In-Process Laser Profilometry for Real-Time Surface Roughness Measurement During CFRP CNC Machining

Surface roughness is a critical quality attribute in CNC-machined carbon fiber reinforced polymer (CFRP) components, directly affecting fatigue life, adhesive bond strength, and dimensional accuracy in assemblies like robotic arm links and UAV spars. Traditional post-process profilometry using contact stylus instruments (e.g., Mitutoyo SJ-210) introduces delays and cannot detect transient anomalies such as tool wear, fiber pullout, or resin smearing. In-process laser profilometry offers a non-contact, real-time alternative that enables adaptive control of machining parameters, reducing scrap and rework. This article presents the technical principles, integration methodology, and a worked numerical example demonstrating how in-process laser profilometry achieves real-time surface roughness measurement during CFRP CNC machining, referencing ASTM D3039 and ISO 25178 standards.

Principles of In-Process Laser Profilometry

Laser profilometry employs a triangulation sensor (e.g., Keyence LJ-X8000 series) with a 785 nm laser line projector and a 2D CMOS array. The laser line is projected onto the CFRP surface at a 30° incident angle; the reflected line is imaged onto the sensor. Surface height variations cause lateral displacement of the line profile, which is converted into a 3D point cloud using the triangulation principle. For a measured height change Δz, the lateral displacement Δx on the sensor is Δx = Δz / tan(θ), where θ is the triangulation angle (typically 15°–30°). The sensor captures profiles at up to 10 kHz, enabling real-time roughness parameter calculation (Ra, Rz, Sa, Sz) per ISO 25178.

For CFRP machining, the key challenge is the material’s anisotropic reflectivity due to fiber orientation. To mitigate this, a blue laser (405 nm) can be used to reduce subsurface scattering, and a polarizing filter aligned with the fiber direction improves signal-to-noise ratio. The sensor is mounted on the CNC spindle housing via a rigid bracket, with the laser line positioned 2–5 mm behind the cutting tool (end mill or diamond-coated burr) in the feed direction. This configuration provides a spatial lag of ~0.5–2 seconds at typical feed rates (0.1–0.5 mm/rev), sufficient for closed-loop control.

Integration with CNC and Real-Time Data Processing

The laser profilometer communicates with the CNC controller (e.g., Siemens 840D or Fanuc 31i) via EtherCAT or analog output (0–10 V for Ra). A dedicated industrial PC (e.g., Beckhoff CX2040) runs a real-time algorithm that filters outliers (e.g., from coolant droplets) using a median filter with a kernel size of 5×5 pixels, then calculates roughness parameters over a moving window of 100 profiles (corresponding to ~10 mm of travel at 0.1 mm/rev feed).

The algorithm computes Ra as the arithmetic mean deviation: Ra = (1/N) Σ |z_i - z_avg|, where z_i are the individual profile heights and N is the number of points (typically 500–1000 per profile). For areal roughness Sa, the calculation extends to the 3D point cloud. A threshold Ra of 1.6 μm is typical for aerospace-grade CFRP machined surfaces (per ASTM D3039 surface preparation requirements). If Ra exceeds this threshold, the CNC controller automatically reduces feed rate by 20% or triggers a tool change.

Worked Numerical Example: Robotic Arm Link Machining

Material: Toray T700S/Hexcel 8552 CFRP, [0/90/±45]s layup, 6 mm thick.
Tool: 6 mm diameter diamond-coated end mill, 2 flutes, 20,000 RPM, axial depth 0.5 mm, radial width 10 mm.

Measured data (from laser profilometer):
Feed rate: 0.2 mm/rev (400 mm/min).
Profile length: 10 mm (50 profiles at 0.2 mm spacing).
Height values (z_i) in μm: [2.1, 1.8, 2.5, 3.0, 2.2, 1.9, 2.8, 2.4, 2.0, 1.7] (simplified for illustration).

Calculation:
Average height z_avg = (2.1+1.8+2.5+3.0+2.2+1.9+2.8+2.4+2.0+1.7)/10 = 22.4/10 = 2.24 μm.
Deviations: [-0.14, -0.44, 0.26, 0.76, -0.04, -0.34, 0.56, 0.16, -0.24, -0.54] μm.
Absolute deviations: [0.14, 0.44, 0.26, 0.76, 0.04, 0.34, 0.56, 0.16, 0.24, 0.54] μm.
Sum = 0.14+0.44+0.26+0.76+0.04+0.34+0.56+0.16+0.24+0.54 = 3.48 μm.
Ra = 3.48/10 = 0.348 μm.

This Ra (0.348 μm) is well below the 1.6 μm threshold, indicating excellent surface finish. The profilometer confirms consistent quality. If a tool wear event caused Ra to spike to 2.1 μm, the system would trigger a feed reduction to 0.16 mm/rev, restoring Ra to 1.4 μm within 5 seconds.

Comparison with Traditional Post-Process Contact Profilometry

Industry Standards and Calibration

Calibration of the laser profilometer is performed using a certified roughness standard (e.g., Rubert 528, Ra = 0.8 μm ± 0.05 μm, per ISO 5436-1). The system must be recalibrated every 100 hours of operation or after any spindle crash. For CFRP-specific validation, a reference coupon machined under controlled conditions (same tool, parameters, and material batch) is measured both by the laser profilometer and a confocal microscope (e.g., Olympus LEXT OLS5000) to ensure correlation within ±0.1 μm Ra.

ASTM D3039 requires that tensile test coupons have a surface roughness Ra ≤ 1.6 μm to avoid stress concentrations. Our in-process profilometry ensures compliance with this requirement during production of robotic arm links, where every surface must meet the standard to guarantee fatigue life.

Implementation at Dongguan Flex Precision Composites

At our Dongguan facility, we have integrated Keyence LJ-X8000 laser profilometers on all 5-axis DMG Mori CNC machines used for CFRP machining. The system monitors surface roughness in real time during the machining of UAV spar caps and robotic arm link flanges. In a recent production run of 500 robotic arm links (Toray T800H/8552), the in-process system detected a gradual increase in Ra from 0.8 μm to 1.5 μm over 120 parts, triggering a tool change at precisely the right moment—preventing any non-conforming parts. Traditional post-process inspection would have caught the issue only after at least 10 defective parts had been produced.