Improving televisual inspections8 May 2019
Darren Wood and Clément Skopinski explain how a new 3D visual inspection technique, to be tested in a nuclear reactor this year, can improve detection of surface defects by leveraging the latest technologies.
INSPECTIONS ARE CRITICAL TO MAINTAINING the integrity of nuclear components, and the tools that are used are equally important. Nuclear operators are performing more televisual examinations than ever before to fulfil ageing management programmes and licence renewal commitments. But new technology can help improve an analyst’s visual inspection toolboxes and reduce subjectivity or bias from visual interpretation.
Though televisual techniques have significantly improved over the past ten years through better sensors, speed, resolution and image processing, visual examinations have remained limited to 2D characterisation, which does not allow an analyst to know the precise details of an imperfection. Until now, visual testing was a “way to see,” rather than a characterisation process. Technologies have been limited to characterising indications across length and width on the component’s surface, but not depth. For remote visual inspections in challenging environments, these limited capabilities can lead to subjectivity or inaccurate interpretations of images and make it difficult for an analyst to properly categorise indications.
Applying multi-element techniques have advanced eddy current and ultrasonic testing for surface and volumetric inspections. Engineers at Framatome’s non-destructive examination (NDE) Technical Centre (NETEC) explored how they could develop multi-element techniques for visual examinations, and whether they could provide operators with a more complete characterisation of an entire region of interest. Unlike eddy current or ultrasonic testing, this would not be constrained by material properties.
During a typical inspection an operator may be faced with images that include corrosion, bumps and scratches, making it complicated to distinguish each component. NETEC sought to develop a tool that would reproduce the way humans move a part around to apply different lighting to better understand the image. To replicate this type of examination engineers would need to develop a visual inspection tool that provided 3D images, rather than 2D.
In Greek mythology, Helios is the god and personification of the sun and is often characterised as all-seeing. This was the vision of the Framatome development team: how could they illuminate a surface and clearly see all of its features, making the image from a visual inspection less subject to interpretation?
The Helios technology uses a dome-shaped enclosure with a matrix of light emitting diodes (LEDs) that creates the multi-element nature of the visual inspection.
This enclosure geometry illuminates a surface at many directions and angles.
NETEC development engineers were aware of labs that had attempted to adapt structured light through a large dome, but were only successful in obtaining 2D images.
NETEC engineers built on previous research to adapt the hardware and develop the algorithms needed to characterise the surface from multiple orientations.
The Helios technology creates the 3D characterisation through a relatively simple process. The operator identifies a region of interest using live camera view and launches image capture, recording multiple images during illumination from all the LEDs, completing in less than five seconds. Next, the operator launches a characterisation routine and within another five seconds, receives the treated images and a 3D characterisation of the acquired area.
Helios can confirm characterisations like a fatigue crack, porosities, chocks or divots and even distinguish true reliefs from apparent colour on a painted surface. This capability is particularly useful to operators as painted colours can affect interpretation.
The accuracy of Helios varies according to environmental conditions, such as working distance or natural reflectance of the matter; however, on a steel sheet with surface contact, it can detect a 10μm opening crack. The 3D characterisation combined with its algorithms can achieve 20μm accuracy, with proper calibration.
As with any technology, especially NDE processes, there are limitations. For example, Helios must come into contact with the surface being inspected. The goal for the technology was to characterise regions of interest, not an entire surface, so if the region of interest is large, multiple image captures are required. Helios must be stationary during the image capture.
Addressing challenges through testing
The first Helios prototype was developed in 2015 for proof of principle testing on large parts. It demonstrated effectiveness depends on LED positioning and field of view. A smaller field of view allows operators to detect thin defects and provide an accurate 3D characterisation.
During testing, the team began with a dome roughly the size of a soccer ball; it was through this larger prototype that the team improved its understanding of the system constraints and essential variables. However, the size was impractical to deploy for inspection in a nuclear power plant. To miniaturise the tool, the NETEC team had to accommodate the lights, camera, enclosure and electrical wiring in a user-friendly prototype.
This was a major challenge. In 2017, Framatome engineers successfully developed an industrialised tool with a diameter similar to a golf ball that meets the accuracy and delivery requirements for nuclear inspections. The tool’s dome is equipped with 96 LEDs; each perfectly aligns on the centre of the surface to illuminate and capture the image. An algorithm analyses the reflectance of the surface, creating a six-coefficient polynomial equation to characterise it.
As nuclear power plants age, the industry is starting to observe wear in hard facings at points of interface for critical components. Until now, one of the alternative methods of characterising this wear required the application of dental impression media, which is difficult in an underwater nuclear environment. Helios may be deployed for this purpose.
Helios was first demonstrated as an experimental device on aeronautical components. The aerospace industry made for an effective early test environment because, like the nuclear industry, it has very rigorous regulatory requirements. But in that industry it is much easier to access the surfaces in question and the tool does not face the harsh conditions of a nuclear power plant.
Early tests sought to measure the loss of height on electron welding on a variety of pieces from aeronautical blades to tubes and flat surfaces. With appropriate calibration, Helios was able to reach 5μm accuracy (compared with a laser reference method).
Framatome now plans to demonstrate Helios in a nuclear reactor vessel by the end of 2019. Engineers will first demonstrate the technology on the hard-faced surfaces of the reactor pressure vessel radial guide blocks to augment current visual inspections and better characterisation the wear observed in these locations.
Framatome is working to expand the capabilities of this technology further to be able to inspect the insides of small tubes, analyse entire surfaces and operate effectively while in motion. As algorithms are optimised and computing power increases, the acquisition and processing times for Helios will decrease.
Automatic defect detection
Now that image processing to develop 3D characterisations has been mastered, NETEC engineers are asking how to add assisted or automated defect detection.
The next step for Helios is to integrate algorithms for automatic defect detection and characterisation, similar to what is being done with eddy current data. A first step is applying quantitative detection criteria for each of the six coefficients and using an alarm if they depart from nominal values.
In addition, Framatome engineers are building a database of images showing a variety of defects to demonstrate the feasibility of using artificial intelligence through deep learning neural networks to identify and characterise visual examination indications. The aim will be to develop robust automated detection in order to assist the operator.
The future of surface inspections will increasingly rely on 3D-image generating tools to reduce individual interpretation, which, in turn, saves resources and time while enhancing safety. The ability to visualise surface defects in 3D has a variety of applications in industries outside nuclear energy. As Helios continues to mature, it may improve operators’ inspection toolboxes in a range of industries.
Author information: Darren Wood, Global product development manager at Framatome; Clément Skopinski, Non-destructive testing surface examination engineer at Framatome