Addressing short term heat peaks Introduction on the shelf life of - - PowerPoint PPT Presentation

1

Results Modelling Introduction Material & Methods Outlook

• K. Schmidt1, T. Lettmann2, R. Stamminger2

1University of Bonn, INRES IPE, Karlrobert-Kreiten Str. 13, 53115 Bonn

CROP.SENSe.net & Nemaplot

2University of Bonn, ILT, Nussallee 5, 53115 Bonn

Addressing short term heat peaks

• n the shelf life of minced meat

Inverse objectives: Modelling the impact of high temperature peaks on microbial growth rates CCM, 4th International Workshop, Bonn, Germany

2

Results Modelling Introduction Material & Methods Outlook

Need for alternatives/extensions?

MAP (“Modified Atmoshere Packaging”) extends shelf life of (minced) meat for a couple of

• days. But how is the remaining

shelf life affected by careless consumer behavior during summer time with up to 50°C in a car? How are microorganism responding to heat events and how the decay kinetics of the product are altered?

Temperature response function Temperature °C

5 10 15 20 25 30 35 40 45 50 55 60 65 70

growth rate h-1, log cfu/g

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

The common Arrhenius equation is not applicable for such temperature ranges

3

Results Modelling Introduction Material & Methods Outlook

Data source

Material and methods MAP meat packages were stored both at constant temperatures in the range from 2 to 20 °C and partly exposed to temperatures of 20/30/50 °C, respectively for 3 hour periods at different points in time. Total viable counts (log cfu/g) were measured directly after delivery and continuously monitored in daily intervals until shelf life exceeded. The constant temperature trials are imposed to identify the basic influence of temperature on microbial growth. The variable temperature trials to identify response functions at higher temperatures.

4

Results Modelling Introduction Material & Methods Outlook

Data source

Constant temperatures 2°C 4°C 7°C 10°C 15°C 20°C Shelf life exceeded

Experimental design

Time

Variable temperatures

97h at 2°C 97h at 4°C 72h at 2°C 47h at 2°C 24h at 4°C 3h at 20°C 3h at 20°C 3h at 30°C 3h at 30°C 3h at 20°C

microbial growth was monitored until end

5

Results Modelling Introduction Material & Methods Outlook

Basic model

Microbial growth at constant temperature time (hours)

50 100 150 200

log cfu/g

3 4 5 6 7 8

7°C

   

 

  

          

  s

e N N N N Log

t

1 log

max

A logistical growth model is proposed as a first parsimonious approach for modelling the growth dynamics. More complex models, as Richards, Gompertz or Baranyi equations are possible, if deviance occur N = total viable counts

6

Results Modelling Introduction Material & Methods Outlook

Temperature response functions

   





d

t r

dt t T d 

 

RT E r

A

e k T d





 

   

  



• pt

T T T T x x

• pt

r

T T Q w and w w x with e T T T T k T d

• pt
• pt

                            

   max 10 2 2 max max max

1 400 40 1 1

max

 

 

bT r

a T d exp  

1. 2. 3. Arrhenius function O„Neill function Exponential function

   

 

  

          

  s

e N N N N Log

t

1 log

max

7

Results Modelling Introduction Material & Methods Outlook

Temperature response functions

Temperature °C

5 10 15 20 25 30 35 40 45 50 55 60 65 70

growth rate h-1, log cfu/g

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 O'Neill <25°C O'Neill >25°C Arrhenius Exponential

8

Results Modelling Introduction Material & Methods Outlook

Thermal transition

Modelling heat transfer of MAP foil Time (hours)

2 4 6 8 10 12

Temperature °C

10 20 30 40 50

MAP foil Meat surface Meat core heating

Ex.: heating up to 50°C for 3 hours & effective temperature the meat has been exposed to

9

Results Modelling Introduction Material & Methods Outlook

Model fitting

time (hours)

100 200 300 400

log cfu/g

3 4 5 6 7 8

2°C 4°C 7°C 10°C 15°C 20°C

R2=0.86

Logistic growth with O„Neill function

time (hours)

100 200 300 400

log cfu/g

3 4 5 6 7 8

2°C 4°C 7°C 10°C 15°C 20°C

R2=0.84

Logistic growth with Arrhenius function

10

Results Modelling Introduction Material & Methods Outlook

Biological times

Biological time

2 4 6 8 10

log cfu/g

3 4 5 6 7 8 9

Model 2°C 4°C 7°C 10°C 15°C 20°C

Transformation of the data to a time invariant scale permit the comparison of different temperatures

