Revised 23 October 2019
The current trend is to produce systems of all sorts with increasing levels of technical complexity. Unfortunately, technical complexity is antithetical to reliability and long life. For example, contemporary automobiles are being produced with many more electronic components of all types—sensors, processors, electric actuators of all forms—and complex wiring harnesses to connect the components. And, of course, the more components there are, the more things that can degrade or fail. The greater the number of interconnections, the larger the number of possible connection faults.
In addition to the proliferation of electronic and electrical components in contemporary systems, there is an explosion in the quantity of software code embedded in the various subsystems. Unfortunately, much of the code is not developed to the highest standards of testing and may be poorly documented to support software maintenance and upgrades. Code at some lower level in a system may have incompatibilities with new code at a higher level but be opaque as to the cause of the problem. Software will be especially problematic as artificial intelligence is incorporated with machine learning capabilities. Determining why some software is behaving anomalously may be nearly impossible when the code has diverged from its baseline on its own.
The tendency to connect modern systems to networks, including the internet, makes them intrinsically vulnerable to attack by malware. Security hardening against such attacks is generally an afterthought, if it is considered at all for consumer systems. I am frightened by the vulnerability of systems like automobiles to hacking, since there are many possible situations with lethal results.
The basic principle of performing failure mode, effects, and criticality analysis (FMECA) for complex systems is often ignored, particularly for systems that are under pressure to be developed at the lowest possible cost. As a result, the systems may fail in a catastrophic mode with no inherent backup. Well-engineered systems fail softly, maintaining as much functionality as possible after a failure. Single point failures are rigorously avoided.
We are not designing contemporary systems to be repaired at a component level. Instead, the current philosophy is to replace entire subsystems (or indeed whole systems) if something fails. In connection with this philosophy, we are not supporting the diagnosis of system problems at a lower level.
The materials incorporated in contemporary systems are a concern. All materials have degradation over time from stress, wear, corrosion, and exposure to various environmental effects. However, the proliferation of plastic components throughout modern systems is likely to cause degradation and failure earlier than the metal components they displace. And plastic parts cannot typically be repaired; they must be replaced with an equivalent plastic part.
The supply of replacement parts is going to be a limiting factor in the life of contemporary complex systems. Every system is dependent on the critical component that is in shortest supply. When the last replacement part in stock is used, the system will have to be retired. Unfortunately, current practice is to produce spare parts only while the system itself is being produced. When manufacturing shifts to a new model that doesn’t use those parts, the inventory of spares is all that will be available. This is going to be especially problematic for the highest technology components, such as electronic devices. Every few years, the manufacturing technology for chips shifts to a new standard and older devices aren’t even able to be produced any longer.
The problems are not limited to physical devices. We are not training new people to maintain older systems. As older maintenance technicians retire, the skills are lost. This is one of the reasons that nuclear power plants scare the living daylights out of me. Nuclear power plants must be maintained continuously to the highest possible standards, or they become extremely dangerous. If something happens (e.g., a pandemic illness) that prevents a sufficient cadre of knowledgeable maintenance personnel from staffing nuclear plants around the clock, there can be a catastrophe.
The overall picture I see is that the highest technology contemporary systems can be projected to have relatively short useful lives, but few people will have anticipated it.