AI and machine learning will shape the coming era of electronics manufacturing.

Ed.: This is the third of an occasional series by the authors of the 2019 iNEMI Roadmap. This information is excerpted from the Smart Manufacturing chapter of the roadmap, available from iNEMI (inemi.org/2019-roadmap-overview).

Smart manufacturing is considered a “journey” that will require hyper-focus to ensure the appropriate technology foundation is established. Several enabling “horizontal” technologies (digital building blocks, data flow, security) are considered the most important to build a strong, agile, and scalable foundation. This article presents digital building blocks, with a focus on artificial intelligence (AI) and machine learning (ML) tools, and digital twins.

Advancements in the development of digital building blocks (interconnected digital technologies) are providing digitization, integration and automation opportunities to realize smart manufacturing benefits. These building blocks will enable electronics manufacturing companies to stay relevant as the era of the digitally connected smart infrastructure is developed and deployed.

Artificial intelligence (AI) and machine learning (ML). AI and ML tools and algorithms can provide improvements in production yields and quality. These tools and algorithms will enable the transformation of traditional processes and manufacturing platforms (processes, equipment, and tools). With respect to PCB assembly (PCBA), AI and ML present several opportunities to aggregate data for the purpose of generating actionable insights into standard processes. These include (but are not limited to):

  1. Preventive maintenance. Collecting historical data on machine performance to develop a baseline set of characteristics on optimal machine performance, and to identify anomalies as they occur.
  2. Production forecasting. Leveraging trends over time on production output versus customer demand, to more accurately plan production cycles.
  3. Quality control. Inspection applications to leverage many variants of ML to fine-tune inspection criteria. Leveraging deep learning, convolutional neural networks and other methods can generate reliable inspection results, with little to no human intervention.

Digital twins. The concept of a digital twin lends itself to on-demand access, monitoring and end-to-end visualization of production and the product lifecycle. By simulating production floors, the PCBA industry will be able to assess attainable projected KPIs (and what changes are required to attain them), forecast production outputs and throughputs through a mix of cyber-physical realities (physical world to virtual world, and back to physical world), and expedite the deployment of personnel and equipment to manufacturing floors worldwide.

Security. As many PCBA industry manufacturers provide services for several customers in a single location, it is critical to closely manage security for internal and external activities. Internally, security is of utmost importance when connecting equipment to networks, managing local data transfer, and leveraging the cloud for data aggregation and computation. A combination of edge/cloud networks, along with protective firewalls and secure gateways, will be essential to fortify a networking architecture. Externally, data will likely have to flow either upstream (from assembler to OSAT), or to suppliers and facilities worldwide, which will encourage a public/private cloud service model.

Deployment. This is the ability to deploy the necessary digital building blocks to realize smart manufacturing at different stages of maturity.

AI and ML will streamline the transition of products across the entire supply chain, leveraging data from one segment to improve operational efficiencies in another (e.g., production time, delivery, logistics). There is also a strong sentiment that the factories of the future will employ numerous employees, whose responsibilities will be augmented with AI and ML, as demonstrated with video analytics systems that can monitor production operations, and flag deviations in processes visually.

Advances in digital twin technologies are accelerating as the potential benefits are communicated to end-users. Also, the enabling technologies (hardware and software platforms) are becoming less expensive.

Standardization of data types and formats will be critical to maintain a consistent digital twin across the entire supply chain (SEMI-OSAT-PCBA). Each segment may have its own preferred guidelines or standards, but these standards should be interoperable among the segments.

The model for gradual penetration of the digital twin will follow aligned with ML and AI, with descriptive, predictive and prescriptive analytics shifting decision-making from individuals to broader system-level views that can holistically suggest and make recommendations on a preferred course of action.

R&D and Implementation Needs

Among the key topics that must be addressed to realize smart manufacturing are:

  • Definition hierarchy – digital twin, AI, ML, deep learning (DL)
  • Education – aspirational versus achievable
  • Talent – data and computer scientists, automation, manufacturing engineering
  • Data sharing and IP concerns
  • Speed of technology deployment versus speed of node introduction mismatch
  • Path for adoption – company/industry best practices, consortium-proposed deployment framework, top 10 smart manufacturing metrics
  • Open collaboration – technology development and pilot environments (e.g., SEMATECH 2.0).

This excerpt from the 2019 iNEMI Roadmap is based on the Smart Manufacturing chapter, co-chaired by Daniel Gamota, PH.D., of Jabil.

Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article