Injection moulding machine and mould temperature monitoring

At a glance

Item	Typical answer
Problem	Process drift (a cooling circuit losing flow, wet material, a failing heater band) is found when parts fail inspection, often hours and thousands of parts later.
What is connected	Per-shot values and states from the moulding machines, barrel zone and mould temperatures, temperature control units, hot-runner controllers, resin dryers, cooling water and energy, through licensed gateways in the hall.
Who uses it	Process technicians and setters, the moulding hall supervisor, quality engineers, maintenance and the plant manager.
Typical first scope	Four to six machines that run the most demanding parts, with their auxiliaries.
Time to first screen	Typically within a week for machines that already expose data over Modbus. Budget four to eight weeks for a pilot with process windows that the process team trusts.

The situation

A moulding hall typically runs 20 to 80 machines of several makes and ages, each with a mould temperature unit or two, often a hot-runner controller, a dryer or a central drying and conveying system, and a take-out robot. The machine control watches its own process and can raise its own alarms, but each machine is an island: nobody sees the hall, setters walk from screen to screen, and the history behind a defect is lost when the machine's internal buffer rolls over.

Most quality problems start as slow drift. A scaled cooling channel loses flow and the mould runs a few degrees hotter, shrinkage changes and parts go out of tolerance. A dryer's desiccant saturates, the dew point rises, and splay appears on parts two hours later. A heater band fails on one barrel zone and the neighbouring zones compensate until they cannot. Each of these shows in the data long before a part fails a gauge.

Today drift is found by inspection, by a setter's experience, or by a customer complaint followed by a search through paper records. The goal is a hall-wide picture of every machine's process against its window, alarms that reach the setter while drift is still small, and a process history per machine and mould that quality can use.

Goals and measures

• Cycle time stability = standard deviation of cycle time over the last 50 shots, against the setter's reference. A rising spread is the earliest sign of trouble in material feed or the mould.
• Process window compliance = shots with cushion, peak injection pressure, injection time and plasticising time all inside their limits ÷ all shots, per shift and per mould.
• Mould temperature deviation = measured return or mould temperature minus the setpoint per circuit, as a maximum per hour.
• Material readiness = share of time the dryer's dew point and hopper temperature were inside limits while the machine produced.
• Scrap rate = rejected shots ÷ total shots, from the machine's reject counter or the robot's reject signal.
• Energy per shot = machine kWh ÷ shots per period, or per kg when shot weight is known, as a long-term efficiency figure per mould.
• Utilisation = time producing ÷ planned time, with setup and mould change counted separately.

A Statistics block in the machine profile's root flow computes the rolling standard deviation and average of cycle_s over a window of shots and saves them as telemetry, so charts and alarm rules use them like any other value. Daily and per-shift figures come from Hermetry's minute, hourly and daily aggregates and the increase of the shot counter.

Architecture

Put one licensed gateway running Hermetry Edge on each row of six to twelve machines. It reads machine controls over Modbus TCP where the control offers it, and the auxiliaries over Modbus RTU on RS485: many temperature control units, hot-runner controllers and dryers have a Modbus option, which is worth ordering with new equipment. Values the machine does not publish, such as cooling water flow per circuit or extra mould thermocouples, are added with transmitters and a small Modbus temperature or analog input module. Each machine and each auxiliary appears as a child device of the gateway. See Hermetry Edge and the licensed devices.

Machine data interfaces vary widely. Where a machine's only interface is one the gateways do not read today, take its signals another way: a cycle-complete or mould-open contact to the gateway's digital inputs, stack-light taps for state, and added sensors for temperatures and water. Ask us about the Edge roadmap for the interface in question before you plan around it.

Component	Role	Notes
Licensed gateway with Edge	Polls machines and auxiliaries, publishes telemetry	One per row. Claimed by scanning its code; one outbound MQTT connection with TLS.
Moulding machine control	Per-shot values, barrel temperatures, states, counters	Modbus TCP where available; otherwise contacts and added sensors.
Temperature control units	Mould circuit setpoint, supply and return temperature, flow, faults	Modbus RTU on the row's RS485 line.
Hot-runner controller	Zone temperatures, output percentage, zone faults	Modbus RTU where offered.
Resin dryer or central drying	Dew point, drying temperature, hopper level	Modbus RTU, or 4-20 mA dew point into an input module.
Cooling water instruments	Supply temperature, flow per circuit or per machine	Vortex or electromagnetic flow sensors with 4-20 mA.
Energy meter per machine	kW and kWh	Modbus RTU; heating bands and drives dominate the load.
Hermetry	Storage, flows, alarms, boards, SCADA	EU, US or Türkiye region, or on your own servers.

