Data science analytics.

Data Analytics

Sonomatic has consistently been at the forefront of data analytics development and utilisation. These areas of specialisation have resulted in two main benefits: maximisation of the amount of quality information we can draw from inspection data, providing clients useful high-value that they can to make key asset decisions.

Sonomatic uses several analytical techniques and tools to achieve this and has developed software to aid data visualisation and our mini digital twin.

Data Trending

Sonomatic have developed a novel approach to data trending by looking at whole datasets and long-term statistical behaviour to consider how corrosion could be affecting components. This approach allows our team to determine behaviour before considering any sub-groupings of data points that are showing similar behaviour.

The data is first normalised, then in cases where nominal thicknesses vary and spurious, results are flagged on upload into our software. The example below shows that there is a noticeable downward trend in the data which is indicative of corrosion.

Data trending graph.

However, it may be the case that this trend has been caused by a subset of the data and Sonomatic’s approach to data trending means that there is the capability to drill further down into the data and investigate all potential groupings to identify which subgroups are showing greater wall loss.

This then allows a corrosion rate per grouping to be calculated giving the client added information and value from the inspection data. In this case, our analysis method identified key subgroups where the corrosion was dominant which lead to the client optimising inspection planning.

Data Comparisons

Data comparisons graph.

When multiple inspections have been performed, cumulative thickness distribution curves can be plotted simultaneously to give a visual representation of any changes between the multiple inspections.

The example shown to the left where a change in corrosion behaviour between the two inspections is shown as 1% of the inspected area in 2020 was measured at 8.6 mm or less compared to 0.03% of the inspected area in 2018.

Comparing inspections in this manner is particularly helpful when looking to quantify any changes to the extent of corroded areas. Changes to the overall minimum can be easily recorded but plotting in the manner described here gives insights into the spatial behaviour of any existing corrosion.

Data Extrapolation

Data extrapolation graph.

Sonomatic’s data analytics capabilities extend to giving added confidence when an inspection has been a sampling approach or, for some reason, the inspection did not achieve the required coverage.

Where a thickness distribution shows abnormal behaviour, a statistical extrapolation can be used to calculate the minimum in uninspected areas.

The abnormal tail is then isolated, and a curve is fitted to this tail to calculate, based on trend and extending relative to the percentage of the component inspected, the overall minimum thickness. The calculation also generates a probability of the minimum being below any alarm limits defined by the client.

Data Profiling

Sonomatic can produce data profiles either in the form of an axial profile, also called a river bottom profile or circumferential polar plots.

Plots of this type give a quick and easy visual representation of any locations of concern as well as being used for comparisons.

In addition, circumferential polar plots give a clear representation of how thickness variations manifest around the circumference of the pipe. The circumferential polar plot example shows data being compared where there was significant corrosion between inspections. There is no limit to the number of historical inspections that can be compared for this type of analysis.