For decades, commercial buildings consisted of four walls and a roof. Over time, simple structures evolved into smart buildings, adding computerized control systems.
Today, the Internet of Things (IoT), big data, and now artificial/automated/altered intelligence are blessing buildings with brains. Commercial structures now boast the ability to recognize, interpret, adapt, and respond. Facilities can program themselves to function at peak efficiency, self-correct processes that have deviated from the norm, communicate with other mechanical systems, collect endless amounts of data, and make informed decisions, all in the name of enhancing the occupant experience. Actions can be triggered without human intervention. Software platforms can be programmed to deliver the optimal thermal load depending on occupancy.
But is AI tangible? How and when should an engineering firm invest in AI? What impact is it expected to have on the design process and project delivery?
Engineered Systems asked seven industry professionals these questions and more for this “Thought Leaders on AI in Buildings” roundtable discussion. Here’s what they had to say.
ES: How does your firm define artificial/automated/altered intelligence (AI)? What do these terms mean to you?
Paul Ehrlich: The term artificial intelligence (or machine learning) is intended to reflect methods for decision-making that utilize a series of models, rules, or algorithms to make what are ideally better decisions. AI can be used for analyzing data, such as the operational information in buildings. It can also be used for improved control of building equipment and systems. Note that these are terms that are often misapplied and used where they probably do not apply. Many of us who are not programmers or data analysts may tend to use the terms to apply to a level of sophistication and analysis that does not truly exist.
Kyle Knudten: We define AI with respect to the building industry as the application of building analytic engines to accumulate information from building automation systems, energy metering systems, lighting control systems, and other building systems with enterprise supervisory systems. We then use that information to identify areas of erroneous, declined, or suboptimal operation and recommend corrective courses of action. That action may be initiated by building operators or automatically by the supervised systems. We see that the application of this technology for smarter buildings has arrived but is still in flux. We expect that future technologies will expand these functions, further penetrating AI into building systems. Our local market is relatively early in the application of these capabilities into building systems, and the lines between what is possible with today’s installed buildings and the promise of what can be is constantly shifting.
Ionel Petrus: To me, soft AI is autonomous and is bounded (by initial constraints) as a way of making decisions. Strong AI is an autonomous, unbounded way of making decisions, which understands the cause of things.
Ken Sinclair: Our industry is a long way from implementing true artificial intelligence. That said, we are using automated intelligence every day. We’re using this mash-up of physical fixed assets and their emotional contents to create new building emotions and identities. This asks the questions: How do we best “look” with rapidly evolving video analytics, “listen” with natural language interaction, and “learn” using the personal assistants that are evolving as part of our edge-bots? How do we use our history of “feeling” temperature, humidity, occupancies, etc. and best combine that all with “thinking” that will come from self-learning? Once complete, we need to work a lot harder on returning that mindful reaction in the creation of anticipatory humanistic relationships.
John Varley: AI is the ability of machines to reason. In the case of buildings, this has been a part of the industry since the thermostat was introduced centuries ago. In the last century, digital controls have greatly expanded the ability of buildings to think. Today, AI incorporates learning into building automation systems. Buildings decide when to wake up and go to sleep. They call for outside help when they encounter difficult situations and tell us how they feel when they are sick or struggling. The level of decision-making skills embodied in building automation systems continues to deepen in scope and sophistication.
Dr. Draguna Vrabie: Artificial intelligence (AI) can be defined as the theory and development of computer systems that can perform diverse tasks that currently require human intelligence. To me, the term AI best describes the learning systems that have not been invented yet. As systems that can perform tasks regarded as artifacts of intelligence are developed, we start thinking of them as mundane technology and no longer as AI. Recent resurgence in AI research and applications is due to machine learning (ML) and deep learning algorithms used for predictive analytics and data-driven decision making. The application space for AI includes diverse domains, such as autonomous vehicles, energy systems, cybersecurity, biological systems, medicine, astrophysics, quantum chemistry, etc. Automated intelligence is largely synonymous with automation and describes hardware and software systems that are capable of learning from measurements and data, making decisions autonomously, and taking actions automatically based on learned knowledge. Automated intelligence as well as altered intelligence is not widely used terminology, and I would discourage their use in this successful, fast-moving, and also overhyped research area.
