- Composite Inspection and AI-Assisted Tool Monitoring
Back in 2020, TWI joined the collaborative ATTIC (Automated TeraherTz Imaging of Composites and tooling profiling) project, to develop a fully automated drilling, monitoring and inspection solution for glass fibre composites used in the aerospace industry.
The low weight of composite materials has seen their use increase in the aerospace industry but, while composite manufacturing can reduce the need for joining, mechanical fastening is still widely used when joins are required.
Mechanical fastening relies on holes being drilled in the parent material, which can lead to defects. Drilling into fibre reinforced composites is made more complex because of their anisotropic, inhomogeneous, and highly abrasive characteristics. The abrasive nature can degrade tools quickly, making it important to be able to detect and monitor tool wear, as well as ensuring the integrity of the joints themselves.
This project sought to address these issues through the use of high resolution laser profilers to measure the level of wear between operations, along with the use of machine learning techniques to optimise tool processing parameters such as drilling speeds. The monitoring was automated as part of the process, reducing the need for operator interpretation.
Once the holes had been drilled, terahertz imaging was used to inspect the parent material around the holes. This process uses electromagnetic radiation with a wavelength in the range of hundreds of microns, to accurately inspect the thick glass fibre composites. The project investigated the efficiency of this technique to detect delaminations, a defect often found when drilling composite parts.
To deploy the developed solutions, our experts integrated two sensors onto a robotic platform so that robotic feedback and sensor data could be combined to create a first of a kind robotic terahertz inspection system (Figure 1). Furthermore, different methods of testing GFRP using terahertz technology were trialled, before the creation of a real-life demonstrator.
- Artificial Intelligence for NDT Scanning of Unknown Geometries using Collaborative Robots
Robotic NDT has traditionally been limited by the need for fixed and portable robots to use pre-planned or fixed paths. Pre-planned robotic paths need to be designed and programmed ahead of use, which takes time, expertise, and increases costs.
To address this issue, speed up inspection times, and remove the need for pre-planned robotic paths, we created a core research project using artificial intelligence to create a cobot-deployed system that allows for a probe to be placed on a component’s surface, where it scans a weld using ultrasonic feedback to ascertain the weld geometry. The system was designed to maintain contact with the component surface as the probe travels across the surface, while adjusting offset and skew based on the ultrasonic feedback.
Not only does this solution speed up inspection times, but it also removes the requirement for a human operator to be placed in potentially dangerous conditions during the inspection. The only time that a human would be closely involved in the inspection would be when they placed the probe and identified the weld geometry to be followed.
Experts at TWI created a prototype system and tested it on a range of different part dimensions and types, demonstrating the viability of the solution (Figures 2-8).
- Automated Underwater Ship Weld Inspection System
Welds made in the hulls of ships create possible structural weaknesses due to any flaws that may be present. With in-service ships being reliant on visual inspection and no systems or procedures in place to inspect flaws below the surface, there was an increased potential for flaws to be missed until it was too late.
We were contacted by The Marine Materials Joining Innovation Centre (MAJIC) in Malaysia to develop the first-ever automated underwater phased array inspection system to fill this important marine industry gap. The solution needed to be capable of inspecting ship hulls with a thickness of 8-20mm under up to 20 metres of water. The system also needed to navigate the side of a ship hull via remote control, find and follow the path of a weld, and provide full weld volume inspection with an offline data analysis capability.
An inspection vehicle was designed with multiple systems, including a 3-probe, 128-element phased array instrument, with 2 probes used for inspecting the weld and 1 probe as a weld follower; laser and camera guidance systems; sonar and depth sensor-based tracking systems; and an inclinometer for posture monitoring.
The system uses an advanced multi-probe phased array technique to provide full volumetric inspection for underwater welds in the hull of a ship, and the vehicle body is made from a marinised aluminium material to minimise corrosion.
Despite the advanced inspection systems, the solution was designed to be easy to use, including global and local guidance systems to assist the operator in navigating the vehicle underwater.
The system was successfully tested both at our own laboratories and also at the Malaysian Institute of Marine Engineering Technology (UniKL-MIMET), where it was able to track a weld with the guide system, acquiring the data for analysis as the location system tracked the vehicle’s path and displayed its posture (Figures 9-10).
- Early Fault Detection for Industrial Machinery
AI was also integral to the UK-funded ‘Digital Monitoring of Ships (DiMOS)’ project which sought to provide predictive solutions for ship-based systems. Our experts joined a consortium of project partners, where we led the development of the application of artificial intelligence-based models to identify and locate developing faults within industrial machinery such as motors, fans, and pumps at case study sites.
The aim was to be able to detect and locate a fault at an early stage, as it developed, using real-time vibration sensor data. AI models were incorporated into the solution to assist with analysis and interpretation of new measurements as they were recorded, alleviating the time-consuming manual interpretation of the vibration data.
The AI system was trained by monitoring healthy machine operating data before detecting changes in the measurements that could be indicate a developing fault. This created a qualitative indication of the operating condition of the asset, although it also took account of risk levels and likely failure modes, meaning that the system not only locates the fault, but also provides a qualitative indication of the fault severity as the condition deteriorates.
The results showed that the system was capable of detecting faults up to three days before asset failure, allowing maintenance to be undertaken in a timely manner, saving on unnecessary maintenance costs and preventing asset damage. The results were verified on a controlled experimental set up with real vibration data and against real data from industry sites across the UK.
Find out More
Our work using artificial intelligence with NDT has already delivered benefits for our Industrial Members. We have used AI to improve predictive maintenance, which has been shown to reduce maintenance costs by 25% and reduce asset breakdowns by 70%. The ensuing costs of a failing asset can be considerable, with some automotive assembly line disruptions costing as much as £10,000 per minute – not to mention safety and reputational implications of an unexpected failure. Our work with AI has also opened up new inspection solutions that have been shown to reduce costs, allow for manufacturing tool monitoring to optimise processes, and improve the safety of human / robot interactions in industrial settings.
AI can optimise NDT inspection procedures - helping speed up processes to lower costs, deliver inspection where it is difficult or dangerous to send humans, and reduce the potential for human error during inspection.
Contact us by emailing contactus@twi.co.uk to discuss how we could help you use AI to solve your NDT challenges and improve your existing processes.