Intelligent inspections — Exploring the world’s first 3D AI recognition tool for concrete structures
Who would have thought that Artificial Intelligence (AI) and drone technology would help humans save millions, if not billions, in building inspection costs, not to mention preventing human injury and improving a myriad of safety, security, and sustainability challenges? Kiwa Sweden and tech company Spotscale are undertaking a partnership that uses AI technology to make these improvements in the most innovative ways.
Kiwa offers real-time, AI-assisted digital inspection services for Intelligent Asset Health Management (AHM). In light of digitalization, we know that Industry 4.0 is becoming a topic well-known to customers looking into efficient asset management leveraged by these new technologies. That’s why we have developed and tailored solutions that slot into digital workflows, including AI, to increase quality and better utilize and visualize data. Kiwa’s AI-driven intelligent inspection of concrete is an example of one of our services.
What the eye cannot see
The Kiwa-Spotscale partnership specifically focuses on the 3D visualization of defects in cement structures, for example, cracks, spalling, corrosion, etc., as documented by drones and drawing upon AI. This is currently the world’s most advanced technology that can map out structures in terms of size or scale and resolution — zooming into tiny details that the eye simply cannot see.
One can also filter down and look at specific-sized cracks in the visualizations. Autonomous recurring inspections utilize AI to trace and monitor the damage or growth of a large number of defects, allowing specialists to predict future structural degradation. This is a really powerful tool for seeing these problem spots in concrete. A further development is to map the inspection findings directly into structural analysis software, allowing for a complete digital workflow to analyze the remaining lifetime of structures.
Power in numbers
According to Milan Poznic, a specialist in non-destructive testing at Kiwa, “You wouldn’t believe how many inspections are still being done manually. I come from a robotics inspection background, and when I saw that this work was still being done manually, I thought, ‘There’s got to be an easier, more efficient way for all involved.’ I have an extreme passion for exploring the unknown and finding answers to questions, so I began my search. My number one drive has always been to do something good for this planet, exploring new technology to make this place a little bit better. So I interviewed our concrete inspectors to determine the target for those kinds of inspections and what inspectors are looking for exactly. And then, in my research, I came across Spotscale; they are unique because they’re super good at processing large amounts of data: image processing and 3D reconstruction. And combined with Kiwa’s expertise in developing inspection systems and concrete structural health analysis, we’ve created a great new solution together. We elevate each other’s strong points in a product that has at least tenfold improvement in inspection efficiency compared to manual inspections.”
Spotscale develops the actual algorithms concerning AI and 3D modeling and the respective software, trained on data from Kiwa, whereas Kiwa ensures the entire solution. Kiwa develops the inspection procedures, does the inspections, has the know-how of the concrete structures, and has the customer base to bring the technology to the market.
An advantage of a digitized workflow is that information becomes accessible to surrounding systems and, thus, creates digital processes where information exchange is essential. Information is further traceable and clearly linked to a time and georeference. New information can easily be added to the model through renewed inspections, which applies not only to surface documentation. Testing with volumetric methods such as ultrasound and radar can also be linked to the model.
Ludvig Emgård, Founder & CEO of Spotscale, was, in fact, a teenager when he first became intrigued by 3D mapping of the built world: “It started with my fascination of how the ancient world evolved historically. I wanted to explore historical maps where the city grows from a few buildings forming a village to a larger, thriving locale. All that anthologizing, the history — it’s how I eventually came to start Spotscale. It led to much research over the years. During this time, I noticed that what is mostly being done in mapping is a once-off representation of the built world, using a plane, car, or human to document the world through images as it is at that very moment. Yet, in this approach, a part of the world is often lost when mapping. This is the area or transition between the satellite and street levels — from the clouds down to the window frame level, where you can see the cracks. I wondered whether we could solve this with technology, this transition between a large-scale image and a very small-scale or more detailed level with drones. So now, many years later, it’s no longer a question of whether maps would become 3D and that we will have sensors and various equipment measuring the world to create a map that is zoomable down to the last detail.”
Spotscale can calculate geometry from mapped images, which is done with the algorithms they have developed in-house. This helps us understand the shape of an object, and when we have the shape, the imagery can be ‘glued’ onto it in a very high resolution. It’s a lot of data processing, but at the end of the day, you have the resolution of the entire object. And for computing purposes, a unique parallelized approach is used where we separate these ‘object parts’ and process them on separate computers. Then it’s all tied together in one piece, enabling the level of detailing and zoomability. So, in the end, there are several layers of detail in the 3D visualization or map, which can be used to find tiny cracks in concrete anywhere on the object. You can ‘switch’ on and off the cracks or ‘toggle’ between visualization layers.
Where sustainability meets efficiency
AI helps to reduce the time needed; if a human had to do these inspections, it would take much longer, and the level of accuracy may be lower. Of course, a human — a specialist — must do the quality assurance and checks regarding the AI results. This person knows how to work with data and technology and can interpret what a specific crack could mean for the stability of the structure. The technology means fewer physical risks for specialists, as drones take photos. There are huge sustainability benefits; with this knowledge, buildings, bridges, and tunnels can be better maintained, leading to a longer lifespan. There is also a cost-saving element in the overall maintenance of structures.
Digital inspections bring greater reliability
What are the benefits of digitalized concrete inspection if you are responsible for operation and maintenance or saftey relating to concrete structures? The table below highlights some of the significant differences and gains compared to a manual inspection.
Where are we headed?
This is just the beginning of AI technology to map the built world. The frequency of image capturing has become faster, meaning we can see such visualizations in record time. The concept of the true ‘digital twin’ is within reach — what is seen in real life is reflected in real-time in a 3D map with almost exact accuracy.
According to Emgård, “In the future, we’ll have drones living in a nest on the roof, and then they will charge themselves, just like a lawnmower; one day, they will fly to one building and scan that building and then go back to their nest. The next day, they will scan other buildings and so on. And when they’re done, they will start again as the world changes in the interim. So that’s what we can expect. Twenty years ago, this was a very futuristic thought. But now, most of the technology to do this is available. So it’s more about the regulations and infrastructure hindering this development. There is the ongoing development of autonomous traffic management systems at every national civil aviation administration to enable better coordination between drones and manned aviation.”
“We intend to scale this technology to more Kiwa countries,” says Poznic. “My idea is that all inspection methods should be digitized, and machine learning and AI could reduce the travel time of experts traveling to each site, making things much more efficient for all involved. In this way, we can use local resources more effectively, like local drone pilots who can be trained and qualified in data collection for Intelligent Inspections by Kiwa. This would mean we could also use experts independent of location. So if you have a small drone company with the hardware we specify, for example, Kiwa can educate you to fly according to our procedures to collect the data. Once trained by us, we’ll give the drone pilot a certification, and they can begin collecting data, which is then uploaded to the cloud and processed by Spotscale. The inspection results are then returned to a 3D copy of the object in question, allowing the customer to visualize the defects on top of it. Additionally, we intend to have various levels of service we can provide, for example, a basic 3D map to something more elaborate that predicts the structure's future if not maintained.”