US 10746654

I'll search for litigation/docket information on this specific patent number, then summarize.

Search results on docket/litigation status

I searched for litigation and Federal Circuit activity involving this patent number specifically. I found no evidence of any CAFC appeal, PTAB proceeding, or district court action involving US 10,746,654. The CAFC September 2026 oral-argument calendar and the March 2026 Federal Circuit roundups that surfaced in my searches contain no QFOOD or beer-spectroscopy matters. My search coverage of docket databases is not exhaustive, so treat "no litigation found" as absence of evidence rather than proof of absence. Google Patents' page for this patent lists family members, citing/cited-by documents, and legal events, but no litigation events — consistent with a patent that has not been asserted.

Note also that the earlier fetched Google Patents page shows the search-result snippet for the granted PDF renders the abstract's mean symbol as μ_R(i) / σ_R(i), whereas the claim text and specification use μ_P(j) / σ_P(j) / σ_R(j). The specification itself uses σ_R(i) in the "for each principal component" passages where σ_R(j) is evidently meant. I am flagging these as literal inconsistencies in the document and am not correcting them.

---

Bibliographic data (US 10,746,654 B2)

---

Abstract (as published)

A method for monitoring a beer sample's correspondence with a reference beer. At least 15 reference beer samples are brewed with the same ingredients and process parameters. Infrared absorption spectra of the reference samples are captured and principal component analysis is performed, ascertaining at least 15 principal components. A factor loading P_R(i,j) is determined for each principal component for each reference sample (i = sample, j = component). Reference values R(i,j) are formed from the factor loadings using the mean and standard deviation of the loadings for each component. A reference interval is formed using n (number of reference samples), m (number of components), σ_R(j) (standard deviation of the reference values of the j-th component), and a non-zero constant k. A spectrum of the beer sample is then captured, its factor loadings P_B(i) computed against the reference principal components, and a characteristic value B formed and compared to the reference interval. If B falls outside the interval, a production fault is indicated.

---

Independent claim — plain-language overview

The patent has one independent claim, claim 1; claims 2–5 depend on it.

Claim 1 — a method with these steps:

1. Reference set: At least 15 reference beer samples of the reference beer are brewed with identical ingredients and identical process parameters.
2. Reference measurement: Infrared absorption spectroscopy captures a measurement signal (absorption spectrum) for each reference sample.
3. PCA: Principal component analysis on those signals, ascertaining at least 15 principal components, and determining a factor loading P_R(i,j) for each component for each reference sample (i = reference sample index, j = component index).
4. Reference value: For each sample and each component, compute R(i,j) = |[P_R(i,j) − μ_P(j)] / σ_P(j)|, where μ_P(j) is the mean and σ_P(j) the standard deviation of all loadings of the j-th component (note the claim renders this as "μ p (j)").
5. Reference interval: Form
   [ (k/n)·ΣᵢΣⱼ R(i,j) − k·Σⱼ σ_R(j) … (k/n)·ΣᵢΣⱼ R(i,j) + k·Σⱼ σ_R(j) ]
   where n = number of reference samples, m = number of components, σ_R(j) = standard deviation of all reference values for the j-th component, and k is a non-zero constant.
6. Unknown-sample measurement: Capture an infrared absorption spectrum of the beer sample to be checked and determine its factor loadings P_B(i) against the reference principal components.
7. Characteristic value: Compute B = k·Σⱼ |[P_B(j) − μ_P(j)] / σ_P(j)| from the sample's loadings, the reference means, and the reference standard deviations.
8. Decision: Compare B to the reference interval; indicate an error in production of the beer sample if B lies outside the interval.

Dependent claims (range/scope limitations):

• Claim 2: n ≥ m; in particular n twice, preferably at least three times, m.
• Claim 3: m ≥ 20, optionally ≥ 30, in particular ≥ 40, preferably ≥ 50.
• Claim 4: k equals the reciprocal of m (k = 1/m).
• Claim 5: The infrared irradiation covers a wave number range between 950 and 3050, in particular 960–2000, preferably 980–1200 cm⁻¹.

---

Worked example given in the specification (for context)

100 reference samples brewed identically; spectra measured on a QFOOD QUANTOS® spectrometer over 980–1200 cm⁻¹, each signal comprising 1000 wave-number/absorbance value pairs; 30 principal components extracted (yielding 3,000 factor loadings). With k = 1/30 the resulting reference interval is [9.0368514 … 33.2298688]. For the tested sample, B = 52.28199576, which lies outside the interval, so a production error is indicated.

---

Uncertainty and caveats

• No docket data found. I could not locate any court or PTAB record for US 10,746,654. The "Active" legal status shown by Google Patents reflects maintenance-fee payment, not litigation outcome. If you need a definitive litigation check, this should be confirmed directly against PACER/Docket Navigator and the CAFC docket, which I cannot query authoritatively here.
• Inventor/assignee data are taken from the patent front page and the recorded assignment; they appear internally consistent.
• Symbol inconsistencies in the granted text (μ_R/σ_R vs. μ_P/σ_P/σ_R) are noted above as they literally appear; I have not silently normalized them.
• The DE 10 2016 009 636 B4 family member is the German counterpart; the WO publication is WO 2018/029088 A1 (dating from the same 2016-08-10 priority).

I'll search for litigation involving US patent 10746654 across patent litigation sources.

Let me search more targeted litigation databases.

Litigation Search Results for U.S. Patent No. 10,746,654

Bottom line

I found no litigation — district court, Section 337 (ITC), PTAB, or CAFC — involving U.S. Patent No. 10,746,654 ("Method for monitoring the correspondence of a beer sample with a reference beer"). The patent appears to be unasserted as of the searches performed. I cannot provide a plaintiff/defendant/case-number list because no case record exists that I could locate.

