Lighting Systems
The cause is rarely where the problem becomes visible.
When the production line stops, the search begins: the root cause is rarely where the problem becomes visible – it is hidden in a multitude of distributed signals and events. The expensive line item is not the repair itself, but the time until the cause is found – on a line with tens to hundreds of thousands of signals, that can take hours.
Fault-origin finding for the production of light: tightly networked machines, controls and processes become one unified real-time diagnosis – 24/7, in passive read-only mode, without touching the production process.
Challenge
Modern production lines consist of tightly networked machines, controls and processes that continuously generate and exchange tens to hundreds of thousands of signals. When a fault occurs, its origin is often hard to locate: the relevant information is distributed across numerous systems, the connections are not immediately visible – and root-cause analysis requires experience and deep plant knowledge.
Solution
The Kauzen AI system continuously analyzes the message traffic of your production environment in a context-aware way. It independently explores the existing OPC-UA data structures, automatically creates interpretable context hierarchies and locally trains hundreds of specialized AI models in your company. A multitude of technical signals becomes comprehensible information – independent of specialized expert knowledge. The connection is minimally invasive: a link to the existing OPC-UA server in pure read-only mode, without changing any control processes.
Where money is lost today
Tens of thousands of signals, hidden cause
Production lines generate tens to hundreds of thousands of signals – from messages and process values to machine and plant states. When a fault occurs, its origin is often hard to locate, because the relevant information is distributed across numerous systems and the connections are not immediately visible.
Result: fault-finding runs across multiple trades – and every hour of unclarified cause costs production.
The alarm shows the symptom, not the sequence
An alarm shows what is off. Which sequence of events led to the fault, and what connections lie behind it, usually remains unclear without context – even when the data exists in the plant.
Result: complex fault states are not systematically traced – the cause remains a guess.
Diagnosis depends on one person
Root-cause analysis requires experience and deep plant knowledge: specialists have to reconstruct fault patterns manually and are pulled away from other important tasks. When that knowledge leaves with the shift change or with the employee, the search starts from scratch with every stop.
Result: high dependence on experts, long diagnosis times, repeated fault-finding.
What we monitor
PLCs & controls
- Program and communication faults
- Response times and takt violations
- Bus and interface faults
Drives & motors
- Overload and overheating
- Load and speed deviations
- Bearing and gearbox behavior
Metrology & quality
- Deviations in measurement signals before they hit the product
- Correlation of process and measurement values
- Traceable measurement point by measurement point
Handling & feeding
- Conveying and feeding faults
- Blockages and material backpressure
- Tracking errors along the line
Power supply
- Voltage dips and grid effects
- Phase and harmonic deviations
- Correlation with fault timestamps
Takt & production data
- OEE losses in real time
- Stop and partial-load events
- Connections across the whole line
Key Benefits
Instant diagnosis
The moment a fault occurs: location and cause at a glance – repair with a target, not guesses.
Deep machine integration
Monitors hierarchical signal structures in real time
Self-learning
Continuously adapts to new production contexts
Machine insights
Never-before-seen insights from your machines – usable in other processes across your plant.
- The smartest on-premises AI on the market
- Full service when IT infrastructure is missing
- Local networks + no cloud lock-in, high data sovereignty
- Modular system *1 or 1000 machines
- Ideal for OPC UA, but open to other interfaces
Implementation Preview
Environment exploration & digital twin
Configuration & NN architecture design
Installation & system integration
Passive neural network training – under optimal conditions
Acceptance & employee training