High-speed ultrasonic rail inspection trades pulse density for track time. At 40 mph, a channel firing 1,000 times a second places one pulse every 17.9 mm (0.70 in.) of rail, against 0.9 mm at a 2 mph walking pace. FRA's 49 CFR 213.237 sets no test speed; it requires a valid search.
That leaves the owner to decide how fast is fast enough. This article explains what changes as a test vehicle speeds up: the pulse-rate ceiling set by sound path, the spacing between pulses, coupling, and the verification workload. It summarizes what the published research does and doesn't show about speed and probability of detection, and ends with what an owner should write into a testing contract.
For how continuous and stop-and-verify testing differ under the rules, see continuous vs stop-and-verify rail testing. For the vehicles themselves, see the role of a railway inspection car.
What is high-speed ultrasonic rail inspection, and what limits its speed?
High-speed ultrasonic rail inspection is vehicle-based pulse-echo testing of rail in track, run fast enough that the vehicle doesn't stop at every indication. The probes sit in fluid-filled wheels or sleds on the rail head. Typical beam sets are 0° for horizontal flaws and loss of the base echo, 37° to 45° for bolt-hole cracks, and 70° for transverse defects in the head, plus side-looking probes. Channel layouts vary by system, so treat that set as typical rather than standard.
A 1999 Colorado State report for the Mountain-Plains Consortium, MPC 99-106, described the cars of that era as "capable of speeds up to 30 mph in a non-stop testing mode", but said that in practice they stopped often to hand-verify and "typically average 6 to 8 mph". Rated speed and test speed are different numbers.
FRA's 2020 continuous-testing rule describes four test processes. Its preamble (85 FR 63362) defines them as follows.
| Process | How FRA describes it (2020 preamble) | Where verification happens |
|---|---|---|
| Portable | An operator pushes a test device over the rail "at a walking pace" | At the device, by the operator |
| Stop-and-verify | A vehicle tests "at a slow speed (normally not exceeding 20 miles per hour)" | The vehicle stops and the operator verifies on the spot |
| Chase car | A lead test vehicle runs ahead of a verification chase car | In the chase car behind the test vehicle |
| Continuous | A "high-speed, vehicle-based, test system runs non-stop along a designated route" | Later, by a field crew, within the 49 CFR 213.240 windows |
Three things limit how fast a vehicle can usefully test:
- Pulse rate. Each channel can only fire as often as the sound path allows, so faster travel means wider gaps between pulses.
- Coupling. The fluid film between wheel and rail has to stay intact, and that gets harder as speed rises.
- Verification. Every suspect still has to be checked by a qualified person, and that work doesn't speed up with the vehicle.
Key point: high-speed testing moves the bottleneck. The vehicle can go faster, but each extra mph thins the data and adds verification work.
How do pulse repetition frequency and sound path cap the pulse rate?
Pulse repetition frequency (PRF) is how many times a second a channel fires. A channel can't fire again until the echoes from its last pulse have returned and died down. Otherwise a late echo arrives in the next listening window and looks like a reflector at the wrong depth. Evident's pulser tutorial puts the trade-off plainly: a high PRF "permits faster scanning", while a low PRF "limits wraparound noise when working with very long sound paths" (Evident, Pulser and Receiver Operation).
In rail, the longest sound path is usually the 0° channel looking at the base. The time it needs depends on two legs: the fluid path inside the wheel, and the rail height.
Time budget for one 0° firing (illustrative inputs)
| Leg | Distance | Velocity | Round trip |
|---|---|---|---|
| Fluid path in the wheel (assumed) | 50 mm | 1,480 m/s (water) | 67.6 µs |
| 136RE rail, head to base | 185.7 mm (7-5/16 in.) | 5,920 m/s (steel, longitudinal) | 62.7 µs |
| First base echo arrives | 130.3 µs | ||
| Allowing 3 base round trips to clear (assumed) | 255.8 µs |
At 130.3 µs the theoretical ceiling is about 7,670 firings a second. If the procedure lets three base round trips decay before the next pulse, the ceiling falls to about 3,910 firings a second. Both the 50 mm fluid path and the three-multiple allowance are illustrative assumptions. Real wheel geometry, gate settings and decay allowances come from the system design and the procedure, so a Level III should confirm them for any real system.