Point list

Per machine, split into a per-shot message and a periodic message, plus the auxiliaries. Register numbers depend on each control and are left out; units and scaling come from its manual.

Point	Telemetry key	Unit	Source	Interval	Notes
Machine state	state	code	Machine control or stack light	on change + 60 s	1 automatic production, 2 semi-automatic, 3 setup, 4 alarm, 5 stopped, 0 off.
Shot counter	shots	count	Machine control	per shot	Cumulative; never reset by the gateway.
Reject counter	rejects	count	Machine control or robot reject signal	per shot
Cycle time	cycle_s	s	Machine control	per shot
Injection time	inj_time_s	s	Machine control	per shot
Plasticising time	plast_time_s	s	Machine control	per shot	Rises with wet or inconsistent material.
Cushion	cushion_mm	mm	Machine control	per shot	Most sensitive to check-ring wear and feed problems.
Peak injection pressure	inj_press_peak_bar	bar	Machine control	per shot	Hydraulic or specific pressure; state which in the label.
Switchover pressure	switch_press_bar	bar	Machine control	per shot
Barrel zone temperatures	barrel_z1_c to barrel_z5_c	°C	Machine control	60 s	And nozzle_c for the nozzle.
Hydraulic oil temperature	oil_temp_c	°C	Machine control	60 s	Hydraulic machines only.
Mould number	mould_no	text	Machine control or setter entry	on change	Links process history to the mould.
Mould circuit temperatures	tcu1_supply_c, tcu1_return_c	°C	Temperature control unit	60 s	One pair per unit or circuit.
Mould circuit flow	tcu1_flow_lmin	l/min	TCU or flow sensor	60 s	Falling flow precedes rising temperature.
TCU setpoint and fault	tcu1_setpoint_c, tcu1_fault	°C, code	Temperature control unit	60 s, on change
Hot-runner zones	hr_z1_c to hr_z8_c	°C	Hot-runner controller	60 s	And hr_fault as a code.
Dryer dew point	dryer_dewpoint_c	°C	Dryer	60 s	Limit per material, for example -40 °C.
Drying temperature	dryer_temp_c	°C	Dryer	60 s
Cooling water supply temperature	cw_supply_c	°C	Sensor on the machine header	60 s
Machine power and energy	p_kw, kwh	kW, kWh	Energy meter	60 s, 15 min
Data link	plc_comm_ok	bool	Gateway Modbus status per device	60 s

Declare keys with label, unit, decimals and range in each profile's Telemetry keys tab. Per-shot values go in one message per shot, time-stamped by the machine or gateway at the end of the cycle, so every value of a shot shares one time stamp and can be read together in tables.

Asset model

• Plant (Site) with time zone and shift pattern.
• Moulding hall (Building or Area) with hall_supervisor.
• Row (Line) with cooling_water_loop and drying_system.
• Moulding machine (Machine) with clamp_force_t, screw_diameter_mm, machine_type (hydraulic, electric, hybrid), criticality.
• Auxiliaries (Machine) under each moulding machine: temperature control units, hot-runner controller, dryer, robot.
• Mould (free kind) with mould_no, cavities, material, customer_part_no, service_interval_shots.

Place each auxiliary's device under its machine, so the machine's status turns red when its TCU or dryer is in alarm, and a machine page shows the whole cell. Relate the mould asset to the machine it is mounted on with your own relation, for example Mounted on, and move the relation at each mould change. Rules on a row act only on that row's machines; an access policy on the hall limits a supervisor to it.

Process windows belong to the mould and part, not to the machine. Keep the limits, for example cushion_min_mm, cushion_max_mm, cycle_max_s and mould_temp_max_c, as server attributes on the machine device and rewrite them at each mould change, by hand on the device or from the MES through the REST API. Alarm rules read the attributes, so one rule set serves every mould.