Patent identity (confirmed from the authoritative record)

• Patent: US 10,746,654 B2 — "Method for monitoring the correspondence of a beer sample with a reference beer"
• Assignee: QFOOD GmbH (Germany)
• Inventors: Holger Klapproth, Robert Seidel, Joachim Haas, Jonathan E. Green
• Application: US 16/324,292 (national phase of PCT/EP2017/069698)
• Priority: 2016‑08‑10 (DE 10 2016 009 636.5)
• Granted: 2020‑08‑18; Status: Active; last maintenance fee (4th yr, small entity) paid 2024‑02‑05
• Family: DE102016009636B4, EP3497430B1, WO2018029088A1

Sources: https://patents.google.com/patent/US10746654/en and the patent PDF at https://patentimages.storage.googleapis.com/fb/37/8f/38c43b43652cdc/US10746654.pdf

What I searched

• Unified Patents litigation portal (https://portal.unifiedpatents.com/litigation/caselist) — no entry for this patent or for QFOOD GmbH as a party.
• CourtListener / RECAP (Free Law Project) — keyword searches for "10746654" returned only unrelated matters (e.g., a Tenth Circuit criminal docket United States v. Murry, 10th Cir. No. 20-1214, and a Middle District of Florida case ParkerVision v. Qualcomm, No. 6:14‑cv‑687‑PGB‑LHP). Neither involves this patent.
• PTAB / USPTO PTAB records — no IPR, PGR, or CBM proceeding identified for this patent.
• CAFC — no appeal identified.
• General web/patent-news searches on the patent number, on "QFOOD GmbH" plus "litigation/lawsuit," and on the patent title — no infringement complaint, docket, or news report surfaced.
• Google Patents family/"Cited by" data — shows only three citing patents (CN110455740B, DK3809118T3, CN116609291A) and ten citations/prior-art references; no litigation section and no opponent/assignee-transfer events beyond the original QFOOD GmbH assignment (REEL/FRAME 051075/0234, recorded 2019‑11‑21).

Important disambiguation (do not confuse)

One search surfaced an IPR discussing a "’764 patent" in connection with product(s) called "Verity" and antedating a reference named "Li" (PTAB document hosted at ptacts.uspto.gov). That is not U.S. Patent 10,746,654. The "’764 patent" in that proceeding refers to a differently numbered patent (a different patent whose number merely ends in 764), and the subject matter, parties, and technology do not match the QFOOD beer-analysis patent. I am flagging this only so the similarly-suffixed number is not mistakenly attributed to 10,746,654, per the instruction not to auto-correct or conflate identifiers.

Caveats on completeness

• Google Patents' litigation tab and the sources I could reach are not exhaustive; a very recent filing (within the last few weeks) or a case where the patent is pleaded only by number in a footnote may not yet be indexed.
• A full PACER docket search by patent number was not directly available to me; the free mirrors (RECAP, Justia Dockets) returned nothing for this number.
• QFOOD GmbH's patent is narrow (a quality-control method using IR spectroscopy plus principal component analysis), which is consistent with a low likelihood of assertion, but absence of indexing is not proof of absence of litigation.

Conclusion: Based on the sources searched, there is no known litigation involving U.S. Patent No. 10,746,654. If you need a definitive PACER-based confirmation, a docket search across all U.S. district courts (e.g., via Docket Alarm, Lex Machina, or PACER's "Patent" case query) would be the appropriate final check.

Proceedings overview

No AIA trial proceedings are on file for US 10746654. The USPTO Open Data Portal structured block records zero IPR, PGR, or CBM proceedings (0 active / 0 claims invalidated / 0 claims sustained / 0 settled / 0 institution denials), and my web searches surfaced no PTAB petition, Final Written Decision, or Federal Circuit appeal involving this patent or its owner, QFOOD GmbH. Bottom line for a defendant: the patent is completely untested at the Board — all five claims stand unadjudicated, and the patent owner has never had to defend it in an AIA trial. There is no invalidated-claim leverage to point to, and equally no estoppel constraining you: you can bring any § 102/§ 103 ground you can find.

---

Proceedings

None. No proceeding numbers exist to report, and I will not construct any.

For completeness, here is what I checked and found:

Caveats on that negative finding: the ODP block is the canonical source and it is authoritative for indexed proceedings; web search is a secondary check and can miss recently filed petitions that have not yet been indexed or published. I found no evidence of a recently filed, unindexed petition either, but absence of search hits is not proof of absence for filings within the last few weeks. If you need certainty, the definitive check is the PTAB E2E / PTAB Center docket search by patent number (https://ptacts.uspto.gov/ptabweb/) and the Patent Center file wrapper for application 16/324,292 (https://patentcenter.uspto.gov/).

Related context that is not a PTAB proceeding but matters to the defensive picture:

• The German priority application DE102016009636.5 issued as DE102016009636B4 (published 2018-07-12); the EP member EP3497430B1 granted 2021-01-06. A granted EP member is opposable at the EPO for 9 months from grant, and I found no EPO opposition data one way or the other — if you are defending in Europe, check the EPO Register directly at https://register.epo.org/ for EP17751070.8.
• Anticipated US expiration: 2037-08-03 (per the ODP record), with the 4th-year maintenance fee paid 2024-02-05 (small entity). Roughly eleven years of term remain, so this is not a patent you can simply wait out.

---

Strategic summary

Claim status: all five claims are UNTESTED. No claim of US 10746654 has been canceled, narrowed, or confirmed by the PTAB. Claim 1 (the sole independent claim) and dependent claims 2, 3, 4, and 5 are all live and in force. Claim 1 recites the core method: brewing ≥15 reference beer samples with identical ingredients/process parameters, capturing IR absorption spectra, running PCA to obtain ≥15 principal components, computing normalized reference values R(i,j) = |(P_R(i,j) − μ_P(j)) / σ_P(j)|, forming the reference interval from k/n ΣΣR(i,j) ± k Σσ_R(j), then computing characteristic value B = k Σ|(P_B(j) − μ_P(j)) / σ_P(j)| for the sample and flagging a production error if B falls outside the interval. Dependent claim 2 adds n ≥ m (twice, preferably three times as great); claim 3 adds m ≥ 20 (optionally 30/40/50); claim 4 sets k = 1/m; claim 5 recites the 950–3050 cm⁻¹ range (preferably 980–1200). Because no IPR exists, there is no narrowing claim construction from a PTAB FWD and no court-friendly "the patent owner already conceded this" record.

Estoppel landscape: none attaches. Section 315(e)(2) estoppel arises only against a petitioner (and its real parties in interest/privies) that obtained a Final Written Decision. With zero IPRs, no party is estopped, and there is no prior art that has been "used up." Every ground that existed on the 2016-08-10 priority date is still available, including: (a) the references the examiner already considered — US5446681 (Exxon, method of estimating property/composition data of a test sample), DE10108712A1 (Warsteiner Brauerei, method for analytical analysis of a beer sample), WO2012167805A1 (Qfood GmbH itself, concentration of an analyte in a liquid sample), and US20150060674A1 (Haffmans, optically determining alcohol/carbohydrate concentration); and (b) the non-patent literature cited on the face of the patent, notably Foyolle et al., "Determination of Major Compounds of Alcoholic Fermentation by Middle-Infrared Spectroscopy: Study of Temperature Effects and Calibration Methods," Applied Spectroscopy, Vol. 50, No. 10 (1996), pp. 1325–1330. None of these has been tested for patentability, which cuts both ways: a challenger is free to run them, but also has no Board precedent telling it which combination the panel finds persuasive.

