Case Studies & Applications

Anonymized examples of deployed projects – what the problem was, how the system was built, and what it delivered.

Robot condition monitoring from joint and cycle data

Problem

A fleet of industrial robots showed sporadic failures with no warning, causing unplanned line stops.

Data / image source

Joint currents, torques, cycle times, and alarm history read directly from the robot controllers.

Solution architecture

Edge collector logging to InfluxDB with anomaly detection models comparing robots and cycles over time.

Integration

Early-warning alerts to the maintenance team and Grafana dashboards per robot and line.

Outcome

Gradual degradation became visible weeks before failure; maintenance is now planned by condition.

Technologies

Robot protocolsInfluxDBGrafanaPythonon-premise anomaly detection

AI-based surface defect inspection

Problem

Manual inspection of machined surfaces was slow, subjective, and missed intermittent defect types.

Data / image source

High-resolution area scan cameras with controlled lighting installed over the line.

Solution architecture

Deep-learning segmentation models running on an industrial GPU PC at the line.

Integration

Pass/fail decisions to the PLC within cycle time; images and results stored in SQL for traceability.

Outcome

Full inspection of every part with consistent criteria and complete image-based quality records.

Technologies

Machine vision camerasdeep learningPLC integrationSQL

Video-based production process monitoring

Problem

Intermittent assembly errors could not be reproduced or explained from end-of-line data alone.

Data / image source

Continuous video streams from cameras over the manual assembly stations.

Solution architecture

Edge device running continuous video inference detecting process steps and deviations.

Integration

Event recording, real-time deviation alerts, and per-shift process statistics on dashboards.

Outcome

Root causes of assembly errors identified from recorded events; process deviations now trigger immediate alerts.

Technologies

Video AI inferenceNVIDIA JetsonRTSP camerasdashboards

CNC machine data collection

Problem

A machining plant had no live visibility into utilization, cycle times, or interruption causes.

Data / image source

Machine states, spindle data, programs, and counters from CNC controls over OPC UA.

Solution architecture

Edge collectors with local buffering feeding InfluxDB and SQL databases on-premise.

Integration

Grafana dashboards for utilization and downtime analysis; data available to MES via API.

Outcome

Objective utilization figures replaced estimates; interruption causes are analyzed weekly from real data.

Technologies

OPC UAInfluxDBGrafanaSQLedge buffering

AI-assisted maintenance knowledge system

Problem

Decades of maintenance knowledge lived in paper records and the memory of a few senior technicians.

Data / image source

Maintenance records, machine manuals, alarm logs, and machine history databases.

Solution architecture

On-premise AI agent with local language models indexing documents and machine data.

Integration

Natural-language interface for technicians; answers reference the underlying documents and records.

Outcome

Troubleshooting information that took hours to find is now retrieved in seconds – and stays in the company.

Technologies

Local language modelson-premise AIdocument indexingSQL

Edge AI inspection on NVIDIA Jetson

Problem

A quality check was needed in a production cell with no stable network and no space for a server.

Data / image source

Two industrial cameras capturing parts directly in the cell.

Solution architecture

Jetson Orin device running optimized detection models locally, fully independent of the cloud.

Integration

Result signals wired to the cell PLC; results synchronized to the plant database when the network allows.

Outcome

Reliable inspection with low latency in a constrained environment – no cloud, no server room.

Technologies

NVIDIA Jetson Orinmodel optimizationPLC integration

Multi-camera quality inspection

Problem

A single inspection point could not see all defect locations on a complex assembly.

Data / image source

Multiple synchronized cameras covering different views of each part.

Solution architecture

One edge system processing all streams with per-view AI models and a combined decision logic.

Integration

Combined pass/fail to the line PLC; per-view images stored for quality analysis.

Outcome

Complete coverage of defect locations in one station instead of several manual checks.

Technologies

Multi-camera acquisitiondeep learningedge AIPLC integration

Production dashboards with InfluxDB and Grafana

Problem

Production reports were assembled manually from spreadsheets days after the fact.

Data / image source

Machine counters, states, and energy data already collected by our data layer.

Solution architecture

InfluxDB time-series storage with Grafana dashboards and automated alerting.

Integration

Shop-floor displays, management views, and automatic threshold alerts to email.

Outcome

Live production visibility for all teams and automated daily reports replacing manual spreadsheets.

Technologies

InfluxDBGrafanaMQTTautomated reporting

Interested in this solution?

Tell us about your machines and processes – we will propose a practical architecture and a pilot you can evaluate.

Discuss your project