High accuracy ultrasonic degradation monitoring

Abstract

This thesis is concerned with maximising the precision of permanently installed ultrasonic time of flight sensors. Numerous sources of uncertainty affecting the measurement precision were considered and a measurement protocol was suggested to minimise variability. The repeatability that can be achieved with the described measurement protocol was verified in simulations and in laboratory corrosion experiments as well as various other experiments. One of the most significant and complex problems affecting the precision, inner wall surface roughness, was also investigated and a signal processing method was proposed to improve the accuracy of estimated wall thickness loss rates by an order of magnitude compared to standard methods.

It was found that the error associated with temperature effects is the most significant among typical experimental sources of uncertainty (e.g. coherent noise and coupling stability). By implementing temperature compensation, it was shown in laboratory experiments that wall thickness can be estimated with a standard deviation of less than 20 nm when temperature is stable (within 0.1 C) using the signal processing protocol described in this thesis. In more realistic corrosion experiments, where temperature changes were of the order of 4 C), it was shown that a wall thickness loss of 1 micron can be detected reliably by applying the same measurement protocol.

Another major issue affecting both accuracy and precision is changing inner wall surface morphology. Ultrasonic wave reflections from rough inner surfaces result in distorted signals. These distortions significantly affect the accuracy of wall thickness estimates. A new signal processing method, Adaptive Cross-Correlation (AXC), was described to mitigate the effects of such distortions. It was shown that AXC reduces measurement errors of wall thickness loss rates by an order of magnitude compared to standard signal processing methods so that mean wall loss can be accurately determined. When wall thickness loss is random and spatially uniform, 90% of wall thickness rates measured using AXC lie within 7.5 ± 18% of the actual slope. This means that with mean corrosion rates of 1 mm/year, the wall thickness estimate with AXC would be of the order of 0.75-1.1 mm/year.

In addition, the feasibility of increasing the accuracy of wall thickness loss rate measurements even further was demonstrated using multiple sensors for measuring a single wall thickness loss rate. It was shown that measurement errors can be decreased to 30% of the variability of a single sensor.

The main findings of this thesis have led to 1) a solid understanding of the numerous factors that affect accuracy and precision of wall thickness loss monitoring, 2) a robust signal acquisition protocol as well as 3) AXC, a post processing technique that improves the monitoring accuracy by an order of magnitude. This will benefit corrosion mitigation around the world, which is estimated to cost a developed nation in excess of 2-5% of its GDP. The presented techniques help to reduce response times to detect industrially actionable corrosion rates of 0.1 mm/year to a few days. They therefore help to minimise the risk of process fluid leakage and increase overall confidence in asset management.