Device setup

Profiles. One profile per machine control family (for example Electric IMM, make A and Hydraulic IMM, make B), one per auxiliary model, and one per gateway. Use Recent message online rules: 60 seconds for auxiliaries; for machines, the periodic message's interval, since per-shot messages stop whenever production stops.

Provisioning. The administrator claims each gateway by scanning its code and places it on its row. Machines and auxiliaries appear as child devices the first time the gateway reports them; name them after the machine number on the floor with a suffix, such as IMM-17 and IMM-17-TCU1.

Settings. Remote device settings on licensed gateways expose serial parameters, the Modbus maps of machines and auxiliaries, polling intervals and input configuration. Build a baseline per row type, dry-run it on one gateway, and roll it to the others with a bulk job in waves; the job pauses itself if fewer than 95 % of answering gateways apply it.

Per shot and periodic. Trigger the per-shot read on the cycle-complete signal where the gateway model supports event-triggered reads; otherwise poll the shot counter every few seconds and read the shot block when it changes. Send temperatures every 60 seconds. If the gateway model supports a deadband, set about 0.5 K on temperatures; otherwise an Only changes block with a dead band in the profile's root flow can do it. Do not dead-band per-shot values: each shot is a quality record.

Time, buffering, firmware. Synchronise gateway clocks; process records are only useful if shots line up with inspection times. Licensed gateways buffer readings while the link is down, within a capacity that depends on the model, and send them with original time stamps; alarm rules then evaluate them as a backfill, marked Backfilled, without paging for the past. Signed gateway firmware rolls out in waves with a soak time; schedule it for mould changes or planned stops.

Dashboards

Board	Audience	Widgets and what they show
Hall live	Supervisor, setters	State card per machine with mould number; stat of shots this hour; alarm table of open process alarms by machine; activity feed of state changes. On a hall screen through a read-only display link.
Machine process	Setters, process technicians	Line charts of cushion, peak injection pressure and cycle time over the last 500 shots, with the current mould's limits in an attributes card beside them; rolling standard deviation of cycle time; gauges for each mould circuit temperature with flow; barrel zone temperatures as a line chart. Opened from the hall board by clicking a machine.
Material and utilities	Process engineer, maintenance	Dew point and drying temperature per dryer; cooling water supply temperature; histogram of cycle time per mould; table of dryers out of limit in the last 24 hours.
Quality history	Quality engineers	Table of per-shot values for a chosen machine and time range, downloadable to Excel; scatter chart of cushion against cycle time; alarm table filtered to process window alarms with times.
Energy	Plant manager	Bar chart of daily kWh per machine; energy per shot by mould as a table.

Entity parameters let the machine process board serve every machine; the viewer picks the machine, or a click on the hall board sets it. Key filters can narrow the hall board to machines running a chosen mould or material.

SCADA screens

Two kinds of screen work well. A hall layout shows every machine where it stands; a cell screen shows one machine with its auxiliaries. The Plastics pack has injection moulding machines, mould temperature units, resin dryers, hopper loaders, gravimetric blenders, regrind granulators and hot-runner controllers; the Factory pack adds robot arms and conveyors, and the Building pack air-cooled chillers for the cooling water plant. See the SCADA library.

• Hall layout: each machine symbol coloured by state (green automatic, blue setup, red alarm, grey stopped), with its mould number and last cycle time as labels.
• Cell screen: machine, two temperature control units with supply and return temperatures on their hoses, hot-runner controller with its hottest and coldest zone, dryer and loader with dew point. A rule turns a hose label red when return minus supply exceeds the circuit's limit.
• Stale data: a rule greys any auxiliary whose plc_comm_ok is false.
• Reuse: draw the cell screen once per machine family and attach it as a view on the machine profile; every machine of the family gets its own view, and Match everything proposes bindings for the auxiliaries by name. Clear the Problems tab before applying.

Rules and alarms

Process window alarms live on the machine profile and read the mould limits from attributes; auxiliary alarms live on the auxiliary profiles.

