COMET project moFOCS (2015-2018)

Modelling, Prognosis, Forecast and Control of Systems

Aims / Research Topics

The moFOCS project focuses on predicting and optimizing process and product parameters in the area of industrial manufacturing. This is accomplished by the development of virtual sensors that employ machine learning and transfer learning methods as well as by the development of evolutionary stochastic optimization methods.

Methods / Software / Proof of Concepts

  • Deep learning models for angle prediction in metal sheet bending
  • Prototype of metal sheet cutting plan optimization that considers constraints on desired product properties
  • Incorporation of domain-specific parallelization patterns to tackle the high computational complexity
  • Learning framework for predicting properties of produced products and learning the causality of used material to resulting product properties
  • Domain specific language for data processing pipelines to manage the large variety of analysis workflows
  • Prototype of a data provenance system for data processing pipelines which ensures the traceability between deployed models, their generation workflow and the according documentation

Selected Publications

  • Transfer Learning [GBS+15, GCN17, GCN16, ZGL+17, GBS+17]
  • Multi-Task Learning [GBS+15, GBS+17]
  • Evolutionary Stochastic Optimization [CZH+16, RC16, CRH17]
  • Learning framework for prediction and causality [CZH+16]
  • Data Processing Workflows [Luf17]
  • Domain-invariant feature learning with recurrent autoencoders for time series prediction [ZZG16]
  • Domain-invariant representation learning for metal sheet bending [ZZG17]

Funding Partner

This project is subsidized in the frame of COMET – Competence Centers for Excellent Technologies by BMVIT, BMDW, State of Upper Austria and its scientific partners. The COMET program is handeled by FFG.

Contact

Michael Zwick

Zwick Michael

Researcher Data Analysis Systems
Phone: +43 50 343 843

Christian Lettner

Lettner Christian

Researcher Data Analysis Systems
Phone: +43 50 343 837

back