Overview
Two reviewers at Very Good/Good, one at Good. This difference remained unresolved.
The reference measurements, held-out failure trials, and capable team support a worthwhile Phase I investigation. Uncertain warning margin, non-independent false-alarm assessment, limited representative module testing, and incomplete operational-impact evaluation constrain confidence in the central claims and hold overall merit at Very Good/Good, supporting the estimated Competitive recommendation.
-
01
Reconcile the five-minute warning target with the available lead time
The panel found the five-minute fused-alarm target uncertain: the proposal cites a 4.2-minute median vent-to-onset interval for overcharge, and the shortest heating-pilot gas lead is 5.1 minutes before confirmation or transport delay. Figure 1 has a caption but no visible trace. Project Description, pp. 7–9 and 11–12; restore Figure 1 beside its caption on p. 8.
-
02
Separate false-alarm tuning from performance assessment
The same 2,000-hour rack run sets the decision threshold and reports the false-alarm rate. The panel noted that even zero alarms over that period gives an approximate 95% upper bound of 1.5 per 1,000 rack-hours, above the stated target. Project Description, Objective 3, “Normal-operation rack” and “Milestone and success measure,” pp. 10–11.
-
03
Address the limits of representative-module validation
The held-out trials use dummy neighbors, and only two full-module tests assess transport. The panel found that these tests leave live-cell localization and performance across module configurations insufficiently validated. Project Description, Objective 3 “Held-out integrated failure trials” and Objective 4, pp. 10–12.
-
04
Explain how the proposed rack can be cycled
The panel questioned whether the listed 5 V-per-channel cycler can run the proposed 16-cell LFP modules for the 2,000-hour rack test; the module wiring and feasible cycler configuration are not explained. Project Description, Objective 3 “Normal-operation rack,” p. 10; Facilities, Equipment and Other Resources, p. 18.
This is the short list. The panel filed three independent reviews, argued across three rounds, and produced nine groups of recommended revisions — including items not shown here. Each finding is tied to the section it came from, so you can check the claim against your own document.
Document intelligence
Detected documents:
- Title block (p. 1)
- Project Summary (p. 2)
- Project Description (pp. 3–13, 11 pages)
- References Cited (p. 14)
- Budget summary (p. 15)
- Budget Justification (pp. 16–17)
- Facilities, Equipment and Other Resources (p. 18)
- Biographical sketches, PI and CTO (p. 19)
- Current and Pending Support (p. 20)
- Data Management Plan (p. 21)
- Project Pitch invitation statement (p. 22)
- Letters of Support, three (p. 23)
- Letters of Commitment, subawardee and consultant (p. 24)
Conditional requirements:
- Project Pitch invitation email (V.A, Single Copy Documents): not supplied for review. Page 22 describes it but is not the email.
- Subaward budget and justification (Line G.5): not supplied for review. The justification says they are included.
- Subawardee letter of commitment stating amount (Line G.5): present.
- Consultant letter of commitment (Line G.3): present.
- Postdoctoral Mentoring Plan: not applicable. The subaward funds a graduate student, not a postdoc.
- Company Commercialization History: not applicable. No prior SBIR/STTR awards are reported.
- Cooperative Research Agreement (STTR only): not applicable.
- Resubmission Change Description: not applicable. No resubmission is indicated.
- Owner statement for non-owned facilities (Facilities section): not supplied for review. Eastbrook and Brightwater facilities are third-party.
- Cover Sheet, Phase I Questionnaire, Certifications, Collaborators & Other Affiliations: not supplied for review.
- Human subjects / vertebrate animals documentation: not applicable.
Document quality: Figure 1 (p. 8) appears as a caption only, with no graphic visible on the page image. Minor text-extraction hyphenation artifacts appear (e.g., "earlywarning"). The budget arithmetic is internally consistent. Line totals, the indirect rate, the 7% fee, and the outsourcing share (15.3%, under one-third) all reconcile with the justification.
Author questions:
- Page 8 shows only the caption for Figure 1, the gas and strain records from the pilot test. Was the figure graphic omitted from this file? Please provide it if the preliminary results are meant to be reviewed from it.
Compliance & Alignment
Status: Compliant with concerns
Verified issues:
- Page 22 placement (concern). The solicitation says "DO NOT upload documents to the 'Supplementary Documents' section except those described," and failure "may" lead to return without review. Page 22 is a narrative labeled "Supplementary Document: Project Pitch Invitation." It is not the invitation email, which belongs in Single Copy Documents. Confirm where page 22 was uploaded, or remove it.
Not verifiable from supplied documents:
- Project Pitch invitation email (dated November 20, 2023, per page 22).
- Cover Sheet and SBIR selection. The printed "SBIR Phase I:" title is a reminder only.
- Phase I Questionnaire and Certifications.
- Collaborators & Other Affiliations.
- Subaward budget and subaward budget justification for Eastbrook.
- Signed owner statement for the Brightwater facility, which is not owned or leased by Larkhaven.
- Signature on the consultant letter: the page image shows a typed name only.
- NSF-approved formats for the biosketches and Current and Pending Support.
- Font, margins and line spacing.
- SAM/UEI and SBC registration, and the one-proposal-per-window limit.
Notes: Submitted before May 15, 2024, so PAPPG 23-1 applies. The following verify compliant:
- Project Summary: three headings and the required opening sentence.
- Project Description: 11 pages, with the required section headers and no URLs.
- Budget: $273,737, under the $275,000 cap. Lines D, E.2, F and G.2 are zero. PI effort is 3 months. The grantee-conference acknowledgement is present. Outsourcing is 15.3%. The 50% indirect rate, 7% fee, $1,000/day consultant rate and $10,000 Boot Camp line all comply.
- Other required items:
- The Data Management Plan contains the required statement.
- There are three letters of support, each with contact information.
- The subaward letter states its amount.
Figure 1 (page 8) shows only a caption, which reviewers may want to see.
