Strategic SMT machine data analysis can be a vital cog in the Industry 4.0 machine.

Creating process improvement with machine data is nothing new. Modern electronics manufacturing facilities are filled with machines that produce rich data about their activities — including detailed information on operations performed, faults encountered during production, material consumed, and products produced.

This data has been used in manufacturing for years to reduce errors and allow operators to monitor more machines. But other than being part of basic line utilization or operational equipment efficiency (OEE) calculations, rich machine data is not widely leveraged for strategic optimization of factory operations. This is a costly oversight in today’s manufacturing environments.

Using rich machine data to improve operational efficiency

Operational efficiency has always been a top priority for manufacturers. But as the world faces unique challenges like shutdowns and chip shortages, it’s become even more crucial. Operational efficiency means increasing the number of defect-free products that can be built in a given time period with the same number of machines and workers. Historically, there are three ways machine data has been used to improve operational efficiency in an EMS factory: real-time operational monitoring, process interlocking, and historical performance tracking.  

However, there is a fourth, emerging category of machine data collection: advanced machine data analytics

This is the practice of examining how machine data values change over time and discovering where the operation most needs improvement. Without exception, there is untapped value to be found in rich machine data. Used properly, data analytics can point to and indicate the resolution of specific problems that will increase the performance metric.

But what, exactly, does the “proper use” of machine data analytics mean?

For electronics manufacturers, it means preparing a one-of-a-kind IIOT framework, connecting to both legacy and new factory machines, streaming the data to a broker for systematic organization, and then to a cloud where it’s structured and usable in real-time. 

But it means much more than this. 

The real value of machine data, like all other data, is in the analytics. This is where the combination of rich machine data, domain expertise, and advanced algorithms provide extraordinary value to manufacturers today. 

High-level performance tracking KPI’s are important for every organization, but leveraging that data to provide insights beyond standard KPI’s is critical. In addition to KPI’s, the underlying rich machine data also more directly enables corrective action by pointing to the specific problems that can be resolved to increase the performance metric, something the ArchFX Platform and Enterprise Action System do remarkably well. 

 

 

Examining how advanced analytics offers more insight than traditional OEE methods.

In the real world, EMS factories are complex operations with a myriad of opportunities for process improvement. Reducing component attrition and increasing line utilization are two of the more common areas where advanced analytics can really drive optimization– and it’s not just hypothetical. Manufacturers who have begun the process of correctly extracting, storing, and analyzing their machine data are solving problems they didn’t even know existed. Once the tools are properly in place, the more they look, the more they find. 

Component Attrition 

Now more than ever, minimizing material attrition is key for all types of manufacturers. EMS factories prioritize eliminating wasted electronic components during assembly that are the result of mispicks during board assembly. It is generally not possible to achieve zero-component attrition, but advanced analytics can help achieve an operational goal of maintaining component attrition below a fixed level such as 0.1% or 0.5%.

Methodologies for achieving these attrition levels include:

  • Realtime monitoring — Reduces component attrition by preventing scenarios where a misconfigured machine wastes several components at high speed before an operator can intervene.
  • Feeder interlocking — Decreases mispicks and cuts down on unnecessary halting of the machine with intelligent, proactive recalibration of feeders rather than simply relying on a planned recalibration interval.
  • Usage-based maintenance — Decreases the number of maintenance events needed to achieve a target mispick rate.
  • Feeder specific analysis — Identifies problematic feeders via advanced analysis techniques then takes them out of service. Since a small number of feeders account for a large percentage of total mispicks, this will dramatically reduce the mispick rate. 
  • Condition-based maintenance — Provides a straightforward way to reduce component attrition rates when monitoring periods are shorter than standard maintenance intervals and it is assumed feeders with elevated mispick rates will not fix themselves. 
  • Predictive maintenance — Advance analysis methods such as regression and statistical inference can be used to predict which feeders will be tomorrow’s worst performers. 

Line Utilization

Maximizing line utilization (LU) is also important for EMS factories. Defined simply as the number of components placed on each line in 24 hours divided by the rated placement capacity of the installed pick-and-place machines on that line, maximizing LU is important because SMT machines are expensive assets and that need to provide maximum return.

Since LU in EMS factories is typically measured directly as the number of placements performed by the pick-and-place machines on each line, machine data collection from those machines can be used to calculate LU automatically.

Real-time monitoring can increase LU by comparing hour-by-hour line performance against a standard target performance that is believed to be achievable and taking immediate corrective action whenever the real-time output of the line drops below the target.

Historical performance tracking looks for trends in LU over time by comparing different lines and identifying opportunities to increase LU. The organization can then focus finite engineering resources on the lowest-performing lines.  However, there is rarely enough information in the LU metric itself to directly enable corrective action without further analysis. 

For example, it is possible to uncover actionable metrics by disaggregating all of the feeders and tracking each one individually.  Outliers can be identified tracking each work order on each line and using advanced machine data analytics, they can be ranked by LU loss and commonalities between low-performance work orders identified

Taking a deeper dive into advanced analytics

Those who have started this journey within their factories know how important the use of their machine data is, and are constantly finding impressive new applications for it. For those who have not yet partnered with domain-specific data experts, it’s not too late. The power of technology can catapult operations leaps and bounds ahead of today’s management practices and illuminate their path forward as well.

For real-world use cases, more detailed insight, and granular explanations of the common uses for machine data in operational improvement at EMS factories, read the full white paper presented at IPC’s APEX 2021 by our CTO, Tim Burke, Strategic Uses of Machine Data.