Control track — four papers
Ordered by readiness. C1 is a submission problem, not a research problem.
C1 · READY — data exists
The numerical substructure echoes the response into the command: feedthrough bias in RTHS monitoring, and its fix
data ✓ · method ✓ · validation ✓ · rig confirm
Any condensed numerical substructure has a force→displacement feedthrough; it plants a zero-lag copy of the measured response inside the command at all frequencies. Every FRF, coherence, damage or stability estimate computed from (command, response) in closed-loop RTHS is biased toward the inverse numerical path — and passes every standard quality gate while being wrong.
experiments (mostly done)
- Same specimen, same estimator: injected broadband (unbiased) vs closed numerical loop (biased) — you have both record classes.
- Bias vs exogenous-to-echo power ratio: intensity sweep 0.02–0.10 el-centro (1 day of runs).
- Zero-lag detector as a classifier over the full record library (done, 10/10 correct).
- Rig confirmation of the exact structural fix (the §1 wiring) — margin 0.29→0.66, authority restored.
results to report
- Closed-form bias limit (M → inverse feedback path), measured convergence to it.
- Why coherence/power gates cannot save you (the anti-lesson for the SHM community).
- Exogenous-tap correction + evidence gating; before/after on real runs incl. preserved detection of true instability.
why it publishes
- Generic: every RTHS with a condensed FEM interface has the term; online monitoring in hybrid testing is a growing subfield walking into this exact trap.
- Complete arc: mechanism → theory → detector → fix → validation, all on hardware data.
reviewer risks
- "Just instrument-variable identification" — position it as the RTHS-specific instance with the D-term structure, the classifier, and the operational fix.
venues: Mech. Systems & Signal Processing · Earthquake Engng & Structural Dynamics · Struct. Control & Health Monitoring
C2 · ONE CAMPAIGN AWAY
Online normalized-coprime stability margin with target-tracking authority: supervised RTHS
method ✓ · runtime ✓ · replays ✓ · live campaign
RTHS stability is traditionally certified a priori (delay margins, passivity, assumed models). This system measures the margin online from operating data and regulates loop authority onto a margin target — arresting incipient instability while preserving test fidelity, with validated false-alarm behavior.
experiments
- Healthy suite: N≥10 runs across intensities/records — authority stays 1.0, zero trips (false-alarm statistics).
- Induced marginality (Pvar ramps): detection latency distribution vs onset speed; arrest vs the unsupervised baseline that diverges.
- Recovery behavior: margin returns → authority recovers, latched after events.
- Head-to-head vs the standard alternatives on identical runs: energy/passivity monitor and amplitude interlock — latency, false alarms, fidelity cost.
results to report
- Detection ≤0.5 s with coherent excitation (already shown in replay); quiet-band information limits stated honestly.
- Fidelity metric while supervising: added tracking error / response distortion vs unsupervised — the cost of safety.
why it publishes
- First (to be verified in the lit review) online coprime-factor margin running on the actual test hardware with closed-loop authority — not a simulation study.
- The bitwise desktop↔target equivalence + staged verification protocol is itself a methods contribution labs can copy.
depends on
- The echo fix wired in (fix_rths_echo_plan.md) — otherwise C1's bias eats this paper's credibility.
venues: Earthquake Engng & Structural Dynamics · IEEE Trans. Control Systems Technology · Engineering Structures
C3 · SYNTHESIS + A/B CAMPAIGN
Damage-robust delay compensation: NCF-constrained lead+notch design across specimen degradation states
optimizer ✓ · designs ✓ · A/B runs · degraded runs
Delay compensators are tuned for the healthy specimen, then the specimen degrades mid-test and the compensator that helped starts hurting (your 1p1z lead is provably unstable on the 5k plant). Synthesis against a family of damage states yields compensation that stays stable and synchronous through degradation.
experiments
- A/B/C on the same protocol: no compensation vs healthy-tuned lead vs robust lead+notch — synchronization (J2/J4/subspace plots), margins, at 7k AND 5k.
- The money run: degradation during the record (Pvar or a specimen driven into softening) — healthy-tuned design destabilizes, robust design rides through.
- Compare against an adaptive compensator from the literature (e.g. ATS) as the third arm.
why it publishes
- RTHS compensation literature optimizes tracking on the nominal plant; explicit robustness-across-damage synthesis with margin constraints is a clean gap.
