“The best part is no part.”
It is one of Elon Musk’s better-known engineering principles, and it sounds almost too simple to deserve serious attention.
It does.
Because hidden inside that short sentence is one of the oldest lessons in engineering.
Every additional part you introduce into a system is another opportunity for something to go wrong.
It may be a tiny gasket.
A valve.
A sensor.
A cable.
A software module.
A database.
An algorithm.
Or an entire processing train.
The size of the component is almost irrelevant.
Every new piece adds complexity.
Every new connection creates another potential failure point.
Every dependency increases the chances that one seemingly insignificant problem will eventually cascade into something much larger.
Complexity has a price.
People often underestimate just how astonishingly complex the oil and gas industry really is.
Take upstream production.
The public tends to imagine little more than a pipe sticking into the ground with oil conveniently flowing out.
Reality could hardly be more different.
Deep underground, hydrocarbons emerge mixed with water, gas, sand and a host of substances that engineers would much rather not have to deal with. They arrive under enormous pressures and temperatures, often in chemically aggressive environments, and must be transformed into stable, transportable streams while maintaining safety, protecting equipment and ensuring continuous operation.
Every stage introduces another layer of engineering.
Another control system.
Another valve.
Another instrument.
Another opportunity for failure.
Then the hydrocarbons arrive at a refinery.
If upstream facilities are sophisticated, refineries are monuments to industrial complexity.
They take a dark, chemically diverse mixture that nature provides and separate, convert, crack, reform, treat and blend it into dozens of highly specialised products that modern society depends upon every single day.
That transformation requires an astonishing amount of equipment.
Pumps.
Compressors.
Heat exchangers.
Reactors.
Distillation columns.
Pipes.
Valves.
Control systems.
Electrical systems.
Safety systems.
The list is almost endless.
Collectively they amount to millions of individual components.
Every single one has a finite lifespan.
Every single one can fail.
Of course, we do not simply hope for the best.
We monitor.
We measure.
We inspect.
Sensors are installed everywhere.
Pressure.
Temperature.
Flow.
Vibration.
Composition.
Position.
Thousands upon thousands of instruments constantly report the condition of the plant.
Yet every sensor is itself another component.
Another device requiring calibration.
Another cable.
Another software interface.
Another possible source of incorrect information.
Then all that information must be collected.
Stored.
Validated.
Transferred.
Processed.
Displayed.
Archived.
Analysed.
Turned into something useful.
Entire digital ecosystems exist solely to tell engineers what the physical plant is doing.
And those digital ecosystems have their own hardware, software, communication networks and maintenance requirements.
They too become part of the complexity.
Maintenance has always attempted to manage this reality.
Calendar-based maintenance reflects decades—sometimes more than a century—of accumulated operational experience. Components are replaced because history suggests they are approaching the end of their reliable service life.
Usage-based maintenance attempts something more sophisticated by monitoring actual operating conditions and intervening when the data indicates deterioration rather than simply following the calendar.
Both approaches have their strengths.
Both have limitations.
Neither eliminates uncertainty.
Neither eliminates human judgement.
Neither eliminates the possibility of failure.
If anything, the more sophisticated our monitoring becomes, the greater the temptation to believe we can manage unlimited complexity simply by collecting more information.
I am not convinced that is true.
At some point, complexity itself becomes the problem.
Not because engineers become less capable.
But because every additional layer introduces fresh interactions, fresh dependencies and fresh opportunities for unexpected behaviour.
That is why I increasingly find myself drawn to simple decision-making governing complex systems.
Not simplistic decision-making.
Simple decision-making.
The complexity already exists inside the physical system.
There is little wisdom in adding unnecessary complexity to the people responsible for operating it.
Sometimes the smartest engineering decision is not to build a more elaborate solution.
It is to eliminate the need for the solution altogether.
Perhaps that is what Musk’s famous phrase really means.
The best part is no part.
Not because parts are bad.
But because every part you never have to build, monitor, maintain or replace is one less opportunity for reality to remind you that complexity always sends an invoice.