Machine Learning

We develop the data analyses in the Jupyter Lab, script the measurement sequences and analyze the data with Tensorflow oder PyTorch. We transfer models and real-time-capable data analyses and machine learning models directly into the analysis operators of Optimizer4D and apply them directly to the machine in real-time operation. The results can be easily transferred to real production environments.

On the subject​

Big Data

With our structure-borne sound sensors, 25 MSamples/s in 16 bit resolution can be recorded in almost all manufacturing processes. With our magnetic sensors, we evaluate 4 MSamples/s in 24-bit resolution to assess the hardness and other properties of ferromagnetic components. 

On the subject​

Visualization of measurement data

The measuring devices are currently equipped with flash memory from 2 TB to 16 TB. Data rates of over 100 MB/s can be recorded continuously.

On the subject​

Network

All data can be stored in parallel on the network. Each
Optimizer4D can be operated completely remotely from a browser.
 

On the subject​

Compression

All data streams can be compressed and pre-processed on the measuring computer. Logs and results are stored in SQL databases and can be correlated with the raw and compressed data.

On the subject​

Documentation/Teamwork

Data and results can be shared from the Jupyter notebooks or via Streamlit. 

On the subject​

Tools for inline data analysis

 Software and smart data analysis are at the heart of our solutions. It is the evaluation of the collected data that gives it meaning. Our system is designed to display complex data transformations intuitively and to be able to apply complex analyses during the measurement.

On the subject​

Python Integration

By integrating Python into our software, we create maximum flexibility for your projects and, of course, access to state-of-the-art machine learning libraries. To keep the entry hurdle as low as possible, we install libraries such as Numpy, Tensorflow, SciPy and many others on the Optimizer4D before delivery. 

On the subject​

Python Operator

The Python Operator is one of our analysis modules for graphically programmable real-time analysis and yet it is completely different. This operator allows you to implement your own analysis approaches and apply them to the data stream in real time.

 

On the subject​

QASS Open Source Libraries

The Analyzer4D software comes with a network interface that can be used to control the measurement sequence, for example. Measurements can be started and stopped. Parameters can be set in the settings and application variables can be defined in the software.

On the subject​

Maschine communication

 
The trigger system is a sequence and communication controller within the Analyzer4D software. This controller is similar to a mini PLC, in which a type of program is executed cyclically, analogous to the PLC of a large machine controller.
​

 

On the subject​

Clustering

Pattern recognition and clustering of sporadically occurring micro-crack signals.

 

On the subject​