MiSIS – Miniaturized system for intelligent signal processing in mechanical engineering

Project

Intelligent signal processing is required in mechanical engineering for process-integrated quality control, for operating life analyses of wear parts and for condition monitoring. For this reason, Fraunhofer IKTS developed a system for intelligent signal processing in the BMBF joint project "MiSIS" together with its partners.

 

DSP and FPGA – heart of the miniaturized system

The base module of the intelligent hardware consists of a digital signal processor (DSP) and a field-programmable gate array (FPGA) along with the required memory, power supply, and peripheral interfaces. The module in the form of a mini PCI express card (30 x 50.95 mm) can be operated in various systems, e.g. in laptops or tablets. The combination of DSP and FPGA allows for the realization of complex, real-time critical algorithms up to acoustic pattern recognition with deep neural networks (DNN) or hidden Markov models (HMM). The base module can be connected to various signal acquisition and signal output modules via flat ribbon cables. This makes it suitable for audio and ultrasonic signal processing as well as processing of other signal sources.

MiSIS – a cooperation with SINUS Messtechnik GmbH and the BTU Cottbus-Senftenberg

In the frame of the project, partner company SINUS Messtechnik GmbH has developed a signal acquisition module for high-quality processing of audio signals in the audible range (up to 24 kHz) with up to four independent channels. With these two modules, the system performance was demonstrated, e.g., in the automated detection of defects in glass bottles in cooperation with the Brandenburg University of Technology Cottbus-Senftenberg.

For this reason, the complete feature analysis (windowing, Fourier analysis, and temporal and spectral smoothing by filter matrices) and the classification algorithms (density calculation for hidden Markov models, search in finite state machines) were implemented in the base module. By transferring these CPU-intensive algorithms to the FPGA, the computing time was significantly decreased, thereby also enabling the feature analysis and classification techniques to be used for high-frequency signals in time-critical applications.

© Fraunhofer IKTS
Base module of the miniaturized system for intelligent signal processing.
© Fraunhofer IKTS
Integration into the measuring system.

Gears, valves, rollers: Quality control via acoustic pattern recognition and AI

Fraunhofer IKTS also successfully uses intelligent signal processing and acoustic pattern recognition for in-line quality control of gears, estimation of remaining lifetime of valves, early fault detection in rollers and bearings in spinning frames, structural health monitoring of airplane materials, and automatic softness testing of tissue products during production. The technology is very well suited to testing machine components or as test units in production plants.

 

Overview: Computing time with regard to signal length

Algorithm DSP [%] FPGA [%]
Feature analysis 9.4 1.1
Density calculation 59 34
Search 32 3.1

 

InfoBox:

Funding body: Bundesministerium für Bildung und Forschung (BMBF)
Project sponsor: VDI/VDE Innovation + Technik GmbH
Funding reference: FKZ 16ES0297
Project partners: SINUS Messtechnik GmbH, BTU Cottbus-Senftenberg