MEP Coordination and how AI is imposing itself

When I had the privilege of participating in the BIM Forum Congress during the first days of November, a topic that arose at the table was how AI is venturing into the BIM Methodology, especially in MEP.
This is how there are currently several initiatives and comments on the web, I must tell you how the algorithms that allow this new technology to help us generate the efficiency that this development lacks. Already at the end of the 70s, a competition called MICROMOUSE began to be held. This consists of getting a mini-robot to solve a maze in the shortest possible time. The space is composed of a 16×16 grid, with 180 mm squares and 50 mm high walls. The mini-robot must be completely autonomous. And during the competition it cannot be programmed. Starting in countries such as Japan, Korea, the United Kingdom, etc., it has been developed without interruption every year. In its 32 version, the dimension was reduced to a 32×32 grid and 2.5 cm walls. With micromouse also reduced. Being the current time record, of only 3,921 seconds.
There are several methods to travel from the start to the finish. Since there are three attempts, the first two are exploratory, generating a score for each square of the matrix. In this way, in the third option it will be followed by the intersections and best expected directions. Thus eliminating travelling a route twice or reaching dead ends. This concept, used in an architectural project, incorporates two more directions, up and down, however it has a lot of similarity with the mouse race. Mathieu Josserand has a very good article about it explaining almost to the smallest detail, how he developed algorithms to work on REVIT models. Indeed, it also generates a matrix that is analysed by AI, and that makes it learn to generate the most efficient routes.
Another example that I can show is the Aumenta Site that is currently in pilot phase for MEP Electrico. The form of this design engine is very similar, however; the process is in parallel recursively, repeating a process several times in the same plant simultaneously.
In this way the tool manages in a short time to find the shortest and cheapest route to build.
But the system, having such an early development, still fails. For this reason, AI must somehow more information and learn from different cases. Augmenta in fact has a link to join the pilot phase, contributing their own projects, to gain experience.
Magic CAD
So much waiting for there to be a friendly robot that I design for you, as if it were a Robotiza of the Jetsons….. Well, it’s still a dream. But this CAD MEP help system is also exploring it by helping with our designs that are hosted in the cloud, so that your AI engine learns what is right, what is not and how to make it more efficient.
And in reality this is how any neural system operates: learning. AI today can be defined as a child of about 2 years old. The one who is still learning to walk. Give some, but not with the accuracy that is required.
Text from its AI Enrolment site: it says: “The recent successes of Artificial Intelligence (AI), Machine Learning (ML) and Deep Learning (DL) in various applications, along with the development of current tools, such as cloud computing, allow the automation of cumbersome, slow and inconvenient jobs, much easier, cheaper, more manageable and faster. At MagiCAD, our primary goal is to make MEP design faster, more efficient and more accurate by introducing automation with AI technologies. Therefore, we are building new tools for designers who use state-of-the-art methods to improve MEP design and modelling step by step.”
Conclusion
Far from stopping here, we must expect developers to make many improvements in a technology that is specifically starting from scratch.
Publishing an article today on the progress of AI in BIM, it may require at least one more year.
But what is in the making has enormous potential, which will improve development times in a huge way. And if between BIM and the traditional design method, there was 45% optimisation over time. Between BIM and AI, between 60% and 70% savings are expected.
I suggest periodically reviewing development sites such as Aumenta or “INEX Space Manager”, a set of tools that allow the automated management of models developed in Revit. These are difficult to configure and maintain and implement, either because of the human capital available to use it or because of the lack of information about BIM and AI. Therefore, they are often underutilised, but are indispensable elements for the development of more advanced processes.
For me, there are many myths about AI; that will dominate the human race, or that will leave many without work. There are even rumours that GPT manipulates us. The truth is that as a result of doubt, ignorance and little information, all this is perhaps possible. But it also opens a door to spend more time on the important things in life, instead of dedicating weeks on a coordinated MEP layout.