Pattern signals. There is no petitioner to pattern — no entity has challenged this patent at the PTAB, and there is no defensive aggregator (Unified Patents, RPX, Open Invention Network, etc.) in the chain. The patent owner, QFOOD GmbH, appears to be an operating German instrumentation company (the specification names its own QFOOD QUANTOS® IR absorption spectrometer as the measurement device), not a litigation-oriented NPE, and there is no visible history of aggressive PTAB appeals or serial assertion. That profile suggests a patent asserted, if at all, to protect a commercial instrument business rather than as a licensing-revenue weapon. The practical inference is that nobody has found this patent worth an IPR yet, so it has neither been hardened nor wounded — it is simply an unknown quantity.

---

Recommended next steps

If you are a defendant facing assertion today:

• Do not represent to a court or to your client that any claim is invalid. There is no FWD to cite. The only public claim-construction and validity record is the prosecution history of application 16/324,292 and the parallel DE/EP prosecution — pull both via Patent Center and the EPO Register.
• Watch your § 315(b) clock: a petition must be filed within one year of service of a complaint alleging infringement of this patent. There is no prior IPR, so you get a clean, full-strength shot; there is no follow-on-petition problem under General Plastic and no § 325(d) "same art previously presented" risk beyond the examiner-cited references, which you should confront head-on if you build a § 103 combination from them.
• The highest-value invalidity theory is likely § 103 over the beer/fermentation-analysis art (Foyolle et al. plus DE10108712A1 and/or US5446681) attacking the recitation of PCA plus a normalized z-score-style deviation metric, which raises an Alice/§ 101 question as well — a mathematical-relationship-plus-method-of-analysis challenge is worth scoping even though § 101 is not an IPR ground. If § 101 is your angle, it must go through a district court or a PGR — and note the PGR window closed 2021-05-18 (9 months after the 2020-08-18 grant), so IPR at the Board is now the only AIA route for prior-art grounds. CBM is unavailable: the transitional program sunset for new petitions on 2020-09-16, and this is a beer-analysis patent that would not qualify as a "covered business method" in any event.
• Because claim 4 sets k = 1/m and the specification's worked example uses k = 1/30 with the interval [9.0368514 . . . 33.2298688], quantify your accused product's B values early. If your instrument's characteristic values sit inside that interval, infringement under a § 112-supported construction becomes very hard to prove; this is the cheapest non-infringement story the patent's own specification hands you.

If you are monitoring rather than defending:

• The absence of PTAB activity is itself the signal — well-asserted patents in active litigation clusters eventually attract IPRs, and this one has attracted none in more than six years from priority. Set an alert on PTAB E2E (https://ptacts.uspto.gov/ptabweb/) and on the Patent Center file wrapper for 16/324,292, and re-check the EPO Register for EP3497430B1 opposition status.

Links used for verification:

• Patent page: https://patents.google.com/patent/US10746654/en
• PTAB proceeding search: https://ptacts.uspto.gov/ptabweb/
• PTAB decisions (public FWDs): https://www.uspto.gov/patents/ptab/decisions
• Patent Center (app 16/324,292): https://patentcenter.uspto.gov/
• EPO Register (EP3497430B1 / EP17751070.8): https://register.epo.org/
• CourtListener (CAFC search): https://www.courtlistener.com/?type=o&court=cafc

Honesty note: I could not locate any PTAB decision, institution order, settlement, or Federal Circuit docket for this patent, so I have not supplied judge panels, petition grounds, or dispositions — inventing those would be far worse than reporting the gap. Treat the "no PTAB activity" finding as high-confidence for indexed proceedings and strong-but-not-certain for the last few weeks of filings.

I'll research the assignment record and cross-references for US 10746654.

Let me cross-reference for litigation, entity status, and any assignment records.

Assignment & Ownership Analysis — US 10,746,654 B2

Patent: Method for monitoring the correspondence of a beer sample with a reference beer
App. No.: 16/324,292 · PCT: PCT/EP2017/069698 (filed 2017-08-03) · DE priority: 10 2016 009 636.5 (2016-08-10)
Granted: 2020-08-18 · Assignee of record: QFOOD GmbH (Freiburg im Breisgau, Germany)

Data caveat up front. The authoritative ownership record is the USPTO Assignment Center (https://assignmentcenter.uspto.gov/). Direct queries to that database returned no retrievable result set for this patent number in my session, so the timeline below rests on the USPTO-sourced assignment entry mirrored in the Google Patents legal-events feed (which reproduces the reel/frame, conveyance, parties, and signing/recording dates) plus corroborating IP databases. I could not retrieve the correspondent-of-record for the single assignment entry — that field is not exposed in the mirrored feed, and I will not guess it. Everything below is labeled accordingly; the one hard cited record is reel 051075/0234.

---

Inventors

Pattern note — departure risk: none observed. The German inventor trio (Klapproth, Seidel, Haas) continues to appear on QFOOD filings with later grant dates (US 10,444,212 granted 2019; US 10,989,631 granted 2021), indicating continued association with the assignee rather than post-filing departure. There is no "all inventors left within 12 months" signature here.

---

Original assignee

QFOOD GmbH — Freiburg im Breisgau, Germany. Private German GmbH (therefore no SEC filings exist — it is not an SEC reporting company, so the 10-K/8-K cross-reference step is inapplicable).

• Primary line of business: development and sale of food/beverage analysis systems, principally the QFOOD QUANTOS FTIR/IR absorption spectrometer, plus associated consumables and analytical services (EUTM filings QFOOD, Qdentity, QUANTOS, IRIS Approved/Scan; Nice classes 01, 09, 42).
• Did it ship a product embodying the claims? Yes, in the practical sense that matters for NPE screening. The specification itself names the "QFOOD QUANTOS®-type infrared absorption spectrometer" as the instrument used to capture the spectra, and the claimed method is a data-analysis workflow run on QUANTOS output. QFOOD markets QUANTOS as a commercial analyzer for on-site beverage quality control.
• Current status: Appears operating. IP activity runs through at least 2019 (trademark filings), US patents granted 2019–2021, and the 4th-year maintenance fee was paid 2024-02-05 (small-entity) — the fee payment confirms a live, maintained asset. No bankruptcy, insolvency, or liquidation evidence was found in my searches (searches for QFOOD insolvency returned nothing).

---

Assignment timeline

There is exactly one recorded assignment in the chain. There are no post-issuance assignments, which is itself the finding: the original assignee still owns the patent.

