Key Differences between Preventable and Predictable Analysis

As the largest consortium of Korean custom machine manufacturers, our main goal is to provide dependable machines to our clients. This goal applies not only to machine users but also to machine manufacturers, as machine profitability drops significantly if warranty issues arise after machine sales.

Many Korean machine builders and users, especially those in the power plant and automotive industries, perform scheduled maintenance to prevent machine failures. In recent years, we have frequently been asked by clients about a “predictive machine maintenance module” embedded within the machine. In many cases, however, they mean preventive maintenance, not predictive maintenance.

What is the difference between preventive and predictive maintenance?

How are they used to extend machine life?

Machines give out indications before they break down.

Maintenance checkpoints, in many cases, include unusual noise, excessive vibration, temperature changes, oil or fluid leakage, electrical issues, and erratic behavior. When we visited a client’s plant manufacturing automotive shock absorbers in Poland in 2016, we witnessed an encyclopedia-sized maintenance logbook created by the client’s technicians, who recorded all machine behavior as part of their scheduled maintenance work. In fact, our client knew more about the machine’s day-to-day behavior than we did.

As a machine manufacturer, we understand how the machine is designed, programmed, and built to perform within defined parameters. However, we know much less about how the machine is used by the client. By analogy, keyboard manufacturers know how each key is supposed to function, but they cannot predict which key will wear out first, as that depends entirely on usage.

Preventive maintenance aims to avoid equipment failure by scheduling regular maintenance and inspections based on historical data and maintenance guidelines provided by the machine manufacturer. The collected maintenance data are then analyzed to determine appropriate timing for machine part replacement.

Preventive analysis is time-based or usage-based (i.e., based on machine operating history). Common preventive analysis includes alarm-triggered analysis from the machine’s PLC program or condition-monitoring data collected by sensors measuring parameters such as vibration, temperature, and alarm frequency.

Regular machine maintenance is key to preventing failures; however, preventive analysis does not necessarily mean that replaced parts have no remaining service life.

Predictive maintenance requires a tool or module known as predictive analysis, which includes data historians, machine learning, or algorithm-based models that estimate the probability of potential failures. Unlike preventive analysis, predictive analysis is forecast-based.

Predictive analysis also uses historical data stored in historians or PLC systems to establish machine behavior patterns. Algorithms or machine learning models then incorporate user specific conditions such as machine usage, plant temperature variations, dust levels, power stability, humidity, and component aging to refine predictions.

When we visited Yale University in 2017, they were using predictive analysis tools to monitor power plant equipment. The power generation industry was among the earliest adopters of predictive analysis as part of its maintenance strategy.