Beyond BIM: Data Driven Construction_Part 2_Structures and Tools

Following on from part 1, this time it is time to talk about the breakdown structures needed for project management, what tools are used for this, how we are able to use the generated data itself to our advantage as I wrote in the first article, and what tools we can use for this.

Project Brakedown Structures (XBS)

A breakdown structure is the decomposition of a whole, organized by levels, which defines the project as a whole.

As we will see below there are several types of structures in a project, the most commonly known is the WBS, which serve to decompose the work to be done in a project, but the structures can be several in a project. From the decomposition of scope, organizational, documents, requirements or information models.

The key for several structures to co-exist in a project is that all the generated structures are mapped between them, which means that one-to-many or many-to-one relationships are established between all the elements of all the structures.

"Each structure has its purpose and all must be interconnected with each other."

Project Management Tools

To be able to create and manage all the structures of a project, there are many digital tools that allow us to do so, System Engineering softwares. An example is the case of Relatics, which is a Project Management tool (Made in Netherlands) that allows not only to keep track of the structures and their relationships, but also to base all the project information on it, requirements, risks, design changes, etc.

It allows traceability and identification of any document, object, element, item in any of the project platforms. A customizable and flexible tool that can be adapted to all types of projects.

A tool of this type is the integration center of a project, mainly for its management, but it is not the only one used, in fact, there is usually an Ecosystem of software that allows it. From a document manager such as Aconex or BIM 360, planning tools such as Primavera P6 or MS Project, or geographic information tools such as ArcGis.

CDE (Common Data Environment)

My particular vision of the CDE is very different from that of 90% of the professionals in the sector or even from how ISO 19650 defines it. According to ISO, a CDE is "A large collaborative space with standards so that we can understand each other and where each of the assets in the process can generate information containers of any type". Which people interpret it as the BIM 360 or the ThinkProject of the day.

My interpretation goes a step further, since a CDE is a set of tools/platforms of a project that contain the project data, and that somehow (which I explain later) we are able to guarantee the unique and unambiguous data. This is known as "Single Source of Truth".

These data environments have been evolving in recent years, and with the technology we have today we are able to connect all of them in a fast and reliable way.

In the beginning we had "Stand Alone" documents, which we stored on our computers and shared with others via email and/or physically. With the risks involved.

In the second step, we moved on to cloud applications that allowed several people in the same project not only to access the same information but also to work on them simultaneously.

A next evolution was the integration of applications, this allowed data to be transferred in real time from one application to another without human intervention, but normally these relationships were 1 to 1, what happened when we had more than one platform?

In the end we arrived at a tremendously complex environment, with a multitude of platforms, which are necessary to connect all of them, and that is why it is necessary to develop a CDE, which allows to store all the generated data.

We already have the information stored.... and now what?

Let's use that data to our advantage. Decision-making in a problem-solving process can be optimized through technology. The success of decisions is based on the quality of the information we have at the time of making them.

What information do we have in a project? With the theoretical knowledge of the professionals, their experience, and of course with the data generated by the project itself (or in the case of knowledge managers of companies with data from previous projects).

And how do we extract that information and make it available to the decision makers in the project?

There are many tools for extracting data from different sources simultaneously, one that I especially like is Frontir.ai from Digital Construction, a Dutch company dedicated to technology in the construction industry. Frontir allows from its platform, to create "Pipelines" to different platforms and in real time extract the data stored in revit files, documents, bbdd and many more, to later work on them, through "querys", reports or graphics.

First of all, the data is regularly extracted from the different platforms to be stored in a Data Lake in the cloud, with something fundamental, the metadata of the structures that I mentioned in the first section to be able to maintain the link between data at all times.

There is no alternative text for this image

Internet source

Finally the information is structured, to finally model it in the Data Warehouse where the Business Intelligence software is able to graph the data in a fully customizable way.

Correlation and Prevalence Matrix

It is necessary to make a data correlation matrix, which serves to define the degree of joint and linear variation of two random variables. It is the normalized covariance in the range and, in addition, it is a way to ignore the variation of each of the variables itself to focus only on the relationship that exists between the two. This allows for a more optimized analysis of the data. It allows us to discard a lot of unnecessary data, and to reduce the degree of uncertainty in the veracity of the data.

The fact that two variables are correlated does not imply that one is the cause of the other, that is, that there is a direct relationship between the two. Correlation does not imply causation.

Once the data has been extracted, there is a fundamental intermediate step, without which the "single source of ruth" guarantee would not be possible. This is the data validation process. In other words, what happens if, after collecting data from two different sources, the value of the same data is different?

This is a matrix in which the types of metadata are listed vertically, and horizontally the source from which they come, and the prevalence of the source is established, and in the event that two data are different, the system returns to the original platform the correction of the data and modifies it.

With an Artificial Intelligence system, it is possible that the system does not require verification of the data in the first instance, but by simple learning it is possible to fill in blank or erroneous data.

Data Visualization - Reporting

There are many tools on the market for the graphical visualization of data. The more than well-known Dashboards, which help to view the data in a descriptive and diagnostic way. In other words, to visualize what happened and to understand, through trends and decision trees, why it happened. In other words "Reporting".

There is one tool that I particularly like, because of its ease of understanding in the interpretation of data, its flexibility and its potential. It is Weaver. https://www.wvr.io/

Each node represents a category of data, and in parentheses represents how many elements are contained in that node. Dynamically, the system establishes the existing relationships between the different nodes. As it is dynamic, if we click on one of the nodes, the element expands and represents what it contains.

Once all the databases are linked, the platform allows, from extracting reports, filtered, or system questions to the system, for data validation. It is a tool with a brutal potential, to perform "Sanity Checks" of the data prior to its analysis, and of course for the analysis itself.

Data Analysis

To talk about Data Driven it is necessary to go beyond "Reporting" and be able to do "Data Analysis" both past and future, with predictive models or even simulations and optimizations for the future. Once the data has been mined, it has to be analyzed if you intend to profit from it. The analysis is done through "questions" that are asked to the data, to know how to interpret them, to detect the causality of the facts.

There are many tools used in the market for data analysis, such as Hadoop, elasticsearch or Apache Spark. There are different types depending on:

1. Open source big data tools. They offer basic infrastructure, servers and storage
2. Big data platforms. This is a higher layer, where all applications with more advanced functionalities are included.
3. Specific vertical applications. These are the ones that focus on the specific needs of each industry, offering collusions that make companies more competitive.

In addition to this classification, big data tools can be differentiated according to the purpose for which they have been designed. Thus, we could differentiate between the following:

• Storage tools.
• Management tools.
• Visualization tools.
• Mining tools.
• Quality tools.
• Analysis tools.
• Security tools.

As I have already mentioned, the effectiveness of big data in an analysis process depends on the accuracy, consistency, relevance, integrity, completeness and updating of the data used. In the absence of any of these factors, data analysis is no longer reliable.

It is not enough to have the best tools; the people responsible for data mining and analysis must use their creativity to identify the underlying question and find answers with low uncertainty.

Data analysis is a very complete discipline, but it only delivers the expected results when guided by the right hands and applied to the right problems.

To close an example of how you can work on a mega project with so many tools/platforms and maintain data traceability thanks to CDE, we see it in the video below. A9 BaHo project in Amsterdam in which I have been immersed for 3 years.

To be continued...Part 3 The Origin of the Data