Robotic Sanding, Grinding, and Polishing: Force Control, Tool Wear, Dust Extraction, and Surface Quality

A robot is taught to sand a part by recording a manual operator's path. The path is played back, and the robot follows it precisely. The first part looks good. The second part is under-sanded on one side. The third part has gouge marks. The path has not changed, but the part has. The part has a casting draft angle that varies by 0.5 mm between units. The robot, following a fixed path, either presses too hard where the part protrudes or too lightly where it recedes. A manual operator feels the resistance and adjusts pressure instinctively. The robot does not. This article covers what it takes to automate sanding, grinding, and polishing, focusing on force control, tool wear, dust management, and surface quality verification.

Define the Surface Outcome Before Selecting the Robot

The first question is not “which robot?” but “what surface result is required?” The answer determines the process:

• Roughness (Ra, Rz): Measured with a profilometer. Required for sealing surfaces, bearing journals, and cosmetic finishes.
• Gloss or reflectance: Measured with a gloss meter. Required for cosmetic and optical surfaces.
• Material removal: The amount of material to be removed, measured by weight or dimension. Required for deburring, gate removal, and dimensional correction.
• Deburring and edge breaking: Removing sharp edges and burrs. The edge radius or chamfer may be specified.
• Visual quality: Appearance standards, including scratch direction, uniformity, and freedom from defects. May be specified by reference samples rather than numerical measurement.

The surface outcome must be measurable. “Smooth” is not a specification. “Ra 1.6 max” is. The buyer should define the measurable quality criteria before evaluating automation, because the criteria determine what process, tool, and verification method are needed.

Why Constant Position Does Not Mean Constant Contact Force

A robot following a fixed path maintains a constant position relative to the part. If the part surface varies (due to casting tolerance, forming variation, or thermal distortion), the distance between the tool and the surface changes. A 0.5 mm variation in part geometry can mean the difference between the tool pressing too hard (gouging the surface) and not pressing at all (no sanding).

Robot repeatability versus accuracy explains the motion parameter that cannot by itself guarantee a surface finish.

Two approaches address this:

• Passive compliance: A spring-loaded or compliant tool holder that allows the tool to float, maintaining contact force through spring pressure. Simple and inexpensive, but the force varies with deflection and cannot be adjusted during operation. Suitable for parts with small, predictable variation.
• Active force control: A force/torque sensor and servo-controlled axis that maintain a constant contact force regardless of part variation. More expensive and complex, but can adjust force in real time and can vary force by location on the part. Suitable for parts with larger or less predictable variation.

Active vs passive compliance

Passive compliance handles geometric variation by allowing the tool to move. It cannot handle changes in the required force (e.g., more force for roughing, less for finishing) without mechanical adjustment. It cannot detect when the tool has lost contact or is pressing too hard.

Active force control can maintain a constant force, vary the force by program, and detect force anomalies (e.g., the tool hitting a fixture edge). It requires a force sensor, a controllable axis (either the robot or a separate linear axis), and the integration to close the control loop. The robot's force control capability, the sensor's performance, and the control bandwidth all affect the result.

Neither approach makes the robot “naturally constant force.” Both require engineering and verification.

Abrasive, Spindle, and Tool Process Constraints

The abrasive tool determines the removal capability:

• Grit size: Coarser grit removes material faster but leaves a rougher surface. Finer grit produces a smoother surface but removes less material per pass.
• Abrasive type: Aluminum oxide, silicon carbide, ceramic, diamond. Each has different wear characteristics and material compatibility.
• Tool form: Disc (random orbital or rotary), belt, wheel, flap wheel, buffing pad. Each produces a different surface pattern and removal profile.
• Spindle speed: Affects the cutting action and the surface finish. Too fast may burn the surface; too slow may be inefficient.
• Feed rate: How fast the tool moves across the surface. Combined with spindle speed and pressure, determines material removal rate.

These parameters come from the tool supplier's technical data and the process verification. They are not generic values that apply to all materials and applications.

Path Overlap, Edges, Corners, and Dwell

The robot path must account for:

• Overlap: Adjacent passes must overlap to avoid leaving ridges between passes. Too little overlap leaves lines; too much wastes time.
• Edges and corners: The tool may round edges unintentionally if the path runs over them. Edges may require a different tool, angle, or path strategy.
• Dwell: If the robot pauses or slows at a position (e.g., at a corner or direction change), the tool removes more material at that position, creating a depression.
• Orientation: The tool's angle relative to the surface affects the contact area and the removal pattern. Complex curved surfaces may require the tool to maintain a specific angle relative to the local surface normal, which requires multi-axis coordination.

Tool Wear and Compensation Strategy

Abrasives wear. A new sanding disc removes material at one rate; a disc that is 50% worn removes at a different rate. If the robot path and force are constant, the surface quality changes as the tool wears.