That ceiling is per firing sequence, not per channel. If several channels share one pulser and fire in turn, they split the budget. If each channel has its own pulser and they fire at the same time, they can interfere acoustically, which is a design problem of its own. Either way, more channels or longer sound paths mean fewer firings per channel per second.
Key point: PRF isn't a free setting. Sound path sets its ceiling, and channel sharing divides it.
How far apart are the pulses at 40 mph? A worked example
Pulse spacing along the rail is simply travel speed divided by the per-channel firing rate. At 40 mph the vehicle covers 17,882 mm (about 58.7 ft) every second.
Worked example, step by step
- Speed: 40 mph × 1,609.344 m/mile ÷ 3,600 s/h = 17.88 m/s = 17,882 mm/s.
- Firing budget: about 3,910 firings a second for the shared sequence (from the table above, with its assumptions).
- Channels sharing that budget: assume 8 channels per rail fire in turn. Each gets 3,910 ÷ 8 = 489 firings a second.
- Pulse spacing per channel: 17,882 ÷ 489 = 36.6 mm (1.44 in.) between pulses.
- The same 8-channel system at 20 mph gives 18.3 mm, and at 10 mph gives 9.1 mm. With only 4 channels sharing, it gives 18.3 mm at 40 mph.
Pulse spacing per channel (mm) at a given per-channel PRF
| Per-channel PRF | 2 mph | 10 mph | 20 mph | 30 mph | 40 mph |
|---|---|---|---|---|---|
| 500 Hz | 1.8 | 8.9 | 17.9 | 26.8 | 35.8 |
| 1,000 Hz | 0.9 | 4.5 | 8.9 | 13.4 | 17.9 |
| 2,000 Hz | 0.4 | 2.2 | 4.5 | 6.7 | 8.9 |
| 4,000 Hz | 0.2 | 1.1 | 2.2 | 3.4 | 4.5 |
What the spacing means for a small reflector. A pulse only finds a flaw if the flaw sits inside the beam at that moment. Take a 70° shear beam about 8.9 mm (0.35 in.) wide at the flaw, a ray-model figure for a 0.5 in., 2.25 MHz probe at 3 in. of sound path. Because the beam is inclined at 70°, a small reflector stays in the beam for about 8.9 ÷ cos 70° = 26 mm of vehicle travel.
| Pulse spacing | Hits on a small reflector (26 mm ÷ spacing) | What that means |
|---|---|---|
| 8.9 mm (2,000 Hz at 40 mph) | about 2.9 | Several hits; a weak echo can still build a pattern |
| 18.3 mm (8 channels at 20 mph) | about 1.4 | One or two hits |
| 36.6 mm (8 channels at 40 mph) | about 0.7 | A single hit at best, and some passes may miss it entirely |
The beam width is a simplified model, not a measured footprint, so treat the hit counts as an illustration of the trend. Larger flaws stay in the beam longer, and recognition logic differs from system to system. The direction holds, though: fewer hits per flaw leave less evidence for an analyst to separate a small defect from noise.
Published figures follow the same arithmetic:
- A 2025 lab study of a high-speed probe head (Chałko et al., Materials 19(1):28) reports that a 6 kHz rate gave a test about every 5.5 mm at 120 km/h (75 mph). The arithmetic is 33.3 m/s ÷ 6,000 = 5.6 mm.
- TTCI's TD-20-007 (May 2020) tested an electromagnetic surface-crack sensor, not UT, on a hi-rail carriage at a constant sampling rate. Sample spacing grew from 4 mm at 3 mph to 25 mm at 20 mph, and the digest attributed a small loss of sensitivity "mainly" to "reduced scan resolution as a result of increased sample spacing".
Key point: pulse spacing is set by speed, PRF and channel sharing together. An owner should ask for the spacing at the contracted maximum speed, not for the PRF alone.
What happens to coupling at speed?