Alignment: Strong
Solicitation-specific rubric:
| ID | Type | Weight | Requirement | Source |
|---|---|---|---|---|
| SC-01 | eligibility | mandatory | Project Pitch invitation received within the past 12 months, with the email uploaded to Single Copy Documents; otherwise Returned Without Review | Eligibility; V.A; V.B Tips |
| SC-02 | eligibility | mandatory | Small business of 500 or fewer employees including affiliates, not majority VCOC/PE/hedge-fund owned; SAM/SBA registration | IV; V.A Registrations |
| SC-03 | eligibility | mandatory | PI primarily employed (at least 51%) by the company; at least 1 calendar month per 6 months; no co-PI | IV; V.C Line A |
| SC-04 | eligibility | mandatory | At least two-thirds of research performed by the company; subawards plus consultants no more than one-third | IV Partnering; V.C G.5 |
| SC-05 | document | mandatory | Budget no more than $275,000; no funds on D, E.2, F or G.2; consultant no more than $1,000/day; fee no more than 7% | III; V.B; V.C |
| SC-06 | document | mandatory | Project Summary with Overview/IM/BI headers; IM opens "This Small Business Innovation Research Phase I project..."; keywords and subtopic | V.C |
| SC-07 | document | mandatory | Project Description of 10–15 pages; IM plan addressing technical AND commercial feasibility, milestones, success criteria | V.C |
| SC-08 | document | mandatory | Budget Justification acknowledges the mandatory grantee conference | V.C Line E.1 |
| SC-09 | document | mandatory | DMP containing the proprietary-data statement | V.C |
| SC-10 | document | mandatory | Signed consultant letter (availability, time, role, rate); separate subaward budget | V.C G.3, G.5 |
| SC-11 | document | mandatory | Cover Sheet, Phase I Questionnaire, Certifications, biosketches, Current and Pending listing this proposal, COA | V.B; V.C |
| SC-12 | additional-merit | scored | Commercial Impact: market, durable advantage, business model, team, NSF as catalyst | VI.A |
| SC-13 | program-priority | scored | BI implementation plan: metrics, partners and their value/engagement, risks, resources | V.C Broader Impacts |
| SC-14 | document | scored | No more than three letters of support with affiliation and contact information | V.C Optional |
| SC-15 | deliverable | mandatory | Phase I final technical report due within 15 days of end date, 15 pages or fewer | VII.C |
Proposal mapping:
| ID | Status | Evidence |
|---|---|---|
| SC-01 | not-verifiable | Invitation dated Nov 20, 2023 (p.22), within the window; email not supplied |
| SC-02 | not-verifiable | 4 employees; angel note (p.6); ownership and registrations not supplied |
| SC-03 | addressed | Full-time PI (p.6); 3.0 of 12 months (pp.15–16); no co-PI (p.1) |
| SC-04 | addressed | $42,000 = 15.3% (p.15) |
| SC-05 | addressed | $273,737; zero D/E.2/F/G.2; $1,000/day; 7% (pp.15–17) |
| SC-06 | addressed | p.2 |
| SC-07 | addressed | pp.3–13 (11 pages); Q1–Q5, objectives with success criteria, milestones (pp.7–12) |
| SC-08 | addressed | p.16 |
| SC-09 | addressed | p.21 |
| SC-10 | partial | Consultant and subawardee letters (p.24); subaward Research.gov budget stated but not supplied |
| SC-11 | not-verifiable | Biosketches (p.19); Current and Pending lists this proposal (p.20); Questionnaire, Certifications and COA not supplied |
| SC-12 | addressed | pp.3–6: 42 interviews, cost model, IP |
| SC-13 | addressed | p.13 table, partners, risks, resources |
| SC-14 | addressed | Three letters with addresses and phones (p.23) |
| SC-15 | addressed | Final report planned in M11–12 (p.12) |
NSF priorities:
- Not applicable: submitted before April 18, 2025.
Rubric caveats:
- The invitation is described as a "Supplementary Document" (p.22), but the solicitation requires it in Additional Single Copy Documents. The upload location should be confirmed.
- The fusion-patent claim is provisional. The fiber routing method originated at Halvern and is unlicensed (p.6), which weakens the "durable advantage" element of SC-12.
- The $10,000 Beat-the-Odds request on G.6 is within the cap.
- NSF 23-515 is archived and replaced by NSF 24-579. The proposal falls within the final 23-515 window.
Read: NSF Research.gov budget, 1 year, with its justification (Proposal, pp. 15–17). Figures are copied from the documents; every sum and comparison is computed, never estimated.
Not verifiable (3)
- The budget's periods are within the maximum durationThe call's maximum duration ("6-12 months") could not be read as a number of months. The call: "projects for 6-12 months".
- Ramona Okonkwo-Laurent: senior personnel months (NSF)Year 1 budgets 3 months; NSF allows more than two months a year only when the budget and justification disclose and justify it (NSF PAPPG 24-1 II.D.2.f(i)(a)): confirm the justification does.
- Tomasz Ferreira-Lind: senior personnel months (NSF)Year 1 budgets 2.5 months; NSF allows more than two months a year only when the budget and justification disclose and justify it (NSF PAPPG 24-1 II.D.2.f(i)(a)): confirm the justification does.
Consistent (19) arithmetic 6 · solicitation and funder 2 · narrative and justification 11
- The entries under A. Senior personnel sum to its total
- The entries under B. Other personnel sum to its total
- The entries under G. Other direct costs sum to its total
- The budget lines sum to the total direct costs
- Direct plus indirect costs equal the total
- F&A equals the rate times the base
- The total requested is within the award ceiling
- Required by the call: Domestic travel to the mandatory NSF SBIR/STTR Phase I grantee conference
Months match the text:
- Ramona Okonkwo-Laurent
- Tomasz Ferreira-Lind
- Daniel Achterberg
Amount matches the text:
- Equipment: Fiber Bragg grating interrogator with optical switch, rented for 9 months
- Subaward: Eastbrook State University, Battery Safety Laboratory
The justification's figure matches the form:
- Senior personnel
- Other personnel
- Fringe benefits
- Equipment
- Travel
- Subawards
Technical analysis
Key technical claims
- Lead of at least 5 min in at least 90% of slow-onset failures (Summary p.2; Obj. 3 p.11). Evidence: the pilot and ref [4]. Questionable. The alarm needs a gas event, so the lead can be no longer than the time from first vent to onset. The proposal's own ref [4] gives a median of 4.2 min for 1C overcharge (p.7). Half of the 16 slow-onset trials are overcharge, yet 15 of 16 must reach at least 5 min. The Obj. 1 criterion (median of at least 6 min for overcharge, p.9) contradicts the same source. The shortest pilot lead, 5.1 min, leaves 6 s of margin before any transport delay.
- Fewer than 1 false alarm per 1,000 h (p.2, p.11). Evidence: 300 bench hours, 4 of 4 nuisance events rejected. Unsupported as designed. The threshold is tuned on the same 2,000 h used to report the rate. Even zero alarms cannot show the target: 95% upper bound = 3.0/2,000 h = 1.5 per 1,000 h. The rate is counted per 12-unit rack. Scaled to a container of about 400 modules (inferred: $38,000 ÷ $95): 400 × 8,760 h ÷ 12,000 unit-hours per alarm ≈ 290 alarms/yr.
- Pilot: gas leads runaway by 5.1 to 9.8 min, strain rises 0.6 to 2.1 min before the vent (p.6, p.8). Partly supported. Six cells, one chemistry, one trigger, an isolated cell, VOC sensor in still air, self-published [10]. The strain signal may be partly heat. At 2 °C/min, 0.6 to 2.1 min = 1.2 to 4.2 °C ≈ 12 to 42 pm ≈ 10 to 35 µε-equivalent, comparable to the 40 to 110 µε reported. No compensation is described.
- One external strand identifies the failing cell under compression (p.5, p.10). Evidence: the CTO's thesis [7]. Questionable. A grating between two cells responds to both. The strain-free references sit at the strand ends, so they cannot separate local heat from strain at the failing cell.
- 1 pm ≈ 1 µε or 0.1 °C (p.10). Supported. Typical 1550 nm FBG: about 1.2 pm/µε and about 10 pm/°C.
- No direct competitors (p.5). Questionable. Rack-level off-gas detectors for single-cell venting were sold for battery storage by 2024 (e.g., Li-ion Tamer). China's GB 38031 requires a 5-min warning, which has driven gas and pressure sensors into EV packs. The gas-plus-localized-strain fusion may still be new, but the proposal gives no citations for the competitive picture.
- $52 per module at volume (p.4). Questionable. The Phase I unit costs $98 in sensors + $210 strand + $85 board = $393 (p.16). The custom strand alone is 4× the target.
Figures and tables
- Fig. 1 (p.8): the page carries the caption only. No plot, axes, units or error information appear, so the claim cannot be checked. As described, it is one trace of six (7.6 min, near the 7.4 min mean), with no reason given for choosing it.
