Robots done right

All photos courtesy of AVR Vision & Robotics

A ceiling-mounted robot using a long-shank rotary bur to remove burrs from holes.

Can robots improve your deburring operation?

Robotic deburring has been discussed since the mid-1970s, but does the process work? Shops need to answer that and other basic questions before considering replacing manual deburring with a “staff” of robots.

First, it is important to distinguish exactly what the user is looking for. Many finishing robots are not truly deburring robots. Some remove casting flash rather than burrs and that difference is important. Robots employed to move parts within deburring cells have abilities far below that required for sophisticated deburring. Robots that polish surfaces (rather than finishing edges) are good applications also, but the term “deburring” is not reflective of what they are doing.

Before considering true robotic deburring, part manufacturers must define their part requirements. That includes drawings and specifications, production rate and quantity expectations, and other essential details that are sometimes not clearly defined.

Part Requirements are Key

Table 1 on page 38 shows the part requirements that determine if robotic deburring is a viable production option. Beyond this information, however, manufacturers must also define the degree of difficulty presented by part geometry. Success in robotic deburring will be determined, in part, by edge and size tolerances, proximity to other edges and surfaces, part complexity, machinability, burr properties and feature accessibility.

For example, a paper by Serefettin Engin, high performance machining specialist for Pratt & Whitney, and other authors, to be presented at the North American Manufacturing Research Conference in May, provides a convenient numerical means of identifying the impact of all these factors, but no formal public system exists to identify the limits of robotic deburring capabilities. Several robot builders use proprietary matrices of issues to understand the limits of their systems to process parts.

While cost reduction is the primary reason to consider robotic deburring, other benefits include safety (removing operators from hazardous work), improved quality, higher production rates and simplified QC.

Like any machine, robots have associated costs, including procurement, shipping, installation, maintenance, supplies, training and labor. Program and process development costs for new parts are often not accounted for in projecting payback periods for robotic deburring. 

Payback is predicated on the ability to impart acceptable edge quality. If the robot falls short, payback is extended, so potential users must define expectations in writing and images and inspect all deburred and affected features before accepting robots. This can be challenging because most companies do not define what a burr is or how it inspects for burrs or other edge conditions. For example, so-called “secondary burrs,” such as burrs formed by countersinks as they remove a large burr, may be acceptable to some but not others. Also, most potential customers want a robot to remove all burrs, but the most common approach is to follow robotic deburring with some amount of manual finishing. 

A robot point-finishes the edge of a cutout with an abrasive-filled, cotton-mounted tool.

A final consideration is the probability the company will continue to produce the part in question. The author worked on a team developing a highly precise robotic deburring operation, but when the system was ready, the part was canceled—as was the need for the robot we spent 3 years developing.

Variation is Critical

While part manufacturers try to account for all production factors when considering robotic deburring, some factors are frequently omitted. Variation is one of those overlooked factors, according to Francois Arrien, vice president for robotic material removal, AVR Vision & Robotics Inc., Montreal. “[Manufacturers] forget about part-to-part variation and variation in burr formation during production,” he said. “They provide nominal part and burr information, but not the variation. The burr on the first part is different than the burr on the 1,000th part because of tool wear and other factors. The part geometry may also vary.”

Table 1: Robotic potential for deburring applications.

Production requirements

Large variety of 3-D part configurations produced each month

Small variety of part configurations

Long runs of same shapes

20 to 50 parts/month

More than 100/month

More than 500/month

Robot potentialTypically not feasiblePossibleReasonableGood potentialPart sizeMiniatureFinger sizeHand sizeHand size and largerEdge quality requirementsLevel 1Level 2Level 3Level 4Level 5Level 6

Burrs

100 percent removal

100 percent removal

100 percent removal

100 percent removal

Microburrs OK

Remove loose material

Inspection (acceptable quality level)

100 percent of edges

100 percent of edges

100 percent of edges

100 percent of edges

Sample

Sample

Inspection magnification

30×

10×

10×

4×

Unaided eye

Unaided eye

Surface finish (Ra)

8μin. (0.2μm)

16μin. (0.4μm)

32μin. (0.8μm)

64μin. (1.6μm)

64μin. (1.6μm)

More than 64μin. (1.6μm)

Sharp edges

Not allowed

Not allowed

Not allowed

Not allowed

No cut hands from handling

Allowed

Max. edge radius allowed

0.001 " (25μm)

0.003 " (75μm)

