The Anatomy of Failure

Machines fail. This is an inevitable fact because in the universe we live in there is gravity, friction, expansion, dissolution and deformation. Things age, tight things loosen, solid things soften. Gears wear, belts break, wheels lose balance.

Failure is something we must detect, predict, stop and prevent — because failure costs money, time, quality and production. But fighting failure without understanding its nature is meaningless.

What is failure, actually? When can we say a machine has failed? When it stops working, or when it doesn't produce what we want?

In a factory, machines and the entire system may appear to be running properly, yet a single bearing slipping in one part of the system can cause major electricity costs. Can we say this system is working perfectly? Failure is not a one-time event — it is an ongoing process. It starts as a micro event, grows and grows, and becomes a destructive phenomenon.

Machines are like living organisms. They may look different but their nature is the same. They consist of many components. Whether a simple washing machine or a space station, they contain similar components performing the same movements — rotation, bending, oscillation, up-down or back-and-forth motion.

Like living organisms, machines need healthcare to diagnose their problems and solve them before it's too late. If machines are like living organisms, how can we detect health problems? How can we communicate with the machine and predict which part of the system is sick or about to become sick?

As you can see from this chart, failure begins shortly after installation — sometimes even during installation. The blue area shows where problems are at a very early stage. During this blue period, you cannot monitor machine health or draw conclusions using traditional maintenance and monitoring options.

After the blue area, failure grows and the cost of fixing the problem becomes quite expensive. At this point, traditional machine maintenance can only help to a degree because failure begins to be felt through heat and audible noise. It may already be too late to intervene, and fixing the problem may become excessively costly. At every stage, machines show all symptoms associated with health problems and provide concrete evidence for diagnosis. With advanced tools to detect these symptoms and analyse their causes, we can keep them healthy at all times.

The traditional view of failure assumes machines and the system are running smoothly as long as machines operate and fulfil their tasks. To express more clearly why this view is wrong, we must focus on the word "optimum." Optimum means the perfect, ideal state of anything or any system. If a system produces efficiently using minimum energy, we can say it is close to optimum — but there are many more parameters to compare.

Optimum level can be determined by comparing cost, efficiency, safety, quality, quantity and other parameters — but if personnel health or lives are at risk, we cannot say the system is operating at optimum. We cannot be satisfied with low energy consumption if our product quality is not at the expected level. To maximise industrial quality and profitability without compromising human life and quality, we must change our perspective on failure and develop solutions accordingly.

Machines will continue to fail. As long as we live in this universe and are subject to its laws, we cannot change that. But we can change our approach to failure to live better in this universe.