Honest about uncertainty
Flags are prompts to inspect, not predictions. We’ll be clear about what the data can and can’t show for each machine.
We’re at the development stage. There is no general release yet, no paying customers, and no published performance results. We’d rather say that plainly than overstate it.
In many plants, machine readings sit in a historian, maintenance notes sit in a work-order system, and the knowledge of what “normal” sounds like sits with experienced technicians. Pulling those together for a single decision — should we look at this pump this week? — is slower than it should be.
MechSignal is being built to make that decision easier: one place to see how each machine is behaving compared with its own history, what has been flagged, and what technicians found when they checked.
We see the software as support for skilled people, not a substitute for them. Technicians make the call; the system helps them decide where to look.
Flags are prompts to inspect, not predictions. We’ll be clear about what the data can and can’t show for each machine.
Performance is measured on the equipment in scope, with technician findings as the reference — not borrowed from elsewhere.
MechSignal reads data. It doesn’t control machines and doesn’t replace protective or safety functions.
Not as a general release. MechSignal is in development, and we’re talking with a small number of organisations about pilots to build and validate it on real equipment.
No, and we don’t claim it can. MechSignal is designed to flag when a machine’s readings move away from its usual operating behaviour, so your team can decide whether to inspect. Some deviations will turn out to be harmless, and some faults may develop without a clear signal in the data available.
We don’t know yet, and we won’t quote figures we haven’t measured. Any benefit depends on your equipment, data quality and maintenance processes. A pilot is how we’d find out, together, with results measured against your own records.
No. MechSignal is a maintenance decision-support tool. It does not replace trip systems, protective relays, emergency stops, guarding, or required inspections, and it is not designed to take any control action on equipment.
Rotating and process equipment such as pumps, motors, fans, compressors and gearboxes are typical starting points, provided relevant signals are available. Suitability depends on the sensors or data connections you have; see the deployment requirements.
Not necessarily. If you already collect suitable data through a historian, SCADA system or PLCs, we’d start there. Where relevant signals are missing, retrofit sensors may be needed. We’ll assess this with you before a pilot.
The planned architecture uses AWS IoT SiteWise for industrial data collection and asset modelling, Amazon S3 for historical storage, Amazon SageMaker AI for anomaly detection models, and Amazon SNS for notifications. This is a plan and may change. See the architecture overview.
You do. Data handling, access, retention and residency terms will be agreed in writing with each pilot partner. [Link to data handling policy when available]
Yes. We’re interested in talking with service providers who maintain equipment across multiple customer sites, as well as manufacturers running their own maintenance teams.
Get in touch — we’re happy to talk through your equipment and whether MechSignal could help.