Alarm	Condition	Delay and hysteresis	Severity	Notified	Escalation
Cushion out of window	cushion_mm < attribute cushion_min_mm or > cushion_max_mm	Debounce 3 shots	Major	Row setter, push	Reminder every 15 min
Cycle time long	cycle_s > attribute cycle_max_s	Debounce 5 shots; hysteresis 0.5 s	Warning	Row setter, in-app	None
Cycle unstable	cycle_s_std > attribute cycle_std_max_s	Debounce 10 shots	Minor	Process technician, push	None
Mould circuit hot	tcu1_return_c > attribute mould_temp_max_c	Debounce 120 s; hysteresis 1 K	Major	Row setter, push	Reminder every 15 min
Mould circuit flow low	tcu1_flow_lmin < attribute flow_min_lmin	Debounce 60 s	Major	Row setter and maintenance, push	Reminder every 30 min
Barrel zone deviation	barrel_z1_c < attribute barrel_z1_min_c or > barrel_z1_max_c, one pair per zone in the same rule (up to 16 comparisons)	Debounce 180 s; hysteresis 2 K	Major	Process technician, push	Reminder every 30 min
Hot-runner zone fault	hr_fault != 0	Debounce 2 samples	Major	Row setter, push	Reminder every 15 min
Wet material	dryer_dewpoint_c > attribute dewpoint_max_c	Debounce 600 s; hysteresis 3 K	Major	Process technician and quality, push and e-mail	Reminder every 30 min
Machine stopped in production hours	Rules card: state not 1 and stays so for 15 min, only in shift hours		Minor	Supervisor, in-app	Rules card raises Major after 45 min
Gateway offline	Inactivity, 5 min		Major	Automation engineer, push	Reminder every 30 min

Process alarms evaluate on each shot, so debounce by shot count rather than by seconds. An alarm opens after its debounce, is acknowledged by the setter who takes it and resolves when the process is back in its window; one alarm per machine and type counts repeats, so a cushion that leaves its window forty times in a shift shows as one alarm with forty occurrences. A higher severity reached by an open alarm notifies again. Write the first checks into each rule's runbook text, for example "check the TCU filter and the hose quick-couplings before changing setpoints".

During a mould change, setup or trial, the process is outside its window by design. Switch the machine's place to Manual on its Rules card for the setup time, which stops time-based rules, and mute the machine for the change window; muted alarms are recorded but page nobody, and the mute ends by itself (at most 90 days). Avoid floods by sending process alarms to the setter of the row, not to the hall; above 20 notifications a minute Hermetry sends a five-minute digest.

Automation and integrations

• Process window from the MES: at each mould change the MES writes the new limits as server attributes on the machine through the REST API, and the alarm rules follow without anyone editing them.
• Quality records: the QMS or MES pulls per-shot values per machine through the REST API with a read-only API key, within the raw window of at most 90 days; quality engineers download shot tables to Excel for an investigation.
• Process alarms to quality: signed alarm.created and alarm.cleared webhooks for process window alarms let the QMS mark the affected time range for containment and review.
• Derived values: the root flow's Statistics block saves cycle_s_std; a Script can compute return-minus-supply ΔT per circuit and save it with Save telemetry, so a single rule watches it.
• Reporting: scheduled, e-mailed reports are not built in today; use tables, Excel and the API.

This design sends no commands to machines or auxiliaries; setpoints stay with the setter at the machine. Sorting and reject gates remain in the machine and robot logic, which can act on a shot within its own cycle time.

Roles and access

Who	Role or group	Can
Setter	User, policy limited to their row	Machine and cell boards, acknowledge process alarms, switch a machine to Manual for setup.
Process technician	User + IoT operator group	All machines, acknowledge, mute a machine during trials.
Quality engineer	User + IoT viewer group	Read every board and the shot history, download tables.
Hall supervisor	User	Hall board, all alarms of the hall.
Automation or process engineering lead	Admin	Profiles, alarm rules, flows, drawings, gateway settings and firmware.
IT owner	Owner	Members, sign-in policy, API keys for MES and QMS.

The audit log records who changed alarm rules, profiles and setting baselines, who muted which machine and when, and who switched a place to Manual, in a hash chain that can be verified and exported. For a customer audit, that record shows the monitoring itself was not altered around a disputed period.