11

Results Modelling Introduction Material & Methods Outlook

Biological time & model fitting

3 4 5 6 7 8 9 2 4 6 8 10 Biological time log cfu/g 2°C 2/12°C 2/30/12°C 2/30/2°C 2/50/12°C 2/50/2°C 2/20/2/20 2/20°C 4/20/4/20°C 4/20°C 4/20/6°C 4/30/6°C 4/20/12°C 4/30/12°C 7/20/6°C 7/50/6°C 7/20/12°C 7/50/12°C 2/20/12°C 2/20/6°C 2/50/6°C 2/50/12°C

12

Results Modelling Introduction Material & Methods Outlook

Validation

Time (hours)

50 100 150 200 250 300

log cfu/g

3 4 5 6 7 8 2/20/12 °C

20 40 60 80 100 120 140

log cfu/g

3 4 5 6 7 8

2/50/2 °C 20 40 60 80 100 120 140

log cfu/g

4 5 6 7 8 4/20/12 °C

Time (hours)

20 40 60 80 100 120 140 160 180 200

log cfu/g

4 5 6 7 8 7/50/12 °C 20 40 60 80 100 120 140 160 180 200

log cfu g-1

4 5 6 7 8

4/20/4/20 °C

20 40 60 80 100 120 140 160 180 200

log cfu g-1

4 5 6 7 8

4/30/6 °C

13

Results Modelling Introduction Material & Methods Outlook

Reduced model

Biological time

1 2 3 4 5

log cfu/g

3 4 5 6 7 8 9

Model 2°C 4°C 7°C 10°C 15°C 20°C CI

Transformed logistic growth model without lag phase and 95% confidence intervals (CI) for constant temperature data

R²=0.88

14

Results Modelling Introduction Material & Methods Outlook

Temperature response function

Temperature °C

5 10 15 20 25

growth rate h-1, log cfu/g

0.00 0.05 0.10 0.15 0.20

O'Neill w. lag Arrhenius w. lag Exponential no lag O'Neill no lag Arrhenius no lag

The embedded temperature response function varies with the chosen model

15

Results Modelling Introduction Material & Methods Outlook

Initial density problem

Biological time

1 2 3 4 5

log cfu/g

3 4 5 6 7 8 9

Model 2°C 4°C 7°C 10°C 15°C 20°C CI

The variance found in the beginning maintains throughout the whole microbial dynamics. When shelf life times exceed, the range of a „biological shelf life time“ is in a proportional uncertainty

16

Results Modelling Introduction Material & Methods Outlook

Conclusions, part I

• A logistic growth model can be sufficiently fitted to microbial

growth data. More complex models did not significantly improve the accuracy.

• All tested temperature response functions were applicable for

the range up to 20°C.

• The O„Neill function only is suitable to address heat peaks and

higher temperatures.

• The growth model & embedded response function must be

seen as one unit for any further application.

• Some deviance in the comparison of prediction and
• bservation indicate that such deterministic, continuous models

are of limited use for prediction. The model behaviour is not flexible enough.

17

Results Modelling Introduction Material & Methods Outlook

Conclusions, part I

• The usually large variance in the initial microbial densities

(cfu/g) values are maintained throughout the growth dynamics

• Any efforts in a more precise modelling are eliminated by this

uncertainty, which makes the prediction of an expiry time vague.

18

Results Modelling Introduction Material & Methods Outlook

Estimating TVC by spectroscopy

Use of sensor technology, here hyperspectral reflectance, to estimate initial microbial density

• f minced meat, N0.
• K. Schmidt, A.-K. Mahlein, U. List, J. Kreyenschmidt

Working hypotheses:

• Total viable counts affects the reflection signature of meat

and can be classified

• Potential disturbances by other factors can be discriminated
• The specific signatures can be fitted to a common model,

changes in parameter vectors can be associated to fine tuned scaling of microbial density

19

Results Modelling Introduction Material & Methods Outlook

Background

Objectives: Estimate the current stress conditions (or microbial contamination) with a reasonable confidence on the basis of spectral reflection of the visual (VIS) and near-infrared (NIR) wave lengths. Using sensor technology to detect plant stress & phenotyping

Wavelength nm

400 500 600 700 800 900 1000

Reflection

0.0 0.1 0.2 0.3 0.4 0.5 0.6

Healthy medium stress heavy stress

20

Results Modelling Introduction Material & Methods Outlook

Material & Methods

3 fresh and 3 older (stored for 24h at 20°C) minced meat packages have been scanned with an ASD Field Spec sensor, Wavelength 400 to 1050 nm (VIS, NIR). Before scanning, half of the meat was separated for lab analysis.

