Environmental test chamber HMI system
An HMI built into the chamber panel: the industrial PC and the PAD run identical features, and the A...
View Details →The Screen
Operators do not walk over with a laptop; they stand at the machine and work on the panel. So the first constraint is not feature count but standing-room usability — a left rail hides 15 modules, a top bar carries language, date, user and lighting, and a floating ball in the corner holds the occasional jobs: screenshots, walkthroughs, videos and alarm logs.
The question every chamber gets is "what is it at now". This is not a reporting tool, it has to be readable from a metre away through the door glass — so the landing view gives two cards: measured value large, setpoint small, deviation flagged in red. Curves and lamps sit below, second tier.
The rail answers "where do I go", the top bar answers "who am I, what language, is the lamp on", the ball handles "something quick right now". Three layers keep any one menu from growing to twenty items; with gloves on, the ball beats digging through a tree.
Three Clients
The specification is blunt: "PAD features are identical to the industrial PC". That is not laziness — a supervisor walking up with a tablet must see exactly what the panel shows, or the two of them are not talking about the same machine. The app goes the other way: it drops high-risk configuration and adds a standby screen plus manual shortcuts.
| Capability | Panel | PAD | Android app |
|---|---|---|---|
| Sign-in and accounts | ✓ | ✓ | ✓ |
| Device monitoring (run / state / faults) | ✓ | ✓ | ✓ |
| Curves (current / history / list / export) | ✓ | ✓ | ✓ |
| Logs (operations / alarms / door events) | ✓ | ✓ | partial |
| Upkeep (parts / maintenance dates) | ✓ | ✓ | ✓ |
| Program editing | ✓ | ✓ | — |
| Operation settings (run / schedule / fixed) | ✓ | ✓ | — |
| Initialisation / parameters / PID / EEV | ✓ | ✓ | — |
| DI/DO configuration and debug mode | ✓ | ✓ | — |
| Standby screen (climate at a glance) | — | — | ✓ |
| Header shortcuts (manuals / alarm highlight) | — | — | ✓ |
▲ Per the specification: panel and PAD align exactly; the app keeps monitoring, curves, upkeep and logs, with configuration left on the machine.
One wrong PID or EEV value can scrap a whole batch inside the chamber. Leaving configuration on the machine keeps the "on site, facing the box" bar in place; the app is for watching and chasing, not tuning. That boundary was set in requirements, not cut later.
Monitoring & Curves
Monitoring splits into running state and status, with faults shown flat and as a list; curves split into energy and climate. The one thing users actually stare at is whether the measured line hugs the setpoint line — hug means the PID is right, drift means go check an actuator.
▲ Faults come in two views: flat answers "how many are broken now", the list answers "when did it start and when did it clear".
Program Editor
A program is the soul of a chamber. Customers are not buying "a box that gets cold"; they are buying "a box that runs this curve for eight hours and then three more times". So the editor has to express three things per step — where to go, how long to stay, and when the clock starts — then stack two levels of looping on top.
Each step records target climate, ramp rate and hold time; steps can be added and edited, and the detail view lists every parameter of that step — changing a recipe means changing one step, not re-laying the program.
Two levels: program loop repeats the whole recipe, step loop repeats one segment. A damp-heat cycle that swings hot to cold fifty times is one loop count, not fifty copied rows.
Time signals close a digital output at a given node to drive a recorder or a linked machine; the wait setting decides when the clock starts — on reaching temperature, or once inside the tolerance band. That single choice decides whether a test is compliant.
PID & EEV
Temperature and humidity fight each other: cooling strips moisture out, humidifying puts heat in. So heat, cool, humidify and dehumidify each get their own PID set, and each set splits again into single-temperature mode and climate mode — controlling temperature alone should not have to humour the humidity loop. On the cooling side sits an EEV stage metering refrigerant.
| Branch | Set one | Set two | Set three | Why separate |
|---|---|---|---|---|
| Heat | Heating PID, single temp | Heating PID, temp + RH | — | Elements respond fast and overshoot; single-temp mode can take a firmer setting |
| Cool | Cooling PID, single temp | Cooling PID, temp + RH | — | Compressors have start delays and lag; settings stay soft to avoid short cycling |
| Humidify | Non-linear PID | Linear PID | Anti-runaway PID | Boiler and ultrasonic behave differently; anti-runaway caps single-shot output |
| Dehumidify | Non-linear PID | Linear PID | Energy-saving PID | Cooling and reheat must cooperate; the saving set trims reheat once target is met |
| EEV valves | Main valve max/min steps in cool and climate modes | Hot bypass: start-stop, EEV and solenoid parameters | Cold bypass: start-stop, EEV and solenoid parameters | Two compressors each carry an EVD set (Rst0–Rst3); step limits stop over-travel |
▲ 12 PID and 11 EEV parameters, all editable, storable and traceable on the panel — tuning is a parameter change, not a code change.
