PTQ Q4 2024 Issue

53

0.4090

52

0.4085

51

0.4080

50

0.4075

49

0.4070

2022-01

2022-02

2022-03

2022-04

2022-05

Original thickness

Original: Gate ip points

Recalculated thickness

Temperature

Figure 3 Apparent increasing thickness due to active corrosion

echoes reflected from a corroding surface typically have a smaller amplitude and are broader (bottom panel). In some cases, the echo splits into two or more peaks.³ Echo deterioration and ToF variability The challenge with echo deterioration due to corroding sur - faces is the accurate identification of ToF reference points, as depicted by the red dots in Figure 2. As ultrasonic ech - oes deteriorate, these reference points shift, causing uncor - related variations in their positions. Consequently, the ToF values change over time, leading to alterations in the thick - ness trends according to Equation 1. These variations often manifest as distinct multiples of the half-wave distance, resulting in step-function-like changes in the thickness trendline. In some instances, these changes produce a steady increase in ToF, which can be misinterpreted as an accumu - lation of deposits on the internal surface of the part being examined, as shown in Figure 3 . Such misinterpretations can lead to critical situations that impact the safe operation of the plant. The blue line in Figure 3 illustrates the apparent thick - ness trend, which is unaffected by temperature changes and results from shifting ToF reference points caused by active corrosion.4 In contrast, the orange line represents the

actual corrosion trend, determined using machine learn - ing techniques to account for variations in the ultrasonic waveform.³ Enhance sensor accuracy Advances in machine learning and the AIoT are helping to improve the limitations of permanent sensors. The actual material velocity at Corrosion Monitoring Locations (CMLs) can be measured directly, estimated using models, or a mix of both. While industry guidelines often describe the rela - tionship between temperature and material velocity as lin - ear, it is usually more complex – non-linear or quasi-linear in certain temperature ranges. Permanent sensors can measure thickness frequently, ranging from minutes to days, which is far more often than the months or years between manual inspections. This fre - quent data collection, often during low-corrosion periods, allows machine learning to find the best temperature coef - ficient for the material at each CML. Additionally, ultrasonic material velocity is affected not just by temperature but also by perpendicular stresses, known as the acoustoelastic effect. This effect is smaller compared to temperature changes and can only be detected with permanent sensors due to their stability and sensitivity. Manual inspections typically miss this effect.

Chart Data

-

Go

1W 1M 2M 6M 1Y 2Y 5Y 09/07/2021

09/07/2023

120

0.288

0.2864

80

Raw thickness and temperature should be positively correlated according to physics!

Anti-correlation of raw thickness and temperature!

40

0.2848

16 May

30 May

13 Jun 27 Jun 11 Jul

25 Jul

8 Aug

22 Aug

5 Sep

19 Sep

3 Oct

17 Oct

31 Oct

Measurement time

Minimum Thickness

TC Thickness Temperature ˚C ˚F All Thickness

Figure 4 Acoustoelastic effect on ultrasonic wave material velocity

102

PTQ Q4 2024

www.digitalrefining.com

Powered by