Introduction

I am Denis Linardi, Civil Engineer of Forte Engineering and Research and for me it is a pleasure to be here to introduce you to our world, to who we are and how we use AI.

So let's start here we have the outline of the presentation of today we will start with who we are and what we do, second we will see how we use AI with an explanation of the development of an AI integrated application, thirdly

we have some analysis on the impact that using AI has on our work, and finally we have some conclusions.

Who We Are: Forte Engineering and Research

So, we are Forte Engineering and Research, we are an innovative startup that was born last year and the founder is my colleague engineer Angelo Forte who can be present here today unfortunately due to some work commitments.

Background and Collaborations

Our application field is on the business of civil structures and infrastructures and we have a academic background since we both have a PhD with international experiences respectively for Angelo in China and for for me in France.

We have a lot of collaborations with different partners, like universities, to major infrastructure operators and also with important engineering firms.

What We Do: Design, Assessments, Inspections, and Monitoring

But what we do? We aim to design civil structures and infrastructures, mainly bridges and cycle pedestrian walkways.

We also do safety assessments that are currently mandatory for infrastructures like bridges, since the collapse of Morandi's bridge in Genova.

And we also perform inspections on this kind of infrastructures, like tunnels, bridges and also structures like buildings. Likewise we perform monitoring of them.

Here we have some beautiful pictures of Pirelli's Tower that is under our monitoring in Milan.

Research & Development and AI Projects

Patented Monitoring and Degradation Prediction

Finally, in our startup there is a sector that is completely dedicated to research and development. For example, we developed a patent for monitoring and provision of degradation on structures and infrastructures. We aim to have in real time the evolution of the degradation and also we would like to try to predict the collapse of some parts of the infrastructures or predict detachments of materials. materials.

AI Applications in Practice and Key Partnerships

Additionally, we are currently developing an AI application for historic masonry identification in collaboration with DDC Lazio and the University of Roma III, and we also developed an AI integrated

application for inspections that we are using for SIA Milan Airports, that is the operator of airports of Milano -Malpensa and Milano -Linate in collaboration with Diesis Informatics for the development of the software part.

AI for Infrastructure Inspections: Context

From this last topic let's link the second part of the presentation based on how we use AI with the application development that find application in the field of inspection inspection of structures and infrastructures like buildings, tunnels and bridges.

Why Traditional Inspection Methods Fall Short

But why we develop this kind of application for inspections? 1First of all, traditional inspection methods are obsolete.

Indeed, currently for performing inspection we go on site for the inspection and we take notes by handwriting on paper about the state of the art of the infrastructure. And this method is surely slow, plus during inspection phase on site we can have some uncomfortable conditions.

Just think about being under a tunnel with the bare minimum of light, or there could be some types of degradation like water infiltration or water flows that can make difficult taking notes by handwriting on paper.

What the AI Tool Is Designed to Improve

Second, we wanted a real -time inspection report compilation. Currently, the report compilation of the inspection is made the day after the on -site inspection, and sometimes also some days after.

Thirdly, we also wanted a fast identification through AI. in terms of type and location of defects and degradation. So, overall, we wanted to reduce time of the inspection phase, from the on -site inspection to the report submission.

Subsequently, we also wanted to reduce subjectivity because even though we are all well -formed professional experts, we can have different point of view respect the same defects, the same type of degradation.

And finally, we also wanted to increase safety during inspection of hard to reach areas, because in infrastructures there could be some areas that are hard to reach, even though if we use drones and in order to avoid damaging drones, we needed a solution.

How the AI Inspection Application Was Developed

But how we developed this application? The training process.

Training Setup: Inspectors, Defect Types, and Image Dataset

We were 10 professional inspectors, we integrated more than 20 types of defects and we analyzed more than 10 ,000 pictures.

Main Challenges in Training Defect Recognition

But why we needed a huge amount of pictures to be analyzed through AI? Because there were some difficulties.

First of all, one difficult hurdle to clear was to let the AI distinguish some defects that have similar characteristics from the point of view of the site. They could appear similarly, like same colors, like same characteristic dimensions. They could appear similar on site whatsoever.

