Hopper cleared before start-up

The problem

Batch traceability rests on a box ticked without evidence.

Batch traceability depends on there being no physical carryover between batches of different origin. Today the responsibility rests on a human 'hopper empty' sign-off with no objective visual evidence, and leftover material from the previous batch can contaminate the start of the new one.

• Switching from clean concrete to mixed C&D waste, or back, means clearing the hopper so the new batch starts clean. In a plant with several origins, that can happen many times a week.
• Today that rests on a human 'hopper empty' sign-off with no objective visual evidence behind it, often under pressure to restart because trucks are waiting and the crusher is the bottleneck.
• Leftover material from the previous batch contaminates the start of the next one: clean concrete aggregate inherits plaster, or a mixed batch inherits an origin it should not carry.
• When the batch is questioned, the investigation finds a signature and no proof — and the management certificate issued on that batch is exposed. It is not bad faith: a signature is not a measurement.

How it fits the IRIS system

Edge with JIDOKA AI — the plant does not start until the cameras confirm and the supervisor signs.

Edge plus JIDOKA AI: industrial cameras with CNNs trained to recognize 'empty hopper vs. leftover material' compare the current state against the reference state. The plant doesn't start until the cameras confirm OK and the supervisor signs off on that evidence.

The camera does not decide alone and does not start anything on its own: it compares the hopper with its reference empty state and the supervisor signs on that evidence. What changes is not who decides, but what the decision rests on. A signature backed by an image holds up in an inspection; a tick does not.

See the full IRIS architecture →

Before and after

The hopper sign-off by eye versus by camera

Aspect	Today	With iLEAN Edge
What 'hopper empty' rests on	A ticked box	Image compared with the reference state
Carryover between origins	Found later, in the batch	Detected before start-up
Start of the new batch	Under suspicion	Released on evidence
Start-up lockout	None	Until OK and signature
Proof for the certificate	A signature	Before and after images, stored
More changeovers per week	More exposure	Same verification each time

Blind sign-off with carryover risk between batches → evidence-based sign-off at each critical point.

Impact estimate

Impact estimate — it scales with your weekly changeovers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• We do not fix a payback range here: the return scales with the number of waste-type changeovers per week, and that varies too much between plants to give a single figure.
• What it protects is per-batch traceability and, with it, every management certificate issued on those batches and every client who relies on them.
• From a blind sign-off with carryover risk to an evidence-based sign-off at each critical point, stored with the batch it released.
• With your weekly changeovers and the cost of one questioned batch, the return is estimated with you before deploying, not promised in advance.

Protects per-batch traceability (and thus the issued management certificate), payback scaling with weekly waste-type changeovers. *Estimate to validate*.

And the fair question from the production manager

“What if the camera locks the plant over a few stones?” — the reference state is defined with you: what counts as empty is a tolerance agreed at commissioning, not the model's opinion. Comparing a hopper against its reference image is an anchored task where the best models drop below 1.5% error [1], and in case of doubt the system escalates to the supervisor, who signs on the image instead of being blocked silently. It stops on a verifiable difference, not on a hunch.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.