Coupling is the fluid film that carries sound from the probe wheel into the rail. If that film breaks, the probe goes quiet. A quiet screen means no data, not sound rail. Test vehicles use fluid-filled wheels with couplant applied at the wheel-rail contact; MPC 99-106 describes the coupling fluid as "usually" oil- or water-based. Gels belong to hand probes, not test cars. For couplant selection and cold-weather issues, see ultrasonic couplant for rail inspection.
Speed works against coupling in several ways:
- Film thickness and pressing force. The Chałko et al. lab study measured the drop in base-echo amplitude against speed, pressing force, film thickness and fluid flow rate. It concluded that controlling flow and force kept the signal usable up to 120 km/h. It also observed that at the top speed "a significant amount of the coupling liquid was splashed by the air streams". That was a rotating test wheel in a lab, not track, and it measured echo loss, not detection.
- Field experience. A 2023 paper on the Russian network (Davydkin, Dymkin and Konshina, E3S Web of Conferences 383) states in its abstract that higher test speed "inevitably leads to the loss of acoustic contact and missing defects".
- Rail surface. Shelling, head checks and debris scatter sound at any speed. At speed there are fewer pulses to bridge the gap. TTCI's TD-16-030 also notes that on worn rail, fixed-angle probes can inspect "the wrong volume of material".
| Coupling problem | What the data shows | Consequence under the rules |
|---|---|---|
| Film breaks for a short length | Base echo drops out on the 0° channel; angle channels go quiet | That length isn't validly searched; the operator decides under 213.237(g) |
| Partial coupling (thin film, bounce) | Base echo drops in amplitude; weak flaws fall below threshold | Detection is reduced without an obvious alarm |
| Shelled or heavily checked head | Noise and loss of signal under the surface damage | Masking; slower retest or hand test may be needed |
| Worn head profile | Beam enters at the wrong angle | Coverage of the intended volume isn't achieved |
Under 49 CFR 213.237(g), if the qualified operator determines that a valid search "could not be made over a particular length of track", that length can't be counted as inspected. Under 213.237(h) the owner must then retest, slow trains to 25 mph, or replace the rail. Under 49 CFR 213.241(c)(6) the record must show the location of any track not tested. Loss of coupling is untested rail, not a false call. See false calls in rail UT for how the two differ.
Key point: at speed, coupling loss is the failure that hides. The test report has to show where the signal dropped, or the owner can't know which rail was never searched.
What does the research say about test speed and probability of detection?
No public TTCI or FRA study was found that measures the probability of detection (POD) of conventional contact ultrasonic rail testing as a function of test speed (search, October 2026). The benchmark data that exist were not reported by speed. The speed studies that exist mostly cover prototypes using other physics.
TTCI benchmarking (1998 runs). TTCI's TD-99-006 (Garcia and Reiff, February 1999) benchmarked six hi-rail ultrasonic systems over 49 known defects on the Rail Defect Test Facility (RDTF) at the Transportation Technology Center near Pueblo, Colorado. Overall, 72% of flaws were detected, with per-run results of 65% to 78%. TDs of 20% of head area or smaller were found 66% of the time, and TDs of 21% or larger 96% of the time. TDs under shells were found about 62% of the time (n = 7), against 77% for TDs with no surface anomaly (n = 33). The digest doesn't report the test speeds of the runs, so it can't show a speed effect. The facility's companion digest, TD-99-002, lists the "ability to test detector cars at speeds up to 40 mph". MPC 99-106 analysed the same six evaluations and also didn't model speed.
