Moisture and protein content in corn starch with NIR spectroscopy

Importance of the topic

Accurate determination of moisture and protein in corn starch is essential for food and confectionery production, where starch influences texture, gelling, drying behavior and shelf life. Rapid, reagent-free methods enable real-time quality control, reduce turnaround time compared with classical wet-chemistry methods, and support process decisions such as drying endpoint and blend acceptance.

Objectives and study overview

This application note evaluates near-infrared spectroscopy (NIRS) for quantifying moisture and protein in corn starch. The study aimed to develop and externally validate predictive models using a commercial OMNIS NIR Analyzer Solid, comparing NIR predictions against reference methods (Karl Fischer titration for water and Kjeldahl for protein).

Methodology

• Samples: 210 corn starch samples were analyzed to build calibration and validation datasets.
• Spectroscopy: Spectra were acquired in diffuse reflection mode over 1000–2250 nm using a large sample holder and 100 mm cup; measurements were rapid (typical acquisition in seconds).
• Reference analyses: Moisture content by Karl Fischer titration; protein content by Kjeldahl digestion.
• Model development and validation: OMNIS Software (Quant Development license) was used for data acquisition, preprocessing and multivariate model development. An external validation procedure was applied to assess predictive performance.

Instrumentation used

• OMNIS NIR Analyzer Solid (near-infrared spectrometer for solids/viscous samples)
• Large holder OMNIS NIR, 100 mm
• Large cup OMNIS NIR, 100 mm (sample vessel for powders/granulates)
• OMNIS Stand-Alone software license and Quant Development software license for building quantification models
• Reference analysis equipment: Karl Fischer titrator and Kjeldahl apparatus (for moisture and protein reference values respectively)

Main results and discussion

NIR models showed strong predictive ability for both parameters. Key figures of merit from external validation:

• Moisture: R2 = 0.977; SEC = 0.21% (absolute), SECV = 0.23%, SEP = 0.28% — indicating excellent agreement with Karl Fischer values and low prediction error suitable for routine QC.
• Protein: R2 = 0.915; SEC = 0.032% (absolute), SECV = 0.033%, SEP = 0.038% — good correlation with Kjeldahl results though with somewhat larger relative uncertainty than moisture, reflecting the smaller absolute concentration range of protein in starch.

Spectral data (selection shown in the study) were sufficient to capture relevant absorptions related to water and organic matrix features. Automated multi-position measurements (rotation of the sample vessel) improved reproducibility for non-homogeneous powder samples. The models benefit from a moderate sample set (210 samples) and external validation, supporting their robustness for similar production and QC contexts.

Benefits and practical applications

• Rapid, chemical-free analysis: results in seconds without reagents or sample destruction.
• At-line or laboratory use: suitable for production-floor QC or central labs, enabling faster decision-making (e.g., drying control, batch release).
• Automation-friendly: instrument and holder design support automated multi-position sampling and integration into process control systems.
• Cost and time savings: reduction in analysis time and elimination of routine wet-chemistry consumables and waste.

Future trends and potential uses

• Model transfer and scalability: expanding calibration sets across multiple production sites and seasons to improve model robustness and transferability.
• Advanced algorithms: use of modern chemometrics and machine learning (e.g., PLS improvements, ensemble models, neural networks) to enhance sensitivity for low-level constituents like protein.
• Inline and continuous monitoring: integration of NIR modules into production lines for real-time process control and feedback loops.
• Miniaturization and cloud deployment: compact spectrometers and cloud-hosted models for distributed QC and centralized model maintenance.
• Combination with complementary techniques: hybrid workflows linking NIR with other sensors or offline confirmatory analytics for comprehensive quality assurance.

Conclusions

The study demonstrates that an OMNIS NIR Analyzer Solid can reliably predict moisture and protein in corn starch with high accuracy and low prediction error. Moisture prediction is particularly robust (R2 ≈ 0.98, SEP ≈ 0.28%), while protein prediction is also acceptable (R2 ≈ 0.92, SEP ≈ 0.038%) given the low absolute protein levels. NIRS offers a fast, reagent-free alternative to Karl Fischer and Kjeldahl methods for routine quality control in starch processing and confectionery applications.

References

• Metrohm Application Note AN-NIR-151: Moisture and protein content in corn starch with NIR spectroscopy. Instrumentation and results as reported in the study.
• Reference methods cited in the study: Karl Fischer titration for water determination; Kjeldahl method for protein determination (standard wet-chemistry techniques used as calibration references).