Wear management strategies:

• Fixed tool life: Replace the abrasive after a defined number of parts or cycles. Simple but may waste abrasive (replacing a disc that still has life) or produce inconsistent quality (using a disc past its useful life).
• Wear logging: Track the cumulative use of each abrasive and trigger replacement at a defined threshold. Requires a tracking system but ensures consistent quality.
• Force or power monitoring: Monitor the spindle power or contact force. As the abrasive wears, the power or force characteristic changes, indicating that replacement is needed.
• Compensation: Adjust the path or force to compensate for wear. For example, increase the number of passes or the contact force as the tool wears. This extends tool life but requires a control system that can make the adjustment.

Robot calibration management can track when tooling, force sensing and programmed references require reassessment after changes.

New vs worn abrasive

The system must be tested with both new and worn abrasives. A system that produces acceptable quality with a new disc may fail with a worn disc. The buyer should test with tools at various wear states and define the replacement trigger based on the quality result, not on a fixed cycle count.

Dust, Sparks, and Extraction Dependencies

Sanding, grinding, and polishing generate dust, sparks, and debris. These affect:

• Sensors: Dust on a force sensor, vision camera, or laser scanner can cause measurement errors or signal loss
• Cables: Abrasive dust is abrasive to cable jackets, accelerating wear
• EHS: Airborne dust is a respiratory hazard. Combustible dust (e.g., aluminum, magnesium) is an explosion hazard.
• Equipment: Dust accumulation on slides, bearings, and motors causes premature wear

Dust extraction is not a robot feature. It is a facility system that must be designed for the specific process. The extraction system must capture dust at the source (near the tool), convey it through ducting, filter it, and discharge clean air. The extraction capacity must match the dust generation rate of the process.

OSHA 29 CFR 1910 (General Industry) and 29 CFR 1926 (Construction) address occupational dust exposure, local exhaust ventilation (LEV), and personal protective equipment (PPE) requirements in the United States. The UK's HSE (Health and Safety Executive) publishes LEV guidance. These standards define the regulatory framework for dust management. They do not design the extraction system for a specific robot cell. The buyer's EHS team and the extraction system supplier must determine the appropriate extraction for the specific process, material, and facility.

Fixture Stiffness and Workpiece Variability

The fixture holds the part while the robot applies force. If the fixture is not stiff enough, the part deflects under the sanding force, changing the contact geometry and potentially producing an uneven surface. If the part is a thin-wall casting or a sheet metal component, the part itself may deflect.

Fixture considerations:

• Stiffness: The fixture must resist the sanding force without significant deflection. The required stiffness depends on the process force and the part's sensitivity to deflection.
• Part variation: If parts vary in size or shape, the fixture must accommodate the variation without compromising holding force. A fixture that holds the nominal part perfectly may not hold the worst-case part.
• Soft areas: If the part has areas that cannot be clamped (e.g., thin walls, finished surfaces), the fixture design must avoid those areas or use compliant pads.

Force control does not eliminate the need for a stiff fixture. If the part moves under the tool, the force control system may interpret the movement as a force change and adjust incorrectly.

Surface Quality Verification Methods

The quality verification method must match the surface specification:

• Profilometer (Ra, Rz): Contact or optical measurement of surface roughness. Provides numerical data for comparison to specification.
• Gloss meter: Measures specular reflectance at a defined angle. Provides numerical gloss units.
• Visual inspection: Comparison to reference samples by a trained inspector. Subjective but may be the only method for complex visual criteria.
• Dimensional measurement: CMM or laser scanner to verify that material removal has not exceeded dimensional tolerances.
• Edge inspection: Measurement of edge radius or chamfer, or visual comparison to reference.

Buyer-defined roughness or visual targets

The quality criteria must come from the buyer's product specification, not from the robot supplier. The robot supplier can verify that the system can achieve a specific roughness under specific conditions, but the buyer defines what roughness is acceptable for the product. If the buyer does not have a measurable specification, one should be developed before automation is attempted. Automating an undefined process produces inconsistent results.

Representative Pilot and Tool-Life Matrix

Before approving a sanding, grinding, or polishing system, the buyer should test:

• New, mid-life, and end-of-life abrasive: To verify that the system produces acceptable quality across the tool's life
• Typical and worst-case parts: To verify that the system handles part variation
• Production cycle time: To verify that the system achieves the required throughput, including tool changes and cleaning
• Dust and cleaning: To verify that the extraction system handles the dust generated during the test

The test matrix should record the surface quality result, the tool condition, the cycle time, the manual intervention required, and the cleaning time for each combination.

The wider robot pilot acceptance criteria framework can be used to document the Go, Modify or Stop decision after surface-finishing trials.