Consider what a nervous system does. It carries commands outward, from the brain to the muscles. It also carries information inward, from the skin, the joints and the organs back to the brain. Anatomists call the outward pathways efferent and the inward pathways afferent. A body needs both. Without the first it cannot act. Without the second it cannot know what its actions are doing to it.
Industrial systems were built with a strong version of the first and a weak version of the second. That imbalance explains a great deal about why machines still fail without warning, and it points to what the next layer of industrial technology has to supply.
Machines that act without feeling
Control technology is a mature field. A modern plant, building or substation is directed by controllers that start and stop equipment, hold setpoints and respond to commands with great precision. Industry even has a standard vocabulary for the layers involved: the ISA-95 model describes how sensing and control at the machine connect upward to supervisory systems, operations management and business planning. This is the efferent half of the picture, and it works.
The inward half is thin by comparison. Controllers measure what they need in order to control: a temperature, a pressure, a flow, a speed. They measure whether the process is on target. They seldom measure how the machine doing the work is holding up. A pump controller knows the discharge pressure. It does not know that the pump's bearing has begun to shed metal into its grease.
The result is a system that can act and cannot feel. A person with no sensation in a hand can still grip, and will not notice when the grip is causing injury. Equipment behaves the same way. It keeps running, and keeps meeting its setpoint, while a bearing overheats or insulation breaks down. Often the first signal that reaches an operator is the failure itself.
Receptors, nerves and a brain
The analogy is useful because it can be made precise. A body senses through receptors specialized for different stimuli: mechanoreceptors for pressure and vibration, thermoreceptors for heat, nociceptors for damage. No single receptor type tells the whole story. The brain builds its sense of the body by combining them.
For a machine, the receptors are sensors mounted on the asset: accelerometers for vibration, microphones and ultrasonic sensors for sound, thermal sensors for heat, current and voltage measurements for electrical behavior. Networks that carry what those sensors find play the part of nerves. The brain is the layer of interpretation that turns signals into a judgment about health, which is what we call machine health intelligence.
It also explains something less obvious. Bodies do not send every raw impulse to the brain. A great deal of processing happens close to the receptors and in the spinal cord, which is why you pull your hand from a hot stove before you consciously register the heat. Machines need the same arrangement. Much of the interpretation has to happen on the device, near the signal, because the raw data is too heavy to ship and some responses cannot wait for a round trip to a data center.
Why sensation was left out
Leaving it out was a reasonable choice for a long time. Sensing the condition of a machine required expensive instruments, and interpreting the result required a specialist. It made sense to apply both only where the stakes were highest, such as large turbines and critical compressors. Everything else was maintained by schedule and by the judgment of experienced people who knew what healthy equipment sounded like.
That judgment is under pressure. Much of the infrastructure people rely on is old, and its owners are stretched. The American Society of Civil Engineers gave the United States an overall grade of C in its 2025 Infrastructure Report Card, the best result in the history of the report and still a mark that describes a system in need of attention. Assets are aging faster than inspection budgets grow.
Meanwhile the economics of sensing have changed. Sensors that capture vibration, sound and temperature are small and inexpensive. Processors that can interpret those signals fit inside a device the size of a hand. The specialist's pattern recognition can, at least in part, be learned by a model and applied to thousands of machines at once.
We've spent decades teaching machines to follow instructions. Now we have a chance to help them communicate their own condition. That's what giving machines a nervous system means to us.
What sensation makes possible
A body that can feel does more than avoid injury. It adjusts. It favors a strained muscle, rests when it is tired and learns its limits over time. Sensation is what makes skilled action possible.
The same would be true of infrastructure. A system that knows the condition of its components can be maintained on evidence instead of on a calendar. It can be operated according to the real state of its equipment, running harder when there is margin and backing off when there is not. It can report honestly on its own resilience, and its operators can plan with knowledge they do not have today.
This is why we describe AFFERENT's work as building a nervous system and not simply as building sensors. A sensor produces a measurement. A nervous system produces awareness, by carrying many signals to a place where they can be understood together, and by turning that understanding into a response that reaches the right person in time.
Signals that are graded, not binary
One more feature of biological sensing is worth borrowing. Pain is not a switch. A mild ache, a sharp twinge and an injury that stops you in your tracks are different signals, and the body treats them differently. Most of the time the nervous system stays quiet about routine sensations altogether. You do not notice your shoes until something is wrong with them.
Industrial monitoring has often worked the other way, with fixed alarm limits that are either off or on. A limit set low enough to catch a fault early fires constantly on harmless variation, and a limit set high enough to stay quiet fires late. A better design grades its signals. It learns what ordinary looks like for each machine, stays silent while behavior stays within that range, and reports a change with an estimate of how far it has moved and how confident the system is. Severity should rise as the evidence does, so a planner can tell the difference between "watch this" and "act this week".
Receptors that adapt
One property of biological receptors is especially instructive for machine sensing. Many of them adapt. Put on a watch and within minutes you stop feeling it; step into a warm room and the warmth fades from attention. Receptors respond strongly to change and weakly to steady conditions. That is not a flaw. It keeps the nervous system from drowning in information that does not require a response, while leaving it sharply alert to anything new.
Machine monitoring needs the same property. A pump that always vibrates at a certain level because of how it was installed is not, by that fact alone, in trouble. A fixed alarm threshold set too low will fire constantly on that pump and be ignored; set too high, it will miss real deterioration on a quieter machine next to it. The better approach is to learn what normal looks like for each machine, under each operating condition, and pay attention to departures from that baseline.
In engineering terms, this means comparing each new measurement with a model of the machine's own history rather than a generic limit. A change of a few percent in a particular frequency band on a machine that has been steady for months can matter more than a large absolute value on a machine that has always run rough. Adaptation, in other words, is what lets a sensing system notice what deserves notice. It is one of the reasons the nervous system is a useful model for machine health rather than a loose metaphor.
Where the metaphor stops
Metaphors earn their keep only if their limits are clear. Machines do not feel anything, and we do not claim they will. A maintenance engineer does not need a pump to be aware of itself. They need it to tell them something useful: what changed, whether it matters, and what deserves attention. In engineering terms, "sensation" means continuous measurement of condition, and "awareness" means a reliable interpretation of that measurement.
There is a second limit. Biological nervous systems were tuned over a very long time. Industrial ones will be built one asset at a time, and they will have to earn trust by being right more often than they are wrong. False alarms are the industrial equivalent of chronic pain: a signal that no longer means anything.
A long project
No one should expect this to happen quickly. Physical infrastructure changes slowly and its operators are careful, as they should be. Sensation will arrive asset by asset, starting where failure is most expensive, and it will spread as it proves itself.
We think the direction is clear all the same. The physical world has had muscles and reflexes for a century. It is time it had senses. You can see how the parts fit together in our introduction to AFFERENT.
