Issue link: https://iconnect007.uberflip.com/i/1228683
48 SMT007 MAGAZINE I APRIL 2020 provide the desired attributes; data flow archi- tecture is considered a prime candidate for le- veraging and cross-industry collaboration to identify optimum solutions (i.e., data synchro- nizers, execution clients). The development and deployment of tech- nologies for data flow are accelerating. Focus on data analytics, and data retention protocols are increasing at a faster rate than first antic- ipated. It is imperative to collect the critical data as well as to establish guidelines to per- form intelligent analysis and to exercise the ap- propriate algorithms to specify data-driven de- cisions. Several topics related to data are under consideration, such as general protocols: • "All" versus "anomaly" data retention practices • Optimization of data storage volumes • Data format guidelines for analytics to drive reactive and predictive technologies • Data quality protocols enabling improvements in time synchronization, compression/uncompression, and blending/merging • Guidelines to optimize data collecting, transferring, storing, and analyzing Data considerations for equipment are: • Defining context data sets for equipment visibility • Improving data accessibility to support functions • Data-enabled transition from reactive to predictive functionality • Data visibility of equipment information (state, health, etc.) Digital Building Blocks The ability to deploy the necessary digital building blocks to realize smart manufacturing is at different stages of maturity. AI and ML A few key attribute needs for AI and ML are data communication standards, data formatting standards, and 3PL tracking solutions. Technol- ogies, such as AI and ML, are seen as enablers to transition to a predictive mode of opera- tion: predictive maintenance, equipment health monitoring, fault prediction, predictive sched- uling, and yield prediction and feedback. This paradigm in AI-enhanced control systems archi- tectures will enable the systems to "learn" from their environment by ingesting and analyzing large data sets. Advanced learning techniques will be developed that improve adaptive mod- el-based control systems and predictive con- trol systems. The continued development and assessment of AI and ML technologies is criti- cal to establish the most robust and well-tuned prediction engines that are required to support emerging production equipment. Digital Twin Technology Advances in digital twin technologies are accelerating as the potential benefits are com- municated to end-users. Also, the costs for enabling technologies (hardware and soft- ware platforms) are becoming less expen- sive. The following are considered key attri- bute needs that will increase adoption and broad-based deployment of the digital twin (product design, product manufacturing, and product performance: digital thread, predic- tive, prescriptive, and systemwide continu- ous data access. Digital twin is a long-term vision that will depend on the implementation of discrete pre- diction capabilities (devices, tools, and algo- rithms) that are subsequently integrated on a common prediction platform. It is generally considered that the digital twin will provide a real-time simulation of facility operations as an extension of the facility operations system. The successful deployment of digital twin in a facility environment will require high-quali- ty data (e.g., accuracy, velocity, dynamic up- dating) to ensure the digital twin is an accu- rate representation of the real-time state of the fab. Also, the realization of this vision will de- pend on the ability to design an architecture that provides the key technologies to operate collaboratively by sharing data and capabili- ties. Ultimately, the success of the digital twin will depend on the ability to develop a path for implementation that provides redundancy and several risk assessment gates.