Assembling an automated microassembly system takes careful planning
Add up the variety of parts, performance specifications and volume requirements involved in automatically assembling a multicomponent micro product and the result is a bewildering sum of possible hardware- and software-based automation engineering solutions.
Yet, according to Jim McClellan, president of McClellan Automation Systems, Bedford, N.H., “automation is not actually an engineering task; it’s a business task.” The key to success, he said, is to focus on the desired result of the automation process, then choose the right automation system.
Development of an effective microassembly automation system takes into account a manufacturer’s business plan, profit expectations, facility resources and the skill level of shop personnel. “If you are a top manager, among the most important things to you are how you are going to compete, what are the features and benefits of your product, how safe it is going to be, what your cost of goods sold is going to be, and how you are going to be competitive in the marketplace,” McClellan said.
Accompanying those factors are the requirements of the assembled product itself. “A specification document with hundreds of pages about a product may be a real concern,” McClellan said, “because buried in the document are the one, two or three things that are absolutely critical.” Rather than assume everything is critical, he said, manufacturers must make some choices. “Every project comes down to a few things,” McClellan continued, “and if you get these things right, the project is going to be a success.”
The majority of the automation systems McClellan Automation develops are for manufacturers in the medical devices, diagnostics and life sciences industries. In those applications, critical requirements usually include product quality, sterile operating conditions and the ability to validate processes.
Meeting those requirements is facilitated through timely input from automation vendors. Forward-thinking manufacturers, McClellan said, involve the vendor in the initial design of the product. That enables early identification of factors critical to meeting quality and cost-control targets. McClellan meets with a client at the beginning of a project to rough out what aspects of product assembly can and can’t be automated. He also looks for strategies that lower assembly costs.
For example, McClellan said, a product designed so access for assembly is required only from one side can be much less expensive to automate than one that requires access from both sides. In a one-sided approach, parts can simply be placed into a device, while two-sided assembly may require additional part-positioning and transport operations. The cost of automating the assembly of a product designed without automation in mind might be half of what it would be if automation was anticipated at the outset, McClellan said.
A typical vision inspection camera setup used to verify component geometry. Image courtesy McClellan Automation Systems.
Karel Rasovsky, business development manager for DWFritz Automation, Tualatin, Ore., agreed that business considerations guide engineering action. “We look at how the assembly automation application is going to be deployed. What are the target production volumes and cost targets? Clients might be looking at automating assembly of a large partmaking operation, but the value of the part itself is very low. The business case for automating that system might not be there.”
Most microassembly automation systems feature some custom-designed components and commercially available technology, such as robots or vision inspection equipment. (See sidebar on opposite page.) “When selecting an off-the-shelf technology platform,” Rasovsky said, “We use a best-of-breed approach, which optimizes the tradeoff of cost versus functionality. The value-add our engineers contribute is to take that technology and innovate around it to design an optimized system.” He pointed out that control software is an increasingly important component of automation systems.
Flexibility for future growth
Flexibility is necessary to enable one automation platform to handle components of different geometries and sizes. “One part might be round, another square, they may be different heights and widths, or be made of different materials,” Rasovsky said. Such differences require adaptable part-handling and inspection methods.
This automated wafer metrology system at DWFritz Automation employs robotic part handling, top and bottom high-resolution cameras, laser interferometers and a profilometer to detect and measure laser-cut features on one side of a silicon wafer relative to features on the wafer's other side. Image courtesy DWFritz Automation.
In many cases, inspection is “the hardest part of automation,” said McClellan. “We look at every project and ask, ‘How are you going to inspect this? How are you going to measure it? How do we know that we have actually accomplished what we wanted to do?” Almost every microassembly automation system the company develops features some form of vision inspection, he said, noting that microscale parts provide a greater challenge. “You have to hold them to see them, so essentially you are using microscopes attached to the cameras to see the part,” explained McClellan.