- Market tables (p.3–4): the arithmetic checks (45×10⁶ kWh × $9.50 = $427.5M; 0.15 × 45 GWh × $11 ÷ 5 = $14.9M; 3,000 × 12 × $95 = $3.4M). The $9.50/kWh figure leaves out interrogators: about 400 strands ÷ 16 channels × $3,200 ≈ $80,000 per container. Sharing one interrogator across more strands by switching divides the 10 Hz rate per strand.
- Success-criteria table (p.11): clearly specified. Its localization row rests on dummy neighbours and its false-alarm row on the tuning data.
- Risk table (p.11), timeline (p.12), Broader Impacts table (p.13), budget tables (p.15–17): planning documents, not evidence.
Methods and analysis
- No independent false-alarm test. A further run with the method frozen is needed.
- Training labels. Positives are labelled "from first vent onward" (p.10), which includes the large gas release after runaway and lets the classifier learn easy cases. Labels should stop at onset. Per-second samples from 30 tests are strongly correlated, and cross-validation grouped by test is not stated.
- Different setting. Positive examples come from accelerating rate calorimetry, which is adiabatic, and from sampling lines. The deployed sensor sits behind a 0.2 L/min pump in a cooled module. Adiabatic conditions also change the vent-to-onset interval, so the Obj. 1 lead-time table will not carry over.
- Internal inconsistency. "Five cells per combination" implies 10 per trigger. The text gives 18 heating and overcharge tests plus 12 nail tests (p.8–9).
- First vent is defined three ways: FTIR dimethyl carbonate above 20 ppm (Obj. 1), audible or pressure-recorded opening (held-out trials), and a pressure sensor (Obj. 4).
- Localization tests (p.10) mostly measure heat from the film heater. Gas-driven swelling needs temperatures past SEI breakdown (80 to 120 °C, p.7), which destroys the cell.
- Obj. 4 has two tests, one per position, on a different cell format, so position effects cannot be separated from test-to-test variation. The 90 s allowance gives 5.1 − 1.5 = 3.6 min, so a 5-min lead remains only if the mini-module lead is at least 6.5 min. Ref [13] reports up to 120 s.
- Nail target (within 10 s): manifold flush alone is 25 mL ÷ 0.2 L/min = 7.5 s, before line transport, sensor response and fiber confirmation.
- Statistics: the exact lower bound for 15 of 16 is about 70% (P(X ≥ 15 | p = 0.70) ≈ 0.026). This is correct and acknowledged.
Feasibility
- Cycler (p.18) cannot run the rack. It provides 5 V and 50 A per channel. A 16-cell, 1 kWh LFP module is about 51 V if wired 16S. Only 1S16P (about 312 Ah) fits, which caps the rate at ≤0.16C, too low for a frequency-response profile. This undermines the main false-alarm dataset.
- Calorimetry time: 54 instrument hours ÷ 18 tests = 3 h per test, including setup, runaway and cooldown.
- Staff: about 11.4 person-months (3.0 + 2.5 + 4.0 + ~1.9) for 56 destructive tests, 20 units, firmware, classifier, rack run and 40 localization tests. The technician's 300 h ÷ 54 cell tests ≈ 5.6 h each, with no time left for the rack.
- Interrogator: rented for 9 months but needed from month 1 to month 11. It is needed at Eastbrook and Larkhaven at the same time in months 2 to 4, and only 4 Eastbrook trips are budgeted.
- Most likely to slip: the units must be built, calibrated and pass a 14-day stability test before month 5 so the 2,000 h (11.9 weeks) finish before the month-8 freeze. Any delay squeezes the held-out trials and Obj. 4.
Technical risks not addressed
- Cell format. Container storage is dominated by large prismatic LFP cells in hard cans held by end plates. The pouch-swelling evidence [9] and the 5 Ah pouch tests may not transfer. Fix: add prismatic cells to Obj. 1 and 2.
- Liquid-cooled, sealed modules have no exhaust airflow to sample. Fix: a sampling concept for these packs.
- Gas-sensor poisoning by HF, organophosphates and siloxanes, and recovery after a vent. A 14-day stability test says little about a 10-year service life. Fix: re-calibrate after exposure; accelerated poisoning tests.
- LFP cells may never reach the 1 °C/s onset criterion. How such trials are scored is undefined.
- Strain transfer. Sideways pad pressure puts little axial strain into unbonded fiber and can split the spectral peak. Fix: bench test of strain transfer under the pad.
- Test environment. An open-lab rack is not a container (temperature extremes, dust, enclosed air).
Strongest technical points
- Quantitative success criteria fixed in advance, the method frozen before the held-out trials, and gas-only lead time reported to show the cost of confirmation (p.11).
- Strong Obj. 1 instrumentation: FTIR plus GC-MS for hydrogen, correction for line delay, and commercial sensors calibrated in the same events (p.9).
- A sound physical basis for fusion (two independent signs of gas generation), and a team that pairs failure analysis with FBG expertise.
- Candid limits on the pilot (p.8) and correct binomial statistics.
Individual reviews
Summary
Larkhaven proposes a module that combines vent-gas sensors with fiber Bragg gratings to detect and locate a failing lithium-ion cell before thermal runaway. Phase I would characterize signals, build prototypes, and test lead time and false alarms.
Intellectual Merit — Strengths
- Fusing gas detection with localized strain has a sound physical rationale and could advance pack-level warning beyond either signal alone (pp. 5, 7).
- FTIR and gas chromatography provide reference measurements; the plan freezes the classifier before held-out failure trials and reports gas-only lead times for comparison (pp. 9–11).
- The pilot results are promising, though limited to six heated NMC cells (p. 8); the PI and CTO bring complementary failure-analysis and optical-sensing expertise (pp. 6, 19).
Intellectual Merit — Weaknesses
- The five-minute alarm target is ambitious relative to the cited 4.2-minute median vent-to-runaway interval for overcharge (p. 7). The shortest pilot gas lead is 5.1 minutes before module transport and confirmation delays; Figure 1 has a caption but no visible trace to examine (p. 8).
- The same 2,000-hour rack run sets thresholds and reports false alarms (p. 10), so that rate lacks independent validation. Even zero alarms would give a 95% upper bound of about 1.5 per 1,000 rack-hours, above the stated target (p. 11).
- Held-out trials use dummy neighbors, and only two full-module tests assess transport (pp. 10–12). These leave uncertainty about localization among live cells and performance across module positions and designs.
Broader Impacts — Strengths
- Earlier, reliable warning could protect responders and storage-site workers (p. 13). Internships, fire-service briefings and a public data subset have specified targets and measures (pp. 13, 21).
Broader Impacts — Weaknesses
- Preventing propagation and lowering insurance or permitting costs remain potential outcomes, not outcomes tested in Phase I (pp. 3–4, 12–13). The proposed public subset omits sensor features needed to independently assess the alarm method (p. 21).
Solicitation alignment
- SC-01: Invitation dated November 2023 is described on p. 22; the email itself was not supplied for review.
- SC-02: Four employees and angel financing are reported (p. 6); ownership and registrations are not supplied for review.
- SC-03: Full-time PI, three months’ effort, no co-PI (pp. 1, 6, 15).
- SC-04: Subaward and consultant costs total 15.3% (p. 15).
- SC-05: $273,737 request; prohibited lines are zero; rate and fee meet limits (pp. 15–17).
- SC-06: Summary has the prescribed headings, opening, keywords and subtopic (p. 2).