• 2019-10-25 to 2019-11-05 (executed, per signing dates) / recorded 2019-11-21 — Reel 051075/0234
  • Conveyance: Assignment (Assignment of Assignors' Interest)
  • Assignor: Holger Klapproth, Robert Seidel, Joachim Haas, and Jonathan E. Green (inventors)
  • Assignee: QFOOD GMBH (Germany)
  • Correspondent: Not retrievable from the mirrored record — flagged as unknown rather than guessed. (Lead, not a finding: the DE sibling DE102016009636 lists the agent of record simply as "Huwer," which is the German prosecution agent, not necessarily the US assignment correspondent.)
  • Context: Routine inventor-to-employer confirmatory assignment — the standard national-phase/employment transfer of rights to the operating company. Not a fire-sale, reorg, securitization, or transfer-to-asserter.

Subsequent events: none recorded. Google Patents legal events for this patent show only the above assignment plus fee/grant/status entries. No security agreement, merger, change of name, license, release, or correction has been recorded.

---

Timeline diagram

timeline
    title Ownership of US 10746654
    2016 : DE priority application filed
    2017 : PCT and US national phase filed
         : QFOOD named original assignee
    2019 : Inventors assign rights to QFOOD
         : Reel 051075/0234 recorded
    2020 : US patent granted
    2024 : 4th year maintenance fee paid

(The 2019 assignment entry is the only recorded conveyance. No later ownership events exist to plot.)

---

NPE / troll-pattern signals

1. Shell-entity transfer — Not present. The only recorded transfer (reel 051075/0234, recorded 2019-11-21) runs into the operating company QFOOD GmbH from its own inventors. There is no downstream LLC, no "IP/Licensing/Holdings/Ventures" successor, and no registered-agent address anywhere in the chain.

2. Known asserter in the chain — Not present. No assignee or assignor at any point matches the Acacia / Marathon / IV / Wi-LAN / Conversant / Vringo / Pendrell / Round Rock / Spangenberg lists or any RPX/Unified high-frequency-plaintiff directory entry. The sole assignee (QFOOD GmbH) is an operating food-analysis manufacturer.

3. Repeat correspondent across the chain — Unclear / not applicable. With only one link in the chain (051075/0234), there is no possibility of recurrence, and the correspondent field for that entry was not retrievable. This signal cannot be scored as present on the evidence available.

4. Cascading transfers — Not present. One transfer total, ~2 years after filing and ~3 years before grant; no chained LLC hops, no common-principal clustering.

5. Pre-litigation transfer — Not present. No infringement suit naming this patent was found in my searches, so there is no litigation for a transfer to precede. The 2019 assignment is temporally tied to national-phase prosecution, not to any filing date.

6. Bankruptcy fire-sale — Not present. No Chapter 7/11 or German insolvency proceeding for QFOOD was found; the assignee paid a maintenance fee in 2024, indicating a solvent, maintained portfolio.

7. Privateering — Not present. No operating company → NPE transfer exists; QFOOD retained the patent.

8. Defensive aggregator (anti-NPE) — Not present. The chain does not terminate at RPX, AST, LOT, Unified, or OIN; it terminates at the original operating assignee.

---

Verdict

Insufficient data — strictly because the record contains only the original assignment (reel 051075/0234, inventors → QFOOD GMBH, recorded 2019-11-21), and the correspondent-of-record for that entry could not be retrieved to complete the analysis.

Substantively, the evidence points the other way from an NPE: the chain is a single, routine inventor-to-employer confirmatory assignment into an operating beverage-analysis company that sells the QUANTOS instrument on which the claimed method is practiced, has kept the patent in force (4th-year fee paid 2024-02-05), and shows zero shell-entity, asserter, cascading-transfer, privateering, or defensive-aggregator signals. No post-issuance transfer to any third party has been recorded — the original assignee still owns the patent.

Verify at: https://assignmentcenter.uspto.gov/ (search patent 10,746,654; expect a single entry, reel 051075/0234). The one item worth physically pulling from the Assignment Center is the correspondent-of-record for that reel/frame, which would close out signal #3 definitively.

I'll research the specific citations for US 10,746,654 to build an accurate prior-art analysis. Let me pull details on the key references.

Let me gather the remaining references.

Prior-Art Analysis — US 10,746,654 B2

Patent under analysis: US 10,746,654 B2, "Method for monitoring the correspondence of a beer sample with a reference beer," QFOOD GmbH, app. 16/324,292, priority DE 10 2016 009 636.5 (2016‑08‑10), granted 2020‑08‑18.

Source of the citation list: The reference list below is exactly the "Patent Citations" set printed on the patent's face and reproduced on its Google Patents record (https://patents.google.com/patent/US10746654/en). Per the strict-instruction rule, I have kept every identifier literally as it appears — including "Gegory John Tanner" (sic), "Crame"/"Cramer", and the EPO/Foyolle spellings — and have not auto-corrected any of them.

Method note / caveat up front. US 10,746,654 has a single independent claim (claim 1) with a very specific mathematical core (≥15 brewed reference samples → IR spectra → PCA of ≥15 components → factor loadings → normalized reference values R(i,j) → a k‑scaled reference interval → characteristic value B → error flag). Anticipation under 35 U.S.C. § 102 requires every element of the claim in a single reference. On the records I could reach, no cited reference discloses that full combination. What the citations mostly establish is the general field (IR/chemometric analysis of beer and other liquids). I therefore rank them by relevance and state, for each, the claim(s) it could bear on and whether it can anticipate.

