused by a laser. In contrast to WAAM, it is suited for very small, highly precise components.
Finite Element Method (FEM) – A numerical simulation method for calculating temperature fields, residual stresses, and distortions in a component – before actual manufacturing begins.
Physics-Informed Neural Networks (PINNs) – Neural networks based not only on measurement data but also on physical laws (e.g. heat conduction equations). They enable precise predictions even with small datasets.
