skip to main content

Moving Toward the "Smart Factory" In Microelectronics Manufacturing

By Michael Armacost, Mingwei Li and James Moyne

The era of smart manufacturing, Industry 4.0 and the smart factory promises significant opportunities to reduce cost, boost productivity and improve quality in microelectronics manufacturing. But it also presents the industry with new, unique challenges.

Smart manufacturing (SM) is a term “generally applied to a movement in manufacturing practices towards integration up and down the supply chain, integration of physical and cyber capabilities, and taking advantage of advanced information for increased flexibility and adaptability.”[1,2,3] It is often equated with “Industry 4.0” (I4.0), a term that originated from a project in the German government that promotes a fourth generation of manufacturing that uses concepts such as cyber-physical systems, virtual copies of real equipment and processes, and decentralized decision making to create a smart factory.[4,5]

THE SMART MANUFACTURING VISION

While the literature base for SM, smart factory and I4.0 is wide and varied, common themes are present that help provide an understanding of the whole SM and I4.0 space. These include:[1]

  • Leveraging big data infrastructures. Data management infrastructures are being enhanced to support improvement in capabilities associated with the "5 V’s"—namely volume, velocity (data collection and analysis rates), veracity (data quality), variety (data merging and consolidation), and value (data analytics).[6] This enhancement is punctuated by the movement to big data architectures such as Hadoop that support (1) storage of data in a serial fashion which is much more “analysis friendly” than traditional relational architectures; (2) parallel and scalable approaches for higher-speed analysis of larger quantities of data; and (3) an open architecture-style environment for development of data management and analysis tools. A key challenge is the migration from existing data management infrastructures and understanding how the data infrastructures coexist in a collaborative environment to support capabilities ranging from real-time online decision making to offline high-fidelity model building.[7]
  • Integrating the supply chain network.Tighter vertical and horizontal integration of systems is a common nt
  • nt of SM and leverages the "variety" data merging and consolidation enhancement in data architectures. From the horizontal integration perspective, the smart factory will be an integral part of the upstream and downstream supply chain, with factory optimization becoming a component of overall supply-chain optimization. The tighter connectivity will allow for leaner operation, better inventory management, higher flexibility of operation, improved response to demand, and better traceability to address issues such as warranty recall investigation. An obvious requirement here is the development of standards for supply-chain data integration that are not industry-specific.
  • Leveraging advanced analytics.The primary benefit of implementing big data infrastructures and practices will be enhancing analytics to support improvement in the quality of existing capabilities such as fault detection and classification (FDC), but also in the realization of advanced predictive capabilities such as virtual metrology and predictive maintenance (PdM). These analytics will leverage increased data "volume" and "veracity" for more robust and maintainable models; "velocity" for more granular models; and "variety" for more causal and predictive models. From the "value" perspective, traditional analytics will become much more effective, leveraging the higher data volumes and data quality to build more robust models. New big data analytics such as deep learning will also emerge to complement more traditional analytics. Additionally, the better integration of data systems will enable these analytics to span much larger domains, such as up and down the supply chain, and incorporate techniques such as "digital thread" for linking analyses to data chains to solve factory-wide or even supply-chain-wide problems.
  • Improving use of cyber-physical systems (CPS). CPS refers to the "tight conjoining of and coordination between computational and physical resources."[9] This is not a new concept as systems that integrate computational and physical resources have been in existence for some time. However future SM systems will continue to improve from a CPS perspective in terms of "adaptability, autonomy, efficiency, functionality, reliability, safety, and usability."
  • Improving the use of real-time simulation through realizing the "digital twin." "A digital twin refers to a digital replica of physical assets, processes and systems that can be used for various purposes."[10] The concept is further refined in the International Roadmap for Devices and Systems (IRDS) as "a state of fab operations where … real-time simulation of all fab operations occurs as an extension of existing system with dynamic updating of simulation model."[6] Many of the predictive applications being developed in the industry today will likely continue to evolve to more directly support this vision.
  • Reliance on a knowledge network. The movement in technology associated with SM and I4.0 requires a corresponding change in the business operation paradigm. As solutions become more complex and consolidate larger domains of data systems and applications, realizing and maintaining these solutions requires a higher degree of cooperation between users, OEMs and analytics solution providers in a structured knowledge network.[6,11] This cooperation enables the required incorporation of subject matter expertise (SME, e.g., process and equipment knowledge) into data-driven (statistical) models for improved model quality and robustness. Issues such as data sharing and partitioning, intellectual property security, and managing solutions in the cloud have all come to the forefront as part of the move to enhance support for this cooperative knowledge network.[6]

REALIZING THE SMART FACTORY VISION IN MICROELECTRONICS MANUFACTURING: Industry-Specific Challenges and Opportunities

A smart factory vision for the microelectronics industry is shown in figure 1.[11,1] Note that the common themes described in the previous section are all present in this high-level representation. However if we dig a little deeper we quickly realize that, while the tenets of SM, I4.0 and smart factory are not industry-specific, each industry has its own unique challenges and opportunities, so industry-specific variations of the smart factory vision emerge.

Figure 1: A smart factory vision for the microelectronics industry.

