Microservice-based machine learning production deployment of an anomaly detection algorithm for smart manufacturing machines

Many production lines produce on high volume with high throughput. An error in production leads to a stop of the production line and that results in high costs and lost revenue. With the increasing number of IoT devices in smart factories, more and more data gets generated and made available. This digitization process provides opportunities to improve and enhance production processes. Therefore, Bosch Connected Industry (BCI) has developed an anomaly detection algorithm for time series data containing process curves generated by IoT devices. The algorithm identifies anomalous behavior and alerts the user to avoid production line stops in a predictive manner. The anomaly detection algorithm will be one module of the software portfolio called Nexeed. The software Nexeed aims to digitize the value chain of smart factories. The goal of this master thesis is to deploy the anomaly detection algorithm into the Bosch production environment given several requirements and constraints. The thesis provides two outcomes. First of all, it provides a suitable software architecture to deploy the anomaly detection model into the Bosch software ecosystem given several requirements and constraints. Secondly, it provides a suitable software development approach to deliver a reliable and functioning minimum viable product while fulfilling all the requirements. The developed software architecture has been tested and evaluated with stress tests, performance tests, and production tests