Patent Summary: Method for Judgment of Cheddar Grade Based on Identification Characteristic Flavor Composition - US Patent 11,293,906 B2

Summary

Here's a detailed summary of US Patent 11,293,906 B2, titled "Method for Judgment of Cheddar Grade Based on Identification Characteristic Flavor Composition."

What the Patent Is About

This patent presents a scientific, instrument-based method for determining the quality grade of cheddar cheese by analyzing its volatile flavor compounds. Instead of relying on subjective sensory evaluation (human taste and smell), the method uses:

• Gas chromatography–mass spectrometry (GC-MS)
• Aroma extract dilution analysis (AEDA)
• Gas chromatography–olfactometry (GC-O)
• K-means clustering
• Support vector machine (SVM) classification

The goal is to identify a small set of "identification characteristic flavor compounds" whose concentrations can be used to objectively and accurately classify cheddar cheese into mild/young, medium, or mature/strong grades.

The Idea Behind the Patent

Traditional cheddar grading relies on sensory evaluation, which is:

• Subjective
• Inconsistent
• Time-consuming
• Lacking in supporting data

This patent proposes a data-driven alternative:

1. Identify the most important flavor compounds in cheddar cheese.
2. Reduce the number of compounds to a manageable set using clustering.
3. Train a machine learning model (SVM) to recognize grade based on compound concentrations.
4. Predict the grade of unknown cheese samples by measuring the same compounds.

The innovation lies in the systematic reduction of flavor compounds to a minimal yet highly predictive set, enabling objective, repeatable, and scalable quality assessment.

Key Points of the Patent

1. Step-by-Step Methodology

• Step 1: Select standard cheddar samples of known grades (mild, medium, mature) and extract volatile compounds using solvent-assisted flavor evaporation (SAFE).
• Step 2: Use AEDA + GC-O to identify compounds with high flavor intensity. This yields 28 characteristic compounds (acids, esters, ketones, lactones, etc.).
• Step 3: Apply k-means clustering to reduce the 28 compounds to a smaller set of identification characteristic flavor compounds (e.g., 6 compounds yield best results).
• Step 4: Measure concentrations of these identification compounds in known-grade cheeses to create a training set for an SVM classifier.
• Step 5: Measure concentrations in unknown cheese, input into trained SVM, and output grade.

2. Key Compounds Identified

The 28 initial compounds include:

• Acids: acetic, butyric, hexanoic, 3-methyl butyric, etc.
• Neutral/alkali: 2,3-butanedione, ethyl caproate, δ-nonalactone, benzaldehyde, etc.

After clustering, the optimal set of 6 compounds (when k=6) is:

• δ-Nonalactone
• 2-Methyl propionic acid
• Ethyl caproate
• Butyric acid
• 2,3-Butanedione
• Pentanoic acid

This set yields >90% accuracy and F1-score in SVM classification.

3. Use of Machine Learning

• K-means clustering reduces dimensionality by grouping similar compounds and selecting those closest to cluster centers.
• Support vector machine (SVM) is trained on concentration data of the selected compounds from known-grade samples.
• Cross-validation (CV=5) ensures model reliability.

4. Optional Refinement

Before clustering, the patent optionally includes:

• Simulation of cheddar cheese system
• Deletion/recombination experiments to eliminate compounds with low aroma contribution (e.g., acetic acid, nonanoic acid).

Claims Summary

The patent includes 8 claims, which define the scope of protection:

Conclusion

This patent replaces subjective cheese grading with a reproducible, instrument-based method that:

• Identifies key flavor compounds via systematic analysis
• Reduces them to a minimal predictive set using clustering
• Uses machine learning to classify cheese grade with high accuracy

It provides a foundation for quality control in the dairy industry, especially for cheddar cheese, and could be extended to other fermented or aged foods.

Here are the basic details for US Patent 11,293,906 B2, based on the document provided:

Patent Number

• US 11,293,906 B2

Title

• Method for Judgment of Cheddar Grade Based on Identification Characteristic Flavor Composition

Date of Patent

• April 5, 2022

Application Number

• 16/267,365

Date Filed

• February 4, 2019

Priority Date

• September 6, 2018 (Chinese Application No. 201811040167.5)

Publication Date (Pre-Grant)

• March 12, 2020 (US 2020/0080976 A1)

Inventors

1. Bei Wang (Beijing, CN)
2. Jing Wang (Beijing, CN)
3. Yanping Cao (Beijing, CN)
4. Li Tan (Beijing, CN)
5. Baoguo Sun (Beijing, CN)
6. Zhennai Yang (Beijing, CN)

Applicant

• Beijing Technology and Business University (Beijing, CN)

Assignee

• Beijing Technology and Business University (Beijing, CN)

Examiner & Attorney

• Primary Examiner: Daniel S. Larkin
• Attorney/Agent: Jiwen Chen; Jacobson Holman PLLC

International Classification (IPC)

• G01N 30/06, G01N 30/72, G01N 30/30, A23C 19/068, G01N 30/86, G01N 30/02, G01N 30/88

Cooperative Patent Classification (CPC)

• G01N 30/7206, G01N 30/8651, G01N 2030/025, G01N 2030/067, G01N 2030/3046, G01N 2030/884

Field of Classification Search

• None (application file used for search history)

Foreign Priority

• China (CN 201811040167.5), filed September 6, 2018

References Cited

Foreign Patents:

• CN 101706490 A (May 2010)
• CN 107788227 A (March 2018)

Other Publications (Non-Patent Literature):

• Chen, Z. et al., "K-Means Clustering with Improved Initial Center" (2009)
• Gan, H.H. et al., "Development and Validation of an APCI-MS GC-MS Approach..." (Food Chemistry, 2016)
• Drake, M.A. et al., "Impact of Fat Reduction on Flavor and Flavor Chemistry of Cheddar Cheeses" (Journal of Dairy Science, 2010)

Number of Claims

• 8 Claims

Number of Drawing Sheets

• 2 Sheets (with Fig. 1 and Fig. 2)