Phil Zito: We view AI as the augmentation of intelligence, whether it be through truly assisted or independent intelligence. It seems that the industry largely equates AI with analytics and/or assisted decision-making. We actually believe this will be the last area that is truly “automated,” and this stems from our industry’s desire to “have control.”
ES: Can you give one instance where your firm has used AI (or is considering using it) to improve your operations?
Knudten: We are integrating building analytic tools into our commissioning process to complement our approach to building commissioning. We have future applications planned for existing building evaluations, continuous building system optimization, and measurement and verification documentation processes.
Ehrlich: We have been involved in research programs that are looking to use AI as part of advanced control sequences. In controls, this is often referred to as “model predictive control” (MPC). In this case, MPC is used to look at the operating characteristics of a group of VAV terminals and then predict what the ideal system static pressure set point should be at the next time interval. While this approach requires more computational effort than current techniques, such as “trim and response,” it has the potential for dramatically improving system performance. But keep in mind that this is still in the research phase and is not yet being applied to commercial buildings.
Varley: Every time we specify building automation systems, we touch AI. Further, AI has become a part of our design process. BIM tools inform us when we create physical clashes in buildings between various components and automatically calculate building thermal loads and pressure drops in air and hydronic distribution systems. These tools translate the design into paper format so they can be built or communicate directly with contractors in the field to answer questions and resolve conflicts. Engineers will soon cede parametric analysis to the machines and focus on higher level design tasks.
ES: What’s the greatest challenge surrounding AI in the HVAC engineering sector?
Vrabie: Mastery of deep learning enables national laboratories to advance the frontiers of scientific research and national security. Information about applications of deep learning techniques at Pacific Northwest National Laboratory (PNNL) is available at https://deeplearning.pnnl.gov/. In the area of energy security and reliability, a PNNL-led team will construct an intelligent, real-time emergency control system to help safeguard the U.S. electric grid by providing effective and fast control actions to system operators in response to large contingencies or extreme events.
Petrus: As it relates to the operating of buildings, AI could help improve the performance of the vast number of existing buildings that have outdated BAS infrastructure. As it relates to the design of building systems, a continuous lack of interest/funding on behalf of the consulting engineers in self-developing AI algorithms exists enhance the design of the HVAC systems. One could imagine a new future in which engineers design AI algorithms in lieu of controls sequences.
Sinclair: Disruption is everywhere, yet, at our core, we still use pneumatics, ethernet, BACnet, and Niagara framework. We are at a disruption point that is equal to what DDC did to pneumatics and what the internet did to DDC. The IoT is disrupting us. Disruption is creating new things that make the old things obsolete. Except our “things” are buildings, and buildings cannot be made obsolete completely, at least not yet, so we need to build elegant, mindful bridges between disruption, innovation, and reality. As scary as this thought is, this is just the dawn of what we will see in the following days of disruptions.
ES: What’s the greatest potential benefit offered by AI in the HVAC engineering sector?
Zito: If AI is assumed to be a form of software that assists the contractor, then you can effectively argue that AI is being used today. Remember, AI is not replacement of the human mind. Many people hear AI and assume that AI is a full-scale replacement of the human. AI, as it is being realized in the field, is already causing a benefit. Assisted code development, automated point discovery, and pattern recognition are already built into some of the major manufacturers’ toolsets.
Petrus: AI has the potential to significantly decrease the energy costs of buildings and provide greater comfort for the occupants and processes inside of the buildings.
Sinclair: “Open” is mostly about opening our minds to the “edge of change” that is upon us. This is a chance to open up with more about our digital transformations keeping us on our crusade to open software, open hardware, and open everything.
Varley: The ethical concerns of AI apply to the building sciences as well. As AI continues to assimilate into the building sciences industry, it should always satisfy Asimov’s Laws: A robot may not injure a human being or, through inaction, allow a human being to come to harm. A robot must obey orders given it by human beings except where such orders would conflict with the First Law. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law. If this is the case, our buildings will continue to serve their human occupants by improving efficiency and comfort. They will have “empathy” for the occupants. What we imagine today will become reality very quickly. The future timescale for the introduction and assimilation of new technology will seem instant by today’s standards.
ES: As an industry, are we ready to talk about AI intelligently, or are we still struggling with our own self-learning?
Knudten: We believe that our industry’s awareness of these systems is still in development. These systems are integrated extensions and evolutions of building automation systems. Additionally, most consultants struggle to have a strong understanding of the innerworkings of building automation systems. Taking this into consideration, many lack an adequate background to fully describe and understand the application of these systems for clients.
