Abstract

Data in the construction industry is heterogeneous, organizations do not work closely together, and construction software is highly specialized for individual users and applications. As a result, knowledge from previous construction projects is often not shared, linked, or transferred to subsequent projects. Additionally, the manual work on-site leads to long and unstable design and construction processes. Based on a review of common challenges in this work a new vision for the Architecture, Engineering, and Construction (AEC) Industry is developed. A platform thinking approach with methods of artificial intelligence (AI) for data preparation and for construction applications can benefit existing companies and support the overall ecosystem with innovative as well as disruptive business models. Additionally, a new ecosystem can emerge. This article shows how artificial intelligence can be established in the AEC Industry. The proposed approach suits applications in the whole value chain of design and construction. The implementation of a platform thinking approach throughout the industry is still missing, but its implementation in parts already shows great benefits. In a current research project, a platform ecosystem will be established and used to implement a number of prototypical applications. Using the proposed approach, construction data can be structured and linked with data from other organizations while simultaneously ensuring legal and technical security for the users. Ultimately, the resulting database will enable various applications within the industry.

1. Introduction

1.1 Characteristics and current challenges in construction

The German construction industry is following a trend of below-average productivity growth with approximately 1,5% until 2020. To overcome this the industry is embracing digitization in pursuit of an intelligent “Construction Site 4.0” similar to Industry 4.0. Digital planning methods such as Building Information Modelling (BIM) already exist and mark a step in the right direction. However, only a small proportion of companies make use of it, and a lack of knowledge often undermines the possible gains. This endangers their competitiveness by missing out on potential new business opportunities. (Baumanns, Freber, Schober, & Kirchner, 2016)

Some of the most significant challenges that the AEC industry faces are in the areas of communication, document management, and interoperability. The reasons for this are an existing tendency to develop many specific separate standards for individual tasks, e.g. only 5-7% rely on the BIM-method. This leads to a decentralized heterogeneous data environment, with a high amount of created data sets that are very hard to access and use. (Froese, Han, & Alldritt, 2007), (Fulford & Standing, 2014), (Roland Berger, 2016), (BauInfoConsult, 2016). In comparison, a recent survey in the UK showed that BIM is already used in19% of construction projects (USP Marketing Consultancy, 2019).

This situation is further strengthened by the complexity of the AEC industry itself, with its many phases and partners from different disciplines (Shen, et al., 2009). The high fragmentation and the demand for new IT-solutions of the AEC industry is especially challenging for small and medium-sized enterprises (SMEs) which make up 99,9% of all German construction companies (Die deutsche Bauindustrie, Gallaher, O’Connor, Dettbarn, & Gilday, 2004, Shen et al., 2009). These companies currently cannot compete with large enterprises in terms of digitization and continue to fall behind  (Tata Consultancy Services (TCS) und Bitkom Research, 2018). Currently, most of the decisions made in German companies are based on gut feel and personal experience and are not data-driven. Only one in ten companies across all industries in 2018 utilized AI to support their decision-making and only 2% base their decisions entirely on AI. (Sopra Steria Consulting, 2018)

There is a need for making data sets from various sources interoperable and interpretable, as well as a web-based collaboration and software solution that can cover entire project lifecycles (Fulford & Standing, 2014).

1.2 Contribution

The highly fragmented industry and the heterogeneous data structures lead to several research questions related to an AI based platform approach.

As stated before, data in the AEC sector is highly heterogeneous and originates from several isolated sources.  Furthermore the AEC industry is fragmented, mostly consisting of SMEs highly depending on each other, yet collaboration is recognized as a significant issue by various sources (see Chapter 1.1).  As mentioned in Chapter 1.1, processes in the AEC industry are dynamic, reliant on external variables and decisions are commonly based on feelings and personal experience instead of data.

The following questions arise from the status quo which is described above:

• How can diverse and heterogeneous data sources be transformed to a machine-interpretable structure?
• How can participants in the AEC industry better collaborate with each other?
• How can the business processes be optimized by data based collaboration? How can this support the implementation of an ecosystem to facilitate new business models?

