The software enables very large volumes of data, such as those produced by modern sensor systems, to be processed, analysed and visually displayed very rapidly
Researchers at Saarland University have developed a free data processing tool that allows rapid evaluation of cyclic sensor signals, pattern recognition and data visualization when processing huge datasets.
The software package – called DAVE – represents the very essence of a large number of research projects carried out by Professor Andreas Schütze and his team of experts in measurement and sensor technology at Saarland University.
DAVE stands for “Data Analysis and Verification/Visualization/Validation Environment” and its source code is publicly available on GitLab.
The free software enables very large volumes of data, such as those produced by modern sensor systems, to be processed, analysed and visually displayed very rapidly, so that researchers can optimize their measurement systems interactively.
A short introduction given in the readme of the repository, says, “DAVE is a MATLAB-based toolbox for the evaluation of, mainly, cyclic sensor signals. It focuses on cycle-based raw data preprocessing, graphical feature extraction, and data annotation, but also provides commonly used machine learning methods (with cycle-specific extensions) to develop data-driven models. This allows for a sleek workflow from start to finish without having to change to third-party tools.”
DAVE helps researchers to rapidly locate the best paths to take
Instead of relying on a conventional and time-consuming trial and error approach, the new software effectively asks the question ‘What happens when…’.
“Whenever we use our gas sensors to measure air pollutants, we are faced with the same old problem of analysing vast volumes of data and of recognizing signal patterns. If we want to continue to make our sensors more sensitive and more selective, we need to know whether very fine modifications to the sensors themselves and to the analysis actually bring about the desired improvements in sensitivity and selectivity. But there are countless ways in which sensors can be modified. We want to be able to identify the best paths as a rapidly as possible, or, equally, to quickly detect and reject the unproductive paths,” explains Professor Schütze.
“The software makes use of machine learning methodologies and enables us to identify patterns rapidly, to evaluate data cleanly and to visualize our results,” he says.
The researchers are now publishing their software tool under a copyleft licence. (Under copyleft rules, any adaptations of the original work, such as changes or enhancements are also bound by the same licence that covers the original work.)
“Anyone may use the open source software, provided that when results are published, the authors make reference to DAVE,” asserts Schütze.
DAVE helps researchers solve measurement problem
The researchers claim that any amount of sensor data can be processed with this software tool.
“DAVE is the opposite of a black box. The software makes the calculations completely transparent. It shows the user that when they alter a particular parameter, it has a specific identifiable consequence. The visualization modules in DAVE also make it easier to optimize a measurement system. The user can run through, test out and visualize different variants, and that helps the user find the most promising variants quickly and efficiently,” explains Manuel Bastuck, a research assistant in the Measurement Technology Lab and the developer of the software.
Schütze added, “Using DAVE as a tool, we were able to rapidly achieve some widely acclaimed results in the field of condition monitoring in ‘Industry 4.0’ applications. The results not only helped to solve the measurement problem itself, but also to configure the measuring system more simply and more cost-effectively.”