- SC-07: Eleven-page description specifies technical and commercial questions and milestones (pp. 3–13).
- SC-08: Grantee-conference attendance is acknowledged (p. 16).
- SC-09: Required proprietary-data statement appears (p. 21).
- SC-10: Commitment letters appear on p. 24; the separate subaward budget was not supplied for review.
- SC-11: Biosketches and current/pending support appear (pp. 19–20); cover sheet, questionnaire, certifications and affiliations were not supplied for review.
- SC-12: Interviews and stakeholder letters support market interest (pp. 4, 23), but the durable advantage depends on unproven performance and prospective fiber-IP licensing (pp. 5–6).
- SC-13: Impacts plan identifies metrics, partners, risks and resources (p. 13).
- SC-14: Three support letters include affiliations and contact details (p. 23).
- SC-15: Final reporting is scheduled for months 11–12 (p. 12).
Overall
This is a significant, original Phase I feasibility study with capable investigators and useful measurements. The lead-time margin and non-independent false-alarm assessment materially limit confidence in its central performance claim.
Summary
Larkhaven proposes to combine vent-gas sensors and fiber Bragg gratings to identify a failing battery cell before thermal runaway. Phase I would characterize failure signals, build modules, and test detection, localization, and false alarms (pp. 2, 8–12).
Intellectual Merit — Strengths
- The fusion concept has a clear physical rationale, and FTIR/GC-MS measurements with transport-delay correction should establish useful reference signals (pp. 7, 9).
- Quantitative milestones, a classifier frozen before held-out trials, and reporting of gas-only versus fused lead time strengthen the feasibility test (pp. 10–11).
- The PI and CTO bring complementary battery-failure and fiber-sensing expertise; relevant facilities and commitments are documented (pp. 18–19, 24).
Intellectual Merit — Weaknesses
- The five-minute target is ambitious relative to the cited 4.2-minute median vent-to-onset interval for overcharge, which accounts for half the slow-onset trials. The six-cell pilot covers heating only, and Figure 1 has a caption but no visible trace to examine (pp. 7–8, 11).
- The same 2,000-hour rack run sets thresholds and reports false alarms; even zero alarms would give an approximate 95% upper bound of 1.5 per 1,000 hours. The specified 5 V cycler also cannot cycle a conventional 16-series-cell LFP module as described (pp. 10–11, 18).
- Held-out localization uses dummy neighbours, while only two full-module tests address transport. Thus, the trials incompletely test performance amid live-cell coupling and module variation (pp. 10–12).
Broader Impacts — Strengths
- Preventing propagation could protect responders and storage-site workers. The public data subset, paid internships, and fire-service briefings have stated output measures and identified partners (p. 13; DMP, p. 21).
Broader Impacts — Weaknesses
- Briefing attendance and knowledge checks measure immediate learning, not whether responders can use alarms operationally; the claimed reductions in fires and insurance costs remain prospective (p. 13).
Solicitation alignment
- SC-01: Page 22 dates an invitation within 12 months; the email was not supplied for review.
- SC-02: Four employees and angel financing are reported (p. 6); ownership and registrations were not supplied for review.
- SC-03: Full-time PI, three months’ effort, no co-PI (pp. 1, 6, 15).
- SC-04: Subaward plus consultant costs are 15.3% (p. 15).
- SC-05: $273,737 request; prohibited lines zero, consultant at $1,000/day, fee 7% (pp. 15–17).
- SC-06: Required summary headings, opening, keywords, and subtopic appear (p. 2).
- SC-07: Eleven-page description states technical and commercial questions and milestones (pp. 3–13).
- SC-08: Conference attendance is acknowledged (p. 16).
- SC-09: Required proprietary-data statement appears (p. 21).
- SC-10: Commitment letters appear (p. 24); consultant signature cannot be verified from the image, and the separate subaward budget was not supplied for review.
- SC-11: Biosketches and current/pending support appear (pp. 19–20); cover sheet, questionnaire, certifications, and COA were not supplied for review.
- SC-12: Interviews and conditional support indicate interest, but the $52 volume cost target is unsubstantiated against $393 in listed prototype sensors, strand, and board (pp. 4, 16, 23).
- SC-13: Broader-impacts activities specify metrics, partner engagement, risks, and resources (p. 13).
- SC-14: Three support letters provide affiliation and contact details (p. 23).
- SC-15: A final report is scheduled for months 11–12 (p. 12).
Overall
This is a promising, well-instrumented Phase I project with an appropriate team. Validation and execution weaknesses materially limit confidence that its lead-time and false-alarm targets can be established as proposed.
Summary
Larkhaven proposes an SBIR Phase I prototype that combines vent-gas sensing with fiber-optic measurements to warn of, and locate, a failing battery cell before thermal runaway. It will characterize cell failures, test a fusion algorithm, and assess module-scale performance and commercial feasibility (pp. 2, 8–12).
Intellectual Merit — Strengths
- Gas sensing plus localized fiber confirmation addresses a consequential technical risk; six pilot tests support feasibility for heated NMC pouch cells, while explicitly leaving other conditions unproved (pp. 7–8).
- FTIR and gas chromatography provide reference measurements, and the plan freezes the algorithm before held-out failure trials (pp. 9–11). The team’s relevant expertise and facilities support execution (pp. 18–19).
Intellectual Merit — Weaknesses
- The five-minute target is ambitious relative to the proposal’s cited 4.2-minute median vent-to-onset interval for overcharge; confirmation and transport can only shorten the available lead (pp. 7, 11–12). Figure 1 has a caption but no visible plot with which to inspect the pilot trace (p. 8).
- False-alarm thresholds are set and evaluated on the same 2,000-hour rack run. Even zero alarms would not establish fewer than one per 1,000 hours with strong confidence (pp. 10–11).
- The listed cycler is rated 5 V per channel, whereas the proposed 16-cell LFP modules may require substantially higher voltage; the rack configuration is not explained (pp. 10, 18). Dummy neighbors in the held-out trials also limit validation of localization among live cells (p. 11).
Broader Impacts — Strengths
- Preventing propagation could benefit firefighters, transit workers, riders, and storage-site neighbors (p. 13). The plan specifies measurable library, internship, and responder-briefing outputs, with named partners and resources (p. 13).
- A public data subset is planned while proprietary sensor features remain protected (Data Management Plan, p. 21).
Broader Impacts — Weaknesses
- Downloads, attendance, and a brief knowledge check measure reach more clearly than sustained safety practice; the plan does not evaluate whether warnings improve response decisions (p. 13).
- Customer and insurer letters express conditional interest, not demonstrated reductions in losses, insurance costs, or permitting burden (p. 23).
Solicitation alignment
- SC-01: A November 2023 invitation is reported (p. 22); the invitation email was not supplied for review.
- SC-02: Four employees and angel financing are reported (p. 6); ownership and registration evidence was not supplied for review.
- SC-03: Full-time PI, three months’ effort, no co-PI (pp. 1, 6, 15).
- SC-04: Subaward plus consultant costs are 15.3% of the request (p. 15).
- SC-05: $273,737 request; prohibited budget lines are zero, consultant rate and fee meet stated limits (pp. 15–17).
- SC-06: Required summary headings, opening, keywords, and subtopic appear (p. 2).
- SC-07: Eleven-page description sets technical and commercial questions and milestones (pp. 3–13).
- SC-08: Conference attendance is acknowledged (p. 16).
