In manufacturing processes, when a piece of tooling begins to wear, operators are often unaware that intervention is needed until a larger interruption occurs. Additional parts of the line may be disrupted, product may be damaged and production may be halted. Had operators been able to identify the individual issue earlier, they could have taken action. Instead, the manufacturer incurs downtime, material waste and lost production.
At Rockwell Automation, software architect Kai Lin and PBCA process engineers Miki Cvijetinovic and Gregory Vance recognised this as a problem for customers and decided to address it. They piloted a solution at the Rockwell Automation manufacturing plant in Twinsburg, Ohio.
The challenge
Rockwell Automation’s electronic assembly operations produce 4000 unique circuit board designs per year, place 9 million components per day, form 23 million solder joints per day, and produce 14 000 panels per day across its facilities. The Twinsburg location has eight manufacturing lines. Like all facilities, it runs many processes that could affect production, and for this pilot, the team focused on the stencil printing process.
A stencil foil has laser-etched apertures under which a circuit board is mated. A squeegee blade then pushes solder paste from the front to the rear of the stencil, filling the apertures with paste that deposits and sticks to the circuit board. At the end of the stroke, the circuit board is transferred and inspected. The next board is printed in the opposite direction with a different squeegee blade. Each part of this process generates performance data.
More complex than meets the eye
Like any seemingly simple manufacturing process, stencil printing has hidden complexities. It involves many independent variables, making it difficult for support staff to detect wear or degrading performance of the stencils and squeegees, especially when that wear is hard to see with the naked eye. The start of the process and long pauses during it also change the effectiveness of paste transfer, making it difficult to analyse the long-term wear of the tooling. Stencils that are thinner or have special coatings have different lifespans, creating further unpredictability. While each step generates performance data, support staff were unable to gather and analyse this vast amount of data, spanning IT, OT and engineering, in real time.
Support staff needed a better way to collect and analyse real-time performance data and be alerted immediately to any equipment anomalies or damage. This was an ideal use case for an anomaly detection algorithm, the computation of which could be calculated and displayed through FactoryTalk DataMosaix seconds after print inspection.
The solution
Rockwell Automation’s project engineer and software teams took an algorithmic approach. They created a patented anomaly detection algorithm, owned by Rockwell Automation, that monitors the tooling in the stencil printing process and sends performance data to support staff. They then implemented notifications so plant personnel could receive immediate alerts on anomalies without performing tedious post-hoc analysis.
First, the engineering team added RFID tags to each stencil and squeegee blade, along with a sensor to detect the direction in which the gantry was moving. The team collected the data and tied it to the serial numbers of the printed circuit boards entering the machines, captured by an existing PLC. The data was also tied to the corresponding stencil and squeegee to obtain performance data on anomalies, defects and solder paste inspection.
The team then turned to FactoryTalk DataMosaix, a SaaS application designed to bring in high volumes of raw data from various sources, contextualise it into one platform, and make it useful for operations management teams. Typical business intelligence tools did not provide enough data relationship modelling along with the ability to perform this contextualisation in real time. Power BI, for example, required someone to perform the analysis manually, which at the Twinsburg plant was conducted monthly. With FactoryTalk DataMosaix, the analysis could be done in real time, a requirement for personnel to be notified of immediate performance anomalies.
FactoryTalk DataMosaix also tracks the entire lifecycle of the squeegee blades and stencils as they move around the plant floor. Typically, when a stencil is used in one line, its data is tracked, but when it is brought back and placed in another line later, the data on its past runtime and performance is not considered, increasing the likelihood that personnel could miss performance issues. Using FactoryTalk DataMosaix, the team plotted the data on a chart that identified the risk of equipment failure. “This analytical method is a really elegant way to look at the performance of the line,” says Lin. “It really speaks to whether the line is getting better or worse.”
Finally, the team built a channel on Microsoft Teams to which anyone at the plant could subscribe to receive real-time notifications of equipment abnormalities or defects. Teams was chosen because it was the business system already in use. Previously, when the solder paste inspection machine detected an issue with a circuit board, the operator had to manually check or clean it, leading to delays or potential errors in judgement. By monitoring the performance of the squeegees and stencils, the algorithm now triggers alerts for unusually high board failure rates, prompting operators to pause and investigate promptly. Operators can quickly identify the equipment causing the failures, whereas previously, the nature of the production environment made it difficult to identify the root of the issue.
The result
There were several instances in which the engineering team and plant personnel received a notification about a stencil anomaly or defect and were able to react accordingly. Instead of waiting for failure, or wastefully replacing stencils based on a calendar date, the algorithm gave operators 30 to 60 days’ advance warning of stencil failure. After replacing or repairing the damaged stencil in each instance, the fail rate improved by 22%.
Machine operators also became more productive. With support personnel alerted to stencil issues before they affected the rest of the production line and meaningful data to work with, operators had fewer distractions and could work more efficiently.
After a four-month evaluation at the Twinsburg facility, Rockwell Automation calculated that the annual labour savings were dramatically higher than the costs, with a return on investment of approximately 200%. With additional use cases planned, the company expects the FactoryTalk DataMosaix return to increase further. The improved fail rate and productivity also correspond to revenue gains, enabling the plant to realise $9 million worth of revenue sooner through less disruption.
A productive future
Engineering and software teams are working to deploy the solution at scale in Twinsburg, as well as at Rockwell Automation plants in Mequon, Monterrey and Singapore, and are looking for other applications within the company’s operations. Because Rockwell Automation owns the algorithm, the potential use cases for customers are extensive. As Lin says, “When you think about what the algorithm provides, then really correlating that back to monitoring performance of tooling in the process, I would see a very wide applicability in discrete manufacturing.”
| Tel: | +27 11 998 1000 |
| Email: | [email protected] |
| www: | www.rockwellautomation.co.za |
| Articles: | More information and articles about Rockwell Automation |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version