FQD-31306 Predicting Food Quality
Marly Verest
FQD-31306 - Predicting Food
Quality
Module 0
Predicting food quality
Quality = satisfying the expectation of the consumer
Q=f(Qint, Qext) focus in this course is on intrinsic quality attributes
Trained sensory panels = used to decompose down components of quality perception into sensory
attributes
Consumer panels = used to rate preference, choice, and liking
Why would we want to predict food quality?
- Fast changing consumer demands
o Safe, healthy, attractive, sustainable, functional
- Continuous product development
o Raw material change
- Increasing demands on food quality
o Consumer wishes, food laws
- Efficient product and process design
o Saving of time, money and resources
Food Quality attributes examples
- Colour
o Components that absorb certain wavelengths
- Smell
o Components that interact with receptors in the nose
- Taste
o Components that interact with receptors on the tongue
- Texture
o Gel strength, polymer networks, particle networks
- Safety
o Absence of pathogenic micro-organisms / toxic chemicals
- Convenience
o Packaging
o Ready to eat
- Shelf-life
o Microbial spoilage, diffusion
- Nutritional value, healthiness
(measurable) Quality Performance Indicators
e.g. colour: carotenoid content or nutritional value: lysine content
1
,FQD-31306 Predicting Food Quality
Marly Verest
2
, FQD-31306 Predicting Food Quality
Marly Verest
Classification in terms of food science:
- Chemical aspects
o Maillard reaction, oxidation
- Biochemical aspects
o Enzymatic browning, proteolysis, lipolysis, hydrolysis,…
- Physical aspects
o Rheology, fracture mechanics, coalescence, coagulation, diffusion, phase change, …
- Microbiological aspects
o Growth of micro-organisms, inactivation
Food Quality Prediction
- Translation of consumer perceptions into measurable and manageable quality attributes
- Identifying most relevant food quality attributes
- Studying of processes underlying the behaviour of food quality attributes
- Predicting the behaviour of food quality attributes
Changes take place in foods, since foods are not static
- What changes are possible thermodynamics
- How fast do changes occur kinetics
- The possible changes but especially their rates affect food quality
Quality Analysis Critical Control Points (QACCP)
How to get grip on the quality changes
- Chain analysis of what actors are doing
- Identifying the process that affect quality
- Identifying the factors that influence the processes
- Identifying what is happening in the food
- Turn the analysis results into a model
Prediction: being able to tell something sensible about future events
Events: results of chemical, biochemical, physical, microbiological reactions in the food
Sensible: outcome should be in the right order of magnitude but some variation is acceptable
3
, FQD-31306 Predicting Food Quality
Marly Verest
Module 1
Structure of mathematical model
Mathematical model = reflects quantitatively a dependence between variables
- Input: independent variable (controllable) = x-value
- Output: dependent (response / not controllable) variable = y-value
Quantitatively = how much does y change when x changes
Parameters = a and b determine the output of the model (y-value)
η = f (θ,ξ)
Finding mathematical models
How to find a mathematical model?
- Theory may predict a certain model (almost never possible, since foods are complex systems)
- Based on experimental observations
- Studying the data pattern (what pattern can be seen)
- Trying to find a mathematical fit without any theoretical background
Overfitting = to many parameters compared to the amount of datapoints
Mathematical models that are relevant for food science problems usually describe changes in time
and/or space:
- Algebraic equations e.g. Stokes’ law, Growth model for micro-organisms
- Differential equations e.g. First-order model
- Partial differential equations (heating of food in a can, increasing temperature, but different
temperatures inside the can)
Models and Errors
Deterministic models = provide an outcome that seems to be without uncertainty BUT: parameters in
models are estimated from experiments that contain unexplainable variation, hence, parameter
estimates are uncertain
Stochastic or probabilistic models = provide range of output values, which reflects the uncertainty of
the prediction same input does not always give the same output but shows variation
η = f (θ,ξ) + ε (error term)
Two sources contributing to total uncertainty separation of the two sources is of importance so
that appropriate measures can be taken
- Variability
o Inherent variation in the system under study
o Cannot be reduced for a given system
- Uncertainty
o Reflects our state of knowledge about the system under study
o Can be reduced by better and more measurements
4