Sizing and data

As an order of magnitude, for 30 machines with an average cycle of 40 seconds running three shifts, two TCUs, a dryer and a hot-runner controller each:

• Per-shot messages: 86,400 ÷ 40 = 2,160 per machine per day, about 65,000 a day.
• Periodic machine messages every 60 s: 30 × 1,440 = 43,200 a day.
• Auxiliaries every 60 s: 30 × 4 × 1,440 = 172,800 a day.
• About 280,000 messages a day, roughly 8.5 million a month. Telemetry and attribute messages count toward the allowance; alarms do not.
• Values: about 650,000 per-shot, 600,000 periodic machine and 1 million auxiliary values a day, around 2.3 million a day and 800 million a year.

Auxiliaries dominate. Reading them every two minutes halves their share, and a deadband on temperatures cuts it further, without losing anything a setter would act on. Per-shot data is the quality record, and as raw readings it is kept for at most 90 days (in the 365- and 1,095-day classes); hourly and daily aggregates last the full class, up to three years. Export per-shot data per order or lot through the API or to Excel within the raw window and keep it in your QMS or own store for as long as your customers require. Storage per value depends on the deployment; ask us to confirm the estimate. Per-gateway message rates stay modest: a row of ten machines sends about 15 per-shot and 50 periodic messages a minute, inside the default 300 a minute.

Rollout plan

Phase	Weeks	Work	Exit criteria
0. Survey	1-3	Data interfaces of every machine and auxiliary; missing sensors (flow per circuit, dew point); state model; first process windows for the pilot moulds.	Signed point list per machine family; sensor purchase list.
1. Pilot	4-9	Four to six machines with auxiliaries. Profiles, per-shot reading, flows, process and cell boards, cell drawing, alarm rules.	Per-shot values match the machine's own shot log; three weeks with alarms the setters accept as useful; at least one drift caught before inspection.
2. Hall	10-16	All machines of the hall, row by row, with baselines per row type; hall layout drawing.	Every machine reporting; process windows set for every running mould.
3. Integration	17-20	MES writes limits at mould change; QMS pulls shot data; webhooks for process alarms.	No manual limit edits in a normal week.
4. Further halls	From 21	Reuse profiles, drawings and rules.	Each hall live within three to four weeks of wiring.

Commissioning checklist

1. Gateway claimed and placed on its row; machine and auxiliaries appear as child devices with floor names and are placed under their machine.
2. One shot's values compared field by field with the machine's own cycle record: cushion, injection time, plasticising time, peak pressure, cycle time.
3. Shot counter increase over an hour equals the machine's own count; no shot is missing or doubled.
4. TCU temperatures and flows compared with the unit displays; a hose temporarily throttled shows falling flow.
5. Dew point compared with the dryer display; units and sign checked.
6. Process window attributes written for the running mould; a deliberate small cushion change during a trial opens and clears the alarm.
7. Mould change procedure rehearsed: Manual on, mute, new limits written, Manual off.
8. Gateway uplink unplugged for 15 minutes in production; afterwards the shot history is complete and backfilled alarms are marked.
9. Cell drawing bound with nothing in the Problems tab; runbook texts written.

Pitfalls

• Polling per-shot values on a timer. A 10-second poll on a 7-second cycle drops shots. Read on the cycle signal or on counter change.
• One window for all moulds. Limits that are not rewritten at mould change produce false alarms or blind spots. Make the limit update part of the setup procedure.
• Alarms during setup. Without Manual and mute, setters learn to ignore the app on day two.
• Mould temperature without flow. Temperature reacts late; flow per circuit shows a blocked channel first.
• Hydraulic and specific pressure mixed. Compare like with like across machines, and label which one each key carries.
• Treating the dashboard as the release decision. Part release follows your quality plan; Hermetry data supports it.

What Hermetry does not replace

Hermetry watches the process and records it; it does not run it. Machine control, setpoints, shot-by-shot reject decisions, temperature control and every safety function (guards, light curtains, emergency stops, heater protection) stay in the machine, its auxiliaries and their safety circuits, and must work when the link or the platform is unavailable. Part release, traceability records required by your customers and the quality management system stay where your quality plan keeps them.