The sensor unit
A single compact unit fits to the machine and carries every channel it measures. It mounts
externally — magnetically, bracketed or cable-tied — and its probes clip to the points that
matter. Nothing is wired into the control panel.
The unit reports roughly every 60 seconds over Wi-Fi. That interval is configurable
remotely, so a machine that needs closer watching can be sampled harder without anyone
visiting the floor. Between reports it keeps counting pulses continuously, so throughput is
totalled rather than sampled.
- Two independent thermocouple inputs for process temperature
- A pulse input for line speed, cycles and output counting
- Three-axis vibration sensing with event detection
- Ambient temperature, humidity, barometric pressure and light level
- Set up from a phone browser — no app required to commission a unit
Machines, sensors and channels
Three words are used precisely throughout the portal. A machine is a piece of
equipment — Extruder 06. A sensor is a unit fitted to it. A
channel is one thing being measured — Zone 3, or line speed.
Limits are set per channel, because Zone 3 and Zone 5 are not the same thing and should
never share a threshold. A machine can carry several sensors, and each sensor can report
many channels, so a large machine is described accurately rather than averaged into a single
number.
Limits and alerts
Every channel takes four thresholds: warning and critical, upper and lower. A zone that runs
too cold is as much a problem as one that runs too hot, and both are caught.
One excursion raises one alert, and will not re-raise while it is unresolved. This matters
more than it sounds — a long drift on a busy machine would otherwise generate a hundred
messages and train everyone to ignore them. When the channel returns inside its limits, a
resolution notice closes the loop.
Notification channels
Alerts go out by WhatsApp, email, in-app notification and phone push. You choose which
people receive which severity, and escalation policies move an unacknowledged alert up the
hierarchy on a delay you set.
- Reply
ACK on WhatsApp to acknowledge and stop escalation
- Reply
SNOOZE 1h through 24h to silence a sensor temporarily
- Reply
STOP to unsubscribe from alerts entirely
- Every acknowledgement is recorded against the person who made it, and where they made it
Product recipes
Different products run at different temperatures. Rather than editing thresholds by hand
every changeover, save a named set of limits per product per machine and switch the line
over in one action.
The active product is also attached to the analytics, so runtime and speed figures are
reported in the context of what was actually being made.
Shifts, breaks and maintenance windows
Alerts respect how the plant actually runs. Scheduled maintenance windows suppress alerts
for a machine that is deliberately down. Shift boundaries and breaks stop a cooling machine
from waking somebody at 6am.
This is the difference between a system that people keep and one they mute after a
fortnight.
Analytics
The same readings that raise alarms build a record of how the plant runs.
- Runtime and downtime, per machine and per day
- Actual speed against the target for the product being run
- OEE — availability and performance, with shift and product context
- Production by shift, so patterns between crews become visible
Reports and export
Scheduled email summaries go out daily, weekly or monthly, with configurable sections so the
report says what your plant cares about. Any channel over any period can be exported as CSV
or JSON — the data is yours, and it leaves in a format a spreadsheet or a data team can use.
Over-the-air firmware updates
Units are updated remotely, without a site visit and without taking a machine out of
service. Your administrator approves which firmware version the fleet is allowed to run, so
updates happen on your schedule rather than ours.
Users, roles and audit
Five roles, from read-only viewer up to administrator — including a notification-only
contact role for people who need the alerts but will never log in. Every configuration
change is written to an audit trail, so who changed a limit and when is never in question.
Example roles and permissions
| Name |
Role |
Can do |
| Daniel Marais | Administrator | Everything: users, limits, products, firmware |
| Priya Naidoo | Operator | Acknowledge alerts, view everything, change nothing |
| Johan Steyn | Viewer | Read-only dashboards and history |
| Thandi Mokoena | Contact | Receives alerts on WhatsApp; has no login at all |
Illustrative data
Integration
API keys and endpoints let an existing MES or ERP read machine and product state, or switch
the active product automatically when a job changes. If your systems already know what is
being made, PulseVault does not need to be told twice.