As a possible solution for the mentioned challenges, this paper is proposing a new AI-based platform approach. Data and information shall be processed and enriched by AI, people and organizations shall collaborate on a web-based platform, and AI-based applications shall empower SMEs by using processed and enriched data.

2. Methodology

A review of the current situation and challenges of the AEC industry was done based on various academic and practical sources. Academic papers where found by searching for trends and opportunities in the AEC industry. By evaluating these and backtracking their sources, a general impression of the worldwide situation of the AEC industry but mostly in North America and Germany was developed. To validate this impression with an emphasis on the German industry, several practical sources were evaluated. These include various surveys in which members of the German AEC industry answered to the state of digitization, use of software and upcoming challenges and were mostly found on statista.com. Other practical sources include publications by consulting company’s evaluation of the current situation, trends, and future needs of the industry.

Overall academic publications were used to generate an overview of the current situation from an academic perspective and were then validated using practical sources like surveys and evaluations by consulting companies.

Based on the identified challenges, possible solutions and their value were derived on a theoretical basis.

3.The idea of an AI construction platform

3.1 Requirements for an AI construction platform

Data management is a fundamental challenge in the AEC industry. Due to a strong, so-called, project-based thinking, conducted projects are rarely documented with the future reuse of structured data in mind. Projects are typically recorded on varying levels of detail, in different formats and utilize different terminology. Therefore, the capability to analyze unstructured data is essential to make the majority of data sources usable.

To secure company-internal process data, the topic of data security is highly relevant. Thus, the resulting platform must ensure a separation of raw data from the developed algorithms. Only the trained algorithms are available in the platform in the form of metadata. In this way, companies can learn from each other without having to release internal data sets.

Due to the low digitization of SMEs, the aim is to provide them easy access to the platform and its applications. Here it is important to design the interactions between man and machine in a way that the effort for the structured digitization of data is reduced and the resulting potentials generate a more significant benefit.

3.2 The resulting AI construction platform

Based on the current challenges, the aim is to enable the medium-sized AEC industry to meet current challenges through AI-based solutions and thus to sustainably strengthen its competitiveness. The overall target, therefore, is to develop a digital platform that enables all parties involved in the AEC industry to process heterogeneous data, regardless of its quality, formatting, and localization, and thus make it usable. Besides, the platform serves as a significant hub for cross-company collaboration, enabling data to be intelligently linked, enriched, and shared. On this basis, it is possible to develop special AI-based applications that automatically and dynamically enable a data-driven decision-making process through collaboration between man and machine. The development of new applications is promoted by a separate developer interface, which allows documentation based on sample data sets, including access via APIs.

These goals give rise to the following three main fields of action:

1. AI-supported data preparation and transformation, regardless of the heterogeneity, quality, formatting and localization of the source data.
2. Development of a digital platform as collaboration hub and possibilities for tool connection, including user and developer interface.
3. Development of AI-supported applications for the automatic and dynamic support of data-driven decision-making processes.

In the abstract, digital platforms can be defined as products, services or technologies that serve as the basis for a variety of companies to offer complementary products, services, and technologies. The structural elements of a platform comprise core components and a periphery. The core represents the actual platform by defining technological- and business-driven „rules of the game“ such as interfaces and processes. The stabilization and reuse of the core components lead to economies of scale and reduces the costs for the variety provided by the second part – the periphery. The periphery shows a high development speed and heterogeneity. The companies in the periphery form an ecosystem of the platform. They do not necessarily enter into business relationships with each other but are often independent participants in the same platform. (Baums, 2015)

3.3 AI construction platform architecture

There are four crucial modules within the platform: The Persistent Layer, the Analytics Runtime Environment, the Service Store, and the Authentication Layer. The four modules are described in figure 1.

Data owners can upload raw data to the Persistence Layer. If required, the persistence layer can perform pre-processing. This is especially the case if the data is not structured and digitized. Pre-processing in the form of structuring can again be done by a service from the Service Store.

