The repository contains the development of applications to detect anomalies in test results, to monitor and predict the test rig condition (forecasting). The overall purpose is to implement predictive maintenance and to identify a need for maintenance in advance of a rig failure or excessive deterioration of its performance. The code also serves to provide advanced analytics of test article performance over time or during the test cycle.
The forecaster app's cloud architecture is built to continuously test, build, deploy and serve the app. The app updates are triggered by the code base modification in the this repository (push-to-master trigger). The build specs can be found in cloudbuild folder and include the pytest testing, Docker image building and pushing to Google Container Registry (GCR).
The app includes file upload boxes to enable new data analysis. The file uploader will ingest and validate the new coming raw data file. In case the raw data file is valid and new indeed (no such data file was found in the raw data storage in the raw data storage, Google Cloud Storage (GCS) test_rig_raw_data bucket), this file will be uploaded to the raw data storage and will trigger execution of the training pipeline.
The following link directs to the app.
streamlit run forecasting/src/serving.py