---

A. The two references that matter most

1. DE 101 08 712 A1 / EP 1 235 066 A1 — Warsteiner Brauerei Haus Cramer

• Full citation: DE 101 08 712 A1, "Verfahren zur analytischen Untersuchung einer Bierprobe" (Method for analytical analysis of a beer sample), Warsteiner Brauerei Haus Crame[r]; priority 2001‑02‑23; published 2002‑09‑12. European counterpart: EP 1 235 066 A1 / B1, "Process for analysing a beer sample," filed 2001‑04‑27, published 2002‑08‑28, granted 2009‑04‑01 (now ceased).
• Source: https://patents.google.com/patent/DE10108712A1/en ; http://data.epo.org/gpi/EP1235066A1-Process-for-analysing-a-beer-sample.html
• Brief description: Spectroscopic measurements are made on a series of samples of a known brand of beer, each sample giving multiple measurement values, to build a reference data matrix that serves as a model. The same spectroscopic measurements are then taken on an unknown ("foreign") sample and compared against the model to decide whether the unknown sample lies inside the model — i.e., whether it is genuinely beer of that brand (the stated problem is brand-substitution/fraud at the point of sale, e.g., in a tavern). The application expressly discusses that natural raw-material and process variation means there is no fixed "standard beer," which is why a statistical model rather than a single fingerprint is built.
• Claims it potentially bears on: Claim 1 (the reference-set + IR-spectroscopy + model-of-the-unknown comparison architecture), and by extension the whole chain of dependent claims.
• § 102 assessment: Likely NOT a full anticipation of claim 1. It shares the conceptual spine (series of same-brand beer samples → spectroscopic reference model → measure an unknown beer → test whether the unknown falls "within" the model). However, on the materials I could review it does not expressly disclose the claim's specific limitations: principal component analysis with ≥15 principal components, the factor loading P_R(i,j) construct, the normalized reference value R(i,j) = |(P_R(i,j) − μ_P(j))/σ_P(j)|, the specific k‑scaled reference interval formula with σ_R(j), or the characteristic value B. A genus of "statistical model / data matrix" is not the claimed species absent disclosure of PCA performed as claimed. Best treated as the closest prior art and a strong § 103 candidate, not a § 102 reference.

2. US 5,446,681 A (and its German counterpart DE 691 28 357 T2) — Exxon Research & Engineering

• Full citation: US 5,446,681 A, "Method of estimating property and/or composition data of a test sample," Gethner, Todd & Brown, Exxon Research and Engineering Co.; priority 1990‑10‑12 (US 596,435); granted 1995‑08‑29. German counterpart: DE 691 28 357 T2, "Verfahren zur Bestimmung von Eigenschaften und/oder Zusammensetzungsdaten einer Probe," published 1998‑07‑16. European counterpart EP 0 552 291 B1.
• Source: https://patents.google.com/patent/US5446681 ; http://data.epo.org/publication-server/rest/v1.2/patents/EP0552291NWA1/document.html ; https://pubchem.ncbi.nlm.nih.gov/patent/EP-0552291-B1
• Brief description: An on-line/computer-controlled spectrometer measures a sample spectrum; statistical analysis of the sample data based on a statistical model using sample calibration data; the system also automatically identifies a sample (statistical and rule-based/expert-system criteria) and triggers sampling when data indicate species not already in the model. IPC includes G01N21/35 (IR) and G01R23/16 (spectral/statistical analysis); it is a foundational chemometrics/spectral-modeling reference (petrochemical context).
• Claims it potentially bears on: Claim 1 — generically, the statistical-model-from-spectra and compare-unknown-against-model limitations.
• § 102 assessment: Not an anticipation. Discloses IR/spectroscopic measurement + statistical (chemometric) modeling + model-based identification of an unknown — but no beer, no brewing of ≥15 reference beer samples under identical ingredients/process parameters, no PCA with ≥15 components as claimed, and none of the R(i,j) / reference-interval / B formulas. It is cited as background for the modeling step; useful for § 103 (motivation to use multivariate statistics on spectra) only.

(Note: DE 691 28 357 T2 is the German family member of the same Exxon disclosure and adds nothing over US 5,446,681.)

---

B. Beer-process / in-line monitoring references (background; no anticipation)

3. US 2016/0369214 A1 — Michael Mosher (Univ. of Northern Colorado)

• Full citation: US 2016/0369214 A1, "In-Line Detection of Chemical Compounds in Beer," Michael Mosher; priority US 62/180,865 filed 2015‑06‑17; published 2016‑12‑22 (granted as US 10,570,357 B2; WO 2017/218039 A1).
• Source: https://patents.google.com/patent/[US10570357B2](/patent/US10570357B2)/en
• Brief description: In-line apparatus/method: a small wort sample is drawn from a fermentation vessel, heated, analyzed by infrared attenuated total reflectance (IR-ATR) spectroscopy, and returned; concentrations of compounds (diacetyl, CO₂, AAL, etc.) are obtained directly from IR-ATR data or via an algorithm based on reaction kinetics/stoichiometry; continuous, non-destructive monitoring during fermentation.
• Claims it potentially bears on: Claim 1 only generically (IR spectroscopy of beer/wort + algorithm-driven evaluation).
• § 102: Not an anticipation. It is an in-line analyte-concentration method (diacetyl/CO₂), not a correspondence-to-reference method; it discloses no reference-sample brewing set, no PCA, no reference interval, no B. Relevance is § 103 context for "IR spectroscopy of beer in a brewery."

4. US 2016/0194586 A1 — Alfa Laval Corporate AB

• Full citation: US 2016/0194586 A1, "Method and apparatus for beer fermentation," Alfa Laval Corporate AB (Nordkvist et al.); priority 2013‑09‑09; published 2016‑07‑07. (PCT counterpart WO 2015/032551.)
• Brief description: Beer-fermentation method using an on-line measuring device to obtain a first extract value representative of the vessel content, and automatically controlling a mixing device (withdraw/re-inject vessel contents) dependent on that value — i.e., process control of fermentation via an on-line measurement.
• Claims it potentially bears on: Claim 1 only as background (in-line measurement during brewing; the "same process parameters" context).
• § 102: Not an anticipation. No spectroscopy-to-PCA correspondence test, no reference interval, no B. Cited as brewing-process/on-line-monitoring background.

---

C. Optical/spectroscopic liquid-analysis references (claim 5 and measurement-step context)

5. US 2015/0060674 A1 — Haffmans B.V.

• Full citation: US 2015/0060674 A1, "Device for optically determining the concentration of alcohol and carbohydrates in a liquid sample," Levels & Martynowicz, Haffmans B.V.; priority NL 2011388 (2013‑09‑05); published 2015‑03‑05 (granted US 10,132,787 B2).
• Source: https://patents.google.com/patent/US20150060674
• Brief description: Device/method with two light sources (750–1000 nm), a spectrometer, and a processing unit; determines an absorption value by comparison to a reference value, and formulates a linear equation from a multitude of reference measurements (simple/multiple linear regression) to compute alcohol and carbohydrate concentrations. Beer is an expressly listed sample type.
• Claims it potentially bears on: Claim 1 superficially (spectroscopic absorption + reference measurements + regression), and the measurement/absorbance aspects underlying claim 5.
• § 102: Not an anticipation. Linear-regression calibration is not the claimed PCA/factor-loading/reference-interval method; no ≥15 brewed reference samples, no R(i,j), no reference interval, no B. § 103 context only.