I/O & Initialisation
Every sensor and actuator in a chamber ends up on one PLC point. Initialisation decides how many channels this machine has; parameter setup decides what each channel means — get either wrong and the numbers coming up are just digits.
A sensor at the outlet reads two degrees below the specimen area — normal. The reference is what this chamber ought to read; the offset is what we add to make it read that. Storing them separately means calibration only moves the offset, and later you can still answer "what standard was this set against".
Tuning a chamber can take days, and a backup is those days preserved. Swapping a controller, flashing firmware or suspecting a mis-edit, one click gets you back to a known state; before shipping, it freezes the tuned set so on-site installation just restores it.
Upkeep & Logs
A chamber runs for days at a time; compressors, heaters and humidifier boilers all have service lives. Upkeep records where this machine has got to; logs record what has happened to it — one looks forward, one looks back.
The granularity is deliberate: parts answer "which component is due", the machine answers "this unit needs a day offline". Two lines running together means neither small parts nor the whole unit slips.
| Log type | What it answers | Presentation | Typical use |
|---|---|---|---|
| Operation log | who changed a parameter | list + detail | replay what preceded a fault |
| Alarm list | when which alarm fired | list + detail | align times with a complaint |
| Alarm statistics | which fault recurs | pie / line | feed model improvements |
| Sign-in log | who was on the machine | list | attribute on a shared unit |
| Door events | was the door opened mid-run | list + detail | judge whether data was disturbed |
Standby Screen
Walk a workshop floor and the question is always "what is this one at". The app switches to a standby page after a period without input, putting measured, setpoint and a period curve together so it reads from two metres — no tapping every machine on the round.
Platform Integration
One line in the specification matters: "terminal operations go to the platform server first, then to the vendor API". That layer is not overhead — it lets three clients share one interface and keeps a PLC point change from rippling through every client.
If three clients each talked to the PLC, "set a temperature" would be written three times, and a point change would mean three edits. With a layer between, the point map lives once and clients only know the interface; new models, new points, new third parties are server changes, not client releases.
A wrong read shows a stale number; a wrong write can disturb a running test. Display data takes a read-only path, operation data gets its own writable path, and every write lands in the log — who moved the setpoint from 85 to 25 is answerable afterwards. That boundary is what separates industrial software from a reporting back office.
FAQ & Delivery
Three ways. Position: the HMI is mounted on the machine and the operator stands at it, so buttons are large, hierarchy is shallow, and mis-taps are expensive — the floating ball and top bar both have to work with gloves on. Real time: climate, state and faults follow PLC points, and curves draw as they sample rather than refresh every few minutes. Authority: the panel not only reads, it writes — PID, EEV, compressor mode, fixed values and schedules — and a wrong write spoils the specimens inside, so every change needs confirmation and a record.
A program is a test recipe: several steps in sequence, each with a target climate and hold time, added or edited individually. Loops come in two levels — a program loop repeats the whole recipe, a step loop repeats one segment, so an alternating test needs no manual re-entry. Time signals close a digital output at a node to drive external equipment or mark a record point. The wait setting decides when the clock starts: on reaching temperature, or once climate settles inside the tolerance band, so a specimen is not timed before it is at condition.
Because the physics differ. Heating runs on elements — fast, prone to overshoot. Cooling runs on a compressor and expansion valve — lagging. Humidifying climbs slowly whether by boiler or ultrasonic. Dehumidifying needs cooling plus reheat and oscillates most easily. One setting across four branches will always fail one of them. So each branch keeps its own set, and each set splits into single-temperature and climate modes: controlling temperature alone need not humour the humidity coupling.
The industrial PC and PAD are identical — they are the machine's own operating surface, where on-site work happens. The Android app adds what the panel cannot: monitoring, curves, upkeep and logs, plus a standby screen that reports climate after a period idle, and header shortcuts for manuals and alarm highlighting. The platform is not a client at all: terminals call the platform server, and the server handles PLC point mapping and the vendor API, with data distribution and schema owned centrally. Multilingual was designed into the schema from the start.
MORE CASES
An HMI built into the chamber panel: the industrial PC and the PAD run identical features, and the A...
View Details →A low-altitude logistics multimodal management system for campus and community scenarios: an on-site...
View Details →A fire-safety management platform for property managers and district fire supervision: admin console...
View Details →Start with our pricing logic, then let's talk requirements — cost structure, technology comparison and acceptance criteria, all laid out in the open.