Automated grinding

What if you could have a robot assist a repetitive manual-labor task such as deburring holes on turbine blades? Steven Somes, president of Cleveland-based startup company Force Robots, asked himself that question, and it led him to develop a unique solution that reduces cycle times and results in more consistent results.

Somes, a controls engineer, was doing automation for a large industrial company when he attended grad school at Case Western Reserve University. There, he became interested in robotics and force control. His goal was to determine what was needed to make an industrial-class robot that could magnify torque while having direct-drive performance – in other words, perfectly smooth operation. To reduce inertia, he’d have to make it as light weight as possible, while still retaining high torsional stiffness.

“And it had to generate only as much force as you’d expect from a person,” Somes adds.
 

From idea to machine

Somes’ quest took 12 years to get from the idea to the prototype machine. He says National Science Foundation grants helped the process, as did funding from the Ohio Third Frontier, an internationally recognized technology-based economic development initiative.

The fruit of Somes’ and his small team’s labors is the Touch Robot, a system that performs precision grinding and machining to polish, deburr, and deflash cast and forged parts. This process has proven extremely difficult to automate, plus metal finishing with hand tools is a difficult task that can risk repetitive stress injury.

Force Robots’ system “combines the precision of a machine with the finesse of the human hand,” Somes says.

It is designed to feel existing part contours, match those to a CAD reference, and work autonomously to remove material to specification.

The Touch Robot is self-contained, portable, and can be carried through a standard-size door. It requires only 120VAC and shop air to operate. A 4-axis material-removal arm and a 2-axis part-positioner are mounted to the 1.2m x 0.8m work table. Dividing the system’s six degrees of freedom between the two coordinated mechanisms is the heart of the device – it preserves a soft touch of the tool arm while allowing heavy castings up to 0.4m long to be manipulated. No force or optical sensing is required for it to work; its feedback is derived from the motor encoders.

Its brushless, slot-less, 24VDC drive motors keep joint speeds below a level where the arm could endanger an operator. This modest force capacity and low-friction, back-drivable joints make it easily overpowered by a human. Joint limits restrict its reach to 0.5m, allowing it to be safely deployed alongside manual metal finishing cells.

Somes compares the system to a labor-saving appliance. “Operators can task it with the heaviest, most difficult material removal work, while they focus on fine finishing and inspection. The key to performance is low friction, perfectly smooth action, just like a human arm,” Somes explains.

To achieve this smooth operation, the Touch Robot uses stainless steel, aircraft-grade cables, pre-stressed and nylon-coated, that wind around a capstan with a 24-to-1 transmission ratio that imparts a magnifying torque and allows ±50° of travel. Cables work better than a belt drive, which has elasticity and needs tensioning, Somes adds.
 

Operation

One button releases the servo control and allows an operator to move the arm. It opposes gravity by knowing the torque to apply and friction to hold the arm in place. The operation begins with the device locating the workpiece with contact measurements, using the material removal tool as the probe. It takes only seconds to determine a part’s height and location, and the CAD model mates to the touch point to determine the motion path.

Machining and grinding passes are made with optimal contact force. With part geometry determining the toolpath, the control software can identify the location and amount of excess material by comparing the tool trajectory to a CAD model reference. Tool trajectories generated automatically focus effort on the part until the measured surface contour matches the CAD specification. The system can deliver results despite process variations that can include part fixturing, the amount of material needing to be removed, and the changing size and efficiency of the material removal tool. Despite initial geometry and positioning variation, the material removal is accurate within a few tenths of a millimeter, due to the common datum reference between measurement and material removal. The absolute location of the workpiece is not important, only that the robot can discover it in the context of its material removal tool.

In the off mode, the robot disconnects the servo and brakes lock the arm in place, while software maintains the tool position.
 

In use

Turbine blade finishing is challenging because the parts require perfect surface smoothness while tolerating the dimensional variations resulting from the casting process. The lack of easily referenced datum surfaces for locating tooling rules out machining to fixed coordinates. However, while part geometry varies widely, the casting artifacts requiring finishing are generally few, such as parting lines, pin blips, and core exit flash. The Touch Robot’s software predetermines strategies for addressing these features on a wide array of parts. An operator generates part programs by filling in blanks on sequential function blocks using a web browser. Manually guiding the robot and part to the desired positions identifies the material removal tool poses and transition waypoints. Areas requiring finishing are designated by annotations on the part’s CAD model within the software. Tool trajectories, contact forces, and performance metrics are dynamically generated at run time with the user-provided parameters, geometry derived from the CAD model, and self-acquired part measurements.

The Eastlake, Ohio foundry of Consolidated Precision Products (CPP) is the first company to use the Touch Robot to remove excess material on difficult-to-reach areas of tough, precision-cast turbine engine components.

The test project involved deflashing or opening up the trailing edge exits of a nozzle ring casting, with each APU disc containing 22 vanes. In grinding small turbine blades, it is tedious for operators to remove the excess material – the flash – so, the robot tracks the edge exit holes where flash needs to be removed. The robot determines contact point and pressure, touches off all 22 airfoils to establish the part height, then grinds away the flash in sequence so the tool is not always powering up and down. Machining removes material to within 0.005" leaving just a small excess for an operator to finish up.

“Since the robot knows where it is grinding – more than a person could control and repeat – you can use aggressive tools, not just abrasive stones,” Somes notes.

CPP makes multiple vane segments with different alloys, soft and hard, and they’re slightly different in size. The robot can distinguish differences in each by the way it cuts.

“This kind of contouring is not easy,” explains Albert Osagie-Erese, senior product engineer with CPP. “We needed a fixture and program to do what was required and Steve was able to develop a program and make it available to us.”

After Somes did a demonstration onsite, other workers were qualified on the robot.

“Two years ago, the workers finished approximately 150 castings per day,” Osagie-Erese says. “With the robot, they can do 175.”

Using a hand deburring tool alone on the 22-vane nozzle ring took 45 minutes to 1 hour, but by using the robot, that is now reduced to 20 minutes per part.

“It’s all about the cycle time,” says Somes. “Three operators can do the operation now, and without the wrist wear-and-tear, reducing repetitive motion injury.”

The production increase is timely, since the customer recently increased its ring nozzle order, up to 1,500 rings per year. “We can increase the number of parts finished without increasing staffing,” Osagie-Erese points out.
 

Tool wear sensing

“Most robots can’t tell if tool is wearing, but as ours wears down it can calculate,” Somes says.

Cycle time goes up, and this tracks to the software. As a tools dulls, the software generates a 1-to-10 scale; when it drops to 3, it’s time to change it.

The carbide burrs with titanium coating now last for 30 or 40 parts, versus only one before. Dry machining with compressed air keeps tools running cooler, so now it’s possible to do two parts instead of just one-third-part per tool.

“From my point of view, this has been a successful project,” Osagie-Erese says.

Force Robots LLC
www.forcerobots.com

About the author: Eric Brothers is senior editor of Aerospace Manufacturing and Design and can be reached at 216.393.0228 or ebrothers@gie.net.