6. WO 2012/167805 A1 — Qfood GmbH (same assignee as the patent)

• Full citation: WO 2012/167805 A1, "Method and apparatus for determining the concentration of an analyte contained in a liquid sample," Qfood GmbH; priority 2011‑06‑09; published 2012‑12‑13.
• Source: https://patents.google.com/patent/WO2012167805A1/en
• Brief description: IR measurement cell for determining analyte concentration in a liquid, with an inert coating on the IR-transparent chamber wall, homogeneous thin fluid layers (e.g., transmission cell with layer thickness ≤30 µm, and MIR/ATR variants), and transmission/reflection/ATR configurations. This is the hardware lineage underlying the patent's QFOOD QUANTOS® spectrometer and its thin-layer IR measurement.
• Claims it potentially bears on: The measurement-condition features (thin fluid layer, IR transmission) that appear in the specification and inform claim 5's IR range.
• § 102: Not an anticipation. Pure instrumentation/apparatus; no reference-beer set, no PCA, no reference interval, no B. Cited because it is the assignee's own earlier IR-cell work.

---

D. Beer-ingredient / beverage-composition references (peripheral)

These were cited, in my assessment, merely to show the general state of the beer/brewing art rather than any analytical method; none can anticipate any claim.

7. US 2007/0254063 A1 — Chemisch En Biochemisch Onderzoekscentrum (CBOK)

• "Use of hop polyphenols in beer." Priority 2006‑04‑07; published 2007‑11‑01. Beer-composition/hop-chemistry disclosure. Bears on no claim; § 102 none.

8. US 2010/0303994 A1 — Guido Aerts

• "Method for brewing beer." Priority 2007‑10‑19; published 2010‑12‑02. Brewing-process disclosure (background for the "same ingredients/same process parameters" limitation). § 102 none.

9. US 2011/0135784 A1 — "Gegory John Tanner" (sic)

• "Barley with low levels of hordeins." Priority 2007‑08‑13; published 2011‑06‑09. Plant/ingredient disclosure. § 102 none.

10. US 2011/0151068 A1 — Roy Kenneth Taylor

• "Improvements in or relating to gassed beverages." Priority 2008‑06‑05; published 2011‑06‑23. Beverage formulation/packaging disclosure. § 102 none.

---

E. Non-patent literature (one item)

11. Foyolle et al., Applied Spectroscopy (1996)

• Full citation: Foyolle et al., "Determination of Major Compounds of Alcoholic Fermentation by Middle-Infrared Spectroscopy: Study of Temperature Effects and Calibration Methods," Applied Spectroscopy, Society for Applied Spectroscopy, 1996, vol. 50, no. 10, pp. 1325–1330, Baltimore.
• Brief description: Uses mid-infrared spectroscopy with calibration methods (including temperature-effect handling) to determine major compounds of alcoholic fermentation — the analytical ancestor of the patent's IR quantification of beer constituents.
• Claims it potentially bears on: The IR-measurement-of-beer/fermentation-compounds aspect of claim 1 and the IR range of claim 5.
• § 102: Not an anticipation. Calibration/quantification of fermentation compounds is not the claimed PCA-based correspondence test; no reference interval, no B.

---

F. Summary ranking and bottom line

Bottom line. Applying the identifiers literally, no reference cited against US 10,746,654 discloses all elements of independent claim 1, so I find no express § 102 anticipation among the listed citations. The genuine § 102/§ 103 battleground is the pairing of (i) the Warsteiner beer-model reference — same field, same "is this unknown beer consistent with our reference?" purpose — with (ii) the Exxon multivariate-spectral-modeling reference and/or (iii) the Foyolle NPL, for the PCA/statistical-calibration step. The novelty of claim 1 appears to reside in the specific factorization/interval mathematics (the normalized reference value R(i,j) and the k‑scaled reference interval around the mean of R, plus the parallel characteristic value B), which the cited art does not show.

Confidence and limits. Records for references 1, 2, 3, 5, 6 and the NPL were reviewed directly (URLs given above). The brief descriptions of the peripheral references (7–10) are based on their titles/abstract-level data as printed in the patent's citation list; I did not pull full texts for all of them, so treat those claim-mapping statements as low-stakes but not exhaustively verified. Where I state a reference does not disclose a feature, that reflects the material I could reach — it is an assessment based on the available disclosure, not a claim of having searched each document end-to-end.

I'll analyze obviousness under § 103 using the prior art cited on this patent's page. Let me ground a few key references first.

Let me ground the remaining key prior-art references.

Obviousness Analysis — U.S. Patent No. 10,746,654 B2 under 35 U.S.C. § 103

Analysis date: April 26, 2026 · Priority date: August 10, 2016 (DE 10 2016 009 636.5) · Field: beer quality monitoring by IR spectroscopy + principal component analysis (PCA)

This builds on the claim-1 element map and the worked example already generated. I do not repeat the full claim text; I map elements to references and then explain motivations to combine. Identifiers are taken literally from the patent page's Citations / Non-Patent Citations sections.

---

1. Legal framework applied

Under Graham v. John Deere and KSR Int'l v. Teleflex, obviousness is assessed from the perspective of a person having ordinary skill in the art (PHOSITA) at the priority date, considering (a) the scope/content of the prior art, (b) differences between prior art and claims, (c) the level of ordinary skill, and (d) secondary considerations. A claim is obvious where the differences are such that the subject matter "as a whole" would have been obvious, and where a PHOSITA would have had a reason to combine references with a reasonable expectation of success — including combining "prior art elements according to known methods to yield predictable results."

Assumed PHOSITA: a food/brewing analytical chemist or process engineer with (i) practical experience with mid-infrared (MIR) spectroscopy of beverages and (ii) working familiarity with chemometrics, including PCA and multivariate statistical process control (MSPC). This is the skill level reflected in the cited art (an Exxon chemometrics team; a brewery's analytical laboratory; academic fermentation-spectroscopy researchers).

Applicant's own admissions on the record (useful to the obviousness case, since a patent's own specification can supply the motivation and the state of the art):

• Factor loadings/scores "can be ascertained by means of a suitable software known per se."
• The constant k "can have any value not equal to zero, in particular the value 1 or the reciprocal value 1/m."
• The method is for use "on a plurality of beer samples produced in a brewery."

These concessions frame the invention as an application of known chemometric tools, which weakens any argument that the mathematical steps alone are inventive.