Inspection technology is not limited to vision systems, though. “Vision can do a lot of good things, but it is sometimes very hard to apply to certain products, so you have to be creative,” McClellan said. “Sometimes you have to measure the thickness of a part using laser interferometers to know that you have manufactured it correctly. We are being stretched into areas of inspection that are well beyond what was traditional in an automated system.”
While automation vendors frequently apply off-the-shelf inspection and part-transport technologies, when commercial equipment is inadequate or not available, vendors will develop their own microassembly automation systems. For example, McClellan said, “O-ring installation is one of the things we are really good at. We are really good at it by default because we didn’t have a lot of success with the automation units that were commercially available. So we made our own.”
Handling the individual components of the product being assembled, especially at the microscale, usually demands custom-engineered solutions. Those solutions include one-of-a-kind end effectors, grippers and vacuum methods, as well as creative approaches, McClellan said. To pick up a tiny part, for example, “you make something sticky or make something have a charge to it. Things like static attraction are either your friend or your enemy.”
However, many traditional methods of transporting parts, including the use of cams, gears and levers, still apply, he added. “It’s not that they disappear entirely, just that they are different.” Those differences can include non-traditional arrangements of downsized automation system components.
A holistic approach
Scientists at the University of Texas at Arlington’s Texas Microfactory are focused on finding “a more systematic way of performing microassembly,” according to director Dr. Harry Stephanou. The work involves a holistic approach that encompasses the entire manufacturing system.
“We have observed companies that, by trial and error, break their microassembly
automation system problem into subsystems,” he said. “One company was developing a fairly complex optical system that had 13 subsystems. They were trying to optimize each one separately. That obviously is not the way to do it, because what happens in subsystem three may impact subsystem five.”
A key tool in the group’s work is computational software it developed called Design for Micromanufacturability, or Dfm2 (see “Last Word,” MICROmanufacturing, November/December 2010). According to Dr. Aditya Das, also with the Texas Microfactory, the software “is a collection of advanced mathematical models connected to each other through an ingenious reduced-order correlation framework to provide an almost-real-time platform to assist the entire manufacturing cycle.”
Collectively, the equations interrelate a large number of input parameters from the manufacturing operation, including data such as design tolerances, machinery specifications, workpiece material information, robotic automation methods and testing criteria. Output is a real-time analytical matrix for overall manufacturing cost, cycle time and yield. “It makes the entire process more systematic and streamlined, as opposed to the chaotic approach that is currently being followed in micromanufacturing scenarios,” Das said. “Our objective is to reduce the number of uncertainties and reduce the risk factor in the manufacturing process.”
The mathematical analysis clarifies tradeoffs between performance and costs. The goal is to find situations where, according to Stephanou, “a slight decrease in performance results in a large decrease in cost, or a slight increase in cost results in a large increase in performance. At the end of the day, that is the bottom line.” (See sidebar on page 33 for information about the Texas Microfactory’s efforts to lessen the cost/benefit tradeoff in nanoscale manufacturing.)
The Texas Microfactory, part of the University of Texas at Arlington, is developing reconfigurable microassembly systems, like this one, that can be assembled relatively simply and quickly. Modules in the system can work together like a robot. Image courtesy Texas Microfactory.
According to Stephanou, the first sector the Texas Microfactory decided to attack is low-volume production. He pointed out that automating the production of low volumes of parts (tens or hundreds instead of millions) at low cost is a challenge. “Some applications die because it is not possible to assemble things in a low-cost way in low volumes.” Stephanou said many startups have to survive a low-volume “valley of death” because, even if they run high-volume production eventually, they have to be able to produce the parts in low volumes to convince investors to support them.
