In the fairy tale "Goldilocks," the girl wanted the porridge to be “neither cold nor hot—just right.”

At the robotic polishing site, engineers face the same dilemma—but at a much higher cost: what we’re striving for is absolute precision in “material removal.”

 

• A little more strength, or staying a second longer: A costly workpiece has developed an irreparable deep pit on its surface, commonly known as "overcut." If it’s an aviation blade or an automotive body panel, that single mistake can render thousands—even tens of thousands—of dollars wasted.
• A little less strength, or one millimeter farther away: The weld seam wasn't ground smooth, and burrs remain—commonly referred to as "under-grinding." This results in the product having to be reworked, and in some cases, even making its way to customers and triggering complaints.

Fine-tuning the polishing process is, at its core, about... Walking a razor-thin tightrope between the precipice of “overcutting” and the abyss of “undergrinding.”

Why does manual polishing have fewer problems? Because humans have eyes and tactile senses, allowing them to provide real-time feedback. In contrast, robots are “blind.”

Today, we’re going to take this steel wire apart and, through an in-depth Q&A, show you how to teach robots to strike the perfect balance.

---

 

Q1: Why does “overcutting” occur?

 

A ： “Overcutting” is usually not caused by the robot being “too strong,” but rather by it being “too sluggish.” or “Too greedy.”

 

During our on-site investigation, we found that 90% of “overcuts” stemmed from the following three hidden logical vulnerabilities:

1. Deadly “Corner Speed Reduction”

 

This is the most common reason.

 

Phenomenon: When grinding smoothly along a straight line, as soon as the workpiece reaches a corner, an arc, or a point where the path reverses direction, a pit immediately appears on the workpiece.

Principle: The laws of physics dictate that for a robot to change its direction of motion, it must first... Slow down 。

If your process setting is: constant pressure (e.g., 30N) + constant rotational speed.

When the robot slows down to take a turn, The grinding time per unit area has increased. 。

Result: The same force, when applied for a longer duration, naturally wears deeper.

Countermeasure:

Smooth transition: Use in robot programs Smooth transition instruction (such as ABB's) Zone , Fanuc's CNT , KUKA's C_DIS ), don't let the robot make sharp right-angle turns; reduce the degree of deceleration.

Process interconnection: The advanced gameplay is to... Robot movement speed and spindle rotation speed or Contact force Perform mapping. If the robot must slow down due to physical limitations, the system should automatically reduce the pressure or spindle speed.

2. Rough “tool entry and exit”

 

Phenomenon: Grinding marks Starting point and End point There are often two deep imprints.

Principle: It’s just like a helicopter landing vertically. If the grinding head “crashes” down perpendicularly onto the workpiece surface, the impact force at the moment of contact is extremely high, and the contact area rapidly increases from small to large, making it very easy to cause instantaneous overcutting.

Countermeasure:

Tangential feed: Imitate an aircraft landing. Let the grinding head follow along the surface of the workpiece. Tangent direction Cut in and gradually apply pressure.

Soft landing: Before contacting the workpiece, enable “position/force hybrid control” and set a relatively low “search speed.” After contact is established, switch to the working pressure.

 

3. Incorrect “path overlap rate”

 

Phenomenon: The surface has an uneven texture resembling a washboard, with some areas being deeper and others shallower.

Principle: To smooth a surface, we need to follow many zigzag paths. If two paths overlap too much, the overlapping area will be polished twice, naturally resulting in a depression.

Countermeasure: Strictly calculate the grinding head. Effective broadband access , a overlap rate of is usually recommended. 20%-30% 。

---

 

Q2: Where does the root cause of “leakage grinding” lie?

 

A ： “Leu Mo” is usually caused by “uncertainty.” The robot is following a standard path, but the real-world situation has changed.

1. The “discreteness” error of the workpiece

This is the biggest pain point, especially for castings and welds.

Scene: The standard parts used in programming are perfect. However, in actual production, the burrs on castings sometimes measure 3 mm high and sometimes 5 mm high; weld seams sometimes shift to the left and sometimes to the right.

Result: The robot is still following the old path, while the burr has “dodged” it—or the burr is too small, and the robot “thinks” it’s making contact when in fact it’s just hovering in midair.

