Inspiration
Aerospace components such as the titanium brackets must conform strict quality and safety standards. Therefore, only detecting manufacturing problems after a part has been completely manufactured leads to quality issues and delays. The tool was created to help process engineers identify problems earlier in the manufacturing process to connect the final strength of the brackets to various printing configurations. This would turn raw manufacturing data into actionable insights using Six Sigma and statistical process control.
What it does
The tool (CPR Aerospace QC) helps determine whether a 3D-printing process is stable and capable of meeting aerospace specifications. It analyzes temperature, pressure, and measured yield strength to identify unusual variation, detect out-of-control production subgroups, and assess process capability. By inputting the necessary spec limits and bin size, the app calculates distributions, normality testing, and process capability indices (Cp, Cpk) to help users recognize potential issues and decide when further investigation or machine calibration may be needed.
How was it built
CPR Aerospace QC was built around a statistical analysis engine that processes manufacturing data from CSV files. Using Python, we implemented Six Sigma methods for calculating X-bar and range control limits, estimating process variation, computing Cp and Cpk, and testing normality. The results are connected the analysis to a web interface so users can review results and receive recommendations. The tool also creates three demonstration datasets representing a capable process, a drifting process, and a process with excessive variation, and added automated tests to validate the calculations.
Challenges
One of the biggest challenges was confirming whether the calculated data would be accurate and correctly displayed in the user-interface app. It took many iterations to successfully connect the dataset calculated in the backend of the tool to the app. While the /review command was useful in ensuring the code itself ran smoothly, it was more difficult to read the calculated data without the app.
Accomplishments that we're proud of
Some of the accomplishments to be proud of include connecting an AI agent to a repo (for the first time), using various techniques to get through the limited tokens in the free Google API plan (switching models, using new scripts, etc.) and applying process control knowledge learned from the course to a visual app.
Lessons Learned
One of the first lessons learned was to carefully read into the code changes that the AI agent made. While all updates were reviewed before pushing to the repo, a more detailed analysis would have helped solve coding issues much sooner (like the difficulty in connecting the dataset to the app). Another lesson learned was applying the calculated data. While the app shows the process capability indices and the normality, these numbers are frankly meaningless if the user cannot apply them to their own standards. In future iterations, it would help to include specific recommendations based on these values.
What's next for CPR Aerospace QC
The next step for CPR Aerospace QC is to make the tool more practical for real manufacturing environments by enabling live monitoring through direct connections to 3D printer data. Future improvements include clearer explanations of out-of-control signals, expanded interactive control charts and capability plots, and stronger validation using additional datasets and real manufacturing measurements. Usability can also be improved by simplifying data uploads, report generation, and sharing results, making it easier for engineering teams to monitor quality and make informed decisions.
Log in or sign up for Devpost to join the conversation.