Design a statistical process monitoring method for the in-line detection of defects, ie Propose a method to model, analyze and reduce the dimensionality of the in-line inspection data (use this real dataset to test and apply the methods you learn in this course).
- You are expected to work directly on the output Dataframe (you can use all variables or a subset of them, you have full freedom in deciding how to use these data to meet the project work objective)
- You are also free to explore image data processing and analysis methods directly and work directly on the images in addition to (or in place of) using the output dataframe
Phase 1: use the phase 1 dataset (collected @ MADE) to design your statical process monitoring method -> present your proposed solution in the Phase I report
report due-date: 23/5/2024 (midnight)
Phase 2: you will be provided with other images of parts with different types of defects -> test your previously designed method on the new data and present the result in the Phase II report (Phase II report is much shorter than Phase I report, and it only includes analysis results and discussions related to the new data)
report due-date: 10/06/2024 (midnight)
- Clearly state all assumptions.
- Clearly describe and motivate any choice in the design of your method.
- You are free to test and compare more than one approach (in that case, a trade-off analysis is expected, as well as a clear statement about the best approach).
- In Phase 2, if you are not happy with the performance of your proposed method, you can also suggest how to tune it or revise it to achieve a more effective defect detection.
- Any creative solution is welcome.