In the Analytics Runtime Environment, AI creates data interoperability and processability. New data sets promote the learning progress of the AI and extend the data interoperability and quality of the preparation. Furthermore, the developer is free to choose his development tools. At the end, his software must be available in the form of a self-contained container that can be persisted in the platform. He accesses the trained algorithms and example data sets for the development of the algorithms.

In the Service Store the developed micro services are managed, and data is offered for sale. A service can bring its logic, persistence, and AI. These developed services allow users to run their data and purchase additional data to increase forecasting capabilities. If compatible, the services can be connected in series. This enables a wide variety of optimization for existing business models.

The Authentication Layer manages authorization throughout all levels of the platform. The platform imposes a clear separation between the raw data and the trained algorithms. On the one hand, data security can be guaranteed and on the other hand, companies can learn from each other without releasing data.

4. Possible applications in AEC with AI

Bobriakov (2019) elaborates eight possible use cases in a data science blog: These reach from predictive analytics (as an accurate simulation before construction and design issue prediction), to a warranty analysis (as analyzing construction project risks and tracking construction equipment’s and assets) to an optimization (as the contractor performance optimization, accurate budgeting, and scheduling). Resulting, he also mentions as two further implications the possibility to support the automisation on construction sites as well as supporting the product development. The last two mentioned use cases will be further analyzed in this paper in the resulting ecosystem and their business models. These use cases so far do not include the challenges with the heterogeneous data sets. Therefore in the following five possible applications of AI in AEC will be explained from design, realization to operation including the implications from heterogeneous data sets:

Finding and labeling equal parts: AI supports the designer in finding, detailing, and cataloging identical elements in the building model. These can be drone or satellite images as well as 2d plans. The items can then be linked by technical object requirements with open data platforms as product information (e.g., prices and availability).

Recognizing design contexts and rules: By identifying design contexts and rules, the designer can be supported to identify relevant next steps, as well as to define modules and rooms. As Chaillou (2019) also shows in an AI architecture approach, rooms and their functionality can be already identified. If this idea is continued further and if cross-discipline contents can also be identified, a room with its possible contents can be designed at a very early stage and presented to the architect and the specialist designers as an initial draft. The resulting design machine can support the overall design. The conversion of the room program into a draft is a complex design task and can be visualized much more broadly with more variants, accelerated and quality assured by the use of AI.

Realization planning: To support execution planning, work package sequences can be predicted by sequential pattern mining. It is also possible to determine the duration of the respective work packages as well as the expected construction time, taking into account past projects. On this basis, logistical measures can be derived (such as required storage areas, materials per calendar week, transport capacities). In this way, routes can be optimized and information passed on to suppliers and manufacturers. Here the points of the two predictive and the optimization applications of Bobriakov (2019) are matching.

Intelligent quality management: A further application can support the construction management and the involved construction companies in finding quality challenges, giving information about these as well as optimizing the plans regarding current changes. The platform services need to work closely with the site management through natural interaction via voice control as well as image recognition as by drones and satellites. This application area is supported by the warranty analysis application of Bobriakov (2019).

Predictive Facility Management: The AI analyses the operating data of the maintenance and malfunction-relevant components connected to the building control system and can thus make predictions on malfunctions and optimum maintenance intervals. Before a component fails, it can be serviced or replaced. The operating times are maximized and thus the downtimes and costs are minimized.

5. Incubating the ecosystem

Based on these applications the overall ecosystem will be incubated by generating new business models. These innovative and disruptive business models are divided in six growing business areas described in the following.

Services for collecting and generating AEC data: The data platform fundamentally changes the view of the value of AEC data. Experience is currently being shared from construction project to construction project. Often only company-internal experiences are collected. If building data is seen as a value, monetary values can be seen in many public places. Images of buildings can serve to improve algorithms to enable optimal architectural design of buildings with their environmental conditions. Furthermore, relevant IoT data can be identified, linked and monetized. In this way, open and existing data platforms can be used and enriched with value.

