Smart fire-safety — community fire IoT platform: every device watched, inspected and verified
A fire-safety management platform for property managers and district fire supervision: admin console, wall dashboard and mini-program in one, covering the community register with map pinning, four kinds of hardware integration, warning and alarm dual thresholds, inspection plans and records, a work-order loop from report to sign-off, and fire-safety training exams — seven modules, thirty-two screens. De-identified presentation.
Multi-vendor device adaptation · dual thresholds · inspection and work-order engine · exam bank · role-based permissions
Jiding scope
Platform architecture, backend and mini-program development
7functional modules in one menu
32screens from register to reports
4hardware families integrated
2threshold levels: warning and alarm
Four layers: field devices feed an integration layer that adapts protocols, scales ranges and judges thresholds; the business layer carries registers and workflows; three clients sit on top
Three untidy ledgers in community fire safety
The hard part is never installing the devices. It is who watches them afterwards — dozens of communities, thousands of points, and only a handful of people.
The asset list lives on paper. Pressure and temperature sit in pump rooms and risers; without live data you learn about a failure when it becomes an incident. A fire panel that alarms inside an empty duty room has effectively not alarmed.
Inspections are signatures, not evidence. Plans live in spreadsheets, records on paper sign-in sheets. When a device was last checked, and whether the finding was closed, means digging through a binder.
Training leaves no trace. A session is held, a sheet is signed, and hours, question banks and scores are not retained. New staff start from zero, and drill records never form a year-on-year file.
Seven modules, thirty-two screens
The platform is organised along four lines: people, devices, process and content. All of it hangs off one backend menu, with today's device status and pending items summarised in charts on the home screen.
Where the work actually is: system admin and device management take almost two thirds of the screens, because permissions and multi-vendor integration are the real cost. Exams and inspections are fewer screens but the most used.
Integration is not "plug it in"
Four very different hardware sources run on the same platform. What decides whether the platform is usable is not whether the protocol connects, but how a number is interpreted and how an exception is graded.
Integrated object
Key fields registered
How the platform judges
BLE devices
Community, MAC address, storage location
Inspection points and proof of presence, linked to inspection records
Serial-server devices
Type / ID, community, location, alarm threshold, warning threshold
Raw value scaled to engineering units, then compared against both levels; thresholds editable in the console
Temperature / pressure / door contact
Temperature alarm and warning values, pressure alarm and warning valves, contact type
Two levels per metric; contact type distinguishes normally-open from normally-closed semantics
Smoke detectors / fire panels
Type / ID, panel, component number, loop number
Located by panel + loop + component; batch import from xlsx
Range configuration
Range type, value, unit
One shared definition, so swapping a sensor does not touch business code
One curve, two levels: the warning goes to the inspection queue, only the alarm escalates into a work order and a notification — which is how you stop an alert that fires every day from being ignored
An easily missed engineering decision: thresholds must be editable by operations, not hard-coded. Every community's network has its own baseline pressure; the same sensor model in two buildings has two different sensible ranges.
From an alarm to a signature
The real risk is not that an alarm cannot be raised. It is that nobody picks it up. So a work order is not a list, it is a path with a start, an owner and an end.
The loop: report and repair live on the same record, and the status only becomes closed after the reporter confirms; the repair note is written back to the asset history for the next fault
Two engineering decisions
Device differences stay in the integration layer. Four vendors send different payloads and use different ranges. Rather than write four business flows, we adapt protocols and normalise the model, so the business layer only knows "device + metric + value". The integration layer is heavier; adding a device family no longer touches business code.
The reporter owns the completion state. The technician saying "it is fixed" does not count; the person who reported confirms on site and only then is the record closed. One extra step, in exchange for every "completed" in the platform being believable — which is exactly what an audit or a post-incident review needs.
The other half: training, drills and notices
Exams and question bank. A paper sets a title and a duration; the bank filters by type, stem and category, with single- and multiple-choice options, scores and answers configured. The mini-program runs the timer and scores on submit.
Courses and drills. Courses are organised in chapters with cover image and attachments (procedures, emergency plans). Fire drills get their own record: date, participants and process, forming a year-on-year file.
Notices and feedback. Announcements, broadcasts and the home banner are managed by content type; pending notifications reach the responsible person; feedback from residents and staff is collected and searchable by user, content and time.
On the administration side, users, roles, departments and menus form the permission skeleton: menu rights and operation rights are ticked separately per role, departments carry an owner, and login records plus IP restrictions cover traceability and access control. Four technical screens (enums, parameters, logs, exceptions) stay closed to ordinary accounts.
What the client got
Device status visible in real time instead of "we go and look every so often"; pressure and temperature exceptions are graded the moment they cross a line.
Inspections that carry a timestamp and a point of presence, instead of a paper signature.
Repairs that are traceable work orders rather than phone calls, with the outcome verified by the reporter.
Training that accumulates: hours, question bank, scores and drill records in one place.
As agreed with the client, this page shows no client or community names and no screenshots of production screens. De-identified improvement figures are available during a scoping call, by comparable scenario.
Evaluating something similar?
Bring three things to the first call: the devices you already have on site (brands, protocols, what they send), who inspects and who repairs today, and which authorities or reports the platform has to answer to. We will first say whether the bottleneck is device access, process or accountability, then propose scope and a phased quote.
MORE CASES
More Related Cases
System Integration OMSmart fire protection, community fire protection IoT
Smart fire-safety platform for residential communities
A fire-safety management platform for property managers and district fire supervision: admin console...