Semi-Quantitative Detection of Borax Adulteration in Wheat Flour Based on Microwave Non-Destructive Testing and...

Area of Science:

• Food Science
• Analytical Chemistry
• Machine Learning

Background:

• Borax adulteration in wheat flour presents a significant food safety hazard.
• Existing rapid, non-destructive screening methods for borax are insufficient.

Purpose of the Study:

• To develop a machine learning-based microwave non-destructive detection method for semi-quantitative identification of borax in wheat flour.
• To create a rapid and robust screening strategy for food safety.

Main Methods:

• Utilized a proprietary microwave detection system (2.5-11.5 GHz) to acquire amplitude attenuation and phase shift spectra from 155 wheat flour samples.
• Employed a hybrid Random Forest-Whale Optimization Algorithm (RF-WOA) for feature selection and hyperparameter optimization.
• Applied hierarchical repeated validation and macro-level metrics for evaluation.

Main Results:

• Achieved an overall classification accuracy of 94.6% and a macro F1 score of 0.95.
• Reduced the feature space from 1800 to approximately 200 dimensions.
• Demonstrated 100% recall for undiluted samples, with minimal misclassification between adjacent adulteration levels and no false negatives for adulterated samples.

Conclusions:

• Microwave sensing combined with RF-WOA offers a rapid, non-destructive, and accurate method for detecting borax adulteration in wheat flour.
• This approach shows significant potential for food safety monitoring and regulatory inspection.
• The developed method provides a robust preliminary screening and grading evaluation strategy.