What the speed studies show
| Study | System | Speeds tested | What it found about speed |
|---|---|---|---|
| Mariani et al., J. Transp. Eng. A 143(5), 2017 | Air-coupled (non-contact) UT prototype, RDTF | 1.6 to 24 km/h (1 to 15 mph) | Good detection at 1.6 and 8 km/h, "much poorer" at 16 and 24 km/h |
| Datta and Lanza di Scalea, Exp. Mech. 63, 2023 | Passive guided waves from wheel-rail contact, no pulser | Field tests at TTC; "possible test speeds as high as 80 mph" | Test speed is one of several parameters scored by ROC curves; the method has no PRF limit |
| TTCI TD-16-050 (Nov 2016) | Phased array roller search unit prototype | Up to 20 mph | Detected TDs under shells, weld and web defects at speeds up to 20 mph; no speed-by-speed POD |
| TTCI TD-20-007 (May 2020) | Electromagnetic surface sensor (not UT) | 3, 6, 15, 20 mph | Small sensitivity loss from wider sample spacing; testing was capped at 20 mph by the carriage |
| Chałko et al. 2025 | Contact UT probe head on a rotating test wheel (lab) | 0 to 135 km/h | Base-echo loss stayed acceptable with flow and force control; no flaw POD |
| Davydkin et al. 2023 | Mobile flaw detection on the Russian network | Not given in the abstract | Speed "inevitably" brings loss of contact and missed defects |
Two cautions apply when reading any POD figure:
- POD belongs to one system at one setting. ASTM E2862-23 ties a demonstrated POD to a specific set of examination parameters. FRA's ORD-26/07 model for handheld rail UT shows mean POD at 25% head area falling from 99.9% to 86.7% as the detection threshold rises from 0.10 to 0.40. A threshold change alone moves POD that much.
- Detection direction depends on physics. In the air-coupled prototype, detection worsened with speed. In the passive method the wheels are the sound source, so the PRF limit doesn't apply. Neither result transfers to a contact UT car.
Key point: the research record supports the mechanisms (spacing, coupling, signal) but gives no speed-versus-POD curve for conventional test cars. Any contract claim that detection is "unchanged at speed" should be backed by a demonstration, not a brochure.
Why does higher test speed raise the need for verification?
A suspect from a high-speed test is not a defect until a qualified person confirms it in the field. 49 CFR 213.240(e)(4) says: "A suspect location is not considered a defect under § 213.113(c) until it has been field-verified by a person qualified under § 213.238." Verification, or prove-up, is usually a hand test with a portable instrument at the marked location. It confirms the type and size that set the remedial action.
FRA's 2020 preamble reports how often each priority of suspect from continuous testing turned out to be real under the waivers that preceded the rule. Priority one suspects were confirmed in about 95% to 99% of cases and priority two in about 65% to 70%. Priority three, the questionable responses, is where most defects were found:
"Data from the existing waivers demonstrates that, although less than two percent of the priority three suspect locations are found to be actual rail defects, priority three suspect locations account for approximately 85 percent of the field-verified defects found as a result of continuous testing." — FRA, Rail Integrity and Track Safety Standards final rule, 85 FR 63362 (October 7, 2020)
That sentence is the case for verification at speed. Most real defects arrive looking like weak, unclear responses. A faster run with wider pulse spacing tends to produce more of those weak responses, and every one must either be verified or protected. Under 213.240(e)(5), a continuous-test suspect not verified in time gets the most restrictive remedial action for its suspected type and size. Response windows are covered in continuous vs stop-and-verify rail testing.
Worked example: why run speed barely moves the average
Inputs (illustrative): 100 track miles, 1 suspect per mile, 6 minutes to verify each suspect.
| Run speed | Moving time | Verifying time | Total | Average speed |
|---|---|---|---|---|
| 15 mph | 400 min | 600 min | 1,000 min | 6.0 mph |
| 20 mph | 300 min | 600 min | 900 min | 6.7 mph |
| 30 mph | 200 min | 600 min | 800 min | 7.5 mph |
| 40 mph | 150 min | 600 min | 750 min | 8.0 mph |
Arithmetic check at 40 mph: 100 ÷ 40 = 2.5 h = 150 min; 100 × 6 = 600 min; 100 miles ÷ 12.5 h = 8.0 mph. Doubling the run speed from 20 to 40 mph raises the average by only 1.3 mph, because verification time doesn't change. Those averages land in the 6 to 8 mph range MPC 99-106 reported in 1999. Continuous testing moves the 600 minutes to a separate crew working afterward, but the work is the same. If the faster run produces more weak suspects per mile, the verification load grows.
Key point: speed saves track time on the run, not on the prove-up. Budget the verification crew before buying the speed.
What should an owner specify about test speed?