- SC-09: Required proprietary-data statement appears (p. 21).
- SC-10: Commitment letters appear (p. 24); the separate subaward budget was not supplied for review.
- SC-11: Biosketches and current/pending support appear (pp. 19–20); cover-sheet materials, questionnaire, certifications, and affiliations were not supplied for review.
- SC-12: Interviews and a business model support market interest, but the cost model and durable advantage remain to be demonstrated (pp. 4–6, 12).
- SC-13: Metrics, partner engagement, risks, and resources are specified (p. 13).
- SC-14: Three stakeholder letters include affiliation and contact details (p. 23).
- SC-15: A final report is scheduled for months 11–12 (p. 12).
Overall
This is a significant, appropriately risky commercialization effort with a capable team and useful milestones. The unresolved lead-time margin and lack of independent false-alarm validation materially limit confidence in its central performance claim.
Panel discussion
I land at Very Good/Good: the gas-and-strain approach has a credible physical rationale, and the reference measurements and held-out trials make this a worthwhile Phase I study. My main concern is whether it can meet the five-minute warning target when p. 7 cites a 4.2-minute median vent-to-runaway interval for overcharge, before module transport and confirmation delays. I also cannot put much weight on the false-alarm target because the same 2,000-hour rack run sets thresholds and measures the reported rate (pp. 10–11).
I land between Very Good and Good: the sensor-fusion rationale, reference measurements, and frozen-classifier plan make this a credible Phase I effort. My main concern is whether it can establish the claimed five-minute warning, given the 4.2-minute median vent-to-onset interval for overcharge and the heating-only pilot on pp. 7–8. I also put weight on the validation design on pp. 10–12: the rack run both sets thresholds and reports false alarms, while localization relies largely on dummy neighbours rather than live-cell coupling.
I land at Good: the safety application is significant, and the page 13 broader-impacts plan names partners, resources, and measurable activities. My main concern is that its measures of downloads, attendance, and knowledge do not show whether warnings improve response decisions. I also think the five-minute warning target needs scrutiny against the 4.2-minute median vent-to-onset interval cited on page 7.
I agree with B that the 5 V-per-channel cycler on p. 18 raises a real execution question for the 16-cell rack run on p. 10, though an explained test configuration could resolve it. C’s point about the p. 13 briefing measures also changes how much weight I give the broader-impacts plan: attendance and knowledge checks do not establish better response decisions. I remain between Very Good and Good because the reference measurements and frozen classifier on pp. 9–11 make this a worthwhile Phase I test, while the short lead-time margin and non-independent false-alarm assessment still limit what it can establish.
I agree with A that the false-alarm claim is not independently tested: the rack run on pp. 10–11 both sets the thresholds and supplies the reported rate. I share C’s concern about the response measures on p. 13, though I would give it less weight than the technical validation gaps because those activities are specified and measurable for Phase I. I would also be careful not to treat the 4.2-minute vent-to-onset median on p. 7 as disproving a five-minute warning, since an earlier strain signal is the premise being tested; the heating-only pilot and limited full-module trials on pp. 8 and 10–12 leave that premise uncertain.
I agree with A and B that the frozen-classifier trials and reference measurements on pp. 9–11 make this a credible Phase I study, even though the same rack run cannot independently establish the false-alarm rate. I would not treat the 4.2-minute vent-to-onset median on p. 7 as proving a five-minute warning impossible, since the fiber signal may precede venting, but it leaves a narrow margin that the proposed trials need to demonstrate. I remain at Good rather than their Very Good/Good because the response-decision benefit is not tested by the page 13 broader-impacts measures, and the letters on p. 23 establish conditional interest rather than safety or commercial outcomes.
I remain between Very Good and Good because the reference measurements and frozen-classifier trials on pp. 9–11 make this a strong Phase I test, but the narrow lead-time margin, non-independent false-alarm assessment, and limited live-cell module testing leave its central performance claim uncertain. Final rating: Very Good/Good
I remain at Very Good/Good because the reference measurements and frozen-classifier trials on pp. 9–11 make this a credible Phase I study, but the heating-only pilot, limited full-module tests, and non-independent false-alarm assessment leave its central performance claims uncertain. Final rating: Very Good/Good
I remain at Good: the reference measurements and frozen-classifier trials on pp. 9–11 make this a credible Phase I study, but the rack run does not independently validate false alarms, and the p. 13 broader-impacts measures do not test whether warnings improve response decisions. Final rating: Good
Panel summary
Panel consensus — The panel considered this a credible Phase I feasibility study supported by a sound sensor-fusion rationale, complementary expertise, and strong reference measurements. The proposed validation does not yet adequately establish warning time, false-alarm performance, or localization in representative live-cell modules. Broader-impacts activities are concrete, but their measures do not establish improved operational response. These moderate limitations support a Very Good/Good merit assessment and an estimated Competitive recommendation.
Intellectual Merit — Consensus strengths
- Gas-and-strain fusion has a credible physical rationale and could improve battery-failure detection and localization.
- FTIR/GC-MS reference measurements, quantitative milestones, a classifier frozen before held-out failure trials, and gas-only comparisons strengthen the feasibility study.
- The investigators combine relevant battery-failure and fiber-sensing expertise.
Intellectual Merit — Consensus weaknesses
- Moderate: The five-minute warning target remains uncertain. The cited overcharge median is 4.2 minutes from venting to onset; the shortest heating-pilot gas lead is 5.1 minutes before transport and confirmation delays. Earlier strain does not itself establish the gas-dependent fused alarm’s lead time. Figure 1 contains no visible trace.
- Moderate: The same 2,000-hour rack run tunes thresholds and measures false alarms. Even zero alarms gives an approximate 95% upper bound of 1.5 per 1,000 rack-hours, above the target.
- Moderate: Dummy neighbors and only two full-module transport tests limit validation of live-cell localization and performance across module configurations.
- Moderate: The 5 V-per-channel cycler requires an explained, feasible configuration for the proposed 16-cell rack testing.
Broader Impacts — Consensus strengths
- Reliable early warning could protect responders, workers, and communities near battery installations.
- Internships, responder briefings, and public-data activities identify partners, resources, and measurable outputs.
Broader Impacts — Consensus weaknesses
- Moderate: Downloads, attendance, and knowledge checks measure reach and immediate learning rather than improved response decisions or sustained safety practice.
- Safety, insurance, and permitting benefits remain prospective; stakeholder letters establish conditional interest rather than demonstrated outcomes.
Solicitation-specific criteria
- SC-01: Not verifiable; invitation timing is reported, but full compliance remains unconfirmed.
- SC-02: Not verifiable; reported company size is consistent with eligibility, but ownership and registration compliance remain unconfirmed.
- SC-03: Addressed: full-time company PI, sufficient effort, and no co-PI.
- SC-04: Addressed: outsourcing is 15.3%, below one-third.
- SC-05: Addressed: request, budget lines, consultant rate, and fee meet limits.
- SC-06: Addressed: required summary structure and content appear.
- SC-07: Addressed: eleven-page description includes technical and commercial feasibility questions, milestones, and success criteria.
- SC-08: Addressed: grantee-conference attendance is acknowledged.
- SC-09: Addressed: required proprietary-data statement appears.
- SC-10: Partially addressed; full compliance is not verifiable.
- SC-11: Partially addressed; full compliance is not verifiable.
- SC-12: Addressed with moderate reservations: interviews and business planning support market interest, but volume costs and durable advantage remain unsubstantiated.
