Skip to content
IRIS NEURALContact

Solutions by typeProduction efficiency

Which machine is actually holding the whole floor up?

The queue grows here and the next machine sits idle

Ten past five in the machining shop. The curved conveyor feeding the machine is piled high with metal casings, stacked and crooked, one on top of another. The signal tower is green: the machine is running, so on paper everything is fine. Twenty metres further on, the conveyor that collects the output is empty end to end and an operator is standing there with nothing to do. Nobody has counted either of those two things.

4 cameras across 3 sectors

This is how you ask for it

«Count how many parts are waiting in front of this machine, and tell me if it goes over forty for more than ten minutes.»

The bottleneck is always found late, when half the floor is waiting and the shift is already lost. Before that there are only impressions: machining says it is drowning, assembly says it is short of material, and both are half right. The argument gets settled by seniority, not by data, because nobody has a figure for how much material is waiting on the floor.

You decide once

Here the organisation decides two things, and draws them: where the area is in which waiting work is counted, and which line the output crosses. Then it puts a number on it: forty parts, ten minutes, tell the shop supervisor. IRIS applies that pair of counts on every shift, hour by hour, and keeps the curve for the whole day. And if the flow changes tomorrow, you move the area and the count is still the same count.

The bottleneck is never where the shouting is. It is where something piles up without anyone measuring it. In machining it is casings stacked on a curved conveyor while the output conveyor runs empty. In bottling it is an overflowing accumulation table with hundreds of bottles locked against each other and two operators pushing them apart by hand. In a sorting hall it is a roller bend where parcels have ridden up over each other while the belt above keeps dropping boxes.

What they share is not the jam. It is that the production system cannot see it. That system knows how many parts each machine has declared; it does not know how many are on the floor, on a conveyor or inside a container waiting their turn.

Here IRIS does not alert: it counts. You draw the area where waiting work builds up and the line the output crosses, and you watch both numbers at once. When one climbs while the other falls, that is your bottleneck, without walking the shop and without fitting a sensor to any machine. And this is set up differently on purpose: asking is for finding, counting is for measuring, and a line cannot be described in a sentence. It has to be drawn, always in the same place, so that today’s number can be compared with yesterday’s.

What changes

Before, the supervisor walked the shop mid-afternoon and decided on what he saw in passing. Now he has two figures on the same screen: how many parts are waiting here and how many are coming out there, hour by hour, each count with its picture. He moves people and material before the next station runs dry, and at month end he can show which machine always piles up, instead of recalling it.

Where we have seen it

  • Afternoon shift in the machining shop. The curved conveyor feeding the machine is packed to the brim with metal casings, stacked and skewed, one on top of another. The signal tower is green: the machine is working, but it cannot keep up. A few metres further on, the roller conveyor that carries the output away is empty end to end. Two operators look at the queue. IRIS counts the parts waiting and the parts coming out.

    CAM-242A mountain of parts has built up in front of one machine

    «Count how many parts are waiting in front of this machine and warn me if it goes over forty for more than ten minutes.»

    See this cameraIndustry

  • Mid-afternoon. The filler’s accumulation table has overflowed: hundreds of juice bottles have packed together until they block each other, and several have fallen on their side. Two operators in white coats and hairnets are leaning over the table, pushing the bottles by hand to break the pile up. Behind them the overhead conveyor keeps bringing product that has nowhere to go. Other colleagues watch from the aisle, at a distance.

    CAM-201Bottles piling up on the accumulation table

    «Tell me when bottles pile up on the accumulation table and stop moving forward.»

    See this cameraIndustry

  • Mid-morning in the sorting hall. On the roller bend the parcels have ridden up on top of each other and stopped moving. An operator in a yellow hi-vis vest has leaned right across the conveyor and pushed both arms in among the rollers to clear it. The red beacon on the post is lit. The panel with the stop button is two steps away, on his right, untouched. The incline belt behind keeps dropping boxes on top. IRIS sees the pile-up, sees the arms inside and times both.

    CAM-275Jammed conveyor, two arms in among the rollers

    «Tell me if somebody puts their hands into the conveyor without stopping it first, and show me the video from when the jam started.»

    See this cameraIndustry

  • Early afternoon in machining. You can barely see the floor: stillages full of parts are stacked everywhere, in front of the machines and across the walkways. At the bench at the back, six people are reworking parts by hand, one after another. The outbound conveyor, meanwhile, runs almost empty. IRIS counts how many stillages are in the area, compares that with the count at the same hour yesterday, and times how long the pile has been growing.

    CAM-248The rework area has swallowed the aisle

    «Tell me if, mid-afternoon, there are more stillages than usual in the rework area.»

    See this cameraIndustry

What it does not do

IRIS counts what is visible from that framing. When the stacks hide what is behind them, one viewpoint is not enough: the same spot gets a second camera from another angle, and the two perspectives together give the count.

And counting parts or containers is an operational measure, not a certified figure. The only thing certified by the Spanish Metrology Centre is our people counting.

Questions

How does IRIS know where the bottleneck is?
By counting two things at once. It counts the parts waiting in front of the machine and the parts leaving on the next conveyor. When one figure climbs while the other falls, that is your bottleneck. No new sensor is needed on the machine and nothing has to be touched in the production system: the count is made on the image, by drawing the area being watched and the line things cross. And every count keeps its picture.
Is this set up by speaking, the way alerts are?
No, and that is the important difference. Asking is for finding; counting is for measuring, and they are not done the same way. A line cannot be described in a sentence: it has to be drawn on the image, always in the same place. That is why counting is set up by drawing a line or an area, and why the number is repeatable: today’s can be compared with last month’s because the line has not moved.
Does IRIS predict when the queue will build up?
No. IRIS does not predict: it counts what is happening while it happens. It sees the queue grow from twenty parts to fifty in twenty minutes and says so straight away, with the figure and the picture. That leaves room to act before the next station runs dry, but the forecast and the decision belong to a person. IRIS supplies the timed figure; the judgement comes from whoever knows the hall.
Does it replace the system that already tracks production?
No. The production system knows how many parts each machine has declared. What it does not know is how many are on the floor, on a conveyor or inside a container waiting their turn. IRIS counts what is visible and puts that figure next to the one that already exists. It is the material nobody has declared yet, and usually the very material that explains the delay. The two numbers together are worth far more than either on its own.

See all the questions

See it for yourself

This isn’t a video. It’s a slice of the real screen.

A summary of the real interface, played hands-free. You see one case from start to finish; what you do not see are the other screens, and there are many.

  1. Four cameras at once
  2. Each understands its own view
  3. The numbers climb by themselves
IRIS NEURAL
30ES
IRISDirectoGrabacionesGISIncidencias3996SituaciónCasosLPRAutomatizacionesIRIS DATA
Ask a question or make a request…Send
IRIS · Infinity Neural

Not seeing yours here?

These words are the ones we have built, not the ones IRIS understands. The list falls short on purpose: you ask for it in a sentence, and sentences do not run out. Tell us what is in front of your cameras and we will tell you whether it understands it, or not yet.

See all solutionsTell us what you see

And the best part

What you bought was a warehouse of images.What you needed was someone to look at them.

Those same cameras that only fill disks today can flag what matters while it is happening. The warehouse stays; it just stops being all you have.

Tell us your problem and we will say whether IRIS understands it, or not yet.

We reply the same working day.