Watson Platform by integrating BIM, the Autodesk® 360 platform, and IBM ®Maximo®. For example, when asked to "give me a list of all pending work orders," AI will understand it is a CMMS query and call MaximoAPI to fetch all pending work orders, collect filtered I.D.s, then link back into the model viewer, and highlight related 3D objects in the viewer. From there, the user could click and get comprehensive BIM properties and access other peripheral information such as sur - roundings. Figure 3 demonstrates a use case that asks the AI to list all switch gear on the first floor and prepare to run a circuit analysis for the selected one. Conclusion The digital twin follows a path of creating virtual copies of physical locations, processes, and assets. It does not have a solid boundary that defines the scope of work since the whole point of the digital twin is to assist end-users to better understand the physical counterpart and get the most value from it. Therefore, the digital twin should be designed and built from the end-user's perspective rather than as a showcase of technical capability. No one can build the perfect digital twin over- night. However, each step taken, provides an opportunity to ensure its place on the roadmap with clear goals in mind. A digital twin can embrace all data resources and services that matter to the final goal, from spreadsheets to the AI assistant. Several years ago, we realized BIM is more like lifecycle data management rather than a modeling tool. Today, we are expanding our vision to create the digital twin that connects it all together.
Artificial Intelligence Integration Picture a scenario designing a generic digital twin platform for mul- tiple clients. The most significant challenge is to collect all their ex - pectations and find a way to fulfill all of them with one platform. The end-users could come from different areas with diverse backgrounds and various jobs. Therefore, expectations will be substantially differ- ent. In 2020, we conducted a survey of people, mostly from facility management, which explained what an ideal digital twin would be like and asked for three ways in which one might help them. The results varied from "I would like to know where is the mech room on the first floor" to "I would like to know which room will be impacted if I shut off this valve." Some questions like "How many entrances does this building have?" or "What is the square footage of the ballroom?" seem irrelevant, but have their real-world use cases. That means if the digital twin is used as a dashboard that serves employees who are not familiar with the building, it will meet their basic need to better understand the building, instead of just answering operation & maintenance (O&M) related questions. Remember part 1 addressed the importance of giving the correct information to the right people and the major challenge to arrange a digital twin platform to meet all these needs. The traditional way was to include buttons and menus on the platform and hope end users could find what matched their need. If new demands were brought up and no existing function matched, corresponding development would start, and a new button added. However, we know the speed of development always lags behind the pace of further requests, which means the plat- form, crowded with buttons, will not be flexible and sustainable from a long-term perspective. This is where AI shines. Its introduction is a promising trend in the AECO industry, and has been gaining popularity in assisting design, construction, and logically, the long-term use case of O&M. With the solid data foundation CDE paved for all upper structures, AI can freely reach each database to grab helpful information, aggregate, analyze and present to the end-user. Then the end-user no longer needs to worry about finding the button or requesting a new feature. Instead, they ask a question in a natural way such as "please show me where the air supply system connects to room 101" or "give me a list of all pend- ing work orders." Trained Natural Language Understanding (NLU) AI receives the question, identifies the intent, capture the keywords, and reaches corresponding databases to grab information. A pilot proof of concept has already been pitched and is in partnership with the IBM
GEORGE BROADBENT is Microdesk’s Vice President of Asset Management and has worked on a variety of projects including the rollout of Microdesk’s Maximo and Revit integration solution, ModelStream. George works closely with key stakeholders to identify strategies for asset management projects and manages the effort to build out new systems.
KAI YIN co-founded China-based ZhiuTech in 2017 and is Vice President of Product Development. The company’s mission is adopting technology for the AECO industry, creating customized solutions and bringing value to clients’ real life work cases. He previously worked as a mechanical engineer at Arup, and technology consultant at Microdesk. Kai earned a bachelor’s degree in civil engineering from Purdue University, and master’s degree in building science from the University of Southern California. JIAYI YAN is the co-founder and senior engineer at Beijing ZhiuTech Co. Ltd. in China. Jiayi graduated from University of Southern California specializing in building science and she is a PhD student of University College London dedi- cated to digital twin topics. She has years of experiences working as a building technology consultant providing technical solutions and implementation plans for world-leading enterprises in the United States and China. Jiayi specializes in BIM, sustainability, and city-level digital twin development in urban regenera- tion from multi-stakeholder perspective.
35
DECEMBER 2021 csengineermag.com
Made with FlippingBook Annual report