Borescope tube ID inspection: video and laser ovality method

A risk-based inspection on a waste-heat boiler has to say something about every tube: 12 m of horizontal 78 mm bore, 6 mm wall, expanded and seal-welded into the tubesheets, with access only from the cold-side tubesheet face. The optical answer is a side-view head on a long insertion tube — one video channel for the state of the surface, one laser ring for geometry. This guide sizes that head: a worked optics example, what the laser ring actually resolves, and the point where eddy current takes over.

What optics find, and what they cannot

An optical head sees the inner surface and nothing behind it. That still covers most of the scope: corrosion scale and deposits, erosion grooves, pitting, dents and ovality, partial blockages, and the visible face of the seal weld. It cannot see remaining wall thickness, subsurface metal, or anything hidden by the ceramic ferrules and refractory on the hot side. Cracking inside the expanded joint is invisible to it by construction, so the plan splits by defect type.

The method below is the optical half: video for condition, laser ring for numbers.

Two sensors, one side-view head

The head combines two channels behind a 90° side-view window. The video channel frames the wall and records condition: scale, erosion, pitting, deposits. The laser channel projects a ring onto the wall perpendicular to the bore axis; the camera sees that ring as an ellipse whose radius, angle by angle, is the bore's radius profile at that station. Fit a circle to the recovered profile and the residuals become dents, the fitted diameters become ovality, and the radius change since the last outage becomes wall loss where the OD is unchanged.

Laser triangulation profilometry inside pipe bores is an established technique: published systems measure deformation and corrosion in bores of the 100 mm nominal class, and this 78 mm example is the same problem slightly tighter. The triangulation core is a compact laser profiler of the kind in the 3D sensors range, with the probe's folded optics putting its line on the wall; the same profile line drives the method in surface defect detection with a 3D laser line profiler. Both cards below are 3D laser profile sensors; the folded head's stand-off and measuring range pick between them.

Optics for the bore: the worked numbers

Treat the imaging side as a close-up macro problem. The wall sits 39 mm from the bore axis — half of 78 mm — and with the head's window and folded mirror, the working distance from lens to wall lands near 50 mm. Frame the whole bore plus a margin of wall: a 90 mm field across. The example sensor is a 1-inch-format 20 MP mono part with 5496 × 3672 pixels of 2.4 µm.

So the head needs a wide-angle lens of about 6.4 mm focal length working at 50 mm. Two checks close the design:

• Sampling: 90 mm across 5496 pixels is 16.4 µm per pixel. Around the bore that is roughly π × 78 mm ≈ 245 mm of arc, about 14,900 pixels of circumference, so 1 mm of arc spans some 61 pixels.
• Lens coverage: a 6.4 mm lens must carry an image circle over the 16 mm diagonal of a 1-inch format. If the lens you hold covers a smaller format, fit the smaller sensor and run the two lines above again.

The calculator below re-runs the two lines for any bore diameter:

For bore-specific optics, the lens options for inside-bore work are covered in 360° inspection lenses for inside-bore defect checks.

What the laser ring resolves

Depth resolution comes from triangulation geometry, not from pixel count. If the projector and camera axes meet at angle θ, a radial depth change δz shifts the ring image by δx = δz × tan θ:

At 16.4 µm per pixel and a line centroid good to 0.2 px, δx ≈ 3.3 µm, which gives δz ≈ 5.7 µm at θ = 30° and δz ≈ 9.1 µm at θ = 20°. Those are idealized figures: speckle, glare from corroded steel and probe tilt set the real error, and they are larger. Design to a validated radial repeatability of 0.1 mm on a reference ring gauge — a factor of ten and more above the ideal number, and enough for this job.

Worked through to decisions:

• Ovality: fitted diameters of 78.6 mm and 77.8 mm against the 78 mm nominal give (78.6 - 77.8) / 78 × 100 = 1.0%, an out-of-round of 0.8 mm that sits far above any realistic error figure here.
• Wall loss: an ID grown by 0.5 mm is 0.5 mm of metal gone where the OD is fixed: 0.5 / 6 = 8.3% of the 6 mm wall, leaving 5.5 mm.
• Dents: a 0.5 mm flat is many times the idealized depth estimate, so it cannot hide.