Another significant challenge was to let the eye understand which parts of the pictures are not involved in the inspection process, because in a picture we can have stuff like boxes, like tubes, shadows, shadows are black, but also infiltration and water flows on the surface of the concrete is black.

So that's why we needed a huge amount of pictures in order to well train the AI and to share to the AI our experience.

Workflow Demo: AI Analysis with Expert Review

Here as a result we have a quick live demonstration as a showcase of how the application works.

The first step is the upload of the picture, second we have the analyze of the AI with the results given by the AI and finally we have the intervent of the professional expert who can confirm or modify what the AI gave to us.

The last phase is the labeling of the sentence always made by the professional expert in which the professional expert can highlight the the type of the material, in this case we have concrete, and the type of the degradation that we have. In this case, we have steel rebar exposure.

Now let's focus, yes?

Model Validation: How Reliability Is Checked

Excuse me, I have a good curiosity, so how do you validate your model? We validate it by a percentage of affidability. So after training the AI, the AI gave us some results and these results are labeled with a percentage. After that we can see if the percentage is correct or not, if the model works or not.

So it's based on our experience. So, after launch the model, we validated by seeing if and compare on what our experience think about the type of degradation that effectively we have respect to what AI gave to us. You're welcome.

So let's focus on how AI... Yes. I have a question. Your solution, I understand, was time for projects, for resource.

But my My question is this, how much time difference between the past, in this moment, with the time that's reduced? Can you repeat the question? The question is, what is the improvement on making it faster, the process of the inspection, thanks to this application, isn't it?

it okay perfect yes okay since the inspection process require required time for on -site inspection plus compilation of the reports plus submissions we can talk about this because after we are in on site we can compile immediately the report and we can submit it to the customer immediately and we will see it is out like I like in

the next slide this aspect but we we we talk about days yes thank you process to argue this and the reasoning? Thanks for your question.

As we have already seen in the demonstration, the first two steps are made by AI, but the two last steps are made by professional experts. The professional expert plays a key role

because can confirm or modify what AI gives to us. And also the labeling process in which we can specify first, the material and second, the type of degradation that we have

is always made by professional experts. So AI, it's like a support tool in order to make it faster, but the intervention of the professional expert is crucial.

Can I ask you to save your questions for the five minutes after the speech, because we need to have a 15 minute speech for the recording,

And then we will ask you if you have questions afterwards, just to conclude the recording of the speech. Thank you. Okay.

Impact of AI on Daily Work and on Customers

Now let's focus on the impact of AI on our work.

Faster, Safer, More Objective Inspections

As aforementioned, we decided to develop this application in order to get some benefits, like have a faster process of inspections, reduce subjectivity during inspection, and also to increase safety during inspection of hard -to -reach areas.

But on the other hand, we also have impact on the customer, because the customer is now able to receive the inspection documentation faster. And this is a crucial point,

because they need to receive the documentation as fast as possible in order to decide which part of the infrastructure needs the intervent earlier, in order to evaluate the cost and in order to decide which type of intervent.

And imagine being the owner of an infrastructure that is kilometers, hundred kilometers, sometimes also thousands of kilometers. This is a strategic point.

Additionally, they receive a structured, clear and objective data, so we determine an innovative system of inspection.

Human-in-the-Loop: Keeping Experts in Control

Now let's focus on the role of AI in our work, as we previously mentioned thanks to the questions. questions.

The first step is the upload of the picture. The second is the analysis of the AI and the results from the AI.

But thirdly, we have the intervention of the professional expert that can confirm or modify what the AI gave to us. And finally, the labeling process of the instance is always made by the professional expert.

1So, to sum up, we are using AI as a tool for the professional expert, but the decision making remains with the professional expert, not to the AI.

Conclusion: Responsible AI and Collaboration

So, in conclusions, our message on using AI is the follows. We all know the importance of AI.

We should use AI because we can get a lot of benefits that we all know. We can reduce human error, we can reduce reduce time, we can process a huge amount of data, but on the other hand we have responsibilities.

Our point of view on it is that decision making responsibility should always remain with the professional expert, not to the AI. We should use AI only for mechanical processes and to save time, but we would like to strongly promote collaboration between people and don't let the use of AI divide us

because for us the most powerful thing that we have is the collaboration between people. And that's all. Thanks for your attention.