Prof. Dr. Heinz-Peter Bürkle

"Development of a software and hardware solution for location-independent and secure operation
control and monitoring of complex production systems"
Funding reference number: ZIM 1KF244161KM2
Term, period of validity, duration: 01.11.12-31.10.14

"Pattern recognition of critical signal sequences at the control level"
Funding reference number: ZIM 16KN042420
Term, period of validity, duration: 01.11.14-31.12.16

"Development of a cyber-physical real-time intrusion detection system in Industry 4.0 automation systems;
Smart Data Aggregator for intelligent signal acquisition of control and sensor signals"
Funding reference number: ZIM-ZF4020503ED6
Term, period of validity, duration: 01.11.16-31.12.18

"Distributed Smart AI Sensor Network with Large Scale Machine Learning Model Management in the Cloud; System on Chip runtime environment for machine learning models with feature extraction based on Sparse FFT"
Funding code: ZIM-ZF 4020508SS8
Term, period of validity, duration: 01.01.19-31.12.20

Machine Learning / AI in Embedded Systems

Challenges:

• Low hardware resources of embedded systems
  • Computing power
  • memory
  • Energy consumption (battery devices)
• Price
• Adaptation of machine learning algorithms
  • Memory, word width
  • Speed
• Tight coupling of digital signal processing (e.g. FFT) to ML algorithms
• Different platforms for training and inference
  • Training in the cloud, inference in the embedded

Machine learning for anamoalization detection of PLC control signals on an
an industrial embedded real-time Linux computer:

Single class SVM on Revolution Pi (real-time Linux) with web server to display
the training and classification results as well as the temporary anomaly rates.

• LTSPICE
• PSPICE
• Other simulation programs