During the height of the , a mathematician named Abraham Wald sat in a room filled with maps and casualty reports, looking at the bullet holes in returning B-17 bombers. The military brass wanted to add armor to the places where the planes were being hit most frequently-the wings, the tail, and the fuselage. It was a logical, data-driven conclusion. If the metal was shredded there, that was where the protection was needed.
The visible “hits” on returning bombers suggested armoring the gray areas. Wald realized the red area (Engine) was the fatal void-planes hit there never returned to be measured.
Wald, however, disagreed with such startling intensity that the room went quiet. He pointed out that they were only looking at the planes that had successfully made it back to the airfield. The bullet holes they saw represented the damage a plane could sustain and still keep flying. The real data, the fatal data, was in the holes they couldn’t see-the ones on the planes currently sitting at the bottom of the English Channel. By armoring the “hit” areas, they were ignoring the engines and the stickpit, the very places where a single bullet meant a total loss.
This is the fundamental trap of all measurement systems. We assume that because we are looking at a number, we are looking at the truth. In reality, we are looking at a selection of truths that someone else decided were worth capturing.
The Cartography of the Invisible
The problem with modern industrial interfaces isn’t that they are confusing; it’s that they are too clear. They present a curated reality that feels complete. When an engineer builds a Human-Machine Interface (HMI) for a complex material handling system, they are performing an act of translation.
They take the messy, vibrating, chaotic reality of a factory floor-where air pressure fluctuates with the humidity and pellets of resin have slightly different friction coefficients depending on the batch-and they turn it into a series of green bars and digital readouts.
The Act of Ignoring
Information is the act of ignoring. To create a signal, you must first define what is noise. The designer of the screen decides that the flow rate, the temperature, and the motor load are the “signals.”
Everything else-the subtle change in the pitch of a vacuum pump, the way the floor vibrates when a filter is slightly clogged, the smell of ozone near a control cabinet-is officially designated as noise. Because it is noise, it is not measured. Because it is not measured, it does not exist in the system.
The Handover on Screen Eleven
Last Thursday, during a handover session at a plastics compounding plant, a controls engineer was walking an operator named Della through a new automated system. He was proud. He had spent months mapping every sensor to a clean, high-resolution display. There were eleven screens in total, ranging from a high-level overview of the silos to the granular details of the micro-component dosing.
“If the flow drops below the set point, the alarm turns red here. We’ve got full visibility on the entire pneumatic conveying chain.”
– The Engineer
Della, who had spent in that building and could tell which extruder was running just by the way the air felt against her skin, watched the screen for a long time. She didn’t look at the graphs. She looked at the pipe behind the engineer.
“Where’s the one that shows me it’s starting to surge?” she asked.
The engineer paused. “Surge is inferred from the flow trend and the blower pressure. If those lines diverge, you know you have a blockage forming.”
Della looked back at the trend line. It was perfectly flat, a beautiful digital lie. She knew the pipe was surging because the rhythmic “thrum-thrum” of the material hitting the elbows had changed its tempo. It was subtle, like a drummer missing a sixteenth note, but to her, it was a scream.
The numbers hadn’t moved yet. The sensors hadn’t “caught up” to the physics. She didn’t say anything else for the rest of the session. She realized that the system was designed to tell her what had happened, while she was interested in what was happening.
The Fountain Pen as a Pressure Valve
I think about this selection bias often when I speak with Mason A.-M., a specialist in the restoration of vintage fountain pens. To most people, a fountain pen is a simple tool: a stick of ink that leaks onto paper. To Mason, it is a sophisticated hydraulic system operating at the edge of failure.
He spends his days looking through a loupe at nibs that are eighty years old, adjusting the “feed”-the small, finned piece of ebonite that sits under the gold point.
“The user thinks they want a smooth write, but what they actually need is a predictable resistance.”
– Mason A.-M., restorative specialist regarding a Parker Vacumatic
A fountain pen is a system of controlled leaks. The feed regulates the exchange of air and ink. As ink leaves the reservoir, air must travel up the feed to take its place. If that exchange isn’t perfectly balanced, the pen either starves and stops writing, or it “burps” a giant blob of ink onto the page.
