Early Detection and Monitoring of Corrosion in the Wide-Area around an Installed Sensor

Early Detection and Monitoring of Corrosion in the Wide-Area around an Installed Sensor

Case study on detection of induced pitting using the novosound Ceilidh.

The Problem

Corrosion is an industrial problem costing billions every year. This requires 24/7 monitoring to enable time and cost-efficient maintenance programs. Technologies such as ultrasonic-guided-wave have been developed for the long-range inspection of pipelines in energy and production. However, the low frequencies of these acoustic systems do not have the sensitivity to detect the earliest signs of corrosion. Alternatively, higher frequency, permanently installed 0-degree compression UT probes can be used. Whilst these improve sensitivity, the drawback is that corrosion which occurs in areas which are not directly under the face of the sensor will not be detected or monitored.

The energy and chemical processing industry is seeking improved efficiency in monitoring aging assets and remote detection/monitoring is becoming a necessity. By reducing the need for fleets of NDT inspectors to access every panel and pipe, their efforts can be more targeted, reducing time taken and reducing costs.

The Solution

novosound’s Ceilidh is a revolutionary monitoring system for corrosion, utilising a sparse acoustic array that is capable of detecting and localising corrosion with sub-millimetre-scale anywhere in its installed area. This is due to the use of high-frequency guided waves which propagate laterally to each of its sensors (rather than in thickness mode). Moreover, machine learning (ML) and artificial intelligence (AI) are used to automate data analysis, deskilling the interpretation of data. novosound Ceilidh identifies the corroded areas at the earliest time, providing enhanced monitoring which lowers cost through increased efficiency of the inspections team.

The Ceilidh Spider

Method

Two sparse arrays (known as novosound Spiders) were connected to the novosound Ceilidh hardware, which was connected to a laptop via Wi-Fi to log the data for processing.

The two spiders were bonded to a 0.5m length of 4” NPS Sch40 steel pipe’s surface using a cyanoacrylate adhesive (Figure 1). The Ceilidh hardware drove each Spider’s central transmitters (Tx) with a 600 kHz burst and baseline data was recorded from every Rx sensor of the associated array.

Ceilidh Arrays Bonded to the Pipe and Connected to Hardware

Corrosion was then induced (Figure 2) at a position on the inside wall of the pipe, and located underneath one Spider using a saltwater bath, copper electrode and a 0.5mA current. This was carried out on two separate occasions for 2 hours, with data recorded between each.

Induced Corrosion Inside the Pipe, Left 2 hours and Right 4 hours.

The above image shows the Corrosion on the Inner Surface of the Pipe and the Sparse Array of Sensors on the Outside of the Pipe. Note that the Corrosion is under one Spider, and adjacent to the other Spider.

The novosound Ceilidh uses a specialised AI/ML algorithm to classify the data gathered from the Spiders before and after induced corrosion to train and test the system to distinguish the corrosion depths.

Rubber casts were made of the corrosion site at both time-points to quantify the actual corrosion depth via optical profilometry. An optical microscope was used to analyse the corrosion depths achieved, with 4 hours of corrosion giving pitting ranging 54 – 188 microns in depth.

The above Image shows the Depth Corroded Spot After 4 Hours of Corrosion Using a 2D Grey Scale Image

Results

The AI/ML model was trained with 70% of the data collected from the Ceilidh system. The remaining data was then used to test the model, which gave a 97% accuracy in classifying between the three states: clean pipe, 2 hours of corrosion, and 4 hours of corrosion. When the two arrays were considered separately, the Spider directly over the corrosion gave 100% accuracy in detecting the corrosion, and the adjacent array gave 87% accuracy (despite not being directly over the corrosion). In comparison, an industry-standard permanently installed ultrasound probe would not be able to detect corrosion adjacent to its active surface.

The above Image Shows the Output of the Ceilidh Software, whereby the Corrosion is Underneath the Spiders (A) and Adjacent (B), showing that whilst Detection is 100% over the Corrosion, it is still 86% when Adjacent. In Comparison, Existing Point Measurement Corrosion Monitors would have 0% Corrosion Detection Adjacent to the Sensor.

Conclusion

This demonstrates the novosound Ceilidh system is capable of detecting sub-millimetre pitted corrosion with an accuracy of 86-100%, covering a large area not directly under the sensor and operating independently of user input. This is a scalable and flexible product which allows for much more extensive area coverage, and technicians/installers do not need to worry about accurate sensor placement. This offering allows you to add the next level of digital integration into a monitoring ecosystem, saving inspection time and adding resolution.

Key Points

Early detection and monitoring of corrosion 

Wide-area monitoring, around the sensor area (NOT just point measurement)

Submillimetre-scale precision with an 86-100% accuracy

Automated data analysis using AI/ML

A scalable and flexible product