Fast, Accurate BIM Classification with Kreo
Few people involved in the design and construction industry would dispute the value of the idea behind BIM modelling; one single model, with all project stakeholders working to the same standards and information.
In theory, Building Information Modeling should be making construction project scheduling , cost estimating and, tendering and execution quicker and more efficient.
But it isn’t.
Because of time constraints affecting the whole project, from the very beginning, BIM 3D models are often missing essential information that totally undermines their value.
Take quantity surveyors, cost estimators cost engineers, for example. For them, one single problem plagues their every experience with 5D BIM: not knowing if the quantities taken from a BIM model are reliable, complete and accurate.
The problem is manifold:
1. Errors in BIM element attributes (misspelled or missing names; clashes between names, family and class attributes; missing materials or specifications; geometry errors.) These mistakes make the elements 'invisible' for quantity take-off tools and software. Finding and fixing these errors is very difficult and lengthy process, making BIM-based quantity take-off impractical for most professionals. Our recent survey showed that only 27% of respondents use this feature of BIM, despite the fact all BIM software solutions provide quantity take-off feature for free.
2. Extracting measurements from the BIM model. Assuming that you accounted for all the BIM elements in the model and correctly identified them, you need to get the measurements that construction quantity surveyors need. The measurement must follow certain professional standards such as the New Rules of Measurement in the UK. This is again very difficult because most BIM models are created by architects and engineers without much regards to how they will be used by contractors and quantity surveyors. In other words, BIM authoring tools provide element-based take-off at best whereas contractors require construction quantity take-off.
3. Most BIM professionals know very little about classification systems and why they are important. According to a recent poll, only 15% of respondents knew about Uniclass. The main benefits of any classification system and Uniclass in particular is to make the data both machine and human readable and easily exchangeable between different parties and software tools. In simple terms, when you adopt a system of classification you and your software can easily find information and use it for a variety of applications (quantity takeoff, finding a particular BIM element or a group, exchanging data with other software tools.)
Right now in most BIM software tools you have to select manually from a mile-long drop-down list of likely System and Product combination to which a particular BIM element belongs. It goes without saying that most practitioners do not do the classification because i) they know very little about it, and ii) they don’t have time, resources and expertise to classify the entire BIM model.
All of the above factors severely limit the usefulness and adoption of 4D Scheduling and 5D Cost estimating by construction professionals.
A human expert evaluating a BIM model can see , at a glance, that there are a certain number of doors, walls, windows, and so on. It’s easy for us to classify elements manually by looking at them because the human brain knows the properties of these elements.
But, while computers have the ability to do this just as well, they need to have the information in a format and structure they can ‘read.’
For a computer to recognise a door, wall or window, that information needs to have been properly and comprehensively input. It needs to have been properly classified, or the computer will just assume that the element doesn’t exist. In short - a computer can only recognise a window, because a human has told it ‘here is a window.’
So, when you subsequently run a quantity takeoff using BIM software , the quantity takeoff only includes elements that it knows exist. You can’t ever rely on the accuracy of your bill of quantities unless you spend countless hours, days, weeks painstakingly, manually identifying BIM elements, extracting measurements and quantities that are not provided by BIM authoring tools such as Revit and Archicad. Even then, we have to weigh in the fact that humans, by nature, make mistakes and get distracted. The impact of human error can be significant and can ultimately lead to dramatic over or under-ordering, even when weeks are spent correcting errors or oversights in the original model.
This has all created a situation where, for just under 90% of typical construction projects, it takes longer than 1 week to create a bill of quantities. And, for over a third, it takes longer than 3 weeks. (Check out our full survey here.)
This is a huge problem that has many knock-on effects. Firstly, it increases the time and cost of bid preparation. Bidding costs is the single most important factor behind low gross and pre-tax profit margins.
Secondly, it restricts us from being able to tweak and experiment to optimise projects during the bidding process. We’re so busy manually preparing Bills of quantities and cost estimates that actually tweaking projects, making expert recommendations, is out of the question - not least because such changes undermine the work you’ve just spent 3 weeks putting together, and send you back to square one.
Making a change with AI
That all sounds pretty gloomy, but we believe the solution is at hand.
With KREO, you can automatically generate an accurate bill of quantities in a matter of minutes instead of weeks. How do we know it’s accurate? Because we use a range AI algorithms to teach and train KREO to ‘read’ and interpret the data regardless of errors or missing information, analyse the geometry and topology of BIM model elements to better identify what they are. You can be sure that all BIM elements are accounted for in each model, with classification to Uniclass or Masterformat standards.
You’ll only be asked to weigh in on the classification of an element when KREO’s confidence level is below a certain threshold and, when you do this, KREO will remember the answer you gave, and use that data to retrain its learning models. This is machine learning in action. Machine learning is a field of computer science which allows computers to 'learn' with data without being explicitly programmed.
The data used and stored by KREO, across all users on our platform, will combine to provide the most comprehensive machine learning framework ever seen in construction scheduling, cost estimating and bid pricing. The more quantity takeoffs are performed on our platform, the more intuitive the platform will become, and the less manual input will be required from human experts like you.
KREO also performs a BIM model audit to identify issues affecting constructability and measurements, helping you pick out any errors and improve the accuracy of the estimates further. The purpose of the audit is not to perform a standard clash detection between design disciplines (architecture, structural and MEP). Rather, it is to find errors in the model that distort the quantity take-off process. Below is an illustration of one of the errors that affect both the constructability and quantity take-off.
With Kreo you can spend a lot less time on looking of information and manual data inputs and focus on value-adding activities - improving cost budgets and project schedules, cost planning and value engineering. This tremendous boost in productivity will help reduce bidding costs and increase the number of bid that you can participate in.