Synapse Group
Predictive maintenance and condition monitoring
Catch degradation before it becomes downtime.
Vibration, temperature, current and pressure on critical assets, with the baseline measured and the alert wired into the CMMS. Without replacing the existing control infrastructure.
Get in touchWhen it shows up
If this is happening, this is the problem.
The schedule runs by calendar or by kilometre, and failures still appear between services.
The bottleneck asset has stopped the line more than once this year without warning.
Vibration readings are taken on a periodic round, filed away, and never cross-checked against what later failed.
Nobody can say how many downtime hours that equipment had last quarter.
How we approach it
What gets installed and how detection works.
- 01
What gets measured
Vibration, bearing temperature, three-phase current, ultrasound, pressure, flow and oil analysis. The variable is chosen for the failure mode that matters on that asset, not for what is easy to install.
- 02
How it is installed
Sensors on the equipment plus whatever the PLC and historian already provide. Replacing the OT infrastructure is not a prerequisite: you start with one asset and connect over OPC UA, Modbus or MQTT.
- 03
How detection works
A baseline of that asset's normal behaviour is built, and deviation is flagged against that pattern rather than a generic threshold. Where there are no labelled failures — the usual case — unsupervised methods are used.
- 04
What reaches the operator
Criticality, probable cause and action, not one more notification. The alert enters the CMMS as a work order, because a notice that never reaches the system where work is planned ends up being a screen nobody looks at.
Where it applies
The industries where this problem weighs most.
The same work changes scale and standard by front, not nature. These are the pages with the context for each one.
What it is measured against
The indicators are set before anything is installed.
A pilot without a baseline is no basis for a scale-up decision. These are the indicators the work is designed against, measured before starting so there is something to compare with.
- Downtime hours
- On the asset in scope, measured before installing so there is a baseline to compare against.
- MTBF and MTTR
- Mean time between failures and to repair, separated by equipment and by failure mode.
- Anticipated detections over total
- What share of the period's events was preceded by an alert. It is the number that says whether the system works or merely watches.
- False positives
- Notices that should not have fired, because they are what teach the team to ignore the system.
Frequently asked
What we get asked before the first meeting.
What is the difference between preventive and predictive maintenance?
Preventive intervenes on a calendar or an hour counter, regardless of the equipment's actual state. Predictive measures that state and estimates degradation, so intervention is scheduled on evidence rather than on turn. In practice predictive does not replace preventive: it reorders it, letting you space out what is healthy and bring forward what is not.
How many sensors are needed?
Fewer than usually assumed. You start with the bottleneck asset and the failure mode that costs most, not by instrumenting the whole plant. A typical pilot covers one machine or one line, and scale-up is decided on the evidence from that perimeter.
Does it work without a recorded failure history?
Yes, and it is the most common industrial case. Unsupervised methods learn the asset's normal behaviour and flag deviation without needing examples of the failure. What they do require is a prior observation period to build that reference, and that is accounted for in the pilot timeline.
Does it integrate with the CMMS we already use?
Yes, and it is part of the normal scope. The alert has to become a work order with its asset, its criticality and its probable cause. Without that integration the system produces information that competes with the maintenance flow instead of feeding it.
Where do you start?
With a 4 to 8 week diagnostic that defines the asset, the baseline, the indicator and the success criterion before anything is installed. A pilot that only proves the sensor transmits is no basis for a scale-up decision.
Contact
Let's talk.
If you have an operational challenge, an innovation project in mind or a question about applying technology to your industry, get in touch.
proyectos@synapsegroup.techWe reply within 48 business hours.