Thinning (Multi-Path Linear Degradation)

  • Updated

The following descriptions and requirements apply to the Thinning (Multi-Path Linear Degradation) Model in 2021.

What Is Thinning (Multi-Path Linear Degradation)?

The purpose of the Thinning (Multi-Path Linear Degradation) Model in Newton is to determine the Predicted Failure Date and Probability of Failure over time. The Thinning (Multi-Path Linear Degradation) Model has a unique approach that models thinning as a set of linear degradations using thickness measurements and modeled corrosion rates.

Why Is the Model Changing?

This is an evolution of the Linear Degradation Model that was originally implemented in Newton. It addresses some issues that were present in the previous model while providing powerful new capabilities. It is important to note that the data inputs from the previous Thinning Model have not changed, so users will not need to make any updates other than to their selected Probability of Failure Model within the Failure Mode card.

Differences from the Thinning (Linear Degradation) Model:

  • Accounts for changes in corrosion rate over time, rather than assuming a constant corrosion rate.
  • Incorporates assessment corrosion rates from failure mechanisms with and without measurement data.
  • Addresses deviation between measured corrosion versus calculated or expected corrosion, for example growth in measured corrosion and loss in expected corrosion.

How Does the New Model Work?

This methodology for calculating the Lifetime Variability Curve (LVC) for Thinning Failure Modes relies on a concept of tracing out corrosion paths for remaining life. Corrosion paths are a series of remaining life values as a function of time. Each corrosion path has an origin and termination point. These two points are form a path. Each segment, or period between measurements/observations, has a distribution of paths. Considering observed data, modeled corrosion rates, and component corrosion, the corrosion paths define a distribution of time to failure.  By calculating many paths, a distribution of the time to failure is collected and subsequently converted with a Weibull model to Probability of Failure.


What Differences Can Users Expect to See?

In reviewing one Condition Monitoring Location (CML) with the Thinning (Linear Degradation) Model applied on a piping circuit in a Reformer unit, users will notice that even with considerable “noise” in readings, the Lifetime Variability Curve (LVC) is depicting next to no degradation occurred on this CML since 1970. It is also worth noting this CML is ranked the 32nd out of 36 total CMLs from lowest to highest Remaining Life, and it is predicted to fail 34 years after the Predicted Failure Date for the asset. It is easy to see the predicted degradation fits all the measurements well as a whole when reviewing the LVC for the Linear Degradation model. However, notice the corrosion rate from the 2012 measurement onward is decreasing more rapidly than the linear model fit. The linear fit contradicts the expected corrosion rate for the failure mechanisms and contradicts the most relevant and recent measurements for this CML. mceclip1.png

The LVC graph is updated for each CML after the Thinning (Multi-Path Linear Degradation) Model is selected in the assessment and the assessment is calculated. The predicted corrosion rate between readings is visible on the LVC graph. This CML is also now ranked as having the lowest Remaining Life, making it the driving CML and setting the Predicted Failure Date 83 years sooner than predicted by the Thinning (Linear Degradation) Model. mceclip2.png

This significant difference in predicted failure date is understood when observing corrosion after the August 2006 measurement has a negative slope that nearly matches the expected corrosion rate in the Multi-Path Linear Degradation LVC graph. This result represents SME and statistical understanding of the rate of degradation. By comparison, the Linear Degradation model fits the measurement data well but does not represent the understanding of degradation for this CML.

Didn't find what you were looking for?

Submit a request