An owner should specify speed as a performance condition tied to a demonstration and a record, not as a single number. Part 213 sets no test speed; the hooks are the valid search in 213.237, the qualified operator in 213.238, the test-type designation in 213.240(c), and the records in 213.241. For the interval itself, see how often must rail be tested and the 49 CFR 213.237 guide.
| Specify | Why | What to ask for in the bid |
|---|---|---|
| A maximum test speed per segment type | Spacing and coupling worsen with speed | Separate ceilings for open track, curves with shelling, turnouts, bridges and grade crossings |
| Demonstrated detection at that speed | POD is system- and setting-specific | Runs over the owner's or bidder's reference rail with known reflectors at the contracted maximum speed |
| Pulse spacing per channel at maximum speed | PRF alone hides channel sharing | Spacing in mm for each channel group, with PRF, channel count and firing scheme |
| Coupling monitoring | Lost coupling hides untested rail | Continuous loss-of-signal flags, with untested lengths listed by milepost |
| Untested-rail handling | 213.237(g) and (h) | A retest plan before the 213.237 limit runs out, and the 213.241(c)(6) record |
| Test type before the run | 213.240(c) | Written designation of stop-and-verify or continuous per segment, and how changes are logged |
| Verification resources | Prove-up sets the real pace | Who verifies (213.238-qualified), the hand-test method, and crew numbers for the 213.240 windows |
| A speed log in the record | Lets the owner audit the run | Vehicle speed recorded against milepost in the test data |
One precedent for slower testing where the risk is higher comes from the NTSB's New Brighton investigation (RAR-08/02). After the accident, the railroad instructed a test speed of no more than 5 mph on critical and open-deck bridges and bridges over navigable waterways, with suspect responses hand-tested. An owner can write similar location rules into its own specification.
Sample contract wording an owner might adapt: "The contractor shall state the maximum test speed for each segment type and the resulting pulse spacing per channel. Detection at that speed shall be demonstrated on reference rail before production testing. Any length where coupling or signal was lost shall be reported by milepost as not tested."
For a broader bid checklist, see the rail inspection RFP guide. For the parts of a test system, see the ultrasonic rail flaw system overview.
What high-speed testing can't do
High-speed ultrasonic rail inspection has the same blind spots as slower vehicle testing, and speed adds some of its own.
- It can't cover the whole rail. FRA's 2014 preamble (79 FR 4234) says the waves "normally scan the entire rail head and web, as well as the portion of the base directly beneath the web". The base flanges are outside that coverage at any speed.
- It can't see past surface damage. Shells and head checks mask the defects beneath them. In TD-99-006, TDs under shells were found about 62% of the time against 77% elsewhere. See rail shelling and spalling.
- It can't replace verification. 213.240(e)(4) makes field verification by a 213.238-qualified person a condition of calling anything a defect.
- It can't make a lost-coupling length count. Rail where the signal dropped hasn't been searched, however fast the car was going.
- The worked numbers here are illustrative. The fluid path, decay allowance, channel count and beam width are assumptions chosen to show the arithmetic. A real system's spacing and detection must come from its own design data and a demonstration.
How EMA Industries applies this
EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners from Southern California, mobilizing nationwide. Its NDT Level III services can help an owner turn the speed questions in this article into specification language: speed ceilings by location, a reference-rail demonstration, pulse spacing, untested-rail reporting and verification resources. To scope a testing program, contact EMA Industries.
Frequently asked questions
How fast do ultrasonic rail test vehicles actually run?
Rated and working speeds differ. MPC 99-106 (1999) described cars capable of 30 mph non-stop that averaged 6 to 8 mph once stops to hand-verify were counted. FRA's 2020 preamble describes stop-and-verify testing as normally not exceeding 20 mph. Continuous testing runs non-stop and moves verification to a later crew.
Does FRA set a maximum test speed for rail flaw detection?
Part 213 sets no test speed for rail flaw detection equipment. 49 CFR 213.237 requires a valid search, defined as a continuous inspection where the equipment performs as intended and a qualified operator interprets the responses. The safe test speed is therefore whatever speed still produces a valid search on that rail, which the owner should demonstrate and specify.
What pulse spacing is good enough for rail UT?