- SC-13: Addressed: metrics, partners, risks, and resources are specified; operational-outcome assessment is limited.
- SC-14: Addressed: three support letters contain affiliation and contact information.
- SC-15: Addressed: final reporting is planned in months 11–12.
Points of disagreement
- Reviewer C weighted operational broader-impact evidence more heavily and retained Good; A and B retained Very Good/Good. This difference remained unresolved.
- Discussion resolved differing interpretations of the warning-time evidence as a moderate feasibility gap, not proof that the approach cannot succeed.
Merit assessment Overall merit: Very Good/Good
Comparative limitations — Competing proposals, portfolio priorities, and funding constraints were unavailable; the recommendation is an estimate, not a rank or funding prediction. Materials not supplied included the invitation email, ownership/registration evidence, separate subaward budget and justification, third-party facility owner statements, Cover Sheet, questionnaire, certifications, and COA. Their absence from the submission is not established. Consultant-signature and format verification remain unresolved. Compliance is “Compliant with concerns,” not verified Non-compliant.
Justification — The reference measurements, held-out failure trials, and capable team support a worthwhile Phase I investigation. Uncertain warning margin, non-independent false-alarm assessment, limited representative module testing, and incomplete operational-impact evaluation constrain confidence in the central claims and hold overall merit at Very Good/Good, supporting the estimated Competitive recommendation.
Two reviewers at Very Good/Good, one at Good.
Panel decision confidence: Moderate2 of 3
Recommended revisions
The panel's weaknesses and gaps, each with how to fix it, grouped by review stage. Where the fix is wording, GrantPanel suggests the words; where it is your research (aims, methods, results), it shows what to write and leaves the writing to you. It never edits your document.
Intellectual Merit — warning-time feasibility
Address next- Reconcile the five-minute warning target with the available lead time The panel found the five-minute fused-alarm target uncertain: the proposal cites a 4.2-minute median vent-to-onset interval for overcharge, and the shortest heating-pilot gas lead is 5.1 minutes before confirmation or transport delay. Figure 1 has a caption but no visible trace.Project Description, pp. 7–9 and 11–12; restore Figure 1 beside its caption on p. 8.How to fix
- Reconcile the cited overcharge interval, pilot range, and gas-dependent fused-alarm target in the success-criterion discussion.
- Supply the missing Figure 1 trace with its axes and units.
NowIn the held-out trials, an alarm naming the correct cell at least five minutes before onset in at least 90% of slow-onset trials (at least 15 of the 16 overcharge and heating trials), and detection within 10 seconds of first vent in at least 7 of the 8 nail penetration trials; and fewer than one false alarm per 1,000 operating hours in the rack run
Answer firstWhat evidence or revised feasibility criterion will you use to assess a five-minute fused alarm in the overcharge trials?
What to writeIn the held-out trials, an alarm naming the correct cell at least five minutes before onset in at least 90% of slow-onset trials (at least 15 of the 16 overcharge and heating trials), and detection within 10 seconds of first vent in at least 7 of the 8 nail penetration trials; and fewer than one false alarm per 1,000 operating hours in the rack run. [Explain how the cited overcharge and pilot intervals, followed by confirmation and module transport, inform the five-minute feasibility decision; revise the target or its interpretation if warranted.]
Intellectual Merit — independent validation
Address next- Separate false-alarm tuning from performance assessment The same 2,000-hour rack run sets the decision threshold and reports the false-alarm rate. The panel noted that even zero alarms over that period gives an approximate 95% upper bound of 1.5 per 1,000 rack-hours, above the stated target.Project Description, Objective 3, “Normal-operation rack” and “Milestone and success measure,” pp. 10–11.How to fix
- Specify which data set thresholds and confirmation margins and which data independently assess false alarms.
- State the unit of operating time and the evidence needed to interpret the reported rate against the target.
NowThe same 2,000 hours of rack operation will be used to set the classifier's decision threshold and the confirmation margins and to report the false-alarm rate, so that the threshold reflects the full range of normal conditions seen in the rack
Answer firstWhat independent normal-operation observations can you use to assess false alarms after the threshold and margins are fixed?
What to write[Specify the data used to set the classifier's decision threshold and confirmation margins, and the independent data used to report the false-alarm rate. Explain how the observation period and unit of operating time support interpretation of the fewer-than-one-per-1,000-operating-hours target.]
Intellectual Merit — representative testing
Address next- Address the limits of representative-module validation The held-out trials use dummy neighbors, and only two full-module tests assess transport. The panel found that these tests leave live-cell localization and performance across module configurations insufficiently validated.Project Description, Objective 3 “Held-out integrated failure trials” and Objective 4, pp. 10–12.How to fix
- Explain what the dummy-neighbor trials can and cannot establish about localization among live cells.
- Specify what comparison or evidence will address variation beyond the two stated module positions.
NowThe three neighboring positions hold thermally matched dummy cells (pouch blanks filled with aluminum foam), so that each trial consumes one live cell
Answer firstHow will you establish or qualify the cell-localization claim when neighboring positions contain live cells?
What to writeThe three neighboring positions hold thermally matched dummy cells (pouch blanks filled with aluminum foam), so that each trial consumes one live cell. [State what localization conclusion these trials permit, and supply the author-selected evidence or comparison needed to address live-cell coupling and module-configuration variation.]
Intellectual Merit — execution feasibility
Address next- Explain how the proposed rack can be cycled The panel questioned whether the listed 5 V-per-channel cycler can run the proposed 16-cell LFP modules for the 2,000-hour rack test; the module wiring and feasible cycler configuration are not explained.Project Description, Objective 3 “Normal-operation rack,” p. 10; Facilities, Equipment and Other Resources, p. 18.How to fix
- State the rack modules’ electrical configuration and the cycler connection used for the specified duty profile.
- Reconcile that configuration with the listed 5 V, 50 A per-channel capability.
NowTwelve commercial LFP modules (16 cells, about 1 kWh each), each fitted with an EW-1 unit, will be cycled on the company's 16-channel cycler through a duty profile taken from a grid storage frequency-response trace, for 2,000 hours (about 12 weeks), in the open laboratory
Answer firstWhat are the modules’ electrical configuration and the cycler connection that will run the stated duty profile?
What to writeTwelve commercial LFP modules (16 cells, about 1 kWh each), each fitted with an EW-1 unit, will be cycled on the company's 16-channel cycler through a duty profile taken from a grid storage frequency-response trace, for 2,000 hours (about 12 weeks), in the open laboratory. [Specify the module wiring, cycler connection, and compatible operating range for this run; revise the rack plan if the listed cycler cannot execute it.]
Broader Impacts — outcome assessment
Address next- Measure an operational outcome of responder briefings The panel found that attendance and immediate knowledge checks measure reach and learning, not whether early-warning information improves response decisions or sustained safety practice.Project Description, Broader Impacts implementation-plan table, “Briefings for fire service personnel” row, p. 13.How to fix
- Identify the response decision or practice the briefings are intended to affect.
- Add an author-chosen measure and timing that can assess that outcome, distinguishing it from attendance and the existing knowledge check.
NowAttendance records; a five-question knowledge check before and after each briefing
Answer firstWhat response decision or practice can the fire-service partner help assess, and when can it be measured?
What to writeAttendance records; a five-question knowledge check before and after each briefing. [Add the measure and timing selected with the fire-service partner to assess use of early-warning information in response decisions or sustained practice.]