Lab analysis of TVC (cfu/g) Specimen (fresh) TVC [cfu/g] TVC [log cfu/g] Specimen (old) TVC [cfu/g] TVC [log cfu/g] 1 605000 5,78 4 1470000000 9,17 2 635000 5,80 5 1530000000 9,18 3 755000 5,88 6 2370000000 9,37 21

Results Modelling Introduction Material & Methods Outlook

Results

Mean spectral signatures

0.1 0.2 0.3 0.4 0.5 0.6 400 600 800 1000

Wavelength nm Reflection

Fresh 1 Fresh 2 Fresh 3 Old 1 Old 2 Old 3

22

Results Modelling Introduction Material & Methods Outlook

Model

 



                  

              

n i nm nm nm nm i nm

i i i i

e e B A F

1

1

   

The model was fitted to the TVC signatures and changes in the parameter vectors were used for density classification. Applied model: Additive Double Weibull function1

1 Registration no 10 2009 0404 944.0 at the

German Patent- and Trade Mark Office

Results Modelling Introduction Material & Methods Outlook

Preliminary results

0.1 0.2 0.3 0.4 0.5 0.6 400 500 600 700 800 900 1000

Wavelength nm Reflection

0.1 0.2 0.3 0.4 0.5 0.6 400 500 600 700 800 900 1000

Wavelength nm Reflection

Fresh meat, R²=0.998 decayed meat, R²=0.989 Blue Llne: Sensosr signal, red: model fit

24

Results Modelling Introduction Material & Methods Outlook

Preliminary results

Parameter

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6

Candidates of parameters for TVC classification A

1.112 1.059 1.051 0.804 0.988 1.034 X

B1

0.356 0.348 0.341 0.255 0.310 0.329 X

nmα1

455.788 455.176 455.363 454.528 456.572 455.526

nmβ1

535.645 535.300 535.175 541.629 540.739 539.330

α1

17.871 17.697 16.917 25.335 22.926 21.638 XX

β1

54.220 52.974 52.064 47.894 41.715 39.716 XX

B2

• 14.214
• 13.715
• 13.549
• 9.502
• 11.734
• 11.804 X

nmα2

572.139 571.898 571.897 571.586 567.717 568.014

nmβ2

573.796 573.522 573.225 574.275 573.895 573.461

α2

38.545 38.076 37.739 38.576 36.687 34.667 X (?)

β2

1.694 1.719 1.715 2.231 1.738 1.710 X (?)

B3

6.945 6.679 6.567 4.384 5.413 5.458 X

nmα3

571.528 571.242 571.230 567.916 565.965 566.069 X

nmβ3

800.538 800.820 801.511 805.104 807.699 809.604 X

α3

32.949 32.456 32.194 34.761 33.484 31.318 X(?)

β3

5.016 5.086 5.131 5.385 5.555 5.613

B4

• 3.179
• 3.073
• 3.061
• 2.515
• 2.870
• 3.112

X (Water content?) nmα4

922.130 922.127 922.123 922.101 922.108 922.095

nmβ4

807.577 808.077 808.273 804.293 804.380 803.454 Required?

α4

1.006 1.025 1.024 1.113 1.058 1.119

X (Water content?) β4

8.287 8.400 8.453 7.901 8.754 8.543 Required?

25

Results Modelling Introduction Material & Methods Outlook

Conclusions, part II

• Results represent the first initial findings of a

concept study

• Spectroscopy did work better than expected
• Procedure itself is fast and quickly analysed
• Differences in spectral signatures are obvious and

appear related to the microbial density

• Even small difference were detectable with respect

to the underlying model

• Changes in the parameter vectors are not clear yet
• Data source is currently too small
• Reliability of the technique has to be tested in future

research

26

Results Modelling Introduction Material & Methods Outlook

Conclusions, part II

Potential application:

• Process control
• Automatical detection of contaminated

charges, high-throughput technology

• Automatical residual shelf life calculations
• Further factors-

during production, transport and delivery.

27

Results Modelling Introduction Material & Methods Outlook

The final slide

Thank you for your attention