Condition Monitoring and Predictive Maintenance are two concepts that often come up in the context of safety engineering. Both are related to ensuring permanent availability of safety-critical equipment, with minimal or even zero interruption. In practice, this translates into a need to ensure prompt and efficient maintenance which resolves – or, ideally, prevents – any defect in a timely manner.
Maintenance is a component of any system’s lifetime and it is critical to ensuring its adequate and safe functioning. However, maintenance costs time and money, and performing, it may requires restricting or interrupting a system’s functioning. Well-founded maintenance is an engineering practice, whereas unnecessary maintenance is not only a waste of resources, but it is also detrimental to the functioning of a system as a whole.
This translates to two basic questions: what procedures are necessary, and when?
Condition Monitoring is one of the most useful methods to provide an answer to these questions. It refers to the continuous monitoring of the equipment’s state and operating parameters, usually through dedicated sensors and monitoring tools.
Its end goal is to identify changes that indicate damage, incorrect configuration or other safety-impacting conditions, so that corrective maintenance repairs can be performed before a failure gets the chance to occur.
Exactly which parameters are monitored depends, of course, on application and equipment: they include, for example, temperature and vibration parameters for electrical drives, or SNR levels for communication equipment. Not only can these parameters indicate an impending failure, but they can also indicate which components are most likely to be at fault, thus enabling engineers to plan and target their maintenance operations with more accuracy.
Data obtained through Condition Monitoring provides valuable information about the current state of a system. But its value is not limited to evaluating an equipment’s condition at a given time. Its evolution can be used to anticipate how an equipment will perform and how it might degrade – and to schedule maintenance according to these expectations.
This is known as Predictive Maintenance and it is based on anticipating the future evolution of a system – in other words, on anticipating what failures may occur and what maintenance needs to be performed in order to prevent them from occurring.
Unlike Preventive Maintenance, where maintenance operations are scheduled based on equipment-specific knowledge, statistics and legal or internal requirements, the Predictive one relies on data about a system’s state and evolution to schedule maintenance operations as they are needed.
Predictive Maintenance enables more efficient, longer-term planning for maintenance operations and makes it easier to define operational maintenance goals and to allocate maintenance resources.
Examining data from hundreds or thousands of sensors, gathered over months or even years, is well beyond the capabilities of human operators. Furthermore, the mathematical models, which describe an equipment’s evolution (and predict potential faults) based on such a wealth of data, are generally prohibitively complex to be used by humans.
Consequently, in recent years, Predictive Maintenance has come to rely increasingly on Machine Learning techniques. Machine Learning refers to a set of statistical techniques, which enable computer systems to learn how to identify and classify patterns in large volumes of data and to make predictions based on it.
Condition Monitoring refers to the process of monitoring a system’s state in order to identify changes, which would indicate damage or an impeding failure. It enables operators to identify and correct problems (through repair and maintenance procedures) before they cause equipment to fail.
Predictive Maintenance refers to planning corrective maintenance based on predictions about the evolution of a system. These predictions are based on data obtained through Condition Monitoring, and on system-specific knowledge.
In other words, Predictive Maintenance is one of the ways in which Condition Monitoring can be leveraged. The two are complementary and refer to different ways of using and acting upon sensor data.
Both are reliable methods to ensure operational safety at every level, including in hazardous areas. However, it is worth iterating that both of them depend on the quality and integrity of sensor data: the quality and safety of the sensors, measurements and transmission chains is critical to their success.