Cameras in Injection-Moulding Units: Short Shots, Cycle Drift and Mould Changes
Your machine controller logs cycle time and alarms. It cannot log the human: how long a stopped machine stood before anyone walked over, how long a mould change really took, or how much scrap never reached the granulator. On a 20-machine Indian moulding floor, that blind spot models out at roughly ₹4,500 per shift.
You run a 200–1,000 worker plastics unit — Naroda, Vatva, Chakan, Hosur, Baddi, Rajkot. You already have 20 to 60 moulding machines, each with a controller that prints cycle counts. This article is about the gap between what those controllers print and what actually happens on the floor at 2 a.m.
Key points
- The controller logs the machine; the camera logs the response. Every controller records "machine stopped at 02:14." Only a camera answers "and the operator reached it at 02:51." That single number — response time — is the cheapest efficiency metric in an Indian moulding shop, and almost nobody measures it.
- Cycle drift is invisible because it is small. A 30-second cycle drifting to 32 seconds costs you 60 shots per machine per 8-hour shift. Nobody notices 2 seconds. A 2026 case study tracking one hydraulic injection-moulding machine over 14 months found that cycle times running above the 45-second target — alongside unplanned downtime and minor stoppages — were the main drag on OEE, and that correcting them moved OEE close to world-class (JMMP, 2026).
- Mould changeover is where the hours hide. SMED is well documented in plastics processing — including a published case study on mould changes in PVC pipe production (Springer, 2026). But you cannot run SMED on a changeover nobody timed, and in most Indian moulding units nobody times it door-to-door.
- Material is the biggest rupee leak. Resin is the dominant variable cost in commodity moulding, so a 1% scrap improvement on 3 tonnes a shift is real money, every shift. Use your own landed ₹/kg — grade, filler and MOQ move it more than any published index.
- Energy is the second, and idling is the villain. In a typical injection-moulding plant, polymer processing is about 66% of energy use, chillers 11% and compressed air 10%; on standard hydraulic machines the base load — energy burned while the machine is idling, including cooling — can be 75% or more of total consumption. Plant energy spend often approaches the cost of direct labour (AMI / Tangram Technology energy-saving guide for injection moulding machines). A barrel sitting at temperature with nobody moulding is close to full cost for zero output. In Gujarat, GERC held FY 2027 tariffs at the previous year's levels (Mercom India, April 2026) — so use the ₹/unit on your own bill, including fuel-price adjustment and duty.
- The law you comply with changed. The Factories Act, 1948 no longer governs your floor. The Occupational Safety, Health and Working Conditions Code, 2020 (Act 37 of 2020) subsumes it along with 12 other central labour statutes and came into force on 21 November 2025 (India Code). The OSH (Central) Rules, 2026 were notified and took effect on 8 May 2026 (SCC Online).
- Almost no Indian moulding unit measures operator response time. The sector is overwhelmingly MSME — around 90% of the membership of the industry's apex body, AIPMA, is from the MSME sector (AIPMA). That is your advantage, not a vendor's.
What the controller sees and what a camera sees
| Event on the floor | Machine controller | Camera + AI |
|---|---|---|
| Cycle time creeping 30s → 32s | Logs it (if anyone opens the log) | Flags the drift with the shift and operator on duty |
| Machine alarms at 02:14 | Logs alarm start and end | Logs when a human actually arrived — the response gap |
| Short shots piling in the reject bin | Blind | Sees the bin filling, sees parts thrown under the machine |
| Mould change | Sees "stop" then "start" | Sees crane wait, hose fitting, water leak, trolley missing, tool hunt |
| Purge material dumped on the floor | Blind | Sees it, and sees whether it reached the granulator |
| Operator running four machines instead of two | Blind | Sees the walking pattern |
| Regrind bin overflowing | Blind | Sees it |
| Toolroom fitter absent during changeover | Blind | Sees it |
This is the honest case for cameras in a moulding unit. Not security. The controller is already a good sensor for the machine. It is a zero sensor for everything human, and in a moulding shop the human is the bottleneck between a stop and a restart.
