Estimating uncertainty when using transient data in steady-state calculations

When using measurement data for monitoring there is often a desire for steady-state analysis. On-line condition monitoring and fault detection systems are typical applications where the traditional way of treating transient data is to remove it using methods that require tuning using thresholds. This paper suggests an alternative approach where the uncertainty estimate in a particular variable is increased in response to the presence of transients and through propagation, varies the uncertainty in the result accordingly. The formulation of the approach is described and applied to two examples from building HVAC systems. The approach is demonstrated to be a pragmatic tool that can be used to increase the robustness of calculations from time series data.