Efficient automation of a low-volume application requires an assembly automation system that can be reconfigured very rapidly. To meet these needs, the Texas Microfactory is developing specific automation modules that can be assembled into a complete system relatively simply and quickly. The modules perform separate tasks. “You may have one for horizontal motion, one for vertical motion, and then a number of stages that perform as if you had a robot,” Stephanou said. “You may be in a situation where motion along one axis requires much less accuracy than another.” As a result, changeover can be fast; Stephanou cited one pick-and-place robot system that can be reconfigured for another task within 2 minutes.
The ultimate goal of the Texas Microfactory is not solely to make
microcomponent assembly faster. “We are very strongly committed to automation because without it, we are chasing pools of cheap labor and it is going to be difficult to keep manufacturing in the U.S.,” Stephanou said. µ
About the author: Bill Kennedy, based in Latrobe, Pa., is a contributing editor for MICROmanufacturing. He has an extensive background as a technical writer. Contact him at (724) 537-6182 or at firstname.lastname@example.org.
Custom and commercial
DWFritz Automation, an automation equipment integrator, provided an example of a microassembly automation system that employs custom-engineered elements and commercially available subsystems.
Built for the electronics industry, the system orients, assembles, laser-welds and inspects small parts within a 3-second cycle time.
An automated laser-welding system (below and above) developed for the electronics industry by DWFritz Automation. The system orients, assembles, laser-welds and inspects small parts within a 3-second cycle time. Inset above shows two 5mm-dia. part components prior to welding. Images courtesy DWFritz Automation.
After an operator loads bulk parts into two vibratory part-feeder bowls, the parts are selected by a custom-designed pick-and-place system, oriented via a Cognex vision system, and then placed in a vacuum nest on a rotating index table.
Prior to laser welding, sensors inspect each assembly for missing or misaligned parts. The laser welder features a Branson laser power supply and diode enclosure. An interlocked metal shroud lowers before an 810nm laser permanently welds the components.
Parts exiting the welder are optically inspected for irregularities by a vision camera, then inspected for planarity by a custom-designed Heidenhain gauge array. Failed assemblies are automatically directed to a reject bin, while verified assemblies are sorted by nest ID (each nest on the rotating table is identified by a number) into bins on a rotating output tray.
A PC running Microsoft.NET C++, with user input via a keyboard and touchscreen, controls the automation system. The process is user-configurable via custom screens at the user interface.
Operator attention is limited to loading raw materials, unloading the output tray and reject bin, and acknowledging any error conditions as reported by the system software.
Cross-scale integration of assembly
The University of Texas at Arlington’s Texas Microfactory is working to integrate nanoscale manufacturing with micro- and macro-scale operations.
“In the real world, you have to be able to span more than one scale,” said director Harry Stephanou.
Dr. Rakesh Murthy, who focuses on nanoscale applications, said, “the main objective is to be able to manufacture products, such as sensors, that are based both at the micro- and nano-scale.” Microscale sensors generally are far more sensitive than the macroscale versions, and nanoscale sensors promise to be even more sensitive and selective.
Carbon nanotubes, with diameters on the order of 20nm to 50nm, are critical elements in the construction of nanoscale sensors. Nanotubes are strong, their conductivity can be easily tuned, and they can be easily made to respond to chemical or biological agents.
Carbon nanotube-based sensors typically are assembled via chemical processes. A large number of nanotubes are suspended in a solution and subjected to external fields, leading to a more or less random distribution on the device. To enable more precise location of the nanotubes, precision microscale robotic tools can position the nanotubes individually. “The downside is that you are doing things one at a time, which is slow,” Murthy said.
To speed the process, the Texas Microfactory employs its experience in microassembly to create many microscale robots. “The advantage is that we can make them in multiple numbers and put them together so that we can employ them in parallel,” Murthy said. This speeds production. “Bigger robots are building smaller robots that can build even smaller devices.”
Typically, the resolution of the robot actuators used to position the nanoparts must be smaller than the 20nm to 50nm diameter of the nanotubes, and the resolution is in the range of 1nm to 5nm.
McClellan Automation Systems
University of Texas at Arlington