In-line sensing for predictive product quality control of plant-based extrusion process


Loading...

Date

2024-09-09

Publication Type

Conference Poster

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Background: The food industry is actively exploring technologies to create sustainable and healthy alternatives to traditional meat products. High Moisture Extrusion (HME) is one of the technologies that has proven its capability to produce meat-like fibrous textures from plant proteins. However, at the present stage, HME has not reached the point of making a substantial contribution to sustainability nor that of consistently producing structures with customized sensory and nutritional characteristics. To achieve those goals, it is imperative to advance the HME processing technology from its current reliance on empirical practices and expert knowledge to a level of enhanced reliability and autonomy. Methods: A sensing method is proposed for in-line characterization of HME products by using computer vision and statistics-based texture analysis. In this sensing system, images of extrudate surface are taken and processed for extracting texture features that describe spatial variation of pixel intensity. The textual features derived from second-order statistics enable to train machine learning models, which can further act as soft sensors to predict product quality. The approach expands the analytical scope of a sensing system, newly conceived for micro-, meso-scale measurements using spectroscopy and slit die rheometry, to classify the extrudate properties at macro-scale. Results: The sensing system was deployed on a pilot-scale extruder, and its feasibility was assessed. The outcomes proved the capacity of method to provide qualitative information on the extruded product, showcasing its effectiveness in tracking variations in both ingredients and process parameters. Conclusions: A vision-based sensing system to measure the quality of HME plant-based meat products is introduced. Pioneering sensing techniques is key to reducing reliance on empirical methods and measuring extrudate quality. Integrating these sensors into control loops, along with actuators, promises to elevate process productivity and scalability.

Publication status

published

External links

Editor

Book title

Journal / series

Volume

Pages / Article No.

Publisher

ETH Zurich, Food Process Engineering Laboratory

Event

22nd World Congress of Food Science and Technology (IUFoST 2024)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

High Moisture Extrusion Cooking; Plant-based meat analogues; Machine learning; Advanced techniques; In-line sensing

Organisational unit

09571 - Mathys, Alexander / Mathys, Alexander check_circle

Notes

Poster Session September 9, 2024

Funding

Related publications and datasets