Internship Parallel Time-Series Anomaly Detection in Production Lines for Philips Personal Health Products

Your role:

Problem description:

Traditional methods of assembling complex products face many challenges because they depend on fixed operations and the expertise of technical staff. Due to the complexity of the assembly process, anomalies can often occur. These anomalies can be classified into three types: point anomaly (a single data point that deviates from normal behavior), contextual anomaly (data that is unusual only within a specific context), and collective anomaly (an unusual pattern formed by a group of otherwise normal data points).

Relevance:

In a parallel system that performs the same type of measurement, anomalies can occur in a different way than in a stand-alone measurement system. In a stand-alone system, a point-anomaly can indicate an issue with the measured product. In a linked system, however, if a point anomaly occurs in multiple linked measurement systems, it can be an indication of an issue with the measurement devices, or even an external issue. This linked approach can ensure better control and understanding of the system behavior, and overall improved efficiency in the assembly process.

Perform periodicity and trend analysis in time series production data;
Synchronize parallel time series measurements done in production lines;
Detect anomalies for linked production system;
Identify important elements that cause anomalies and influence performance;
Create and test ML algorithms for predicting future measurements.

You’re the right fit if:

You are studying towards your bachelor or master in Industrial Engineering, Data Science, Mechanical Engineering or related study;
You have strong analytical and problem solving skills;
You are able to work within a scope and a given timeframe;
You are a team player, contributing to effective team work and offering a point of view.