EMDS 3 3 1 Conceptual Modeling printmqm.in.tum.de/teaching/EMDS/ws1112/slides/EMDS_3_3_1...

23
Model-Based Systems & Qualitative Reasoning Group of the Technical University of Munich WS 11/12 EMDS 3 70 1 The topic 2 Decision support systems 3 Modeling 3.3 Advanced modeling Outline Compositional modeling: requirements Conceptual modeling: Why? How? Qualitative modeling: Why? Limitations? Automated model composition

Transcript of EMDS 3 3 1 Conceptual Modeling printmqm.in.tum.de/teaching/EMDS/ws1112/slides/EMDS_3_3_1...

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 70

1 The topic2 Decision support systems3 Modeling

3.3 Advanced modeling

Outline

Compositional modeling: requirements Conceptual modeling: Why? How? Qualitative modeling: Why? Limitations? Automated model composition

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 71 71

Ecological Modeling and Decision Support Systems

Requirements

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 72

Model Structure (Townsend et al. 08)

Adult vultures in year t, Nt

SurvivalMaturation

and survival

Rate at which carcasses are

eaten, F

Probability of a carcass

containing diclofenac, C

Effect of diclofenac

Baseline survival, S

Rate at which carcasses are

eaten, F

Probability of a carcass

containing diclofenac, C

Effect of diclofenac

Baseline survival, S

Adult vultures in year t-1, Nt-1

Vulture births in year t-5

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 73

Re-arranged Structure of a Model of Vulture Population

Adult vultures in year t, Nt

Survival Maturation and survival

Adult vultures in year t-1, Nt-1

Vulture births in year t-5

Baseline survival, S

Probability of carcasses with diclofenac, C

Effect of diclofenac

Rate at which carcasses are

eaten, F

Why re-arranged? Reuse of model elements

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 74

Requirements

Compositional modeling– Complex model: aggregation of elementary model fragments– Requires conceptual modeling

Conceptual modeling– Represent objects, relationships, interactions explicitly

Qualitative modeling– Models capturing partial knowledge and information

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 75

1 The topic2 Decision support systems3 Modeling

3.3 Advanced modeling3.3.1 Conceptual modeling

Outline

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 76 76

Ecological Modeling and Decision Support Systems

Processes - Motivating Examples

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 77

Process-oriented Modeling - Part 1

Identify and model elementary, independentphenomena/interactions: “processes”

Populationsize

Reproduction

Death Immigration

Emigration

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 78

Process-oriented Modeling – Part 2

Identify preconditions for the process to happen– objects– object relations– quantity conditions

DiclCarcassesProb.: C>0

VulturePopN>0

SameLocation(VulturePop, DiclCarsasses)

DiclPoisoning

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 79

Process-oriented Modeling – Part 2 Cont’d

Identify preconditions for the process to happen– objects– object relations– quantity conditions

Resourcesamount>0

Populationdensity > d0

Accessible(Population, Resources)

Reproduction

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 80

Process-oriented Modeling – Part 3

DiclCarcasses

Vulture Pop

DiclCarcassesProb.: C>0

VulturePopN>0

SameLocation(VulturePop, DiclCarsasses)

DiclPoisoning

Identify and describe impact on objects/relations

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 81

Process-oriented Modeling – Part 3 Cont’d

Resources

Population

Resourcesamount>0

Populationdensity > d0

Accessible(Population, Resources)

Reproduction

Identify and describe impact on objects/relations

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 82

Process-oriented Modeling – Part 4

Describe effects on quantities

Resources

Population

Resourcesamount>0

Populationdensity > d0

Accessible(Population, Resources)

Reproduction

“ dN/dt = r·N ”?

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 83 83

Ecological Modeling and Decision Support Systems

Processes – More Formally

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 84

Process-Oriented Modeling

Model Fragment (Process) Conditions Structure (objects,

object relations) Quantities

Effects Structure Quantities

STRUCT-CONDS QUANT-CONDS

STRUCT-EFFECTS QUANT-EFFECTS

„Model“ in the„classical“ sense

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 85

Formal Basis for Reasoning about Models

Process: a logical formula deduction consistency check model composition model-based diagnosis model-based therapy …

STRUCT-CONDS QUANT-CONDS

STRUCT-EFFECTS QUANT-EFFECTS

Conceptual layer explanation education

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 86 86

Ecological Modeling and Decision Support Systems

More Examples and Demonstration

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 87

Example: Chloramin Reactions

Water treatment: protocol of knowledge acquisition from experts

Preconditions– Structural

(substances)– Quantity

Effects– Structural

(new substances)– Quantity

CHLORAMINE REACTION

NH4+ + HClO NH2Cl + H2O + H+ \monochloramines (1)

NH2Cl + HClO NHCl2 + H2O \dichloramines (2)

NHCl2 + HClO NCl3 + H2O \trichloramines (3)

The pH determines the stability of chloramine compounds. So in excess chlorine concentration,

it’s possible to say that monochloramine stability go down, and it decomposed to:

2NH2Cl + HClO N2 + 3HCl + H2O (4)

If the pH is favorable to dichloramines existence, it is decomposed as:

NH2Cl N2 + 2HCl + Cl2 (5)

Cl2 + H2O HClO + HCl (6)

If the pH is favorable to dichloramines and monochloramines existence, then (4) will be the

predominant reaction. So the water with this conditions won’t have trichloramine (NCl3), because

they will be reduced to dichloramine form, as:

NCl3 + H2O NHCl2 + HClO (7)

Finally, to simplifying the reaction it can be said that amonium is totally destroyed by chlorine as:

2NH3 + 3Cl2 6HCl + N2 (8)

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 88

Demonstration: SIMGEN for Tutoring

Runs a simulation answers queries about the simulation

- „WHAT happens?“: values, active processes about the domain theory

- „WHY does it happen?“: preconditions Example: different cups containing a liquid

under different conditions

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 89 89

Ecological Modeling and Decision Support Systems

Requirement: Context Independence

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 90

Context-independent Models

Compositional modeling re-usable model fragments re-usable in different contexts context-independent model fragments

refer to local variables only attributes of involved objects state all preconditions Otherwise: applied in wrong context specify effect locally “dN/dt = r·N” ???

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 91

Effects Cannot be Stated by Derivatives

dN/dt = r*N

Populationsize

Reproduction

Death Immigration

Emigration

?Solution ch. 2.3.3

Model-Based Systems & Qualitative ReasoningGroup of the Technical University of Munich WS 11/12 EMDS 3 92

Modeling Assumptions

Context-independent models In the ideal sense: impossible Complete preconditions? Unstated negative preconditions: “… and X must not be present”

Resourcesamount>0

Populationdensity > d0

Accessible(Population, Resources)

ReproductionResources

Population