Manufacturing
We help you achieve continuous improvement, predictable
quality, and a line that runs like a Swiss watch.
Modern plants already record almost everything — every setpoint, every sensor reading, stored and timestamped.
Let’s put that data to good use.
Algorithms and software for your adaptive manufacturing systems
An optimised, adaptive production line runs more efficiently, at higher quality despite varying feedstock, and with less downtime. We build the model for your process, design the required optimisation algorithms, and write software to advise the operator. After rigorous testing algorithms could tweak the process autonomously. Where we come in:
Got a production line?
Would you like to make it smarter? Book a chat, let’s figure it out.
Optimisation and scheduling
Even small optimisations can bring substantial savings, for example in resource costs and in reduced production waste. We help you find the bottlenecks by setting up optimisation and simulation models. Our expertise with statistical tools and mathematical optimisation can be used in static settings or digital twins integrated into real-time data feeds.
Digisalix has extensive experience across the chain. Buying electricity on the market, scheduling production, optimising best operator shifts, tuning process controls, and predicting the quality of the finished product.
)
Condition monitoring
A modern production line is often well instrumented. That makes it a good place to do condition monitoring and anomaly detection properly: motor current, vibration, temperature, pressure, and the controls around them, read together into cloud storage.
Together with our customers, Digisalix works from the data, process diagrams, and operator experience: condition monitoring that directs attention, evidence-based remaining-life estimates, and properly calibrated thresholds where they suffice. For a cheap part with a spare on the shelf, running it to failure can be the right answer.
)
Predictive maintenance
The question is when a component is likely to fail, and more importantly what to do about it: repair now at a cost, or risk the wait for the next scheduled shutdown. Some parts don’t fail suddenly, they just get dirty. Then the operator is balancing between a cleaning break and slipping quality. With some modelling work the scheduling becomes exact maths with profit-maximising answers.
Sometimes the hard part is patching together the data, like the failure and maintenance history. The maintenance log is usually free text in another system, so “what broke, and when did it start” has no clean answer yet. Matching the failures to maintenance records and signals preceding this is where these projects tend to get slow, and turn into data archaeology more than modelling. Establishing early whether your maintenance history can carry a model at all is crucial.
)
Automated quality inspection
Optimising a process without reliable quality feedback is like riding a bike blindfolded. Every manufacturing process has its own quality control regime. For complex machines like an excavator it might mean a full day at a test range. For paper, a visual scan for stains. For others, a slow lab test that holds up the line. An algorithm grades the thousandth unit by the same standard as the first, at three in the morning, in the sixth week of a run.
We’ve worked on predicting lab test results to get feedback on process state in real time. We’ve taught cameras where to look. We’ve built the tools to curate training data for machine learning algorithms.
If the sensors and data are there, we can make sense of them.
)