whoami
Lukáš Raszka
I work on data-driven methods for industry, from machine learning and machine vision to predictive maintenance, digital twins and industrial analytics, with a focus on results that hold up in real operation.
PhD student, Department of Industrial Systems Management, Faculty of Materials Science and Technology, VSB – Technical University of Ostrava, Czech Republic.
Research
My work asks a practical question: when does a model that scores well in the lab keep working on a real machine, a new plant or a different year? I evaluate methods under deployment conditions and report how uncertain the answer is.
The topics below are only a selection of what I work on. Every machine, process and dataset is unique, so I gladly take on problems that are not listed here, even when the solution is not yet clear.
Reliable machine learning for industrial time series
Anomaly detection and forecasting with calibrated uncertainty, evaluated under distribution shift, data leakage and limited labels, so reported accuracy reflects what happens in operation.
Predictive maintenance and condition monitoring
- Vibration-based anomaly detection for CNC machining, including commissioning on new machines without labelled faults
- Remaining useful life estimation and transfer between machines
- From tool-wear prediction to cost-aware replacement decisions
- Virtual sensors and fault detection for wind turbines, photovoltaics and HVAC systems
Machine vision and quality inspection
Camera-based and 3D-scanning inspection, focused on small, frequently changing production batches where fixed inspection setups are impractical, and in-situ imaging for additive manufacturing.
Digital twins and decision support
Digital twins that connect measured data with simulation, and decision support for remanufacturing and end-of-life recovery that carries uncertainty through to the decision.
Industrial data and cybersecurity
Benchmarks and data audits for anomaly detection on industrial control systems, and measurement of exposure and exploitation across industrial security advisories.
Industry, energy and sustainability analytics
Data-driven studies of manufacturing, energy and emissions, with colleagues from the economics and management side of the department.
Projects
Vision-based leak detection in pressurised gas cartridges
A machine-vision system estimates the leak rate of a submerged component from the number and size of the bubbles escaping from it. It targets sealing defects in mass-produced cartridges used in consumer and safety products.
Machine vision, bubble analysis, leak testing, quality control
Defect detection for small production batches
Planned work on flexible inspection of parts made in batches of a few thousand pieces, comparing camera-based and 3D-scanning approaches.
Machine vision, 3D scanning, flexible inspection
Team
Our group is part of the Department of Industrial Systems Management, which combines automation and industrial IT with the management, economics and ecology of industrial companies.
Ondřej Grycz, Ph.D.
Hardware: sensor and electronics design, embedded systems
Lukáš Raszka
Machine learning, machine vision, data acquisition and analytics
Colleagues and students
Data analytics, machine learning and industrial applications
Management and economics colleagues
Industrial management, economics and ecology
What the team builds
- Custom embedded electronics: sensor integration, PCB design with signal conditioning, firmware for microcontrollers and single-board computers
- Edge computing: models running directly on devices for real-time response
- Connectivity through OPC UA, Modbus, MQTT, LoRaWAN and BLE, including older machines
- Data pipelines, time-series storage and real-time dashboards with alerts
Publications
Complete list on ORCID, Google Scholar and ResearchGate. The items below are preprints that have not yet been peer reviewed.
Raszka, L., Grycz, O., Vilamová, Š., Ameir, O., Frischer, R. Shift, shortcuts and calibration: a deployment-realistic evaluation of unsupervised vibration-based anomaly detection in CNC machining. SSRN 7610685, preprint.
Raszka, L., Grycz, O. Zero-shot time-series foundation models for battery degradation forecasting: a leakage-aware multi-dataset benchmark with conformal uncertainty. SSRN 7661699, preprint.
Raszka, L., Grycz, O., Vilamová, Š., Ameir, O. Native or inherited? Remediation gaps and exploitation signals across 3,960 CISA ICS advisories (2010–2026). Preprint.
Collaboration
We are open to collaboration in many forms: joint research and projects, shared experiments and data, co-authored publications, studies, prototypes and pilot projects. Our scope is broad, and we are happy to work with both academic partners and companies.
Academic partners
Joint research and project proposals, shared experiments and data, co-authored publications and exchange of ideas across disciplines.
Companies
Studies, prototypes and pilot projects tailored to your process and equipment, from measurement and inspection to data analysis and digital tools.
The topics listed on this page are only a part of what we do. Every machine, process and dataset is unique, so we gladly take on problems that are not described here, even when the solution is not yet clear.
Write to me at lukas.raszka@vsb.cz and I will gladly go through everything with you. No question is a bad question, so send whatever you have, even a rough idea.
Contact
Department of Industrial Systems Management
Faculty of Materials Science and Technology
VSB – Technical University of Ostrava
17. listopadu 2172/15, 708 00 Ostrava-Poruba, Czech Republic