A statistical strategy to assess cleaning level of surfaces using fluorescence spectroscopy and Wilks' ratio

- A statistical strategy combining fluorescence spectroscopy, multivariate analysis and Wilks' ratio is proposed.
- The method was tested both off-line and on-line having riboflavin as a (controlled) contaminant.
- Wilks' ratio signals unusual recordings based on shifts in variance and covariance structure described in in-control data.

There is a high demand for techniques able to monitor on-line, in real-time, the bio-contamination level of contact surfaces in the food industry. Such techniques could help to react promptly whenever failures in the cleaning or sanitation operations occur, keep the safety parameters in control at any time during production, and ultimately tailor the operations towards more sustainable and efficient practices. However, monitoring surface areas such as conveyor belts comes with a distinct set of challenges from the construction materials used in food processing equipment such as compositional-heterogeneity, background signals and continuous changes due to wear and tear. In this work we demonstrate the potential of front-face fluorescence spectroscopy in combination with Wilks' ratio statistics for monitoring large surface areas fouled under industrial working conditions. The technique was tested in both off-line and on-line mode, for a polymer-based conveyor surface, which presents an intrinsic natural variation across its running length and which was contaminated artificially for a proof of principle. Results show that any potential contamination will shift the variance and covariance structure of the in-control fluorescence landscapes modeled with PARAFAC, and detected this shift as a deviation from the reference clean state in a Wilks' ratio based monitoring charts.