Machine learning - the entrance!

Data are the raw material of the 21st century

What can be put into practice in machine learning?

In the industrial environment, the possible applications of these technologies are also possible on a broad scale, but have not yet been consistently implemented or are only in the implementation phase.
There are two main areas of application: Knowledge Discovery and Predictive Analytics. Industrial Knowledge Discovery aims to identify causal interdependencies from data in order to derive specific measures.
The most frequently asked questions in this context are:"Which process parameters lead to high quality" and "What is the reason for the recently increased rejection rate? Here, established, interpretable methods such as learning and analysing decision trees and multivariate regression methods (with integrated feature selection) can be used. For example, during a laser welding process, the cause for non-optimal welding seams was identified as an oil film residue that was too thick on the raw material (cooperation between AMS Engineering GmbH and SCCH). The integration of the experts into the knowledge discovery process is essential for the success of these methods.
Machine-learning methods can be used to assess the quality of the models, but not the value of the correlations found and the hidden potential for improvement. This enables users to ensure that the data is properly prepared and the algorithms are correctly applied. On the other hand, an automatic distinction between causal relationships and often irrelevant statistical correlations is very difficult and manual intervention is often required. The latter is still a very active field of research and the methods developed so far are not yet sufficiently stable for productive industrial use.

Generate predictions

We go beyond the pure generation of knowledge to tasks in predictive analytics. Their aim is to generate prediction models for future events. The field of industrial application of such models is very diverse. It ranges from virtual sensors, error detection, diagnosis and prediction to prediction of critical quality attributes and model predictive control at the lower levels of the automation pyramid (see Fig. 2). On the upper levels of the pyramid, these methods are used, for example, to support logistical processes or to better predict material requirements and the required storage. The spectrum of methods used here is very broad and includes well known methods of machine learning such as neural networks and support vector machines and related methods that originate in areas such as system identification or control engineering. Particularly promising are hybrid approaches in which expert knowledge about the system and the process serves as a basis for an analytical model, whose potential systematic shortcomings can be corrected with a data-driven approach.

The full article can be found here: