In the 2012 National Bridge Inventory, 61,038 bridges and culverts had a load restriction (a load posting, in FHWA terms), about 10% of the inventory, and 82% of them belonged to local governments. According to NCHRP Synthesis 453, which analyzed that data, 93% of the restricted structures had been rated by computation alone, while load tests were used for less than 1% of all load ratings. A restricted bridge can be strengthened or replaced, but there is also a cheaper option, which is checking its rating against how the structure responds to a real load. It is also the least used.

So the questions for bridge load rating with measured data are two. When is a measurement worth more than a weight limit, and how long does a single measurement remain valid?

Computed ratings and load restrictions

Three quarters of the restricted structures in the NCHRP analysis were in fair or good general condition. A load restriction does not necessarily mean that a bridge is in bad shape, but that the engineers rating it did not have enough information to prove otherwise.

The design live load is unknown for 57% of restricted structures, and many locally owned bridges have no plans at all. When information is missing, the rating engineer has to fill the gaps with conservative assumptions. Composite action is ignored unless shear connectors are documented, bearings are modeled as free, and live load distribution is taken from code factors instead of the physical deck. Older steel stringer bridges are the most affected, with 22,481 restricted structures, more than one in five of all steel stringer bridges in the inventory. Each assumption is reasonable on its own, but when they are all applied together, even a sound span can end up with a rating factor below 1.0.

FHWA requires the weight limit to be in place within 30 days of the load rating. Since the FAST Act, bridges within reach of the Interstate System must also be rated for emergency vehicles up to 86,000 lb (the EV2 and EV3 models), and states have added emergency vehicle limits on many rural bridges.

Diagnostic load testing

The procedure is described in Section 8 of the AASHTO Manual for Bridge Evaluation (MBE). A truck with measured axle weights crosses the bridge along chosen paths, while sensors record strain or deflection at the critical members. The engineer then compares the measured response with what the model predicts for the same load, and from that comparison calculates an adjustment factor, K = 1 + Ka·Kb, which is applied to the analytical rating factor. Ka is the benefit observed in the test. Kb, between 0 and 1, is how much of that benefit the evaluator considers reliable at rating-level loads.

It is well known that older bridges usually deflect less under a test truck than their models predict. Girders share load better than what is assumed by code distribution factors, decks act compositely with girders that have no shear connectors on paper, and bearings modeled as free carry part of the moment. In a Kentucky study, researchers load-tested 14 bridges without plans. Most of them were below the restriction threshold on paper, and after the tests the researchers recommended removing the restriction on all 14. In a study published in the ASCE Journal of Performance of Constructed Facilities, the inventory rating factor of a rural steel girder bridge rose from 1.5 to 4.1 for the test truck.

Of course, a test truck is not the rating load. Unintended composite action and bearing restraint can be present during the test and disappear at higher loads, or after a few more winters of corrosion at the deck–girder interface. Therefore, if the extra stiffness cannot be shown to persist, the evaluator reduces the benefit, which is why the adjustment factor from a single test is often much smaller than the difference measured between model and bridge.

Continuous data after the test

Continuous monitoring is not necessary on every tested bridge. On a short slab or timber span with better load distribution than the model and no ongoing deterioration, the test alone is sufficient. One case where monitoring would be necessary is a bridge whose higher rating is due to a mechanism that can change over time. The second is a bridge that is changing itself, because of deterioration or a recent rehabilitation.

Bearing restraint is the clearest example of the first case. Elastomeric pads stiffen in cold weather, and corroded rocker or sliding bearings lock and release with the seasons, so the restraint measured in a winter test may be absent at summer temperatures. With a tiltmeter at each bearing line, engineers can follow the end rotations under traffic for the whole year. If the rotation under heavy trucks grows during the summer, part of the K factor obtained in winter depended on temperature, and engineers need to reconsider the rating. The same applies to composite action on bridges with deteriorating decks, and to any span after a rehabilitation whose benefit was included in the new rating.

The deflection profiles recorded during the test, analyzed in the Static Deflection Tool in MyMove, become the baseline. The weight of each truck crossing the bridge after the test is unknown, but over a month the distribution of deflections is stable. So if the highest deflections start to drift at comparable temperatures, either the bridge or the traffic has changed, and both are relevant for the rating. A DECK002-X dynamic displacement sensor at mid-span and tiltmeters at the supports measure the same global response as the test, so each month of traffic can be compared with the baseline. AXE-STD-LR-2 accelerometers and the Modal Analysis Tool add a second check that does not depend on heavy trucks, because a shift in the fundamental frequency means that the stiffness of the bridge has changed.

In the Iowa-led pooled fund study TPF-5(219), researchers concluded that the most practical use of monitoring data is as a tuning factor on the rating, not as its basis. Continuous monitoring does not replace the engineer who calculates the rating, but it gives that engineer enough information to know whether the adjustment factor is still valid, instead of assuming it once.

Frequently Asked Questions

Can a weight limit be removed with a load test on a bridge in poor condition?

Rarely. A diagnostic test measures stiffness, and section loss or cracked members reduce capacity long before they change stiffness. When the restriction follows documented deterioration, the test results usually match the computed rating.

Is deflection enough, or does a diagnostic test need strain gauges?

The MBE accepts either strain or deflection as the measured response. Strain at the critical section relates directly to the member being rated, while deflection and end rotation capture global behavior and load distribution with fewer sensors. Most test programs combine a few strain gauges with displacement at mid-span.

How long should sensors stay on a bridge after a load test?

At least one full seasonal cycle, so the baseline covers the temperature range of bearings and deck. After that, the bridges worth keeping instrumented are those where the restriction was removed because of bearing restraint or unintended composite action.