Trailing the uncertainties …in analytical equipment with Marie Top Mygind (FOSS)

This technical presentation delves into the multifaceted challenge of managing uncertainties in analytical equipment used for food analysis and quality control.

Marie Top Mygind presents a detailed case study of the FoodScan development project, which aimed to create a next-generation near-infrared transmission analyzer capable of measuring meat and dairy products with minimal measurement variation.

The presentation systematically breaks down the sources of analytical error and their relative contributions to overall instrument performance. A critical insight involves the relationship between sample homogeneity and measurement repeatability—inhomogeneous samples introduce variability as the measurement spot encounters different compositional areas, requiring multiple measurements to achieve reliable results.

The framework identifies three primary error categories: repeatability error (influenced by sample presentation and system noise), transferability error (affecting consistency across multiple instruments at different customer sites), and accuracy error (comparing instrument predictions to chemical laboratory references).

The presentation includes practical examples demonstrating how sample redistribution affects measurement outcomes and discusses the application of classical sampling theory to minimize errors. Notably, accuracy error consistently emerges as the largest contributor to total analytical error, partly due to environmental factors affecting sample condition during transport to reference laboratories and inherent variability within reference laboratory procedures themselves.

The speaker emphasizes that instrument developers must balance design optimization with cost considerations, particularly regarding measurement frequency and sample handling protocols. Quality control strategies include ensuring tight manufacturing tolerances, implementing standardized algorithms, and developing robust predictive models.

The presentation concludes that perceived customer performance depends primarily on controlling the most significant error sources, with particular attention to accuracy validation against reference standards.

By understanding how different error components combine and interact, manufacturers can strategically prioritize improvements that maximize customer satisfaction and instrument reliability across diverse operational environments and product types.

From the meeting "Certain when Uncertain".