- Everything is already built: optimizer, analytic margin maps, both plants, the winner design.
venues: Engineering Structures · Mech. Systems & Signal Processing · Smart Structures and Systems
C4 · SHORT PAPER / LETTER
The loop that actually binds: inner vs outer stability margins in RTHS, and monitoring the one that matters
finding ✓ · offline wrap ✓ · online b_out · demo run
The transfer-system loop everyone monitors had margin 0.59; the hybrid (numerical-closed) loop had 0.40 at 38 Hz — and your real divergences were outer-loop events the inner monitor is structurally blind to. Because the numerical side is exactly known software, the outer margin comes free by wrapping the already-estimated plant FRF.
experiments
- Implement numerical_ncf_plan.md WP-B (publish-only b_out); replay library first.
- One controlled outer-loop instability (stiff FEM / raised k ratio) where b_out falls while inner margin stays healthy — the decisive figure.
- Impedance upgrade: static-k wrap vs measured K̂(ω) wrap accuracy.
why it publishes
- Simple, sharp, and actionable for every RTHS lab: monitor the hybrid loop, not just the actuator loop; here is how, at zero identification cost.
- Natural companion/sequel to C2 — or its strongest section if merged.
venues: Struct. Control & Health Monitoring · Mechanics Research Communications (letter) · or fold into C2
Structural track — the column-base assembly is the story
A steel column anchored to a concrete base is not "a column": it is an exposed base-plate connection —
base plate, anchor rods, grout, concrete breakout cones. Design practice models it as fixed or pinned; reality is semi-rigid,
pinching, and degrading, and its behavior sets first-story drift, period, and collapse margin. Your physical substructure
contains the one component the profession models worst. That is the thesis of every card below.
S1 · FIRST STRUCTURAL PAPER
How much fixity does a real anchored base have — and what does the building above feel?
rig ✓ · instrumentation · campaign · analysis
Same ground motions, same numerical superstructure, three base conditions: the numerical model's fixed-base assumption, its pinned-base assumption, and the REAL connection in the loop. Quantify the rotational stiffness the connection actually delivers, how it degrades with demand, and the system-level consequences (drift concentration, period elongation, force redistribution) of both idealizations.
experiments
- Characterization: low-amplitude BLWN + moment-rotation loops → initial rotational stiffness, slip/gap thresholds (compare AISC Design Guide 1 / Eurocode component-method predictions).
- Hybrid runs at 3 intensities × 3 records × {real base | numerical fixed | numerical pinned superstructure variants}.
- Fixity evolution: rotational stiffness identified per window across each run (the k̂ tracker generalizes).
questions answered
- Where between pinned and fixed does a code-designed exposed base plate actually sit, dynamically?
- At what demand does fixity start to drop, and is it recoverable (slip) or permanent (anchor yield/grout)?
- Which idealization is conservative for drift, and by how much — does it flip with intensity?
instrumentation to add
- 2× LVDT/laser at base plate edges → uplift + rotation.
- Strain gauges on 2–4 anchor rods (axial force history).
- Rosettes on base plate; crack mapping / camera on grout & concrete.
reviewer risks
- Single column, no axial-load variation — scope the claim to shear-dominated exterior columns; state boundary conditions precisely.
venues: J. Structural Engineering (ASCE) · Engineering Structures · J. Constructional Steel Research
S2 · THE SIGNATURE COMBO
Deterioration under realistic seismic histories: mainshock–aftershock sequences with a live stiffness meter
tracker ✓ · protocol design · campaign · fragility fit
Run the same physical assembly through recorded sequences (mainshock, then real aftershocks) while the spectral k/m/c tracker logs stiffness continuously. Deliver degradation-vs-demand-history curves that no cyclic protocol can produce: does damage accumulate with peak demand, with dissipated energy, or with the ORDER of events? What residual capacity is left after a design-level mainshock?
experiments
- Sequences: MS-only vs MS+AS1+AS2 (recorded pairs, e.g. Mainshock/aftershock sets from PEER), matched-intensity singles as controls.
- Order swap: strong-then-weak vs weak-then-strong at equal cumulative energy — history dependence isolated.
- Continuous k̂(t) + per-event moment-rotation autopsies + anchor strain ratchet.
results to report
- Stiffness-degradation curves vs cumulative measures (peak drift, dissipated energy, cycles above threshold) — which collapses the data best.
- Residual-capacity statement per damage state — directly usable for post-earthquake tagging of columns.
why it publishes
- Aftershock capacity of steel base connections is a live code question with almost no component-level dynamic data.