Compliance — permitted document placement
The solicitation requires- Resolve the Project Pitch invitation document’s placement Page 22 is labeled “Supplementary Document: Project Pitch Invitation” but contains a narrative rather than the invitation email. The solicitation requires the email in Single Copy Documents and restricts what may be uploaded as Supplementary Documents; the actual upload location cannot be verified from this packet.Page 22, “Supplementary Document: Project Pitch Invitation”; submission document placement.How to fix
- Check where page 22 was uploaded and remove it if it occupies an unpermitted Supplementary Documents slot.
- Verify that the actual invitation email is uploaded as a PDF in the required Single Copy Documents location.
NowLarkhaven Sensing submitted a Project Pitch on this project in October 2023 and received an invitation from the cognizant NSF SBIR/STTR Program Director for Energy Technologies to submit a full Phase I proposal on November 20, 2023. The invitation email is uploaded with this proposal
Answer firstWhere in Research.gov were the page 22 narrative and the actual invitation email uploaded?
What to write[Verify the actual Project Pitch invitation email is uploaded as a PDF in Single Copy Documents. Remove this narrative from Supplementary Documents if it was uploaded there in place of, or in addition to, permitted material.]
Commercial Impact — cost credibility
Address next- Substantiate the projected volume cost Under the solicitation’s Commercial Impact criterion, the panel reserved judgment on volume economics. The proposal projects a $52-per-module bill of materials, while the listed Phase I unit prices for sensors, fiber strand, and board alone total $393.Project Description, Commercial Opportunity “Business economics and market drivers,” p. 4; Commercial feasibility work, p. 12.How to fix
- Show the author’s basis for the volume-cost assumptions against the listed prototype component prices.
- Explain how the planned supplier-quotation cost model will test the $52 target without treating it as established.
NowAt volume (10,000 modules a year) our cost model gives a bill of materials of $52 per module for the gas array, electronics and fiber strand, plus $1,900 per shared 16-channel interrogator
Answer firstWhat documented volume-price assumptions bridge the Phase I component prices to the $52-per-module target?
What to writeAt volume (10,000 modules a year) our cost model gives a bill of materials of $52 per module for the gas array, electronics and fiber strand, plus $1,900 per shared 16-channel interrogator. [Provide the source and assumptions for the volume component costs, reconcile them with the listed Phase I unit prices, and distinguish current estimates from the supplier quotations to be obtained during Phase I.]
Commercial Impact — durable advantage
Address next- Qualify the claimed durable advantage while fiber licensing remains prospective The panel’s Commercial Impact reservation includes the durability of the competitive advantage: the fusion patent application is provisional, and licensing of the Halvern-origin fiber-routing method has not yet been arranged.Project Description, “Intellectual property,” p. 6; Commercial Opportunity “Competition,” p. 5.How to fix
- Distinguish intellectual property already controlled by Larkhaven from the Halvern method for which discussions are planned.
- Explain the author’s basis for a durable advantage if the prospective license is not obtained.
NowTogether, the patent position, the library and the performance data form the barrier to entry described in the Commercial Opportunity section
Answer firstWhich rights to the fiber-routing method does Larkhaven currently hold, and how would the competitive position change without a Halvern license?
What to writeTogether, the patent position, the library and the performance data form the barrier to entry described in the Commercial Opportunity section. [Distinguish the rights Larkhaven currently holds from the Halvern-origin fiber-routing method for which licensing discussions are planned, and explain how that status affects the claimed durable advantage.]
Broader Impacts — calibration of claims
Strengthens the proposal- Distinguish prospective safety and economic benefits from demonstrated outcomes The panel found that reduced fires, insurance costs, and permitting burden are prospective benefits; the stakeholder letters express conditional interest, not demonstrated reductions.Project Description, Broader Impacts “Societal and economic benefit,” p. 13.How to fix
- Qualify the stated prevention and cost-reduction outcomes as potential benefits.
- Keep those outcomes distinct from the Phase I tests and conditional stakeholder interest.
NowAn early-warning module that prevents propagating fires protects firefighters, who face toxic and flammable vent gases when they respond to battery fires, along with transit riders, depot workers and neighbors of storage sites. By lowering the expected loss from a fire, it also lowers the insurance and permitting costs that add to the price of storage
Suggested wordingIf an early-warning module helps prevent propagating fires, it could protect firefighters, who face toxic and flammable vent gases when they respond to battery fires, along with transit riders, depot workers and neighbors of storage sites. If it lowers the expected loss from a fire, it could also reduce insurance and permitting costs that add to the price of storage.
Supporting research
Prior funded work
Related awards found in NSF Award Search, NIH RePORTER, USAspending.gov and SBIR.gov awards (file of October 21, 2026), closest first (searched: battery thermal runaway detection; lithium-ion vent gas; fiber Bragg grating batteries; energy storage fire safety). Use them to judge novelty and positioning, not as a verdict: a related award is context for how the proposal differs, not evidence it should not be funded.
DOE award: In-Module Multi-Gas Sensing for Early Detection of Lithium-Ion Cell Venting Elena Varnholt, Quarry Lane Instruments, LLC · $206,500 · 2022–2023 · SBIR Phase I Develops a multi-gas sensor installed inside battery modules to detect cell venting before thermal runaway. It overlaps closely with the proposal's gas array but uses no fiber confirmation or cell localization.
NSF award: SBIR Phase I: Fiber-Optic Strain Monitoring of Battery Modules for Grid Storage Owen Breckinridge, Lumastrand Photonics Inc. · $274,980 · 2023–2024 · SBIR Phase I Uses distributed fiber strain sensing to monitor swelling and state of health in grid storage modules. It shares the proposal's fiber approach but does not use gas sensing or target early warning of runaway.
DOW award: Shipboard Lithium-Ion Battery Off-Gas Detection and Isolation Patrice Lemaitre-Odum, Gullhaven Safety Systems Inc. · $1,149,000 · 2021–2023 · SBIR Phase II Develops off-gas detection that isolates failing battery modules aboard ships. It is a further-developed gas-only early-warning product in an adjacent market, relevant to the proposal's competition analysis.
NSF award: Collaborative Research: Mechanisms of Gas Evolution Preceding Thermal Runaway in Lithium-Ion Cells Harriet Osgood-Pryce, Ashcombe State University · $389,214 · 2022–2025 · Chemical process systems (CBET) Studies the chemistry and timing of gas evolution before runaway with calorimetry and spectroscopy. Its results overlap the proposal's Objective 1 signature library, which applies similar measurements to commercial sensor selection.
NASA award: Multi-Sensor Thermal Runaway Detection for Spacecraft Battery Packs Graciela Fennimore-Ash, Tamsin Ridge Technologies · $149,850 · 2023–2024 · STTR Phase I Combines gas, pressure and temperature sensing to detect thermal runaway in sealed spacecraft batteries. It addresses multi-sensor fusion in a different environment and does not use fiber sensing.
DOT award: Thermal Event Detection for Battery-Electric Transit Bus Depots Simon Achebe-Hart, Carrow Valley University · $612,000 · 2023–2025 · Transit safety research Evaluates detection and suppression approaches for battery fires in bus depots. It overlaps with the proposal's transit market but tests depot-level systems rather than cell-level warning.