Converting the ring profile to coordinates is the step described in 2.5D height map to 3D point cloud; the interpolation choices behind sub-pixel depth are compared in sub-pixel depth sampling: bilinear interpolation or triangulation. To attach each tube's findings to plant coordinates, follow the approach in corrosion coordinate mapping for internal vessel inspection.

Exposure, focus and capture along the tube

Capture at stations: stop, shoot, advance, with the same station list for every tube so the runs compare. If you image during motion instead, blur sets the exposure: at a pull speed of 50 mm/s, two pixels of blur is 2 × 16.4 µm = 33 µm, so exposure must stay at or under 33 µm / 50 mm/s = 0.66 ms. Stationary capture carries no such limit, and a second of integration makes the LED budget easy.

Focus depth is the second constraint. With the permissible circle of confusion at two pixels (c = 4.8 µm) and m = 0.147:

At f/8 that is 2 × 8 × 4.8 µm × 1.147 / 0.0216 ≈ 4.1 mm, enough to hold focus across ±2 mm of probe decentering, the realistic positioning error at this diameter. At f/4 it halves to about 2.0 mm. Run f/5.6 to f/8 and let close-coupled LEDs carry the light: a 12 m bore of dark steel is a lighting problem first, and the layout options are laid out in cylindrical bore inspection lighting. The brightness difference between a live stream and a long still exposure is explained in video exposure and frame rate.

Set-up: camera, lens, light and controller

The imaging channel wants the class in the card below: 20 MP mono, 1-inch format, USB3, C-mount for the wide-angle 6.4 mm lens. At 16.4 µm per pixel that sensor delivers the 14,900-pixel circumference of the worked example.

The triangulation channel takes one of the two profiler cards above. A narrow bandpass filter matched to the laser wavelength, fitted to the triangulation camera, strips the video channel's LED light out of the ring images and lets both channels run at once.

Video-channel illumination is an LED ring light at the tip. A vision controller beside the tubesheet logs both streams against the station list; the camera finder filters the imaging side by format, resolution and interface, and the full range sits under area scan cameras.

Where optics stop, and what takes over

The hard boundary is the hot face. Refractory cladding and the ceramic ferrules in the tube mouths cover the wall beneath them, so an optical pass from the cold side cannot clear that metal, whatever its resolution. Cover the hidden lengths volumetrically: remote-field eddy current runs from the tube ID, reads past the expansion and along the full 12 m, and is the standard answer for tube-to-tubesheet cracking. IRIS returns a wall-thickness profile on water-filled tubes but finds no cracking, and the ferruled length stays outside its reach too. On the cold face itself, clean and dye-penetrant the visible seal-weld toe; a helium leak test earns its place only where shell-side access exists.

Hardware follows the environment if the route ever moves toward the hot end: match the probe's temperature rating to the wall before any approach, which is the subject of high-temperature borescope options for hot pipe. Where the run is tighter or bent than a straight 78 mm path, laser optic tube inspection for heater tubes walks through matching laser-optic heads to tube size.

The decision for this boiler, stated plainly: optical head down every tube from the cold face for corrosion, erosion, ovality and dents; dye penetrant on every seal-weld face in reach; RFET for cracking and wall trend through the expansions; IRIS only where the water handling and the ferrules allow.

Checks before the outage

• Reference ring gauge at the nominal bore with a machined 0.5 mm step and about 1% ovality: the head must recover both on the bench and again at the tubesheet face.
• Decenter test: offset the probe ±2 mm off axis and confirm the fitted diameter moves less than your acceptance band; the f/8 depth of field above is sized for exactly this.
• Depth zero: capture a flat reference before each session and log it, the same logic as in depth camera flat plane distortion.
• Station list: identical stations and order for every tube, so outage-to-outage trends line up.

Parts and next step

If you are sizing a head for an upcoming outage, start the camera configurator with the bore diameter and working distance from your drawings; the two formula lines above fall straight out. The applications team will check lens coverage and the profiler choice against your tube layout — ask through the contact page. More inspection set-ups of this kind are collected on the applications topic page.