- The online tracker turns each run into a continuous measurement, not a before/after pair — that is the novelty multiplier.
consumables
- 3–6 nominally identical specimens (anchor groups are the sacrificial part; the column can likely be reused early on).
venues: Earthquake Engng & Structural Dynamics · Bull. Earthquake Engineering · J. Structural Engineering
S3 · THE PRACTICE-FACING COMPARISON
Do standard cyclic protocols predict what an earthquake actually does to this connection?
rig ✓ · twin specimens · both protocols · comparison
Twin specimens: one tested under the standard quasi-static protocol (ATC-24 / AISC 341 loading), one under RTHS with real ground-motion demand at matched peak drift. Compare damage progression, dissipated energy, deterioration rates, and failure mode. If they differ materially, that is a finding about the protocol; if they match, that is a validation the profession also wants.
experiments
- 2–3 twin pairs (repeatability!), matched peak rotation; document cycle counts, energy, k̂ trajectories, failure surfaces.
- Secondary: rate effect isolation — replay the RTHS displacement history quasi-statically on a third twin (separates rate from sequence).
why it publishes
- Protocol-vs-reality comparisons exist for beams and BRBs; base connections with real concrete anchorage are underrepresented, and this is exactly what RTHS is FOR.
- Either outcome is a paper — rare and valuable property.
venues: J. Structural Engineering (ASCE) · Earthquake Spectra · Engineering Structures
S4 · THE CAPSTONE
Hybrid incremental dynamic analysis: fragility of the anchored base from escalating real tests
needs S1+S2 first · OSI safety net ✓
Escalate intensity run-over-run (hybrid IDA) until defined damage states — slip, anchor yield, stiffness −30%, loss of moment capacity — and fit fragility functions for the component from physical data. The OSI safety supervision (B1) is what makes running this close to failure repeatedly a controlled activity rather than a gamble.
experiments
- IM ladder per specimen (e.g. Sa scaled 0.5×→2.5× design) across ≥3 records; damage-state detection from k̂ thresholds + physical inspection.
- Cross-check numerical: does a calibrated deteriorating spring model, fed the S1/S2 parameters, reproduce the fragilities?
why it publishes
- Component fragilities for exposed base plates are mostly judgment + monotonic pushes today; dynamic, sequence-aware physical fragility data is a genuine contribution to performance-based assessment.
venues: Earthquake Spectra · Bull. Earthquake Engineering · Structural Safety
S5 · METHODS-ADJACENT
How much does the hybrid answer depend on the numerical half?
vHS twin ✓ · variants · runs
Same physical column, same record; vary only the numerical substructure: number of floors, damping model, condensation choices, feedthrough handling, time-scale. Quantify sensitivity of the physical demand and of the identified connection behavior to numerical modeling decisions — practical guidance for anyone designing an RTHS campaign.
experiments
- Numerical variant matrix (3 floors vs 8; Rayleigh vs modal damping 2–5%; with/without interface flexibility term) × 2 records, physical side untouched.
- Report demand deltas (drift, rotation, energy into specimen) and margin/fidelity deltas.
why it publishes
- Cheap (no consumables — stay below damage thresholds), fills a guidance gap, and doubles as your own uncertainty quantification for S1–S4.
venues: Struct. Control & Health Monitoring · Engineering Structures (technical note)
Campaign matrix — runs feed multiple papers
Design every campaign to feed at least two cards. Specimens are the scarce resource; data is not.
| campaign | runs / consumables | feeds | prerequisite |
|---|
| Echo A/B + intensity sweep | ~15 runs · 0 specimens |
C1C2 | echo tap wired (fix_rths_echo_plan) |
| Healthy supervision suite + Pvar marginality ladder | ~25 runs · 0 specimens |
C2C4 | gate + shadow validation |
| Compensator A/B/C at 7k and 5k | ~18 runs · 0–1 specimens |
C3C2 | — |
| Base characterization + fixity variants | ~20 runs · 1 specimen |
S1S5 | base instrumentation installed |
| Numerical-variant matrix (low amplitude) | ~16 runs · 0 specimens |
S5C4 | — |
| MS–AS sequences + order swaps | ~12 runs · 3–4 specimens |
S2B1S4 | S1 characterization baseline |
| Twin protocol comparison | ~8 runs · 4–6 specimens |
S3 | quasi-static fixture path |
| Hybrid IDA ladders | ~15 runs · 3–5 specimens |
S4B1 | C2 supervision active; S2 damage states defined |
Repeatability is your credibility budget. Reserve twin runs early (same record, same intensity,
fresh vs same specimen) and report the scatter in every structural paper. RTHS skeptics' first question is always
"how repeatable is the loop?" — you have vHS + margins + tracking-error metrics to answer it quantitatively; make that
a standard figure in every paper.