In our industry, there is a natural temptation to demand absolute certainty from predictive technologies. Operators and financial planners understandably want a simple, deterministic answer: "On exactly what day will this asset fail?" While a single, fixed date may seem reassuring, presenting predictions in this way is not only unrealistic, but also potentially misleading.
In the real world, engineering assets operate under highly variable conditions, and predictions without quantified uncertainty can lead to poor planning and costly mistakes. Truly robust predictive maintenance systems must move beyond deterministic outputs and communicate not only what they predict, but also how certain they are about those predictions.
Uncertainty is an unavoidable reality of physical operations. It enters our models from a wide range of sources, including the inherent accuracy limitations of sensors, fluctuations in data quality, operational variability, changing model assumptions and evolving environmental conditions.
To make this uncertainty useful and actionable, we must quantify it using advanced mathematical techniques, including Bayesian approaches, probabilistic models and uncertainty quantification methods based on probability distributions.
When calculating the Remaining Useful Life (RUL) of a critical asset, the result should always be presented as a confidence interval rather than a single, fixed date. For example, telling an operations team, "This asset is expected to fail within 12 to 18 months, with 90% confidence," is far more valuable than predicting a single failure date.
This range provides operators with a realistic, risk-informed framework for planning budgets, scheduling outages, and coordinating maintenance activities without being caught off guard by unexpected events.
Furthermore, a meaningful assessment of asset health requires us to evaluate overall operational risk, not just physical condition. To achieve this, risk scores should combine the asset’s physical condition, its statistical probability of failure and the real-world consequences should that failure occurr.
Our confidence in these risk-informed assessments naturally increases when multiple, independent data indicators point to the same physical conclusion and when those conclusions are supported by physics-based engineering models. By embracing probabilistic predictions and replacing deterministic assumptions with clearly defined confidence intervals, we can move beyond guesswork and enable operators to make smarter, more informed decisions.
This will bring together HBM, Brüel & Kjær, nCode, ReliaSoft, MicroStrain and Discom brands, helping you innovate faster for a cleaner, healthier, and more productive world.
