Download PDF: Combination of technologies improves friction stir welding
NESC developed several innovative tools and techniques during its evaluation to find the root causes of low tensile strength and low topography anomaly (LTA) in welds formed using a solid-state welding process called self-reactive friction stir welding (SRFSW).
The evaluation team used a combination of machine learning, statistical modeling, and physically-based simulation to help improve the welding process and solve both problems, removing constraints placed on the flight hardware.
Development of LTA detection technology
Several techniques are needed to identify the root cause of LTA observed in low tensile strength welds and weld fracture surfaces.
- Deep learning for LTA detection: The NESC team developed a machine learning model to detect and segment LTA in weld images. The model was trained on images annotated by metallurgical experts, and majority consensus was obtained to resolve disagreements. The team then developed standard operating procedures to accompany image capture to improve robustness and reduce bias. The model built on NASA’s previous work to develop a foundational model for specialized microscopy analysis by pre-training it on more than 100,000 microscopy images. This step was crucial to link process parameters and LTA occurrence in an objective and unbiased manner.
The research team trained a neural network to detect LTA from images of fracture surfaces and eliminated the problem of manual LTA identification by pre-training the encoder on a large NASA dataset of microscopic images called MicroNet.
- Integrated data ingestion framework: SRFSW is a complex process with many interacting variables. Welding processes generate large amounts of data of various data types, including dozens of tabular process parameters, dozens of continuous data streams from production tools, cross-sectional images of fractures and welds, and mechanical test lab data. A Python-based framework was developed to automatically ingest, validate, and compile these diverse data into a single master spreadsheet and database. This tool reduced manual effort, minimized transcription errors, and improved data quality for downstream analysis. The team provided tools for continued use by stakeholders.
- Data analysis web applications: New web-based visualization and analysis tools have enabled engineers and subject matter experts to rapidly explore integrated datasets for faster hypothesis testing and more intuitive insight generation throughout investigations.
- Space-filling design of experiments: Because SRFSW involves complex nonlinear relationships between process parameters, the team realized that traditional factorial design was insufficient and implemented a space-filling design of experiments (DOE) to efficiently explore the entire parameter space. These data-trained machine learning models capture the underlying welding behavior. The team also developed a software tool to generate such designs and shared it with stakeholders.
- Physics-based SRFSW simulation: Creating a computational model of the SRFSW process simulates welding conditions, microstructure evolution, and resulting properties, providing insight into aspects of the welding process that are not accessible with physical sensors. This deepened understanding and led to improvements.
Identifying the root cause of LTA
Using these tools and analysis, the team identified two root causes of the decline in LTA and tensile strength.
- Overly aggressive post-weld surface treatment during production will reduce weld strength.
- Welding power input outside the optimal range will result in inconsistent welds and increase the risk of LTA.
The process model helped define the target welding power input window and recommended how to adjust the key control parameters to ensure that goal was achieved. Follow-up manufacturing testing confirmed that these adjustments can be implemented with high accuracy, eliminating both low-strength welds and LTA.
Friction stir welding
SRFSW involves inserting a rotating pin into the seam between two metal plates, which generates heat through friction to fuse the sheets together without melting the materials. This technique produces stronger joints than traditional welding and allows the use of high-performance but traditionally unweldable alloys, such as aluminum 2219.
SRFSW technology uses friction to agitate the material at the molecular level and does not use a blowtorch or solder.
NASA’s Friction Stir Welding Laboratory is located at NASA’s Michaud Vertical Assembly Center in New Orleans and is used to join the major components of the SLS rocket.
/Open to the public. This material from the original organization/author may be of a contemporary nature and has been edited for clarity, style, and length. Mirage.News does not take any institutional position or position, and all views, positions, and conclusions expressed herein are those of the authors alone. Read the full text here.
