Rot-Weiß

• 

  Scheduling optimization for test order

• 

  Which processes take up most of the time during testing?

• 

  data-driven recalibration optimization

• 

  Error heuristics for tests

• 

  Different tests have different time durations

Inspiration

What it does

We are optimizing the time that hardware testing takes. Through analysis how long each part of the testing takes, we developed the following approaches (with decreasing impact):

1. check if tests are unnecessary (never fails or only fails IF another test fails) --> that was not the case in the fiven dataset
2. optmize order in which the tests are conducted (_test duration_, _likelihood that test fails_) -> _order of test_ We reached a time saving of 7.8 hrs!
3. optimize calibration. only calibrate when necessary by detecting calibration error in data We have a PoC that calbration errors increase differently after each calibration and thus it makes sense to adapt the calibration to the real error

How we built it

Data processing was done using python with the provided scripts as a starting point

Challenges we ran into

Many plots were uninteresting because most values were normal-distributed, so we had to tweak the machine simulator a bit

Accomplishments that we're proud of

We already have a real proof of concept that optimizing the order the tests are conducted on-the-fly can reduce testing time. And we visualized it in a nice video

What we learned

Creating animated graphs :)

What's next for Rot-Weiß

Testing the approaches with real world data in order to give an estimation of the time saving potential!

Built With

• matlibplot
• numpy
• python
• scikit-learn