Machine Learning Models for Food Analog Functional Property Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for evaluating food analogs are impractical, inaccurate, and inefficient, particularly when testing large numbers of prototypes, as they rely on subjective sensory panels and insufficiently capture the nuances of functional properties like texture and mouthfeel.
Innovation Solution
A method that determines functional property feature values from raw measurement signals, using machine learning models to predict manufacturing variables that produce prototypes with target characteristics, enabling more accurate and precise comparisons and predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensory panels are used to evaluate food analogs, then the evaluation process is simple to implement, but the accuracy and precision of functional property measurement deteriorates
Solution Approach 1:
The patent replaces the mechanical/conventional sensory panel evaluation system with an automated instrument-based measurement system. Texture analyzers, rheometers, and other functional property measurement devices objectively quantify properties like texture, mouthfeel, and viscoelasticity, eliminating subjective human perception and providing precise, reproducible data.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between raw measurement data and functional property evaluations. The ML models process complex measurement signals, extract meaningful features, and predict functional properties, bridging the gap between raw data and actionable insights while handling complexity systematically.
2Productivity
If subjective sensory panels evaluate large numbers of prototypes, then the evaluation coverage is extensive, but the time required and efficiency deteriorates
Solution Approach 1:
The patent implements self-service through automated measurement systems that independently collect, process, and analyze functional property data without requiring human panelists. Instruments automatically test multiple prototypes sequentially, and the system self-manages data processing through integrated software and machine learning algorithms, dramatically increasing throughput.
Solution Approach 2:
The patent enables continuous evaluation through automated instrumentation that can rapidly sequentially test multiple prototypes without interruption. The system maintains continuous operation from sample loading through data analysis, eliminating the start-stop nature of manual sensory panels and maintaining productive activity throughout the evaluation process.
3Measurement precision
If conventional methods evaluate functional properties, then the measurement process is straightforward, but the ability to capture nuances of texture and mouthfeel deteriorates
Solution Approach 1:
The patent segments complex functional properties into multiple measurable components. Instead of relying on overall subjective impressions, the system separates texture, mouthfeel, viscoelasticity, and other properties into distinct measurement parameters, each captured by specific instruments and analyzed independently to reveal nuanced characteristics.
Solution Approach 2:
The patent transitions from low-dimensional subjective ratings to high-dimensional objective measurements. By capturing multiple functional properties simultaneously across different physical dimensions (force, deformation, time, frequency), the system creates a comprehensive multi-dimensional profile of texture and mouthfeel that conventional single-parameter methods cannot resolve.
Data Source
AI summary
In variants, a method for sample evaluation to mimic target properties can include: determining functional property feature values for a target and determining variable values for a prototype based on the functional property feature values for the target. In variants, the method can optionally include: determining functional property feature values for a first prototype, processing functional property features, comparing prototype functional property feature values to target functional property feature values, training a variable value model, training a characterization model, and/or any other suitable steps.