0.005 " (125μm)

0.010 " (250μm)

0.015 " to 0.030 " (375μm to 750μm)

Not required

Allowable dimensional change from deburring

Less than 0.000050 " (1.25μm)

Less than 0.0001 " (2.5μm)

Less than 0.0002 " (5μm)

Less than 0.0005 " (12.5μm)

Less than 0.001 " (25.4μm)

Less than 0.0005 " (12.5μm)

Table courtesy of L. Gillespie

Arrien noted that some customers wait until after the system is ready to be delivered to notify AVR of surface finish requirements near part edges. “When that happens, we have to start over developing a process with gentler brushing or use other approaches,” he said. “That takes more time and frustrates both parties.”

Arrien noted that simple robotic deburring systems may require as little as 16 weeks to design, build and prove-out, but more complex systems may take a year. Simple systems may cost as low as $150,000, and multiple copies of the same system are even less expensive. These multiple copies are used by some aerospace manufacturers that want operators to produce a complete part without passing it to other departments. The deburring robot is positioned beside the blade grinders, providing each machinist with a miniature cell to complete the part. AVR’s 30 engineers produce about 15 deburring and finishing robotic cells each year as well as automated surface vision inspection systems.

Parts with hundreds of features and tight tolerances may require $750,000 “bulletproof” robotic deburring systems. For those parts, AVR uses vision and, in some instances, force feedback and compliant tools, which flex out of the way rather than gouge the part when they come across an unexpectedly large burr or a change in part dimension. Vision systems identify changes in part geometry and varying burr sizes in a closed-loop process. While vision can help understand what is at an edge and allow the robot to compensate, in some instances vision is used merely to assure that the burr in, say, each of 600 holes has been satisfactorily removed and, if not, the robot returns to assure 100 percent removal.

Robots are fast but typically not precise, so they need a sensor or compliance control to provide repeatable edge conditions. Without such control, it is common for robots to generate chatter when cutting with rotary burs. The robot can easily lose contact with edges when X-Y paths change quickly or when operating at high feed rates. These relatively large position errors leave some areas of a part with burrs and other areas damaged by the tool digging too deeply into the part. 

Compliance prevents most of these issues. With active compliance (denoting advanced control of the pneumatic operations of the compliant device), researchers have shown robots can limit position errors for simple shapes after a few seconds of operation to about 0.0008 ", according to Chang-Hoon Kim and Jae H. Chung, Stevens Institute of Technology, in their paper published in the Journal of Robotic Systems 22. However, that is not a realistic positioning capability for mass-produced parts with complex shapes.

Some programs allow the user to change the desired cutting forces of the deburring tool at predefined areas of the part, rather than traditional force-feedback methods that are continuously engaged. Two deburring robot builders indicate their robots can achieve an absolute accuracy of ±0.020 " and repeatability of 0.002 ".

A robot holds a part against a large abrasive-filled nylon brush to finish edges.

Keeping Control

Controllers can be used to manually write code for robots. Most have a teach function that allows the user, by putting a tool or same-size cylinder in the robot gripper, to touch points along the contour in the orientation needed to deburr. The robot interpolates points in between the taught points. This approach does not require knowledge of CAD/CAM and is often used for deburring. Simple parts may only require the entry of about 200 touch points, but complex, high-precision parts may require thousands of points.

In-House Solutions Inc., Cambridge, Ontario, provides Robotmaster CAD/CAM-based robot-ready programs for deburring robots. The program allows immediate program path correction and optimization for changing offsets for different deburring tool diameters, tilting the tool for optimal surface contact and checking for typical robot errors, including wrist overtravel and singularity points where specific wrist and tool combinations must be avoided, according to Tyler Robertson, Robotmaster applications engineering. Furthermore, the part can be positioned inside the robot’s work space, ensuring proper robot reach and posture, including allowing the robot to travel down a track to finish long parts. The software was developed by Jabex Technologies Inc., Montreal, which produces the native robot language for ABB, Staubli, Motoman and other common robots.

Small Innovations

Success in any deburring operation is often the result of small innovations. The same applies to robots. For example, a deburring tool was developed by Italian researchers that, when placed on the end of a robot, prevents damage from the tool digging into a part surface. Ball bearings at the top and bottom of the tool guide the part contour. The bearing even works on chamfers, although chatter is still a concern. The burr must be oriented in the correct face for this approach to work and it does leave some smaller burrs, but that may be adequate for many products—particularly if the edges are brushed afterwards. The tool is limited by part geometry (tight changes in direction, cutouts and related areas smaller than the cutter bearings).