Data monetization: On the data platform, data can be exchanged and sold on a data marketplace. AEC companies can differentiate between internal company data and data that can be sold. Satellite images, data on the subsoil or material information on building components can be sold without having to release company-internal data.

Platform business model: The operation of the platform represents another business model. Via access to trained algorithms or metadata as well as access to applications in the store, fees can be charged on a pro-rata basis. This ensures the continuous operation of the platform.

Technical application development: Technical software developers use open APIs to build applications that access the metadata. These applications can be offered in the store of the platform to an AEC company. In this way, existing business models can be continuously improved and highly manual, repetitive processes can be reduced.

Development of services in AEC companies: By reducing waste in processes, AEC companies free up personnel. With the staff freed up, further additional services for the customer can be made possible. The client receives a personal consultant who advises him on decisions and shows alternatives via data-driven evaluations. The site manager is trained or additionally supported in the field of mediation and communication. Safety and health consultants accompany trades, pay attention to the correct use of applications and optimize their processes with regard to ergonomic aspects.

Innovative business models: Finally, completely new, disruptive business models can emerge based on the platform. Construction machines are no longer rented per construction project; for example, an excavator is instead rented according to the number of lifts used. Furthermore, new construction machinery and materials can be developed on the basis of transparent data. Construction workers are trained to master several professions and jump between construction sites according to current needs.

The value of a platform increases with each additional member, as the new customer can potentially come into contact with any producer. In comparison, in the pipeline model, only one new customer relationship is gained. The network effects are mainly driven by the cost-effective use of platforms (procurement of a smartphone, Internet costs, etc.) and the use of cloud computing services. By rising members, the ecosystem will get further support (Laine, Alhava, Peltokorpi, & Seppänen, 2017).

6. Conclusion

This paper proposes a new AI-based platform approach to empower SMEs of the AEC industry. Chapter 1 highlighted existing challenges the AEC sector is currently facing, these are mostly heterogeneous data and therefore a lack of interoperability and comparability. Further, the industry is fragmented and consists of 99,9% SMEs with collaboration and communication challenges. Lastly, decisions in the dynamic processes are, for the most part, based on feelings and personal experience and not on data or AI methods. As a result of this analysis four research questions were stated, how can data be made machine-interpretable, how can participants collaborate, which processes can be optimized by data collaboration and which new business models can emerge and enable the ecosystem.

Chapter 2 introduces an AI-based platform for the AEC industry. It involves AI-supported data preparation and transformation of heterogeneous data sets of varying quality, the development of a digital platform serving as a hub for collaboration between users and developers. Lastly, the development of AI-supported automatic and dynamic applications promotes a data-driven decision process for users. It was further defined that the baseline requirements for AI acceptance were ease of use, data security by only providing meta-data and a culture shift in terms of data usage. Around these requirements, a new system architecture prototype was developed.

Chapter 3 introduces use cases as proof of concept and four cases to show the possibilities AI offers for process optimization.

1. Pattern recognition to simplify construction drawings in terms of complexity and buildablility
2. AI based design assistance to support designers with data-driven decision-making within the process.
3. Scheduling, logistics and tendering based on past project data and AI predictions are implemented to determine the value of buildings in regards to cost, quality and time.
4. Provide AI based assistance for practitioners on site in the construction process an facility management.

The last research question was explored in Chapter 4 in providing an overview of possible new business opportunities that can be incubated in the AEC industry ecosystem. The platform ecosystem and business models offer new possibilities for services specializing in AEC data collection and generation, companies can decide to monetize their metadata. Application development and the development of services in AEC companies may serve as new innovative and potentially disruptive business models.

This paper shows how an AI-based platform approach may serve as a possible solution for current challenges in the AEC industry. With this new approach, companies get simple access to AI solutions for data interoperability issues, can overcome problems of high fragmentation by working together on a platform-based collaboration hub, and can base their decisions on data by relying on new AI-based applications. This new ecosystem leads to the empowerment of the AEC industry and may provide disruptive new business models for the future. In a current research project, the ideas of this platform are analyzed in an international project team, containing AEC companies, software developers as well as research institutes. More information to this project can be found on www.sdac.tech.

