Failure Modes in Complex Systems

For want of a rail?

Complex systems fail in complex ways. Moreover, the scope of a catastrophic failure of a complex system is commensurate with the scope of the complex system.
-- J.N. Nielson's post, Complex Systems and Complex Failures



Failure Modes in Complex Systems

Systems lie along a complexity gradient ranging from simple to complex.

At the extreme simple end, we have simple machines: inclined planes, levers, pulleys and wheels with axles. The middle covers quite a spread, from windmills to sailboats to internal combustion engines to assembly lines to factories to computers. 

We don't often think of it, this way, but living organisms - including ourselves - outpace our most ambitious artifacts. These might be contenders for the extremely complex end. But no. Consider communities of these organisms, ecosystems of these communities, the geo/biosphere in its entirety, with its solar and meteoric inputs.

The global industrial economy fits in there somewhere. Less complex than the terran system as a whole. More complex than any given ecosystem.

The question I'm considering in this post is how different levels react to trauma.

Clearly, you can't take a sledge hammer to many of our artifactual machines. Bust holes in them, and they falter and fail in fairly short order. And up to a certain point, the more complex they are, the harder they fall. That sledge can take a lot of abuse. If the head mushrooms, still usable; if the handle breaks, repair or replaceable. But the computer we're hypothetically bashing? Like Bruce Lee, I could take it out with this pinkie!

Beyond a certain level of complexity, many to most systems must be resilient, robust, self-repairing, self-healing... all words those which can handle a good deal more than might appear. Most such systems are evolved, though we have begun to learn to apply them to our own creations.

But there are limits.

If the resilient mechanisms of a system are overwhelmed - or if key mechanisms fail permanently - the system goes down. Permanently.

Worse, both Complexity and Chaos Theory state that the failure of complex systems is inevitable.


Internal Failure Modes

Three general modes can bring a system down from within. I'll informally call them node, runaway and domino failure modes.

Node failure is deep doo-doo. Nodes are sub-systems upon which the function of many other aspects of system functionality depend. A node goes down, and all dependent, 'downstream' sub-systems fail with it. Worse, nodes are often inter-dependent. Failure of one node can spread to others in a process called contagion; essentially a critical case of the domino mode mentioned below.

A human health example of node failure is ventricular fibrillation. The heart, for whatever reason, begins to spasm out-of-sync, and fails to pump sufficient blood flow. If it can be brought back into sync (via defibrillation techniques), a full recovery is possible. But if not - and the heart fails permanently - all other bodily systems fail in short order for lack of blood perfusion.

The power grid is a node within a modern, industrial economy. It falters, and if not brought relatively quickly back on line, all dependent systems - banking, communications, IT, etc. - exceed their back-up arrangements and fail.

Runaway failure occurs when positive feedback - that which ramps a system into more 'extreme' states - is insufficiently damped by governance mechanisms. These counter positive feedback loops with negative feedback. This can happen even in an otherwise healthy system. Typically, a system in runaway mode will suffer increasingly wild oscillations until a vital node gives way.
 
 "Galloping Gertie" - the first Tacoma Narrows Bridge, which dramatically tore itself apart in 1940 - is an example of runaway failure. The process, described as "phenomenon in which several degrees of freedom of a structure become coupled in an unstable oscillation" is very hard to rule out of any complex system, whether designed or evolved. Degrees of freedom, couplings and even initial oscillations are extremely difficult to predict or detect until well underway. Complexity guarantees that.

A prediction of Peak Everything theory is that market feedback cycles will begin to oscillate wildly in post-peak environments. Supply (particularly of energy / fossil fuels) in post peak situations has plenty of room to fluctuate in response to demand. Demand, however, while driven upward by exponential growth of any market component, is coupled with forces of 'demand destruction' (loss of those willing or able to pay higher prices). Demand fluctuates more and more wildly, with production attempting to follow jerkily in its wake. Both supply and demand 'players' begin to drop from the field as market conditions exceed their ability to cope. Add in risky financial 'instruments', and the finance node is threatened. Banks begin to fail...

Domino Failure is sequential failures radiating outward from a single trigger event among dependent elements (again, even if they are otherwise functioning 'normally'). This is the "For want of a nail... ...the Kingdom was lost" mode. Failure of a small element can initiate such cascading failures. It may 'burn out', or it may take down a node (see Doo-Doo, above).

Again, the power grid is an example (see Blackout Inevitability and Electric Sustainability). In the case of the 2003 Northeast Blackout, it all began with a software bug. The sequence of events cascaded until large swathes of Canada and New England were without power.

The Global Financial Crisis of 2007-8 is another example. The (predicted) burst of the housing bubble set dominoes falling in all directions, some foreseen and others not. Analysis continues to this day. Unfortunately, so do many of the practices which led to the crisis, and the bursting fracking bubble may well give us another shot at total disaster along similar lines.

Again, dependencies within a complex system are extremely difficult to map, much less predict. Those who correctly predict the broad outlines of disaster are typically ignored both before ('pessimists on the lunatic fringe') and after ('who knew?' and 'hindsight quarterbacks'). So it goes.


External Failure Modes

Systems can fail for 'external' reasons, as well.

Blunt trauma failure is something I hesitate to call a 'mode', though any system is susceptible to it. Here, something big and bad happens - a nuclear exchange, for example - that blows a hole in important fabric. It may be random - as from solar flare - or targeted - as in ICBMs or sabotage. Blunt trauma can activate any of the other modes, or merely destroy the system as a whole.

Running out of steam failure happens when the resources upon which a system runs fall below some necessary threshold. In particular, energy, space and material resources. Run low of any essential, and the whole thing winds down. Run out of any essential and the whole thing grinds to a halt. This type of failure follows a trajectory (curve) which can range from 'Seneca's Cliff' to the 'Staircase Descent'.


Timescales of Failure

Whether total, systemic failure is 'catastrophic' (my school) or 'catabolic' (Greer, Kunstler, et al), I suspect is a matter of the timescale over which failure is viewed.

The graph is similar. A long period of growth is followed by a relatively shorter period of decline or 'collapse' (Senaca's Cliff).

For myself, I tend to think of high to low collapse occurring within a single lifetime as both catastrophic and probable. Our current situation has been collapsing - by my estimation - for decades, now, and will accelerate to whatever bottom very soon.

Yet both means and trajectory look very similar, whether they play over a shorter or longer term. From the perspective of the individual, a lifetime is all one has. Generations may experience the long descent. Cultures may experience the rise and fall of civilizations. Our species has experienced ages - the paleo- and neo-lithic, bronze, iron, industrial and information ages.

From a long enough perspective, the rise and (presumed) fall of agricultural civilizations over roughly 10K years - is but a blip. Look close enough and any process plays out in gradual, slow-mo. Thus, whether collapse is fast or slow is a subjective matter.

Whether it plays out within our or our children's lifetimes, however, is a practical matter. One prepares for 'contractions' differently than from SHTF.

Ya lays yer money and ya takes yer chances.