When data learns to act

Modern SMT lines are already highly automated today. Printers, placement machines, and inspection systems perform many process steps autonomously, continuously generating data on processes, material flows, and quality. Powerful software detects deviations, calculates optimization suggestions, and reports material requirements. Nevertheless, one crucial step is often still left to humans: the evaluation of this information, as well as the decisions based on it and their implementation. Bernhard Fritz, Head of Global Marketing, ASMPT SMT Solutions

For the first time, three developments are coming together:

Until now, at least one component was usually missing for largely autonomous manufacturing. Sometimes the data set was incomplete; other times, software solutions were not yet able to reliably evaluate complex interrelationships. And robotic systems were simply too inflexible or too expensive for many tasks.

Today, however, the situation has changed fundamentally. Modern SMT lines already provide a large portion of the information needed for automated decisions. At the same time, powerful AI systems and data analytics enable the intelligent evaluation of this data, the identification of correlations, and the derivation of recommendations for action. Added to this is a new generation of collaborative and autonomous mobile robots that are flexible enough to handle standard manufacturing tasks around the clock. This makes what was long considered a vision of the future a tangible reality for the first time.

The path to autonomous manufacturing

A look at today’s SMT manufacturing facilities shows that much of this is already a reality. ASMPT has been developing and integrating the necessary technologies for years.

“In my view, the exciting development isn’t that completely new technologies are suddenly emerging. Rather, we’re seeing existing solutions increasingly converge. Many building blocks of autonomous manufacturing already exist – now begins the phase in which they are intelligently networked with one another,” explains Thomas Bliem, Vice President of R&D at ASMPT SMT Solutions.

A good example of this is the automatic download of the production program during a product changeover on the SMT line via the new printed circuit board, as implemented by ASMPT. The product information is transmitted from machine to machine along the entire SMT line via the standardized IPC-HERMES-9852 interface. Each station – from the DEK solder paste printer to the Process Lens SPI system and the SIPLACE placement machines – automatically detects whether a product change has occurred and, if necessary, automatically loads the appropriate production program. Even family setups and product changes across setups can thus be carried out largely without manual intervention.

At the start of the line, the DEK TQ printing platform already handles numerous tasks that were previously performed entirely by operators. These include high-speed stencil cleaning, automatic paste transfer during stencil changes, software-controlled placement of support pins, and automatic offset correction using a closed-loop process with the SPI system. All of this reduces the workload on employees while ensuring more stable processes. With a new, innovative solution for automated setup changes, ASMPT further simplifies the process significantly. Instead of replacing the stencil and squeegee separately, the operator simply replaces the old stencil – including the squeegee positioned on it by the print head – with a prepared setup. In this setup, the squeegee and paste transfer unit are already connected to the stencil via an adapter. All subsequent steps are performed automatically. This reduces the operating time at the printer from about four minutes to just about 30 seconds.

The situation is similar with material flow in smart manufacturing. Modern software from ASMPT now handles many tasks that previously had to be planned and carried out entirely manually. It continuously tracks material inventories and storage locations, as well as future demand on the SMT line. Based on this, a demand-driven replenishment time is determined, and the corresponding transport orders are initiated automatically. In combination with autonomous mobile robots, this results in an increasingly automated material flow based on the 4R principle: the right material in the right quantity at the right place at the right time.

As demand for AI servers rises, printed circuit boards are becoming increasingly complex and the number of components continues to grow. This inevitably leads to higher material consumption, more reel changes, and more splicing operations. ASMPT’s SIPLACE Smart Feeders already support 1-mmpitch tapes. As a result, significantly more components fit on a single reel, meaning reels need to be changed and spliced less frequently. This noticeably reduces the amount of manual labor required. ASMPT is taking the next logical step in collaboration with a specialized robotics partner: In the future, splicing robots will autonomously handle the splicing operations that are still required on the SIPLACE placement platforms. Automatic splicing systems have already proven themselves in practical use, especially in high-volume/low-mix production. Since they simultaneously reduce time-consuming manual routine tasks, sources of error, and line downtime, they often pay for themselves within one to two years, depending on the production process.

AI connects what was previously separate

The true strength of AI does not lie in making individual machines or systems smarter. It is only through their interaction that the potential of these solutions is realized. For example, if the Process Lens SPI system detects a gradual deterioration in the printing process, the WORKS software can automatically readjust the printing parameters in a closed-loop control system. If an AOI system detects recurring defects, the causes can be traced back to the respective process step by linking inspection, printing, and assembly data with the software. With the help of data analytics and AI, the WORKS software can not only evaluate this information but also derive the appropriate measures and initiate their implementation. In this way, the analysis evolves into largely automated process control.

This also opens up new possibilities at the shop floor level. For example, software solutions from ASMPT use AI to analyze production data from different lines, compare it with theoretically optimal processes, and reveal previously hidden potential for optimization. AI-powered assistance systems such as ASMPT’s Virtual Assist also provide voice-controlled support to service technicians during troubleshooting and systematically guide them through complex maintenance or repair tasks.

Process knowledge becomes a decisive success factor

However, this requires far more than just artificial intelligence. For software to make decisions independently, it must first understand how the individual processes in SMT manufacturing are interconnected. Solder paste printing, material supply, placement, inspection, and logistics all influence one another. A seemingly minor deviation in one process step can have noticeable effects in a completely different part of the process.

This understanding does not develop overnight. For decades, ASMPT has been working with customers and technology partners to develop solutions for virtually all areas of SMT manufacturing. This close collaboration has fostered a deep understanding of processes that today extends far beyond individual machines or software functions.

“Process knowledge cannot simply be programmed. It is built up over many years – together with our customers and from experience across a wide variety of manufacturing environments. That is precisely why artificial intelligence and data analytics will continue to be only as good as the process understanding on which they are based,” explains Thomas Bliem, Vice President of R&D at ASMPT SMT Solutions. “This is precisely where ASMPT’s particular strength lies. No other manufacturer is so deeply and broadly rooted in the SMT process. This knowledge, gained from countless projects and manufacturing environments, forms the basis of our solutions and is incorporated into our products every day.”

Only the interplay of this process knowledge, the manufacturing data available today, and the capabilities of modern data analytics, AI, and robotics creates the foundation for the next level of automation. This understanding of complex manufacturing processes is crucial to determining how far automation can actually be developed.

The next wave of automation has already begun

The future of SMT manufacturing will therefore not be shaped by a single spectacular innovation. Many of the technologies required for this have long since found their way into practical use. Today, data is generated throughout the entire manufacturing process. AI can derive correlations from this data, and modern robotics is increasingly taking over tasks that were previously performed manually. What previously often existed side by side is gradually merging into a seamless, integrated system.

The decisive advance, therefore, does not lie solely in faster machines or more powerful software solutions. Rather, what is crucial is that information is immediately translated into action. If the system detects a deviation, it can not only identify its cause but also initiate the appropriate corrective action in the next step.

In the future, AI will not only make decisions for individual machines or lines but will also coordinate production capacities, material flows, and resources across the entire manufacturing process. In this way, smart manufacturing is evolving from a networked production environment into a system that autonomously controls an ever-increasing number of processes. Routine tasks are increasingly being automated, allowing process engineers and operators to focus more on optimization, monitoring, and the continuous improvement of manufacturing.

“For us, one thing is clear: The future does not belong to AI alone, but to intelligent collaboration between humans and AI. Humans set the guidelines, define decision-making parameters, and always retain control when it comes to critical decisions,” explains Thomas Bliem.

The groundwork for this has long been laid. Now it’s a matter of sensibly integrating the existing technologies.