News and events

Finding Faults in Aircraft Wings and Human Knees

Professor Mike Holmes, Head of the Graduate Research School, and Alison Naylor went to meet with Professor Lik-Kwan Shark to talk about the work currently going on in the Applied Digital Signal and Image Processing Research Centre & the Centre for Advanced Digital Manufacturing Technology.

Can you give me a brief outline of your area of research?

My area of work is very much cross-disciplinary; mainly applying signal and image processing for different disciplines, sectors and domains. Lancashire is one of the most important high value manufacturing regions in the UK and therefore we are working on signal and image processing with manufacturing industry. In addition, we have a number of leading hospitals in the area – Christies, Blackpool and Preston hospitals – with whom we work on new medical diagnostic methods, derived from signal and image processing. Essentially the problems of signal and image processing are the same in both areas because it is all data. These two areas are our main source of funding.

It is data from a variety of different sources and then processing it to extract information about that system or process?

It could be processing, or it could be the visualisation of the data and then making intelligent decisions based on the data.

Lik-Kwan Shark

The other area is more on the interactive and immersive digital environment using our 3-D visualisation cave display technology. This is particularly useful for environmental monitoring, and remote sensing, because they are all dealing with huge data sets that require processing.

Can you illustrate this by some examples; maybe from the industrial engineering side?

On the manufacturing side, my research has a strong focus on ‘non-destructive testing’, which means evaluating and analysing a structure's integrity without destroying it in the process. You cannot cut through a component to see if it has any holes because it will no longer function as a serviceable component. The whole idea is to “see” inside the structure without destroying it so that it is still usable after testing. This is important in high value manufacturing where the parts are safety critical. For example in industries such as aerospace, submarines, nuclear, or even in the food industry. It is to do with the quality control of the product that will go on to be used.

If you take, for example, an aircraft frame where you want to see whether there are any cracks in the frame; would you be using X-rays, ultrasonics or combination?

The aircraft frame is a primary structure of an aircraft and its integrity is critical! Typically, ultrasonics, X-rays and a variety of inspection methods are used in order to detect defects. Obviously, such a structure is huge. You need to be able to examine it in a rapid manner otherwise, the manufacturing process will be slowed down. For example you make a wing and it takes you maybe a week, to go through each pixel of the ultrasound image or each pixel of the X-ray image to say whether it is sound or not. That is too long, because by this time another aircraft wing will be in line and it might have the same manufacturing error. You not only need to know the quality of the finished product but you also want to feedback to the production line so you can refine of the process and make the next one better.

Are you interested then in taking a wing and applying algorithms that can identify things like cracks or defects in the frame, so you don’t have to search through pixel by pixel? The algorithm is doing the analysis for you.

That is correct. This is automatic analysis.

The other thing is to do what we call data fusion; mixing data from several different techniques or modalities. For example we might acquire an X-ray and a ultrasound image of a component, apply our algorithm to overlay the two images and to compare the two images and from that you could identify more easily and quickly a defect in the structure.

Another technology we use is called shearography. When you examine a component there are two ways you can examine it. The “unloaded situation” means that you take an image of the component but it is not under load or stress. Shearography puts the component under load or stress. An optical interference pattern is then taken of the surface deformation of the component.

So you are looking for surface rather than volume defects?

No, it can pick up sub-surface defects as well, because some materials such as composite materials have known elasticities and it is possible to identify sub-surface defects. We can also look at the temperature distribution on the surface of the component. Regions containing a defect will have non-uniform distributions and can be detected.

You can essentially overlay data from a variety of modalities and use your algorithm to identify areas where there are potential failures or problems?

Yes. We can also use CAD (Computer Aided Design) data. CAD data is another modality because CAD data also tells us what the ideal component should be like. The whole idea is that different data sets can be superimposed on top of each other to reveal defects.

So somebody presses a button and a lot of analysis and comparison happens automatically and then you are looking at an image of an aircraft wing and it could be highlighting where there is something wrong?

Yes. That is the whole idea. The inspection process is speeded up, so manufacturers can have a higher throughput and identify if there are problems in the component manufacture. This also reduces environmental impact by eliminating scraps.

Moving on to your work in medical research, how does signal and image processing improve medical treatment?

Most people think that hospitals and aircraft manufacturing are two totally different businesses but from my point of view they have a lot of similarities. One is looking at components in aircraft and the other is looking at components of the peoples' bodies. They are both trying to identify structural problems. Some of the equipment used is very similar, in terms of modality; manufacturers and hospitals both use ultrasound and X-rays. For us dealing with these two sectors together whilst unique, results in cross fertilisation, not just in signal and image processing but even in inspection methods, the processes involved, and the regulatory aspects.

We have applied our techniques to the monitoring of human knee joints. The idea came from a technique used generally in the engineering industry, called acoustic emission. Acoustic emission uses a microphone-like sensor attached to the structure you want to investigate and when the structure is loaded it will generate sound. The sound generated is in the ultrasound range. In the engineering sector it has been used for monitoring structures such as bridges, water tanks, buildings, aircraft engines, etc. We are trying to translate this particular technology for use in humans. The idea is that you mount the same sensor on a joint and, when the joint is loaded by moving, sounds are generated. We did some trials comparing the healthy knee joints of students with those of OA (osteoarthritis) patients. We observed the differences between a very young knee and a very old OA knee based on the amount of sound emitted. We can now track the deterioration of knee health with age.

If you know knees are degenerating then treatment can start at an earlier stage.

I want to ask you about the Centre for Advanced Digital Manufacturing Technology – what is happening there and how is your expertise feeding into that?

Our advanced manufacturing research focuses on data-driven digital manufacturing execution. By working with the local manufacturing sector, we have established the world first Tele-immersive Digital Manufacturing platform at the Burnley campus. It consists of reconfigurable plug-and-play work cells linked by conveyer belts, it is equipped with various digital tracking and sensing technologies to provide multiple data threads such as process performance, product deviation and energy consumption, and it is supported by Cisco industrial network to allow remote control and monitoring. It represents our vision for the affordable factory of the future.

Tele-immersive digital manufacturing execution stems from my research in applying signal and image processing to aerospace and medical sectors and development of immersive and interactive digital environments. It provides a new and exciting arena for us to explore the power of data processing by dealing with much more data modalities and much larger data sets.