No public standard sets a pulse spacing for rail testing. Work it out from the beam's footprint on the smallest flaw the owner needs to find and the number of hits the system needs to recognize it. Then prove it on reference rail at the contracted speed. As a rough guide, the worked example above shows a small reflector getting about three hits at 8.9 mm spacing and less than one at 36.6 mm.
Is the 25 mph in 49 CFR 213.237 a test speed?
The 25 mph in 213.237(h)(2) is a train speed, not a test speed. It is one of three options when a valid search couldn't be made: retest, limit trains to 25 mph over that rail until a valid search is made, or replace the rail. It says nothing about how fast a test vehicle may run.
Can a faster continuous test replace hand verification?
Continuous testing changes when verification happens, not whether it happens. Under 213.240, suspects must be field-verified by a 213.238-qualified person within 36 hours (for suspects that would need remedial action A, A2 or B) or 84 hours. Until then a suspect isn't a defect; if the window passes unverified, it gets the most restrictive remedial action for its suspected type and size.
Sources
- 49 CFR 213.237, Inspection of rail: eCFR (text read on Cornell LII, October 2026)
- 49 CFR 213.238, Qualified operator: eCFR
- 49 CFR 213.240, Continuous rail testing: eCFR (text read on Cornell LII)
- 49 CFR 213.241, Inspection records: eCFR
- FRA, Rail Integrity and Track Safety Standards; Final Rule, 85 FR 63362, October 7, 2020: govinfo
- FRA, Track Safety Standards; Improving Rail Integrity; Final Rule, 79 FR 4234, January 24, 2014: govinfo PDF
- Garcia, G. A. and Reiff, R. P., Benchmarking Results of North American Rail-Flaw Detection Technologies, TTCI Technology Digest TD-99-006, February 1999: PDF
- Garcia, G. A. and Reiff, R. P., Rail-Defect Test Facility: A Tool for Evaluating Defect-Detection Technologies, TTCI TD-99-002, January 1999: PDF
- Jeffrey, B. D. and Peterson, M. L., Assessment of Rail Flaw Inspection Data, MPC 99-106, Colorado State University, August 1999: PDF
- Witte, M. and Poudel, A., High-Speed Rail Flaw Detection Using Phased Array Ultrasonics, TTCI TD-16-030, July 2016: PDF
- Witte, M. and Poudel, A., Characterizing Rail Flaws Using Phased Array Ultrasound, TTCI TD-16-050, November 2016: PDF
- Poudel, A. and Witte, M., Evaluation of In-Motion ECHO-Rail RCD Measurement System, TTCI TD-20-007, May 2020: PDF
- Mariani, S., Nguyen, T. V., Zhu, X. and Lanza di Scalea, F., Field Test Performance of Noncontact Ultrasonic Rail Inspection System, J. Transp. Eng. Part A 143(5), 2017: doi.org/10.1061/JTEPBS.0000026
- Datta, D. and Lanza di Scalea, F., Influence of Varying Operational Parameters on the Defect Detection Performance of a High-Speed Ultrasonic Rail Inspection System During Field Tests, Experimental Mechanics 63:1043–1054, 2023: doi.org/10.1007/s11340-023-00972-9
- Chałko, L., Antolik, Ł., Rucki, M. and Trochta, M., Possibility of High-Speed Ultrasonic Detection of the Internal Material Defects in Rails, Materials 19(1):28, December 2025: doi.org/10.3390/ma19010028
- Davydkin, A., Dymkin, G. Ya. and Konshina, V. N., Influence of speed of rail flaw detection on the reliability of its results, E3S Web of Conferences 383:01003, 2023: doi.org/10.1051/e3sconf/202338301003
- Bruzek, Einbinder and Tunna (ENSCO), MAPOD for handheld rail UT, DOT/FRA/ORD-26/07, July 2026: ROSA P
- ASTM E2862-23, Standard Practice for Probability of Detection Analysis for Hit/Miss Data: ASTM store
- Evident, Pulser and Receiver Operation (NDT tutorial): ims.evidentscientific.com
- NTSB, Derailment of Norfolk Southern Railway Company Train 68QB119 with Release of Hazardous Materials and Fire, New Brighton, Pennsylvania, October 20, 2006, RAR-08/02: PDF