The rupee model: one shift, 20 machines
Assume 20 machines averaging 250 tonne clamp, 8-hour shifts, PP and PP-filled parts.
For machine-hour rate this model uses ₹375/hour — a working assumption, not an industry benchmark. Indian commercial moulders often price off a thumb rule in the range of ₹10–15 per tonne of clamp per 8-hour shift, which puts a 250T machine near this figure. Substitute your own costed rate before you use any number below.
| Leak | Assumption | Loss per shift |
|---|---|---|
| Cycle drift | 2s on a 30s cycle, 8 of 20 machines | ~480 shots ≈ 256 machine-min ≈ ₹1,600 |
| Unattended alarms | 3 stops/shift, 25 min each before anyone reacts | 75 machine-min ≈ ₹470 |
| Slow mould changes | 4 changes/shift, 45 min avoidable each | 180 machine-min ≈ ₹1,125 |
| Scrap and lost regrind | 3 t/shift throughput, 2% → 1% scrap, ~₹45/kg net after regrind credit | 30 kg ≈ ₹1,350 |
| Total | ≈ ₹4,500 per shift |
Two shifts, 26 days: roughly ₹2.3 lakh a month on a floor that looks like it is running fine. You will not capture all of it. Capturing a third of it still clears the cost of a camera-and-analytics layer on most 20-machine floors — see what a factory camera AI system costs in India and does camera AI pay for itself for the other side of that arithmetic.
Change the four assumptions to your own numbers before you believe this table. The point is the structure, not our figures.
The night-shift problem, specifically
Between 11 p.m. and 6 a.m. one operator often covers six to ten machines, the supervisor is in the cabin, and a machine that alarms on short shot or ejector fault can stand for 40 minutes. Nothing in your MIS shows this the next morning except a slightly low shot count that gets explained away.
A camera watching a bank of machines, plus a simple rule — "machine idle, no human within frame for N minutes" — turns that into a WhatsApp message to the supervisor at 02:16 instead of a mystery at 08:00. This is the single highest-value trigger in a moulding unit and the easiest to justify to a night-shift team, because it makes their shift less painful, not more policed.
If women work your third shift, the position has changed. Section 43 of the OSH Code, 2020 ("Employment of women") permits women to work in all establishments and at night, subject to consent and prescribed conditions, replacing the old blanket restriction. Under the OSH (Central) Rules, 2026, where a woman is employed before 6 a.m. or beyond 7 p.m. the employer must obtain her written consent and ensure the workplace is well lit including entry, exit and pathways, has toilets, washrooms and drinking water nearby, provides safe pick-up and drop, and is equipped with CCTV surveillance (SCC Online summary of the Central Rules).
Read that scope carefully: the Central Rules bind establishments in the central sphere, and state governments frame their own rules under the same Code. Your actual obligation is set by your state's rules and your factory licence conditions — see which Indian states require CCTV for women on night shift and the retention checklist.
Where the cameras go in a moulding shop
Five positions cover most of the value:
- Down the machine aisle, so each camera sees 4–6 operator stations and the space between them. This is the response-time camera.
- At the mould-change bay / crane path. Changeover analysis needs the whole choreography, not the machine.
- Under-machine and reject-bin view for the scrap leak, at a low angle. Scrap piling on the floor is visible; scrap in a bin is not.
- Granulator and regrind area. Purge that never reaches the granulator is pure material loss.
- Finished-goods and mould store exit. The theft path in a plastics unit is usually granules, copper cooling fittings and mould inserts, not finished parts.
General placement logic is in where to place cameras on a shop floor and sizing in how many cameras for a factory floor.
If you already have DVR cameras covering aisles, do not rip them out. Analytics can often run on the existing feed — running AI analytics on existing CCTV.
What cameras will not do in a moulding unit
Be clear about this before you spend.