---

2. Prior art landscape mapped to claim 1 (from the page's Citation list)

The single most on-point reference is DE 101 08 712 A1 (Warsteiner Brauerei Haus Cramer KG; EP 1 235 066 A1/B1). Its abstract recites: spectroscopic measurements of "a series of beer samples of a known given make of beer with a number of measurement values for each sample used to create a reference data matrix. The reference matrix is used to compare measurements of unknown samples (18) made using a spectrometer (11). If the measurements lie within the limits of the determined model then it can be recognized as being of the given make of beer." The EP specification expressly reduces the multivariate dataset "nach der Methode der sogenannten Hauptkomponentenanalyse (PCA)" into a model used for the comparison. (https://patents.google.com/patent/DE10108712A1/en ; http://data.epo.org/gpi/EP1235066A1-Process-for-analysing-a-beer-sample.html)

The closest statistical reference is US 5,446,681 A (Exxon, priority 1990-10-12). It discloses on-line spectral measurement, multivariate calibration by PCA/PCR, and — critically — "comparing calibration sample spectra as to whether or not the measured spectrum is within the range of the calibration sample spectra," followed by isolating a test sample if it "is not within the correlation between said calibration sample spectra and the property and/or composition data." (https://patents.google.com/patent/US5446681/en ; EP 0 552 291 B1 PDF: https://patentimages.storage.googleapis.com/49/b6/16/17ecc2b80ef2d1/EP0552291B1.pdf) DE 691 28 357 T2 is the German-language member of the same Exxon family and carries the same disclosure.

---

3. Combination 1 (primary § 103 rejection): Warsteiner '712 + Exxon '681 + Foyolle [sic] et al.

Where each claim-1 element comes from:

• Warsteiner '712 supplies the architecture: a reference model built from spectroscopic measurements of a series of same-brand beer samples (element a), the beer/spectrometer context (b), PCA to reduce the multivariate data into that model (c), spectroscopic measurement of a test sample (f), and the decision step of testing whether the test sample "lies within the model" (h).
• Exxon '681 supplies the statistical detection mechanism the patent actually claims: a PCA-based model with a quantitative check of whether an unknown spectrum is "within the range of the calibration sample spectra," with automatic flagging of samples that fall outside the model. This converts Warsteiner's qualitative "within the model" determination into the numeric reference interval + characteristic value B of elements (e) and (g).
• Foyolle [sic] et al. supplies the spectroscopic feasibility in fermentation/beer matrices: "The potential of Fourier transform middle-infrared spectroscopy has been demonstrated for the quantitative analysis of substrates (glucose and fructose) and metabolites (glycerol and ethanol) involved in alcoholic fermentation," using PLS regression with calibration/validation/prediction sets. (Applied Spectroscopy 50(10):1325–1330; https://journals.sagepub.com/doi/10.1366/0003702963904872)

Motivation to combine (articulated):

1. Same field, same problem, same solution type. Warsteiner and Exxon both address objectively verifying whether an unknown liquid sample conforms to a reference population. Warsteiner's own specification states that brewery-lab pattern recognition "erfolgt subjektiv" (is subjective) and that a spectral model is needed to make the determination objective and reproducible on-site. Exxon addresses the identical need — objectively flagging a sample that does not fit a calibration model.
2. Reasonable expectation of success. Warsteiner expressly directs the reader to PCA as the reduction method, and Exxon expressly directs the reader to PCA/PCR as the predictive model underpinning the "within the range?" check. A PHOSITA combining them would expect the Exxon check to work directly on the Warsteiner model because both operate on the same object (a matrix of spectra reduced to principal components/scores).
3. Predictable implementation. Elements (d) and (g) are the textbook normalization of PCA scores: subtract the mean and divide by the standard deviation of the loadings for each component. The specification itself calls the software for computing loadings "known per se," and this Z-score/autoscaling step is a routine, well-understood chemometric operation. The reference interval of element (e) is a symmetric control band built from the mean of the normalized reference values plus/minus k standard deviations summed over components — ordinary MSPC control-limit construction, with k an arbitrary non-zero scaling constant (the patent concedes any non-zero value, including 1, works).

Net: Every element of claim 1 is disclosed or would have been obvious as a predictable use of a known technique (PCA-based model checking) in a known beer-spectroscopy context, with no teaching away and no unexpected result. The result (a scalar "good/bad" indication) is exactly the predictable output of the combined teachings.

---

4. Combination 2 (alternative): Exxon '681 + Foyolle [sic] et al. + Warsteiner '712

Reordering with Exxon as primary is viable because Exxon is the broadest disclosure of the claimed statistical method (PCA model + "within the range?" check + flagging outliers). Foyolle [sic] et al. motivates applying MIR multivariate analysis specifically to alcoholic-fermentation/beer liquids, and Warsteiner motivates confining the population to a single beer variety/brand and using the model for conformity rather than quantification. This is the classic "adapt a known process to a known, analogous material" rationale (KSR): a PHOSITA would have applied Exxon's PCA outlier check to beer spectra with a reasonable expectation of success because Foyolle [sic] et al. had already shown MIR spectra of fermentation media to be well-behaved under multivariate regression.

---

5. Combination 3: Qfood's own WO 2012/167805 A1 + Warsteiner '712 + Exxon '681

WO 2012/167805 A1 (Qfood GmbH — the same applicant) discloses "Method and apparatus for determining the concentration of an analyte contained in a liquid sample" using infrared absorption/transmission measurement, a cuvette/measuring chamber, an inert coating on the chamber wall, and — pertinent to the specification's layer-thickness discussion — a transmission measuring chamber "with a layer thickness of ≤30 µm" and MIR measurement. (https://patents.google.com/patent/WO2012167805A1/en)

Motivation: a PHOSITA seeking to implement the Warsteiner/Exxon conformity test on a commercial IR instrument would naturally adopt the applicant's own thin-layer MIR cell with its inert coating (which addresses the fouling/matrix problems noted in WO '805) because it is the same applicant's optimized measurement front end for exactly this class of measurement. Note that the ≥15-sample count, the PCA computation, and the interval/B statistics are not supplied by WO '805; they come from the Warsteiner/Exxon combination and common chemometrics.

Caveat: WO '805 is the applicant's own earlier work. It is available as prior art for § 103 purposes (it published 2012-12-13, before the 2016 priority date), but because it shares an owner, a PHOSITA-motivation narrative that leans on WO '805 is strongest for the instrumentation side (layer thickness, MIR cell) rather than the claimed algorithm.

---

6. Combination 4: Mosher '214 + Warsteiner '712 + Exxon '681 (+ Alfa Laval '586)

US 2016/0369214 A1 (Mosher; granted as US 10,570,357 B2) teaches in-line detection of chemical compounds in beer: a wort sample is removed from a fermentation vessel, analyzed by IR-ATR spectroscopy, and the concentrations are ascertained "directly from IR-ATR spectroscopy data, or indirectly from applying an algorithm to the IR-ATR spectroscopy data." (https://patents.google.com/patent/[US10570357B2](/patent/US10570357B2)/en) US 2016/0194586 A1 (Alfa Laval / Nordkvist) similarly teaches an on-line measuring device for extract value with automatic control of the fermentation vessel.