Another approach is to use robotic waterjet deburring. When the only part surface requirement is to remove loose or easily removable material at part edges (Table 1, Level 6), watertight robots equipped with waterjets can provide the necessary edge quality. Unlike waterjet cutting heads, which commonly operate at 50,000 psi and higher, these systems use pressures as low as 3,000 to 10,000 psi. The water blast does not cut but rather knocks off loose material. Because the pressures are so low, the waterjet does not chamfer or affect any dimensions on most parts. However, the waterjet may blunt or dull edges of soft materials. 

Waterjet deburring reaches deep features and difficult-to-reach locations. Dirt, chips and other loose residue are removed along with the thin burrs. Robots also have the potential to combine waterjet deburring and brushing or other hard tools, if water effects on toolholders can be accommodated.

Another approach is to use on-machine deburring. Machining and deburring the part in one operation reduces handling damage, errors and delays. While some companies still move parts multiple times to utilize machines that are less expensive to operate, getting completed parts out the door in the quickest time is becoming the norm.

Robots are good at doing the same thing to every part. That makes them useful for long part runs. Keep in mind that long runs may mean doing as few as 10 or 20 of the same parts each month for long periods of time. Specifically, parts like jet engine turbines will have runs that last for years, even though production quantities may be small in later years of the contract.

A long-shank rotary bur deburring holes.

Not for Everyone

The author estimates there are approximately 300 to 500 deburring robots operating in the U.S. today. That represents less than 4 percent of all the deburring machines used in U.S. manufacturing. The author also estimates that up to 3,000 deflashing and trimming robots are used in the U.S. on cast and plastic parts.

Robotic deburring can be useful in certain applications, but it is not for everyone. Consider the following statement from AVR’s Arrien: “Try all the simpler solutions before you try to use robotics. Robotic deburring has been touted as a great time saver that provides perfect quality. That is not the reality for many applications. You also will want to pick a system integrator that has provided deburring robots for your type of parts. Aerospace regulations limit what you can do and the same is true in the medical manufacturing arena. You may get a system that works, but is not usable for regulatory or other reasons.”

Robotic controls continue to advance and, as they do, the capabilities of robotic deburring machines will increase. Researchers have been developing better controls for the past 25 years using force feedback, vision, acoustic emission and fuzzy logic. Robots have become more predictable and cheaper to install. 

Precision edge finishing is challenging for all processes. Robots can be an effective tool for commercial edge requirements, but precision applications will require a lot of local ingenuity. Their growth in deburring applications has been slow, but it is increasing. CTE

About the Author: Dr. LaRoux K. Gillespie has a 40-year history with precision part production as an engineer and manager. He is the author of 12 books on deburring and over 220 technical reports and articles on precision machining. He can be e-mailed at laroux1@earthlink.com.

Contributors

AVR Vision & Robotics Inc.
www.avr-vr.com
(514) 788-1420

In-House Solutions Inc.
(519) 658-1471
www.inhousesolustions.com

Key questions about robotic deburring 

Questions parts manufacturers should ask builders when considering robotic deburring:

• Can I talk with an existing robot user to understand the issues they faced?
• How many machined part deburring—not deflashing—robots have you installed?
• Are tool center-point registers or other means available to accommodate automatic tool offsets after predetermined intervals?
• What curvilinear path limitations does the robot have?
• What sensor inputs will the robot accept to control path and cutting conditions?
• How accurate is the robot?
• How repeatable is the robot?
• How much does the robot drift in an 8-hour period?
• How does the operator change tools?
• Do your robots produce true circular paths, if required?
• Will the system always translate and rotate data to allow deburring of similar features on other axes or in other planes to save programming time?
• Do all robot axes respond automatically to keep tools perpendicular to workpiece?
• What offline programming is available?
• Do all the edges have to be finished to justify the installation or only specific ones? 

Questions the robot builder should ask the company considering robotic deburring:

• Is your intended use dedicated to a single part or many different parts?
• How repeatable are your burr thicknesses?
• How repeatable must the finished edges be?
• How heavy are your parts and how heavy are the toolholder/tool combinations? 
• Will your own CNC system do the same thing you are asking the robot to do?
• What skill levels are required to operate and maintain the robot and tooling?
• What metals will be deburred?

—L. Gillespie