Literature review (deep research)
Research area & central claim
The proposal sits at the meeting point of lithium-ion battery safety, low-cost gas sensing and fiber-optic strain sensing. Its central claim is that confirming a vent-gas event with a localized swelling signal from an external fiber strand can keep the minutes of warning that gas sensing gives while rejecting the nuisance alarms that have kept gas sensing out of packs, and that no product does this today. Review cutoff: May 15, 2024 (close of the final NSF 23-515 submission window).
State of the art at submission
- Tenbrook et al., 2020, Journal of Battery Safety Engineering: established that first venting precedes runaway by a median of several minutes in heating and overcharge, but by under 30 seconds in nail penetration.
- Brevik et al., 2021, Electrochemical Hazards Letters: showed that first-vent gases are dominated by carbon dioxide and carbonate vapor, with much less carbon monoxide and hydrogen from LFP cells than from NMC cells.
- Ferreira-Lind et al., 2021, Journal of Applied Fiber Grating Sensing: demonstrated that one grating strand between pouch cells can resolve the swelling of a single cell in a compressed module.
- Mbeki-Stroud et al., 2021, Journal of Pack-Level Battery Engineering: measured gas transport delays of 30 to 120 seconds and strong dilution in forced-air cooled modules.
- Thibodeaux-Lane et al., 2023, Electrochemical Hazards Letters: tested four commercially available off-gas detectors, two of them designed for installation inside battery modules, in module-level propagation tests and found warnings 4 to 9 minutes before runaway.
- Strandberg et al., 2022, Journal of Pack-Level Battery Engineering: surveyed in-pack venting sensors (pressure, VOC and combined units) offered by battery management suppliers and found them in a growing share of new commercial designs.
- Marquez-Hale et al., 2021, Journal of Applied Fiber Grating Sensing: localized a heated cell in an eight-cell module using a single external grating strand and temperature alone, without strain.
Competing approaches
- Commercial off-gas detectors (Thibodeaux-Lane et al., 2023, Electrochemical Hazards Letters): single- and multi-gas detectors, including in-module units, already give minutes of warning in propagation tests. The proposal describes these products only as room-level single-gas detectors and does not compare its target lead time or false-alarm rate with them.
- In-pack sensing from battery management suppliers (Strandberg et al., 2022, Journal of Pack-Level Battery Engineering): pressure and VOC sensors integrated into new pack designs. The proposal expects such suppliers to add sensing in the future rather than treating them as present competitors.
- Temperature-only fiber localization (Marquez-Hale et al., 2021, Journal of Applied Fiber Grating Sensing): a simpler fiber approach that the proposal's strain-based confirmation must outperform in lead time.
- In-module pressure sensing (Okafor-Lindqvist et al., 2023, Battery Diagnostics Letters): detects venting from the pressure rise inside a sealed module, with no gas sensors. The proposal argues that gas gives the earliest signal and a localized fiber event the confirmation.
Open problems the proposal addresses
- Nuisance alarms from low-cost gas sensors in real battery environments, which limit their use for automatic response.
- Locating the failing cell within a module from sensors outside the cells.
- Retaining vent-based lead time at full module scale with forced-air cooling.
- Weak gas signatures from LFP cells.
Citation gaps at submission
- Thibodeaux-Lane et al., 2023, Electrochemical Hazards Letters: shows that commercial off-gas detectors, including in-module units, already provide several minutes of warning. This bears directly on the proposal's statement that there are no direct competitors and on its claimed differentiation.
- Ilves et al., 2023, Battery Diagnostics Letters: shows that demonstrating a false-alarm rate below one per 1,000 hours requires several thousand hours of normal operation held out from threshold tuning. This bears on the proposal's plan to set thresholds and report the false-alarm rate on the same 2,000 hours of rack data.
- Marquez-Hale et al., 2021, Journal of Applied Fiber Grating Sensing: prior work on single-strand fiber localization of failing cells, relevant to novelty and to the proposal's intellectual property position.
In a real report, each work ends with a link to its own page (DOI, journal, PubMed or arXiv), so you can check it in one click. The works in this sample are invented, so they have none.
Retrospective developments (not used for scoring)
None reviewed.
References lookup
The proposal cites 18 references. Each was looked up in Semantic Scholar (Allen Institute for AI), OpenAlex and Crossref: 16 were found, and the reviewers read their abstracts. Any that no index matched are marked.
- [1] Staged decomposition reactions in lithium-ion cells during thermal abuse: a calorimetric review (2018) — Marchbanks, L., Oduya, P., Sterne, K. · Thermal Runaway Science Quarterly
- [2] Venting and ignition of large-format pouch cells under external heating (2019) — Hallworth, D., Pietrangeli, S. · Journal of Battery Safety Engineering
- [3] Composition of first-vent gases from NMC and LFP cells under three abuse triggers (2021) — Brevik, A., Manotti, C., Seol, J. · Electrochemical Hazards Letters
- [4] Time between first venting and thermal runaway in heating, overcharge and penetration tests of 20 commercial cells (2020) — Tenbrook, R., Achebe-Moss, F. · Journal of Battery Safety Engineering
- [5] Metal-oxide gas sensors for electrolyte vapor detection in battery enclosures (2022) — Quist, H., Delamere, J. · Journal of Chemiresistive Sensing
- [6] Cross-sensitivity and humidity drift of low-cost carbon monoxide and hydrogen sensors in battery rooms (2023) — Ravensworth, P., Ilunga, T., Kask, M. · Journal of Chemiresistive Sensing
- [7] Distributed fiber Bragg grating sensing of pouch-cell swelling in a compressed battery module (2021) — Ferreira-Lind, T., Halloran-Voss, E. · Journal of Applied Fiber Grating Sensing
- [8] Surface-mounted fiber Bragg gratings for temperature mapping of battery packs (2020) — Sandoval-Prieto, M., Whitlow, G. · Journal of Applied Fiber Grating Sensing
- [9] Cell swelling as an early indicator of internal faults in lithium-ion pouch cells (2022) — Nakamura-Holt, R., Ebbesen, L. · Battery Diagnostics Letters
- [10] Gas and strain signatures preceding venting in overheated 5 Ah pouch cells: a pilot study (2024) — Okonkwo-Laurent, R., Ferreira-Lind, T. · Proceedings of the Energy Storage Fire Safety Symposium
- [11] Nuisance alarms in gas detection systems for stationary battery installations (2019) — Castellane, V., Rudd-Amari, S. · not found in the indexes
- [12] Gradient-boosted classifiers for battery fault detection from heterogeneous sensor data (2023) — Hartigan, O., Baptiste-Lowe, N., Fenwick, D. · Battery Diagnostics Letters
- [13] Gas transport and dilution in forced-air cooled battery modules during cell venting (2021) — Mbeki-Stroud, A., Coyle, B. · Journal of Pack-Level Battery Engineering
- [14] U.S. stationary storage deployments and safety spending outlook, 2023-2030 (2023) — Delacourt-Ames, J. · Grid Storage Market Review
- [15] Root causes of thermal events in fielded stationary storage systems, 2017-2021 (2022) — Prewitt, S., Langholm, E. · Journal of Battery Safety Engineering
- [16] Loss trends in battery energy storage property claims (2023) — Abernathy-Kwon, T. · not found in the indexes
- [17] Thermal runaway onset criteria in accelerating rate calorimetry of lithium-ion cells (2017) — Yarrow, P., Danjuma, K. · Thermal Runaway Science Quarterly
- [18] Ageing effects on vent composition and onset temperature in NMC cells (2018) — Wexley-Barr, C. · Electrochemical Hazards Letters