Speaker
Description
Biscuit dough mixing is a critical stage in the breadmaking process that affects biscuit quality. The mixing step must ensure a uniform distribution of ingredients (water, sucrose, fat, and flour). Developing a monitoring sensor for the biscuit dough mixing process is crucial for effective quality control. Most online systems are based on the indirect survey of biscuit dough rheological properties changes using torque and consistency probes. Non-invasive near-infrared spectroscopy (NIRS) offers a possibility for fast measurement times that simultaneously deliver critical sample properties like water, sucrose, fat, and flour. Developing the NIR sensor with a fiber optic probe positioned inside the mixer is important for further automation of the industrial mixing process.
In the present study, we investigate the ability of the diffuse reflectance NIR spectroscopy to monitor the real biscuit doughs data in the wavelength range (1100–2500 nm). The presence of a vertical outlier and some leverage points in that data set may cause an unreliable calibration model with misquantified components. As a solution, we propose using Partial Robust M-regression (PRM) (Serneels et al., 2005) to determine fat, flour, sucrose, and moisture content. PRM has been combined with different pre-processing techniques such as first derivative Savitzky Golay algorithm (D1), standard normal variate (SNV), multiplicative signal correction (MSC), and the combinations of MSC and SNV with first derivative (D1+SNV, D1+MSC). The obtained results are comparable with pre-processing in the literature using the first differences according to Marx and Eilers. The optimal pre-processing for every component (water, sucrose, fat, and flour) was found. PRM regression improves the performance compared with the results of classical PLS1 regression.
Together with the facility of NIR technology to be implemented in the process engineering, these improvements make it ideal for the quality control of the bread dough mixing process and other food manufacturing processes.
Keywords: NIR, outliers, leverage points, PRM
Acknowledgements: First Author gratefully acknowledges receiving funding from programme COST ACTION, Horizon Europe, European Commission, CA19145 “European Network for assuring food integrity using non-destructive spectral sensors”.
REFERENCES
Serneels, S., Croux, C., Filzmoser, P., Van Espen, P. J. 2005. Partial robust M-regression. Chemometrics and Intelligent Laboratory Systems, 79(1-2), 55-64.
