The terms prognosis and diagnosis have been the subject of confusion for decades. These terms are related, but each have completely different meanings. The first thought that comes to mind when the words prognosis and diagnosis are mentioned, is usually when discussing health. These words are commonly used by medical doctors to describe someone’s health or condition. However, they can also be used by maintenance professionals to describe the health of complex systems and equipment. During its life in service, a system or a component can exist in a state of health anywhere from perfectly serviceable to functionally failed.
In this article, we will discuss the terms prognosis and diagnosis as they relate within the digital data world. We will also touch upon the difference between prognosis and diagnosis and how technology is helping to achieve a more efficient degree of prognosis and diagnosis of complex systems.
What is Diagnosis?
When a system has functionally failed, the equipment is clearly faulty. This is the domain of diagnostics and troubleshooting and in this scenario, a diagnosis examines the symptoms of an evident issue or problem.
Functional failures demand reactive maintenance – essentially some form of urgent action after the failure has occurred. Reactive maintenance is, by definition, a surprise.
Functional failures are undesirable because of their disruptive effect on productivity when the equipment goes down. Reactive maintenance is undesirable because of its disruptive effects on the maintenance teams responsible for fixing the problems.
Generally speaking, it is preferable to identify impending faults before they become functional failures – and that is called prognosis.
What is Prognosis?
A prognosis is a future prediction – in this case, anticipation of a near-future failure. Essentially, a prognosis attempts to pin-point a looming failure while the system is behaving normally. A prognosis allows pre-emptive maintenance to take place, in an effort to avoid the looming failure.
There are several approaches to pre-emptive maintenance, including: Predictive Maintenance (PdM), and Condition-Based Maintenance (CBM). The main difference between PdM and CBM is the presence of symptoms impending system failure.
CBM includes early indicators of impending failure emerging, or symptoms of impending failure. Historically, CBM has used temperature and vibration data to predict failures quite successfully, but digital error messages present a different challenge. This includes error messages by the machine alerting to an issue. Such messages are produced in high volume, and not all messages signify impending failures. There are nuisance (invalid) and spurious (incorrect) messages and service alerts that may not be failures - but may indicate required service is needed. Successful CBM isolates the meaningful messages, detects patterns that indicate looming failures, identifies the likely cause, and thereby enables corrective maintenance of the system to be scheduled.
A true CBM system will not only identify patterns in the low-level indicators preceding failure, but also will provide pre-emptive troubleshooting guidance where there is ambiguity as to the cause.
PdM is a tougher challenge because it must predict a failure before the machine itself is starting to show any sign of trouble. The currency of PdM is “leading indicators”, each being a set of circumstances that reliably predict a future failure. PdM requires data flowing from a normally operating machine, in addition to other helpful data sources, and divine its own signals to identify a component that will fail in the future.
Acquiring leading indicators is serious work requiring big data, intelligent sampling, and domain expertise. Each one is a science project unto itself. But once established, a leading indicator becomes something that can be monitored, confirmed when triggered, and acted upon.
How Can Technology Help?
To quote Nate Silver, author of the book The Signal And The Noise, “What matters most is not pattern detection capabilities, but pattern analysis capabilities.”. Finding patterns is easy in any data-rich environment. The key is in determining which patterns represent signal, and which are just noise. Relevant signals will be there but so will a whole host of irrelevant patterns that look very much like the relevant ones.
With both CBM and PdM, detection of the potential failure is the first part of the puzzle. Knowing exactly what maintenance action to take is the final part. Between these two parts is that crucial analysis, and therein lies a tradeoff between ambiguity and precision that affects efficiency. This is where technology can help.
Every leading indicator of a potential failure will have some degree of ambiguity. The lower the ambiguity, the more precise the corrective maintenance action. But in many cases, achieving low ambiguity purely through data analytics can be challenging, time-consuming, and expensive. Often the information needed to sway the decision can be found in data that is not sensed, such as sound or visual observations, special tests, or maintenance history. In those cases, solutions like SpotLightâ by ATP CaseBank can take the PdM or CBM alert and guide the reliability team or maintenance crew to the best maintenance action.
Where technology like SpotLight really shines is with troubleshooting and diagnostics. SpotLight has a diagnostic database and diagnostic reasoning engine that work together to gather symptoms, causes, and solutions for equipment defects in order to optimize the troubleshooting process. Given a CBM alert in the form of a detected symptom or a PdM leading indicator, SpotLight can guide a technician to eliminate alternative possibilities and identify the maintenance action required.
The words prognosis and diagnosis are more than simply terms for health professionals. Maintenance teams use them on a regular basis, and they are even more relevant within the world of digital data being emitted from complex mechanical devices and equipment. Luckily, with advancements in technology, the ability to define both in relation to a critical maintenance issue is becoming easier, contributing to more efficient maintenance operations and overall safety, in many cases.