Arch AI
Turn Insights Into Scalable, Guided-Actions
With more than 15,000 machines across 100 or more sites, the complexity of doing so was staggering. Flex partnered with Arch Systems to extract data from new and existing machines, creating a universal solution powered by standardized data. Successful implementation of the solution enabled real-time dashboards, Global KPIs, custom reports, and advanced analytics, leading Flex to significant utilization and productivity gains and ROI across its global organization.
Flex is a $24B/year manufacturing and supply chain services company that employs 160,000 people in over 30 countries. They work with more than 1,000 customers and over 16,000 suppliers. They have deep domain expertise across many industries and markets, including 4,000 designers and engineers and 20 design centers, which they leverage to help their diverse customers. Put simply, Flex is in nearly every industry imaginable.
The machines used inside manufacturing facilities vary dramatically. With more than 5,000 high-value machines in over 100 sites, Flex’s SMT landscape uses machines that come from a wide variety of different vendors, purchased in different years, and running different operating systems and software versions. Flex had the same challenge across Mechanicals lines, augmented by a wide variety of legacy machines that provide no modern way to extract data.
With a company of Flex’s size and complexity, standardized key performance indicators (KPIs) are indispensable. Flex knew it needed KPIs to manage complex global operations for a wide variety of tactical and strategic reasons. High-value assets, like those used in surface mount technology (SMT), account for millions of dollars in machinery and hundreds of thousands of dollars in annual depreciation per line.
Tactically, Flex wanted to increase day-to-day operational efficiency. Strategically, they wanted to increase long-term machine capacity utilization and knew the advanced insights allowing them to do so were hidden in untapped complex machine data. With individual sites implementing local, customized solutions to get data, Flex recognized the need to consolidate into a single global solution that worked at every site.
Like many prudent companies, Flex tried both building and buying a solution. They began an exhaustive effort to build a standard practice for providing needed visibility and insight. First, they leveraged a standard business intelligence (BI) tool to build a top-level global KPI dashboard that would address the ROI problem with a subset of machine data. Next, they created a carefully designed process using spreadsheets and semi-manual data collection to a data warehousing solution from various machines that could be further aggregated and massaged by a central team to power the global dashboard. However, report generation proved to be slow and labor-intensive. While the reports were valuable, there were significant delays in data generation.
Even more problematic, the manual data-gathering processes inevitably led to inaccuracies that could affect strategic decisions. Even if Flex successfully implemented an efficient global process to produce top-level metrics, they knew they were leaving most of the rich data in the machines behind. They hadn’t yet solved the complex IT/OT problem of extracting and publishing global data for more powerful industry 4.0 use cases such as ML/AI and advanced analytics.
Seeking to simplify the global IT/ OT problem, Flex teams spent years working with vendors, standards bodies, and internal teams to standardize their machines, accelerate the adoption of new protocols, and develop rich data aggregation solutions that could pave a path over the remaining challenges. The team of expert engineers and operators ultimately determined that none of the solutions delivered the desired results – results they knew were possible. The effort was halted and they returned to the simpler KPI system they had previously built, putting progress towards advanced insights on hold once again.
A program was formed with Flex IT, its global asset management team, site operations, and Arch Systems to deploy the ArchFX platform. The project to connect and analyze global SMT lines included 3 primary pieces:
Flex deployed hardware or software connectors from Arch to heterogeneous machines. For SMT machines, these were largely pure software connectors focusing initially on integrating ASM, Fuji and Panasonic pick-and-place (P&P) machines and moving next to Vitrox and Koh Young inspection machines. For Mechanicals machines, Arch’s IOTile hardware would be used to extract machine data to be standardized at the broker.
Arch and Flex deployed three centrally-managed regional data brokers; one in the Americas, one in Europe, and one in Asia. This provided a scalable IT environment to reach all factories, supporting real-time dashboards, custom reports, and Global KPIs.
Arch’s web-based Global KPIs Solution and its advanced analytics mapped the data to desired actions for targeted use cases. Chosen KPIs were calculated for every SMT machine, line, area, site, and globally across the world in real-time.
Steps 1 and 2 completed the first major deliverable. The teams had collaborated to build an enterprise data broker, and it had been successfully deployed within Flex’s global operations and provided unprecedented visibility into operations. All data was stored in the cloud data lake to enable global utilization statistics, providing a real-time view of operations across different sites.