Microelectronics manufacturing is a very unique industry characterized by high process precision and dynamics; process and equipment complexity; high degrees of intellectual property (IP) in equipment, processes and analytical solutions; and a business model that focuses on developing and maintaining fab-wide solutions.[1] These characteristics result in unique requirements (or at least re-prioritization of requirements) and challenges in realizing the microelectronics smart factory. The primary unique requirements include:

  • Incorporating SME in computational process control (CPC), the forefront of solutions. Incorporating knowledge of the process, equipment and product in the development and maintenance of solutions is a tenet of SM in general; however, it is of premier importance in realizing the microelectronics SM vision. CPC is defined as "the integration of process, equipment and analytics domain expertise" in microelectronics analytics solutions.[11] The importance of CPC in microelectronics SM comes from the complexity, precision and dynamics associated with processes and equipment as noted above, but also from the large number of context changes (e.g., product change, maintenance event, or different upstream product routes) associated with the production environment. In a purely statistical analysis world, these complexities would result in a need to partition data streams in order to understand the impact of each context change, process drift, etc. This, in turn, would result in changing the "big data" source into a large number of "small data" sets with insufficient precise data in each set to support good models. Incorporating CPC elements of process, equipment and product SME allows quality models to be developed, verified, and especially importantly, maintained with less data. It also allows for the intelligent merging of these small data sets when the relationships between the different contexts and dynamic situations are understood.
  • Providing an analytics roadmap.The highly complex and precise production environment in microelectronics, combined with cost and production pressures, has resulted in a heavy focus on analytics to support enhancement of existing solutions such as next-generation FDC, and the realization of new solutions, such as virtual metrology and PdM. While there is a strong literature base in the industry of specific analytics being applied successfully to point solutions, it often is not clear how and when specific analytic types should be employed. This often results in a focus on the elegance of the analytic (e.g., deep learning or purely statistical techniques) over the practicality, extensibility and robustness of the solution, and a lack of emphasis on CPC. As a first step to address this issue, some SM literature efforts have tried to define the analytics capabilities in terms of dimensions and apply these dimensions to the needs of particular applications, as shown in figure 2.[1,12]
  • Maintaining data and IP security.While the opportunities in SM are significant, this new paradigm of operation brings with it a risk of maintaining security in the face of higher levels of integration, data production and management, and information sharing for collaboration. While this is a challenge for SM in general, it is especially acute in microelectronics manufacturing where there is significant IP in process, equipment and analysis solutions. In fact, it is noted in the IRDS that system and information security is one of the primary challenges hindering the advancement of smart manufacturing and I4.0 concepts in the microelectronics industry.[6] Aspects of this issue vary widely, ranging from concerns such as protection of IP in collaborative activities to introduction of malware through a USB hookup. One specific area where security is severely limiting SM evolution is data-sharing environments such as "the cloud." These environments allow data from multiple sources (including potentially multiple companies) to be centrally located so that analytics can be applied in a scalable fashion. However, cloud-based data and IP partitioning risks and solutions are not well-defined, leading many manufacturers to completely avoid these solution tools, and instead choose to execute SM activities completely and exclusively within the fab. With the help of the IRDS, a roadmap to address the data and IP security issue will eventually be charted that first identifies the issues, a solution baseline, and standards needed to move forward.[6] Until that time, security will likely be the main issue governing the progress of SM in our industry.

Figure 2: Defining the dimensions of analytics approaches and how they map to microelectronics manufacturing applications. (Phenomenological models are physical model forms, representing process knowledge, that are tuned with statistical data.)

LOOKING AHEAD

The era of the "smart factory" is upon us, punctuated by a focus on integration, data and analytics. Terms like "cyber-physical system" and "digital twin" are now commonplace in discussions of the next generation of manufacturing, regardless of industry. The microelectronics manufacturing industry continues to be a leader in migrating to the smart factory. The complexity and precision of our processes and equipment present numerous challenges that are opportunities for improvement through employing smart factory concepts. A key to realizing these opportunities will be our ability to utilize a secure knowledge network to incorporate subject matter expertise in developing and maintaining solutions.

For additional information, contact michael_d_armacost@amat.com

[1] J. Moyne and J. Iskandar, "Big Data Analytics for Smart Manufacturing: Case Studies in Semiconductor Manufacturing," Processes Journal, Vol. 5, No. 3, July 2017. Available online: http://www.mdpi.com/2227-9717/5/3/39/htm
[2] Wikipedia: Smart Manufacturing. Available online: https://en.wikipedia.org/wiki/Smart_manufacturing.
[3] Davis, J., Edgar, T; Porter, J., Bernaden, J., and Sarli, M. Smart manufacturing, manufacturing intelligence and demand-dynamic performance, Computers & Chemical Engineering, 2012, vol. 47, pp. 145–156..
[4] Project of the Future: Industry 4.0 (Germany Ministry of Education and Research, http://www.bmbf.de/en/19955.php.)
[5] Kagermann, H.; Wahlster, W. INDUSTRIE 4.0 Smart Manufacturing for the Future. Germany Trade and Invest, 2016.
[6] International Roadmap for Devices and Systems (IRDS): Factory Integration White Paper, 2016 edition. Available online: http://irds.ieee.org/images/files/pdf/2016_FI.pdf.
[7] J. Moyne, J. Samantaray and M. Armacost "Big Data Capabilities Applied to Semiconductor Manufacturing Advanced Process Control," IEEE Transactions on Semiconductor Manufacturing, Vol. 29, No. 4, November 2016, pp. 283-291
[8] Najafabadi, M. N., et al. Deep learning applications and challenges in big data analytics. Journal of Big Data (2015) 2:1.
[9] Cyber-Physical Systems (CPS) Program Solicitation NSF 10-515. Available online: https://www.nsf.gov/pubs/2010/nsf10515/nsf10515.htm.
[10] Wikipedia: Digital Twin. Available online: https://en.wikipedia.org/wiki/Digital_twin
[11] Hasserjian, K. Emerging Trends in IC Manufacturing Analytics and Decision Making. Keynote. Advanced Process Control Conference XXVII, October 2016. Available online via: http://apcconference.com.
[12] Lopez, F.; et. al. Categorization of anomalies in smart manufacturing systems to support the selection of detection mechanisms. IEEE Robotics and Automation Letters (RA-L), August 2017