- A camera cannot see a short shot reliably at speed. At a 20–30 second cycle with parts dropping into a bin, visual defect detection on flash, silver streaks, sink marks or dimensional drift is a machine-vision problem — dedicated lighting, fixed part presentation, sometimes a robot — not a ceiling CCTV problem. Anyone who tells you a shop-floor dome will grade your parts is selling.
- A camera cannot diagnose why the cycle drifted. Barrel temperature, chiller performance, hydraulic oil temperature, resin moisture and mould cooling are the actual causes. The camera tells you the drift is real and who was on shift. Your process engineer tells you why.
- A camera does not replace machine guarding. Interlocked guards, light curtains and two-hand controls — the subject of ISO 20430:2020, the safety-requirements standard for injection moulding machines — are the safety system. A camera is evidence and audit, not a protective device, and no clause of that standard asks for one. Related: machine guarding and danger-zone detection.
- A camera does not fix a changeover that is slow for structural reasons. If the mould arrives without a preheated water manifold and the crane is shared across the bay, video will only show you the same 90 minutes with better clarity. Video is the input to SMED, not SMED.
- It will create a data obligation. Worker footage is personal data. The DPDP Rules, 2025 were notified on 14 November 2025, with an eighteen-month phased compliance period — putting the substantive obligations (notice, security safeguards, retention limits, breach reporting) in mid-2027 (MeitY / PIB). Retention policy is cheap to set now and expensive to retrofit — see DPDP and worker CCTV.
- Cameras used only for blame will be defeated. Lenses get taped, angles get nudged, the granulator gets moved. If the first output of the system is a punishment, you have bought an expensive way to teach your night shift about blind spots.
A sensible first phase
Start with one machine bank and the mould-change bay — six to eight cameras — and measure exactly two things for 30 days: response time to a stop, and actual changeover duration door-to-door. Do not add PPE, attendance or theft rules in month one. Both metrics are new information for almost every Indian moulding unit, both convert directly to rupees, and both are defensible in front of your shift in-charge.
If those two numbers move, expand. If they do not, you have spent a small amount to learn something true about your floor. Budget the switch and power correctly at the start — PoE budgeting for factory cameras — and pick storage before you pick brands: NVR vs cloud VMS in India.
FAQ
Can a CCTV camera detect short shots in injection moulding? Not reliably from a ceiling position at production speed. Short-shot detection needs controlled lighting and consistent part presentation — a machine-vision station, not shop-floor CCTV. What a floor camera does catch is the consequence: rejects accumulating, parts dumped under the machine, and the operator running the machine while knowingly producing scrap.
How much can an Indian moulding unit realistically save with camera analytics? On a 20-machine floor, the modelled leak from cycle drift, unattended alarms, slow changeovers and scrap is roughly ₹4,500 per shift. Recovering a third of it is about ₹75,000 a month across two shifts. This is a model, not a measured result — run it with your own machine-hour rate, resin cost and changeover count before committing to any number.
Do I still comply with the Factories Act, 1948 for my moulding plant? No. The Factories Act, 1948 was subsumed into the Occupational Safety, Health and Working Conditions Code, 2020, in force from 21 November 2025, with the OSH (Central) Rules, 2026 effective 8 May 2026. State rules under the Code are still being aligned. Our compliance playbook for 2026 tracks what actually applies.
Is CCTV mandatory in my moulding plant? Not as a general rule. There is no blanket law requiring cameras in Indian factories. The specific hook is night work by women: the OSH (Central) Rules, 2026 list CCTV surveillance among the conditions for employing women before 6 a.m. or after 7 p.m., and several states impose similar conditions. Customer and tender terms are a separate matter, and are contract, not law.
What is the single most useful camera alert in a moulding shop? "Machine stopped, no operator present for N minutes" on the night shift. It uses the machine's own stop signal plus one visual check, needs no defect model, and directly attacks the loss that no controller log can quantify — human response time.
Do I need new cameras or can I use my existing DVR system? Often the existing aisle cameras are adequate for response-time and changeover measurement, which is where the money is. Defect-adjacent and low-light night views usually need replacement. Assess before you buy — retrofit AI on existing CCTV.