Motivation: Mosher and Alfa Laval supply the "indicate an error during the production of the beer sample" step's practical context — continuous, in-line quality monitoring during brewing with an algorithm applied to IR data. Warsteiner supplies the beer-model/PCA step; Exxon supplies the "within the model?" statistical flag. A brewery-facing PHOSITA would have been motivated to combine these because monitoring beer as it is produced (rather than authenticating it at the tap, Warsteiner's stated objective) was an express goal in the art — Mosher's background section frames the problem as "controlling the fermentation process" and obtaining "consistency and quality of the beer produced."

---

7. Why the specific mathematical limitations do not save claim 1

The examiner's likely position (and the one supported by the record) is that elements (d)–(g) are routine data-processing steps with predictable results:

1. Z-score normalization of PCA scores (element d) is a conventional chemometric preprocessing step; the patent itself calls score computation "known per se."
2. The aggregation B = k·Σ|normalized deviation| (element g) is a simple sum of absolute standardized residuals — the same family as conventional multivariate residual/outlier indices. Selecting the absolute-value form and a scalar constant k is an arbitrary mathematical choice; the patent concedes k may be any non-zero number.
3. The reference interval (element e) is a mean ± k·(sum of component standard deviations) band — standard statistical process-control limit construction, applied to the normalized reference values.
4. "Brewed with the same ingredients and the same process parameters" and the "at least 15" thresholds (elements a, c) are data-collection/quantity choices that a PHOSITA would select as a matter of routine experimental design to obtain a statistically meaningful model; the specification gives no criticality or unexpected result tied to the number 15 itself.

Under KSR, "a combination of familiar elements according to known methods is likely to be obvious when it does no more than yield predictable results." Nothing in the patent points to a result that is more than the expected sum of its known parts. The only asserted advantage — capturing "spectral components caused by beer ingredients present only in low concentrations … to a greater extent" — is a naturally flowing consequence of normalizing by the loading standard deviation, not an unpredictable discovery, and would have been appreciated by any PHOSITA performing autoscaling.

---

8. The dependent claims

• Claim 2 (n ≥ m; in particular 2×, preferably 3×): Routine design choice for statistical robustness of a PCA model (more samples than variables/components avoids overfitting). Obvious over any of Combinations 1–4.
• Claim 3 (m ≥ 20, ≥ 30, ≥ 40, ≥ 50): Optimization of a known parameter (number of retained principal components). The specification's own example uses 30; choosing a component count in the 20–50 range to capture minor spectral features is an expected, iterative optimization. Obvious.
• Claim 4 (k = 1/m): Purely an arithmetic averaging choice (mean of the normalized deviations instead of a sum). The specification itself equates k = 1/m with taking an arithmetic mean. Obvious as a mathematical convenience.
• Claim 5 (wave number range 950–3050; in particular 960–2000; preferably 980–1200 cm⁻¹): Foyolle [sic] et al. and WO 2012/167805 A1 both work in the MIR region where the characteristic absorption of sugars/fermentation compounds occurs; the 980–1200 cm⁻¹ "fingerprint" region is where carbohydrate C–O/C–O–C bands appear (well-established in the literature). Selecting a sub-range of a known informative MIR window is a predictable optimization. Obvious in view of Combination 1 or 3.

---

9. Counterarguments a patentee would raise (and their weaknesses)

1. "Warsteiner is brand-authentication, not production monitoring." True but not dispositive: the claim's final step ("indicate an error during the production") merely relabels the same comparison outcome. Mosher/Alfa Laval supply the in-production context (Combination 4), and the technical operation — compare a spectrum to a reference model and flag a mismatch — is identical.
2. "No single reference discloses the exact interval and B formulas." Correct, but § 103 permits a combination of references, and these limitations are routine statistics. The applicant's own characterization of the score software as "known per se" and of k as arbitrary cuts against non-obviousness.
3. "Unexpected sensitivity to trace components." This is the only quasi-secondary-consideration argument available, but the specification offers no comparative data showing an unexpected result versus ordinary autoscaling; the effect is inherent in the known normalization. It therefore likely fails as evidence of non-obviousness.
4. Data-gathering steps (brewing/measuring) may carry limited patentable weight. The reference-generation steps (a)–(b) may be treated as obtaining information; the patentee's stronger (though still weak) ground is that the interval/B formulation is not literally shown in one reference.

---

10. Bottom line

Substantial and well-supported § 103 rejections are available against all five claims:

• Strongest combination: DE 101 08 712 A1 (Warsteiner) + US 5,446,681 (Exxon) + Foyolle [sic] et al. (1996) (plus the ordinary knowledge of a chemometrician for the Z-score normalization, control-interval, and summation steps). Warsteiner supplies the beer + PCA + "within the model" architecture; Exxon supplies the PCA-model "within the range?" flagging mechanism; Foyolle [sic] et al. supplies MIR feasibility for fermentation/beer matrices. Motivation: identical problem (objective, reproducible conformity verification) and use of the same PCA-based tool, with a reasonable expectation of success.
• Corroborating combinations: WO 2012/167805 A1 (Qfood) for the thin-layer MIR cell (spec's ≤30 µm; supports claim 5), and US 2016/0369214 A1 (Mosher) / US 2016/0194586 A1 (Alfa Laval) for the in-line, during-production monitoring context.
• Dependent claims 2–5 add only routine ranges, ratios, and an arithmetic averaging choice, each independently obvious or obvious in view of the specification's own example.

Caveat on confidence: I could retrieve the full text and abstract of Warsteiner '712/EP 1 235 066, Exxon '681/EP 0 552 291, WO '805, and Mosher '214, but I did not retrieve the complete claim sets and figures of DE 101 08 712 A1 or US 5,446,681. The mappings above should be confirmed against those full texts (particularly whether either reference expressly describes normalizing scores by their standard deviation and summing absolute deviations). Also, as noted in the earlier sections, the granted text contains literal symbol inconsistencies (μ_R/σ_R vs. μ_P/σ_P/σ_R) that could affect claim construction and thus the precise metes-and-bounds of the § 103 comparison; I have not normalized them.

Identifier note (not auto-corrected): the patent cites the NPL reference as "Foyolle et al." The corresponding published article is by Fayolle, Picque, Perret, Latrille & Corrieu, Applied Spectroscopy 50(10):1325–1330 (1996). Whether these are the same document or the citation contains a spelling error in the original is not resolved here; I have flagged rather than corrected it.