Semola and semolina quality control using NIR spectroscopy

Significance of the topic

Good pasta quality depends critically on the physico-chemical properties of semola and semolina produced from durum wheat. Rapid, reliable assessment of protein, moisture, ash, gluten and water absorption in these flours is essential for raw material selection, process control and final product consistency. Near-infrared spectroscopy (NIRS) provides a nondestructive, high-throughput alternative to time-consuming reference analyses and enables inline or at-line quality control in milling and pasta production chains.

Objectives and study overview

This application note evaluated the capability of an OMNIS NIR Analyzer Solid to predict key quality attributes of semola and semolina. The goals were to demonstrate representative spectral acquisition of powdered samples, develop calibration models using OMNIS Software, and compare NIR predictions against standard laboratory reference methods to assess accuracy and precision for routine quality control.

Methodology

• Samples: Commercial semola and semolina flour samples were measured as received.
• Spectral acquisition: Measurements were performed in diffuse reflection mode using a large (100 mm) sample cup. Samples were rotated during acquisition to capture multiple sample positions and improve representativeness for heterogeneous powders.
• Calibration and validation: OMNIS Software was used for spectral preprocessing, model building and cross-validation. Figures of merit reported include coefficient of determination (R2), standard error of calibration (SEC) and standard error of cross-validation (SECV).
• Reference methods: Conventional laboratory methods used for reference data included Kjeldahl for protein, Karl Fischer for moisture, gravimetric ash determination, ISO methods for gluten, and farinograph-based water absorption (see References).

Used instrumentation

• OMNIS NIR Analyzer Solid — near-infrared spectrometer configured for solids and viscous samples, operated in reflection mode.
• Large holder and 100 mm large cup accessories for stable positioning and automated multi-position rotation of sample vessels (glass bottom constructed from impact-resistant Gorilla Glass).
• OMNIS Software with Quant Development license for model creation, and stand-alone OMNIS license for instrument control and data acquisition.

Key results and discussion

The OMNIS NIR Analyzer produced robust calibrations for several parameters, with best performance observed for moisture and protein and variable results for ash, gluten and water absorption.

• Semolina moisture: Excellent predictive performance with R2 = 0.949, SEC = 0.29% and SECV = 0.31% — indicating highly reliable moisture control via NIR.
• Semolina protein: Strong correlation with reference values (R2 = 0.907, SEC = 0.60%, SECV = 0.63%), suitable for routine protein screening.
• Semolina ash: Moderate performance (R2 = 0.734, SEC = 0.14%), acceptable for coarse ash screening but with less robustness than moisture and protein.
• Semolina gluten: Weaker model performance (R2 = 0.550, SEC = 1.71%, SECV = 1.87%), suggesting NIR-based gluten estimates are less precise and may require expanded datasets or alternate preprocessing to improve accuracy.
• Semola protein: Very good results (R2 = 0.933, SEC = 0.14%), confirming NIR suitability for protein assessment in coarser semola samples.
• Semola summary (table of merit): For semola samples (number of spectra reported per parameter), moisture (range 12.5–15.1%, N=146) R2 = 0.834 (SEC = 0.20%), ash (0.74–1.12%, N=99) R2 = 0.787 (SEC = 0.04%), gluten (23–32%, N=43) R2 = 0.815 (SEC = 0.67%), and water absorption (53.7–58.8%, N=37) R2 = 0.82 (SEC = 0.37%). These results show generally good calibration quality across parameters for semola when sufficient calibration samples are available.

Overall, moisture and protein predictions are highly reliable and support rapid screening and lot acceptance. Ash and water absorption can be estimated with moderate confidence. Gluten prediction in semolina showed limited accuracy in this dataset and should be applied cautiously or supplemented with laboratory testing during model improvement.

Benefits and practical applications

• Speed: Measurements take less than 10 seconds per sample, enabling high throughput at-line or in the QC lab.
• Non-destructive and reagent-free: NIR avoids chemical consumables and simplifies sample handling.
• Representativeness: Multi-position automated rotation reduces variability from powder heterogeneity.
• Integration: The OMNIS platform supports automation and connection to other analytical techniques or manufacturing systems for process control.
• Cost-effectiveness: After model development, routine NIR screening reduces reliance on slower, consumable-intensive ISO reference methods for many routine decisions.

Future trends and possibilities

• Model enhancement: Increasing sample diversity and using advanced preprocessing or multivariate algorithms can improve weaker predictions such as gluten content.
• Inline monitoring: Transitioning from at-line to inline NIR monitoring could enable real-time control of milling and blending operations to maintain consistent flour quality.
• Hybrid analytics: Combining NIR with other sensors (e.g., moisture probes, near-line rheology) and multivariate process models will improve prediction robustness for functional parameters like water absorption and dough behavior.
• Digital QA workflows: Integration of calibrated NIR instruments into plant-wide data systems will support batch tracing, supplier evaluation and automated release criteria based on spectral fingerprints.

Conclusion

This study demonstrates that the OMNIS NIR Analyzer Solid can rapidly and non-destructively quantify critical quality attributes of semola and semolina flours with high accuracy for moisture and protein, and acceptable performance for ash and water absorption. Gluten prediction from the current dataset was less robust and should be treated cautiously until models are expanded. NIRS offers a practical route to accelerate QC workflows, reduce laboratory burden and enable more responsive process control in pasta production chains.

References

• Protein reference: ISO 1871 — Kjeldahl method.
• Moisture reference: ISO 760 — Karl Fischer method.
• Ash reference: ISO 2171 — Determination of ash content.
• Gluten reference: ISO 21415 — Methods for determining gluten.
• Water absorption reference: ISO 5530-1 — Determination of water absorption using a farinograph.