A technical object is what its design has decided it should be. The plan, the specifications, the tolerances fix what it is. Operation is the pursuit in time of this initial definition. Breakdown is the event by which the object temporarily ceases to be itself. Maintenance restores identity, cancels drift, restores nominal state. To maintain is to preserve what has been prescribed.
This thesis is comfortable. It is also false.
Simondon describes concretization as the process by which the technical object integrates its own constraints until it achieves coherence of its parts. But maintenance does not appear in his analysis as a constitutive category. It remains subordinated to what design has already resolved. What this architecture leaves suspended: the hypothesis that the industrial object, once produced, would have an identity stable enough that subsequent interventions could be read as the conservation of the same.
Consider a Boeing 747 in service for thirty years. The original fuselage is still in place. Almost nothing else. The engines have been removed, overhauled, sometimes replaced by more efficient variants. The avionics have gone through four generations. The electrical harnesses have been retraced. The maintenance procedures themselves have been rewritten several times, under the effect of regulation, feedback from the global fleet, service bulletins from the manufacturer. The object that takes off in 2024 is not the object delivered in 1994. It bears the same registration number. It responds to the same type certificate. It has almost no material in common with the first exemplar that left the Everett factory.
Administrative identity says: it is the same aircraft. Material identity says: it is another object. Functional identity says: it is a third. None of these three identities coincides with the other two.
What appears here is not the paradox of Theseus. It is something more precise. Maintenance is not a conservation operation. It is a distributed reconception operation, performed by a collective whose members have never met, over durations that exceed any individual career.
Each intervention settles a question that the original design had not posed: what must be preserved, what can be substituted, what must be modified to remain compliant with requirements that did not exist at the time of the initial plan. The technician who replaces a spar according to a procedure revised in 2018 is part of the object's design. The certification agent who validates a life extension is part of the object's design. None of these gestures was inscribed in the plans signed thirty years earlier. Together, they redraw what the object is.
Maintenance has access to information that design could not have. Alloys did not exist. Standards were not written. Failure modes had not been observed, for the precise reason that they require the duration of operation that the first designer could only model. Maintenance designs with this information.
Canguilhem had opposed machine and organism: the organism integrates its maintenance into its operation, the machine receives it from outside. This distinction holds for the isolated machine. It yields for the long-duration industrial object, which aggregates around itself a maintenance ecosystem so integrated into its existence that there is no longer sense in describing the object without it. The 747 does not function with its maintenance. It is what its maintenance has made of it.
Two regimes of identity are superimposed in every industrial object that endures. The specified identity is what the original document fixed. Dated, signed, archived. It does not change. The operational identity is what emerges from the chain of interventions, which reformulates itself at each revision cycle, and which is the only one to which the object effectively corresponds at a given moment. These two identities are not linked by a conservation relation. Operational identity inherits from the specified as from an initial constraint, but it distances itself continuously, through displacements which, taken one by one, are always presented as minor, and whose integral, over three decades, sketches another object.
At each replacement, the object's anterior trajectory is partially rewritten. The substituted part carries away with it the history of its own failure. The new part inaugurates a duration whose counter does not align with that of the rest.
The aging object does not have an age. It has a distribution of ages.
What is called its commissioning date is only a boundary condition among others. What gives the object its continuous identity is not its matter. It is the document that accompanies it and that records its interventions. The industrial object that endures is first a file that has been maintained.
The designer works on an object that does not yet exist. The maintainer works on an object that will soon no longer exist in this form. Between the two, there has never been a stable object on which one would have handed over to the other. The stability of the industrial technical object is a retrospective construction. It is the effect of an archiving that has preserved the name, the number, the certificate. It is not a physical property of the object.
Design names what the object will claim to be. Maintenance determines what it will have effectively become. Between the two, the object, which belongs properly to neither one nor the other.
