for the entire enterprise. Furthermore, the current trend of Artificial Intelligence (AI), and Machine Learning (ML) can be used behind the curtain to aggregate the wisdom from all DDMS clients in the cloud to benefit each other. Details of that will be shared later. How does the DDMS Bridge the Gaps? Compared to another industry, say, manufacturing, the adoption and utilization of current trending technologies in the AECO industry are lagging behind. The bright side is there are a lot of mature technolo- gies to take advantage of including AI, ML, Cloud-based computation, and big data. Variations of other algorithms and technologies should also be combined in the package to form the core functionalities that address the current issues among PIM to AIM transition and asset data management. Several key features of the DDMS will be discussed in the following content. FM Data Dictionary Health Assessment Main issues mentioned before are the ambiguous and inconsistent data structure and naming conventions. All CMMS implementation should start with a clean, consistent, and meaningful data dictionary, rather than fix all the troublesome data in the future. Therefore, the DDMS should apply various algorithms to evaluate FM data dictionary health - Semantical ambiguity occurs everywhere in a data dictionary draft. Semantical issues are inevitable especially for some large- scale organization or agencies. The same asset or attribute can be named differently in different systems or by different stakeholders (i.e., “hot water pump” vs “pump-hot water.” Figure 2 is a good example of this kind of issue). Fuzzy analysis is applied to identify non-exact matches of assets to spot the potential duplications of assets and/or attribute naming. Different semantic patterns can be automatically recognized and applied to corresponding fuzzy match- ing algorithms. Based on successful past project experience, this analysis can greatly help firms purify their data dictionary in a timely manner, especially for heavy Excel users. from different angles, including: • Semantical Analysis (Fuzzy analysis)
asset (i.e., CWS-GEN-CHILL). Missing one acronym will create an incomplete asset code. Meanwhile, it will create confusion if two as- sets have the same acronym or one asset has two acronyms. Again, if team members are utilizing Excel spreadsheets to manage and develop data dictionary, these data glitches will be everywhere.
• Classification Consistency Analysis
- The classification and consistency check ensures all the clas- sifications under different systems and subsystems have the same set of attributes (again, Figure 2 is a very good example to show how a same classification can be developed differently). These health checks refine and consolidate the potential data duplica- tions, as well as standardize the attributes and classifications. When all the asset information is aggregated from Excel spreadsheets or other formats, the DDMS goes through all the asset information, including categories, systems, subsystems, assets, child assets, attributes, as well as all properties for attributes (domain, type, unit, etc.). All detected issues will be provided to the user in an interactive decision-making approach (as shown in Figure 4). This process provides a tremendous advantage for the data or facility manager who doesn’t have to go through every entry to purify the data dictionary. Cloud Knowledge Community & Smart Recommendation “Knowledge on the cloud” is the term not only being heard more and more, but actually already in use for a long time. From Wiki, Quora, Pinterest, to streaming media like YouTube and TikTok, users share their thoughts and knowledge and wisdom to form a community that can absorb every member’s contribution and in turn benefit each other. A single user’s thumb up or down will impact the piece of contents’ popularity among the entire knowledge tank. To many, this is not something new, however, to the AECO industry, it is. Every user and individual has different stories and approaches to manage their systems. This is a double-edged sword since it also means when developing their FM data dictionary, every person works in a silo. The DDMS believes in “you don’t know what you don’t know” and tries to capture and aggregate all the knowledge and wisdom from multiple users and analyze how industries are building their data requirements. The DDMS applies cloud-based ML utilizing backend labeling to tagged asset classifications and attributes from different entities. Through the learning and training of the aggregated information and wisdom, the DDMS establishes a powerful knowledge base for smart recommendations of which attributes should be considered for a spe- cific classification, as illustrated in Figure 5. When more users and clients involve and interact with the platform, it learns their prefer- ences and captures the importance of each attribute as it relates to its
• Grammar & Spelling Check
- As the name implies, this function is designed to ensure cor- rect spelling and meaningful input. For example, “Installation Data” should be “Installation Date.” While “Data” is a correctly spelled word, it is the wrong word to use in the context of the example. Therefore, the DDMS will provide the suggestion to correct this type of error.
• Completeness & Uniqueness Analysis
- Completeness and uniqueness sometimes need to be highlighted based on users’ requirements. A common example is an acronym could be very important when users or asset managers use the ag- gregation of a series of acronyms to form a unique code for each
Figure 4: DDMS provides an interactive option for decision-making
12
csengineermag.com
october 2020
Made with FlippingBook Annual report