This first ROI-generating solution allowed Flex to continue funding and expanding its use of Arch. They repeated the data collection playbook across additional factory machines in the SMT line, across Mechanical areas, and with quality testing machines. Mechanical machines were integrated into their MES systems, providing machine data automatically and closing process loops that previously required manual clicks or even pen and paper.
Once the rollout of the Global KPIs program was completed, Flex could see the largest opportunities for improvement across their hundreds of lines and thousands of machines. Next, Flex had to decide how to make improvements. Could the ArchFX system go beyond identifying problems to finding root causes and suggesting fixes?
Historically, companies that identified problems using KPIs often failed to find solutions because only a small fraction of the available machine data was collected and made available globally. This reduced global data set would show that a line had low utilization, but did not provide enough context to determine why. To identify the root causes or problems, an SMT expert would need to go back to the rich data stored on the machine and use their experience to figure out what happened.
Arch took a different approach. In addition to calculating top-level KPIs, ArchFX collected all of the rich machine data that a human SMT expert would need to discover root causes. Arch collected millions of data points from each machine, providing detailed descriptions of every operation it performed and the errors it encountered.
Now Arch had to determine how to make use of this rich data. Could advanced analytics automatically determine the top problems to address on an SMT line? Many previous attempts by others to feed rich data into generic “AI” or “machine learning” algorithms never worked well because the problems were too varied and complex for the AI to solve.
Arch instead asked, “How would the best SMT expert solve this problem today, working on a single line in a single factory?” The answer was clear: combine SMT expert methodology with advanced big-data algorithms carried out automatically on all lines at once.
The first major success of this approach was the development of an advanced analytics module called Sessions Analysis. This analytics module uses advanced algorithms to study every product cycle on all SMT lines. From this rich machine data, Session Analysis automatically knows which cycles should be identical because they are building the same product and provides an apples-to-apples comparison across thousands of occurrences of theoretically identical machine operations. This allows for automatic inference and builds the picture of a “golden cycle” for each product and on each line without any manual input.
Available in the Arch SMT Utilization Suite, Sessions Analysis is an advanced analytics module that uses advanced algorithms to study every product cycle on all SMT lines. From this rich machine data, Session Analysis automatically knows which cycles should be identical because they are building the same product and provides an apples-to-apples comparison across thousands of occurrences of theoretically identical machine operations.
Based on this key data, Sessions Analysis classifies all of the structural losses that prevent the line from consistently achieving the golden cycle time. These include such losses as workload imbalances between machines, underutilized placement heads, loading and transfer delays, bottleneck steps outside chip mounting, etc. Sessions Analysis then separates out all transient operational losses on top of the structural operational losses and classifies them based on the rich state information and error codes reported from the machines.
The success of Sessions Analysis in complementing top-level utilization KPIs with detailed, actionable root causes motivated Flex and Arch to repeat the same playbook across other aspects of the SMT line. Flex is starting to find similar success looking at advanced global feeder analysis to reduce component attrition and nozzle analysis to identify P&P maintenance problems before they result in attrition and yield issues.
Combining rich machine data with the advanced algorithms provided by the ArchFX platform provided novel powerful insights, and teams began to accurately identify previously undiscovered significant actionable improvements. Dave Dunne VP, Asset Management at Flex said, “Arch brings rich data-based analytics, visualisations, and intelligent actionable insights, at speed, real-time, that can allow Flex to truly optimize its operations and performance, to a different level.”
ArchFX provided a robust platform for industry 4.0 implementation. By the end of the year, Flex had installed Arch solutions in over 22 sites across SMT & Mechanicals production lines with close to 1,000 machines connected across 200 lines globally. Myckel Haghnazari, Director of Information Technology at Flex, said, “As we move from site to site and region to region, we now have capabilities that are easily put in place.”
In addition to feeding utilization analytics, the growing data lake provides a foundation for developing new insights. Data scientists and operational experts at Flex and Arch can now form hypotheses about the root causes of problems and opportunities for increased performance and retrospectively validate them using the historical data lake.
The ArchFX Platform empowers electronics manufacturers to reclaim 20-60% in valuable machine performance and line utilization losses. Every day, Arch extracts data from thousands
of SMT machines worldwide, accelerating Industry 4.0 and enabling real-time visibility and actionable insights into some of the most complex problems manufacturers face today.
Get the latest news and content about AI in Manufacturing and ROI-driven processes.