Ehrlich: While it is encouraging to see robust discussion about new technologies such as AI and ML, it is largely still in the stage where we have more excitement (even hype) than real deliverables. But it is encouraging to see actions such as the recent order from the White House on this topic: http://bit.ly/WhitehouseAI. In general, the HVAC industry moves slower than other, more technologically innovative industries. I would expect that in time it will be used in our industry, but I would expect that this will be a slow, long-term process.
Sinclair: As an industry, we need to grow younger to truly understand AI. I’ve found only one method of growing younger, which is to look at change through the eyes of our trusted younger mentors. Using their eyes and minds, we can quickly grow younger in ours and better understand the change that is upon us. To younger minds, this evolution is not a change, it’s simply their understanding of the present problems. Our resistance to change is caused by the necessary mind shift we must make to view clearly what has changed. These younger mentors’ eyes come with a clear understanding of present digital dynamics, which makes these mentors indispensable in our acceptance of change. Most of my younger mentors have their own trusted younger mentors and they tell me, “You will not believe what these kids are thinking.” In the ongoing struggle for “open” in our changing connection communities, we need to open 2019 sharing through the eyes of those who have grown up to only know the digital era.
Vrabie: The topic of AI and machine learning brings two main sentiments: (1) Big data and AI are the latest new shiny things, so we should be doing them, and (2) we don’t understand or trust AI. It sounds like magic, so we should wait and see how successful it can be in other domains. Research scientists and engineers are definitely ready to talk about how to apply machine learning in AI systems. Theoretical development of current AI state-of-the-art methods had started long back in 1950. With the relatively recent advancement and availability of computing resources (graphical processing units), data availability/deluge, and open source machine learning libraries, these AI-based methodologies are ready to be utilized in diverse application domains including building energy systems.
ES: If you had to pigeonhole your firm’s interest/investment in AI over the next five years as unlikely, neutral, or likely, which would you choose and why?
Ehrlich: Our firm tends to be actively involved with research organizations, helping them to understand the industry and demonstrate how new technologies can and should be adopted. So, our involvement tends to be on the early side of the technology adoption curve. We view this as a promising new technology, but we’re cautious about the lack of understanding for what it really can and can’t provide.
Knudten: Likely. We are currently deploying these technologies into our project delivery and have every expectation that we will extend the reach of these technologies further into our approach.
Petrus: Highly likely. Our firm prides itself on being at the forefront of the industry. We’re now proposing a new technology that uses AI algorithms for the design and control of building systems without the use of standard proprietary controllers. We are not just talking the talk, we are walking the walk.
Sinclair: None of these words are a great fit. The words we should use are inevitable and unavoidable. Automated intelligence is everywhere and growing daily. Even though our products and services are very real, we virtually market them. We need to enter today’s virtual world and instill emotion using vivid graphics and words of want to play on the emotion of our clients.
Emotion is the noun and verb used to describe the creation and depiction of a mindful interactive relationship — a conversation if you will — between users and their devices and services. It’s a virtual identity, a feeling, a learning, an interactive piece, a virtual brick-and-mortar that hosts the emotion.
The building blocks of our digital transformation are a mindset, people, process, and tools. Note that tools, which equal our technology, devices, and services, is the last item.
We, as an industry, tend to focus on tools first, then try to trowel over mindset, people, and process. This is not working.
Varley: Likely. Our business won’t survive without the skills to apply AI for our clients.
Vrabie: Scientists and engineers at national laboratories are already using machine learning and developing AI systems for many national interest applications. With the increasing availability of higher quality datasets, more AI algorithms will be evaluated in the buildings domain to extract the knowledge for building operation and performance optimization. This will reduce the dependency on empirical and manual process and facilitate the deployment of advanced control algorithms.
Zito: We are a training provider. As such, we have to make bets on where the industry is moving and train our students accordingly. Our current IT and programming courses are being revamped to focus increasingly on data models and API’s. These are the two areas I would recommend that any layperson in our field focus on. Our focus is going to be teaching professionals how to be enterprise and system architects. The ability to understand a business outcome and then piece together technology solutions to achieve that outcome will be the most invaluable skill. The reality is all of the technology we need (assisted Cx automated decision-making, etc.) has been in existence for a long time in other markets. The problem is we, as an industry, are extremely slow to adopt technology.