There are no sensors in a fountain pen. There is no HMI. There is only the physical relationship between surface tension, gravity, and atmospheric pressure. When Mason repairs a pen, he isn’t looking for a “green light.” He is feeling for the way the ink “breathes.”
The Geometry of the Silent Factory
In the world of industrial material handling, the stakes of what we choose to ignore are significantly higher than a blot of ink. When you are moving powders, granules, or flakes across a plant that spans several thousand square meters, the material behaves in ways that defy simple logic.
It bridges in silos. It clogs in long-distance pneumatic lines. It drifts in weight accuracy if the dosing screw isn’t perfectly calibrated to the specific bulk density of that morning’s shipment.
Industrial Extensibility
This is why the engineering philosophy at
Zhangjiagang Yifan Machinery Co., Ltd.
focuses on control platforms built on PLC and HMI systems that plant personnel already know, such as Siemens or Mitsubishi.
The “smart” factory promised us that sensors would solve this. But a sensor is just a witness, and witnesses are notoriously unreliable when they are only asked “yes” or “no” questions. If a system is built on a rigid platform that doesn’t allow for the operator’s sensory input, it creates a dangerous gap.
It isn’t just about brand recognition; it’s about extensibility. If an operator like Della knows that a specific sound precedes a surge, the system needs to be flexible enough to allow that intuition to eventually be quantified. You can’t automate what you haven’t first observed with your own ears.
The 0.1% Paradox
We are obsessed with accuracy. We talk about batch accuracy of plus or minus 0.1 percent as the gold standard. And it is. In a world where recipe drift can lead to thousands of dollars in off-spec scrap, that precision is vital.
But there is a paradox here: the more precise the measurement, the more we tend to trust the screen over the machine. Yesterday, I spent in the dentist’s chair. My dentist has a high-definition monitor that shows a 3D scan of my molars.
He was pointing at a pixelated grey area, explaining that the structural integrity looked “within parameters.” I tried to tell him that every time I chewed on the left side, I felt a sharp, metallic zip of lightning. I told him my tooth disagreed with his pixels.
He was looking at the “measured” data. I was experiencing the “unmeasured” reality. We were both right, but only one of us was in pain. Eventually, he sighed, took a manual explorer tool-a simple metal hook-and poked the spot I mentioned.
The Human as the Sensor of the Gaps
Every measurement system is a decision about what will be ignored, and the ignored part does not stop influencing outcomes. When we design centralized feeding systems that move material across sixty countries and hundreds of different production environments, we have to account for the “noise.”
A sealed, automated network is a triumph of engineering because it eliminates dust, reduces manual labor by sixty percent, and protects the environment. But the “sealed” nature of it also removes the operator from the process.
The Risk of Dependency
They can no longer see the material moving. They can no longer smell the dust. They are entirely dependent on the HMI.
The Loss of Agency
They stop listening to the pipes because the screen tells them everything is fine. They have lost the “ear” for the machine.
If that HMI is a “closed” loop that only reports what the sensors are programmed to see, the operator becomes a prisoner of the green light. They lose their agency. They stop listening to the pipes because the screen tells them everything is fine.
Then, when the system eventually fails-because all physical systems eventually fail-they are caught off guard. They have lost the “ear” for the machine.
We need to stop judging interfaces solely on their clarity. We should start judging them on their humility. A humble interface acknowledges that it is only showing a fraction of the truth.
It leaves room for the “Dellas” of the world to say, “I don’t care what the flow trend says; the suction scale is breathing wrong.” When we build these massive, 20-year-lifespan systems, we aren’t just installing hardware. We are installing a worldview.
If that worldview doesn’t have a way to translate the operator’s “pitch change” into a system update, then we haven’t built a smart factory. We’ve just built a very expensive way to be surprised by the obvious.
Measurement is a tool for understanding, but we must never mistake the instrumented world for the world itself. The most important signal in the factory might be the one that hasn’t been given a name yet.