Sizing a Fiber Bragg Grating Wing Array That Survives a Broken Fiber

How many gratings you can put on one fiber depends on the interrogator's spectral window and the wavelength slot each grating needs. Whether the digital twin still works after a fiber breaks depends on something else: which end of the fiber the interrogator reads from. A wing strain array sized only for channel count tends to lose its observability the first time a fiber cracks at a fastener row. The sizing below follows a single wing spar fiber all the way through, and every number in it is an estimate you should replace with your own hardware data.

The spectral budget sets the grating count

Every grating in a wavelength-multiplexed line needs its own slot in the spectrum, and that slot has to cover the full Bragg shift from mechanical strain plus thermal soak, plus a guard band so neighbouring peaks never cross. The design parameter here is a slot of 4 nm per grating, which holds limit-load strain over the full operating temperature range with some guard left over. Take a usable interrogator window of roughly 40 nm (estimate) and you get about 10 gratings per fiber (estimate). The error bar comes almost entirely from the slot, not the window. If cold soak at altitude plus a gust load case widens the real excursion beyond what you assumed, the count drops by around two gratings (estimate), and peaks that cross mid-flight produce a swapped sensor identity. The twin will turn that into a wrong deflection shape without raising any flag.

What one break does to observability

The failure mode that drives the architecture is a fiber break where the line crosses a ply drop, a rib cap or a fastener row: low strain margin, high stress concentration, and the spot most likely to see repair damage. With single-ended interrogation, every grating beyond the break goes dark at once. A midspan fracture on our spar fiber takes out roughly half the ten gratings (estimate), and they are the outboard ones, which carry the most information about tip deflection and torsion. The modal or inverse finite element reconstruction in the twin needs at least as many independent strain readings as the mode shapes it keeps, so losing the outboard half usually leaves the problem underdetermined exactly where aeroelastic tracking matters most. If you read the same fiber from both ends on two independent interrogator channels, each end sees the gratings up to the break, so you lose only the grating at the fracture point. The spectral budget doesn't change, because both ends see the same wavelengths.

The decision rule: put the two ends of each loop on separate interrogator cards, separate power buses and separate harness routes, then check that after the worst single failure, whether a fiber break, a card or a bus, the surviving gratings still outnumber the modes the reconstruction keeps, with margin, and loop only the fibers where that check fails.

Where the hardware cost lands

Double-ended loops don't add gratings. They add interrogator channels, connectors and a second route through the wing root, and that second route usually costs more than the electronics because it needs its own pass-through and its own clamping. The interrogator cards and their optical switches go through environmental qualification for airborne equipment such as DO-160, so the vibration and temperature profiles you test against have to match the zone each card sits in. Splitting the two ends across zones also gives you thermal and vibration diversity for free. Software has to handle the degraded state as well: when one end reports a break, the twin should mark the confidence of the affected region instead of quietly interpolating across it, so that a maintenance decision based on that file knows which strain values were measured and which were inferred.

Can the model fill in for a lost sensor instead of adding a second interrogator end?

Yes, but only for a few scattered dropouts and only if the remaining gratings still constrain the retained modes. A break that removes the whole outboard half of a fiber takes away the measurements the model would need to make that inference, so on primary load paths a second end is cheaper than an unobservable tip.

Back to AircraftTwins · Domains

← Back to AircraftTwins
Engineering & Architecture Brief

Sizing a Fiber Bragg Grating Wing Array That Survives a Broken Fiber

How many gratings you can put on one fiber depends on the interrogator's spectral window and the wavelength slot each grating needs. Whether the digital twin still works after a fiber breaks depends on something else: which end of the fiber the interrogator reads from. A wing strain array sized only for channel count tends to lose its observability the first time a fiber cracks at a fastener row. The sizing below follows a single wing spar fiber all the way through, and every number in it is an estimate you should replace with your own hardware data.

The spectral budget sets the grating count

Every grating in a wavelength-multiplexed line needs its own slot in the spectrum, and that slot has to cover the full Bragg shift from mechanical strain plus thermal soak, plus a guard band so neighbouring peaks never cross. The design parameter here is a slot of 4 nm per grating, which holds limit-load strain over the full operating temperature range with some guard left over. Take a usable interrogator window of roughly 40 nm (estimate) and you get about 10 gratings per fiber (estimate). The error bar comes almost entirely from the slot, not the window. If cold soak at altitude plus a gust load case widens the real excursion beyond what you assumed, the count drops by around two gratings (estimate), and peaks that cross mid-flight produce a swapped sensor identity. The twin will turn that into a wrong deflection shape without raising any flag.

What one break does to observability

The failure mode that drives the architecture is a fiber break where the line crosses a ply drop, a rib cap or a fastener row: low strain margin, high stress concentration, and the spot most likely to see repair damage. With single-ended interrogation, every grating beyond the break goes dark at once. A midspan fracture on our spar fiber takes out roughly half the ten gratings (estimate), and they are the outboard ones, which carry the most information about tip deflection and torsion. The modal or inverse finite element reconstruction in the twin needs at least as many independent strain readings as the mode shapes it keeps, so losing the outboard half usually leaves the problem underdetermined exactly where aeroelastic tracking matters most. If you read the same fiber from both ends on two independent interrogator channels, each end sees the gratings up to the break, so you lose only the grating at the fracture point. The spectral budget doesn't change, because both ends see the same wavelengths.

The decision rule: put the two ends of each loop on separate interrogator cards, separate power buses and separate harness routes, then check that after the worst single failure, whether a fiber break, a card or a bus, the surviving gratings still outnumber the modes the reconstruction keeps, with margin, and loop only the fibers where that check fails.

Where the hardware cost lands

Double-ended loops don't add gratings. They add interrogator channels, connectors and a second route through the wing root, and that second route usually costs more than the electronics because it needs its own pass-through and its own clamping. The interrogator cards and their optical switches go through environmental qualification for airborne equipment such as DO-160, so the vibration and temperature profiles you test against have to match the zone each card sits in. Splitting the two ends across zones also gives you thermal and vibration diversity for free. Software has to handle the degraded state as well: when one end reports a break, the twin should mark the confidence of the affected region instead of quietly interpolating across it, so that a maintenance decision based on that file knows which strain values were measured and which were inferred.

Technical FAQ: Can the model fill in for a lost sensor instead of adding a second interrogator end?

Yes, but only for a few scattered dropouts and only if the remaining gratings still constrain the retained modes. A break that removes the whole outboard half of a fiber takes away the measurements the model would need to make that inference, so on primary load paths a second end is cheaper than an unobservable tip.

Related Architecture Specifications (AIRSPACEGATEWAY): evtolorchestration.com · evtolsync.com · swarmevtol.com

More Technical Articles on aircrafttwins.com

Predicting Composite Delamination in eVTOL Airframes via Real-Time Structural Telemetry →Linking Continuous Flight Load Logs to Predictive MRO Workorders in Commercial Fleets →