Quellen und Referenzen:

BauInfoConsult. (2016). Bauunternehmer: Welche der folgenden Arten von Spezialsoftware nutzen Sie? Abgerufen am 02. 07 2019 von https://de-statista-com.eaccess.ub.tum.de/statistik/daten/studie/709809/umfrage/genutzte-spezialsoftware-von-bauunternehmern-in-deutschland/?fbclid=IwAR2qNYUFOu4Gp-oiDiiITHlFa9SybL7c-qQ4f3RZ71pyuTa8I2mYeobjR1Q.

Baumanns, T., Freber, P.-S., Schober, K.-S., & Kirchner, F. (2016). Bauwirtschaft im Wandel Trends und Potenziale bis 2020. Munich: Roland Berger GmbH & UniCredit Bank AG.

Baums, A. (2015). Abgerufen am 02. 07 2019 von Digitale Plattformen – DNA der Industrie 4.0: http://plattform-maerkte.de/dna/

Bobriakov, I. (25. 06 2019). Top 8 Data Science Use Cases in Construction. Abgerufen am 02. 07 2019 von https://medium.com/activewizards-machine-learning-company/top-8-data-science-use-cases-in-construction-9ce8035e936f

Chaillou, S. (2019). AI + Architecture | Towards a New Approach. Harvard GSD.

Die deutsche Bauindustrie. (kein Datum). Unternehmen und Umsätze im deutschen Bauhauptgewerbe 2016. Abgerufen am 13. 06 2019 von https://www.bauindustrie.de/zahlen-fakten/statistik-anschaulich/struktur/unternehmensstruktur/

Froese, T., Han, Z., & Alldritt, M. (2007). Study of information technology development for the Canadian construction industry. Canadian Journal of Civil Engineering , 34 (7), 817-829.

Fulford, R., & Standing, C. (2014). Construction industry productivity and the potential for collaborative practice. nternational Journal of Project Management 32 , S. 315-326.

Gallaher, M. P., O’Connor, A. C., Dettbarn, J. L., & Gilday, L. T. (2004). Cost analysis of inadequate interoperability in the U.S. capital facilitie industry. U.S. Department of Commerce Technology Administration, National Institute of Standards and Technology, NIST REport No. GCR 04-867.

Laine, E., Alhava, O., Peltokorpi, A., & Seppänen, O. (2017). Platform Ecosystems: Unlocking the Subcontractors’ Business Model Opportunities. 25th Annual Conference of the International Group for Lean Construction. Heraklion, Greece.

Roland Berger. (2016). Bau: Wie schätzen Sie den Umsetzungsgrad der Digitalisierung für folgende Bereiche in Ihrem Unternehmen ein? Abgerufen am 02. 07 2019 von https://de-statista-com.eaccess.ub.tum.de/statistik/daten/studie/605291/umfrage/umsetzungsgrad-der-digitalisierung-in-der-bauindustrie-in-der-dach-region/

Shen, W., Hao, Q., Mak, H., Neelamkavil, J., Xie, H., Dickinson, J. K., et al. (September 2009). Systems integration and collaboration in architecture, engineering, construction and facilities management: a review.

Sopra Steria Consulting. (2018). Potenzialanalyse Agil Entscheiden 2018. Abgerufen am 02. 07 2019 von https://www.soprasteria.at/docs/librariesprovider33/Studien/sopra-steria-consulting-potenzialanalyse-agil-entscheiden.pdf?sfvrsn=0

Tata Consultancy Services (TCS) und Bitkom Research. (2018). Unterwegs zu digitalen Welten Deutschland startet in die technologische Zukunft. Abgerufen am 02. 07 2019 von https://downloads.studie-digitalisierung.de/2018/de/Trendstudie_TCS_2018_Bericht_DE.pdf

USP Marketing Consultancy. (2019). European Architectural Barometer. Abgerufen am 02. 07 2019 von https://www.usp-mc.nl/files/brochure-european-architectural-barometer_1543245060_1794e6e0.pdf