Methoden und Tools

Our Methods and Tools

Tools in frequent use

Our experienced data experts master a variety of tools and methods – the means to unlock your data treasure chest. One of the most important tools has less to do with data than with know-how: honest communication at eye level. After all, successful data-driven innovation is a joint effort between data and domain experts. Feel free to take a look inside our toolbox!

With anomaly detection as a generalist approach, we identify previously unknown occurrences and patterns in existing process or machine data. Algorithms learn to independently recognize the normal state and report deviations, such as defects or other issues, to downstream systems or directly to the operator at an early stage.

In addition to physical-technical data, image data analysis provides further relevant information for quality control or process monitoring in production as well as wear detection on components or tools. It is also the basis for the digitization of documents. Deep learning approaches such as transfer learning allow to create comprehensive solutions even with small amounts of data.

By means of process data analysis, we generate added value and create transparency through expert aggregation and visualization of data. We develop individual KPIs for process optimization in order to improve the overall equipment effectiveness (OEE).

Data pre-processing and cleaning are often the key to make previously invisible correlations visible. We also contribute our technical-physical expertise to the analysis.

This way we discover process-specific correlations in the data that are relevant to the use case. This is often the most suitable initial approach. Artificial intelligence may come into play further down the road.

The raw data is often very noisy and the correlations are unknown. Even a label in the data which specifies the use case is often only of limited assistance. Outlier detection and appropriate filtering make correlations more visible. In order to find the right filters and algorithms, data and domain expertise must be combined.

Machine Learning and Artificial Intelligence only come into play when developing a predictive model. After all, without the basic preliminary work, it's "garbage in, garbage out".