Countermeasure:

Must go “floating”: Use Floating grinding head (passively controlled) or constant-force grinding system (actively controlled) Give the robot a “tolerance range” (such as a floating stroke of ±5 mm). As long as the error stays within this range, the floating mechanism will “push” the grinding head toward the workpiece, ensuring constant and reliable contact.

 

2. The Invisible “Abrasive Wear”

 

Scene: The newly replaced grinding wheel is working quite well. After grinding 50 products, I noticed that it’s starting to miss spots.

Principle: The grinding wheel diameter has decreased (for example, from 100 mm to 95 mm), or the abrasive belt has become thinner. However, the robot’s TCP (Tool Center Point) is still calculated based on the original 100 mm; in reality, the tool no longer touches the workpiece.

Countermeasure:

Software compensation: Establish Wear model For every meter ground, the robot automatically compensates by 0.01 mm in the negative Z-axis direction.

Hardware calibration: Set up a “tool setter” or simply a reference plate. At regular intervals, have the robot press against the sensor to recalculate the actual length of the tool and automatically update the coordinate system.

 

3. “Positioning error” of the fixture

Scene: The worker didn't place the material correctly, or the fixture itself was designed with gaps, causing the workpiece to be off by 1 mm.

Result: The robot “carved the boat to find the sword,” resulting in areas that needed to be ground remaining unground.

Countermeasure: Optimize the fixture design by adding locating pins; or use a simple method before grinding. Laser rangefinder sensor Scan the key points and proceed. Workpiece coordinate system offset 。

---

 

Q3: How can we achieve “neither overcutting nor undergrinding”?

A ： To achieve a fundamental cure, we must move from “blind men touching an elephant” to “aiming precisely at the target.” We need to introduce “perception.”

Option 1: Force Control Technology—Empowering Tactile Sensation

This is currently the most mainstream and mature solution.

Core logic: Instead of commanding the robot to “go to position X,” we now command it to “maintain force F.”

How to address overcutting? When encountering a protrusion (high point), in order to maintain constant pressure (e.g., 20N), the force control system will automatically adjust the robot/grinding head. Backward rather than forcing it through, thereby avoiding overcutting.

How to solve grinding leakage? When a depression (low point) is encountered, the force control system will automatically adjust the robot/grinding head. Forward extension “Track” the workpiece surface until a pressure of 20N is reached, thereby preventing under-grinding.

Applicability: Suitable for sanding almost all curved surfaces. It’s the “Anti-lock Braking System (ABS)” for polishing processes.

 

Option 2: 3D Visual Scanning—Empowering Vision

This is the “trump card” for addressing extremely poor consistency in incoming materials (such as large castings).

Core logic: “Look first, then act.”

The workpiece is in place, and the 3D camera takes a photo to generate a point cloud model.

The software automatically compares the point cloud with a standard CAD model to identify the exact location, height, and allowance of burrs.

Dynamically generated path: The software tells the robot: “This burr is larger than usual—you’ll need to make two extra passes,” or “This time, the position is off by 2 mm to the left—your path needs to adjust accordingly.”

Applicability: Workpieces with high value and large deviations—such as wind turbine blades and high-speed rail bogies. Although the cost is high, this investment is worthwhile compared to scrapping a workpiece worth hundreds of thousands.

 

Option 3: In-machine Measurement and Closed-Loop Control

Core logic: “Grind it out first, then take a look.”

After polishing is complete, the robot switches to a probe (or uses a force sensor) to contact the inspection points. If it detects that the dimensions are still insufficient (indicating under-grinding), it immediately and automatically performs an additional fine-grinding pass.
 

---

 

“Overcutting” and “undergrinding,” though seemingly two extremes, are actually two sides of the same issue: a lack of adaptability.

Traditional automation is “rigid”—it assumes the world is perfect. But grinding deals with the most imperfect workpieces. To address this issue, don’t try to eliminate all errors in the workpiece—doing so would be prohibitively expensive—but rather, focus on... Enhancing the “adaptability” of robotic systems 。

Technically: Optimize the tool entry and exit paths as well as corner speeds to eliminate overcutting caused by kinematics.

Hardware: Must use Floating/Force Control The device eliminates missed grinding caused by workpiece errors.

Management: Do it well Wear Compensation Eliminate the variables introduced by time.

 

By achieving these three points, your robot will no longer be a reckless “brute” who only causes trouble, but rather a well-balanced “craftsman” who combines strength with finesse.