Simulation and Optimization of Construction Processes (2020 – 2021)
Summary of the lectures
Simulation and Optimization
Summary of the Lectures Simulation and Optimization of Construction Processes
Lecture 1: Introduction
What is simulation?
Simulation is the practice of: Building a computerized model that imitates the behaviour of a real and
complex system (or process) over time and from a given point of view.
Simulation models are used to: Investigate, analyse, predict, and optimise the performance of the
system under various conditions (decision variables).
Why simulation? Simulation is useful since construction processes are:
There are the following reasons to simulate: - Complex
- Duration and scheduling of activities - Of a time-varying nature
- Material inflow and outflow - In largely uncontrolled (random) environment
- Safety-related issues (evacuation) - Interdependent
- Influenced by many external factors
- Unique but repetitive
What is optimisation?
Optimisation is the practice of: Using simulation models to find the set of controllable variables that
yield the best result/performance from a certain perspective under certain constraints.
A model is a representation of reality based on a set of available information It is used to capture,
analyse, and predict the behaviour and characteristics of a real system.
The basic definition of a real system:
- A process that changes a set of inputs to a set of outputs
- An integrated set of components working towards a goal
- A set of interacting objects working towards a goal
The basic definition of an environment:
- Context (World) that hosts the model
- External factors that affect the system
- Boundaries between the system and the environment
In a model, there are different components with a different definition. An activity is a continuous
process over time, while an event is a point in time when something is happening (for example that
start of emptying the bucket the excavator when the truck arrived). An event triggers a change in the
system, while the system state is a point in time when all activities are frozen (a snapshot of the
system). In this example, the amount of soil can also be classified under objects.
Model Components:
(1) Entities/objects
(2) Attributes of entities (e.g., bucket size)
(3) Events (e.g., truck arrival)
(4) Activities (e.g., excavator digging)
(5) System state (snapshot of the system)
(6) Initial state (e.g., number of excavators)
Page 1 of 20
,Simulation and Optimization of Construction Processes (2020 – 2021)
Summary of the lectures
Simulation models can be classified into static and dynamic models. A static model is for example
tossing a coin, time is not influencing the process/outcome, it is about winning the game. In a dynamic
model, the system changes when the time moves ahead.
Static Model: Dynamic Model:
- Time-invariant system - System behaviour is a function of time
- E.g., Monte-Carlo simulation - E.g., Model of water retaining behind a
- E.g., Estimating the probability of dam
winning a game
The simulation models can also be classified into deterministic and stochastic models. In a
deterministic model, the if-then rule is applied. For example, if I throw a dice and six is the output then
I will do this (the same thing is always done for a given outcome). A deterministic model is often used
in construction simulation. In a stochastic model, you might get a different output with the same input.
The stochastic model is more accurate than a deterministic model.
Deterministic Model: Stochastic Model:
- Output is determined with certainty - Model generates different outputs from
- Input 𝑥 leads always to output 𝑦 the same input in different runs/iterations
- Output is determined by a probability
function
In simulation models, there can be made a difference between a continuous state and discrete state,
and between continuous-time and discrete-time.
Continuous state is for example filling a glass with water and taking snapshots, there are infinite
possibilities (no predefined or limited number of states) you can stop at any point in time and take a
snapshot.
Discrete state is for example the game of hopscotch (hinkelen), you jump from square to square and
there is a predefined number of squares (1 to 10). If the process is frozen at any point in time, you can
only be in one of the predefined squares (squares 1 to 10). There are no infinite possibilities.
Continuous time is used for System Dynamics (SD). There are no intervals, but time is running
continuously.
Discrete time is used for Agent-based Simulation (ABS) and Discrete Event Simulation (DES). You only
look at the system at certain time intervals (fixed intervals or intervals determined by events) For
example check the state of the system every two minutes, between the two minutes is not simulated.
A simulation built in the following seven steps:
(1) Observation of real system
(2) Determine the purpose of modelling
(3) Develop a conceptual model
(4) Data collection
(5) Build a simulation model
(6) Model verification
(7) Model validation
Verification building the model correctly (correctly implemented with good input and structure).
Purpose: ensure the conceptual model is reflected accurately in the computerised representation.
Page 2 of 20
, Simulation and Optimization of Construction Processes (2020 – 2021)
Summary of the lectures
Validation building the correct model (an accurate representation of the real system). Purpose: ensure
that the simulation model is a “fair” representation of the real system (considering the modelling
purpose)
Two approaches to validate simulation models: subjective and objective methods.
Observable System Non-observable System
Subjective - Comparison using graphical displays - Explore model behaviour
Approach - Explore model behaviour - Comparing to other models
Objective - Comparison using statistical tests - Comparison to other models using
Approach statistical tests
Subjective methods:
- Face validation: the user/expert analyses the results of the model. Sensitivity analysis can be
used.
- Graphical/Animations: Graphically, the results of the simulation are compared to common
sense/historic data.
- Turing Test: Testing whether or not a human agent can determine the difference between the
results from the actual system and the model.
Objective methods:
- Predictive testing: The strongest method of validation. The predicted results from the
simulation are tested in the real system. Is this always feasible?
- Statistical Analysis: Using historical data to compare the performance of the mode.
o Confidence interval
o T-test
Operation and process simulation:
- A construction operation results in the placement of a specific segment of construction that
has some technologic processes and work tasks (for example earthwork operation).
- A construction process is the unique collection of work tasks related to each other through a
technology structure and sequence (for example digging process).
Depending on the nature of the processes and the purpose, three types of simulation are common in
the construction industry: Discrete Event Simulation (DES), Agent-based Simulation (ABS), and System
Dynamics (SD)
Page 3 of 20
Summary of the lectures
Simulation and Optimization
Summary of the Lectures Simulation and Optimization of Construction Processes
Lecture 1: Introduction
What is simulation?
Simulation is the practice of: Building a computerized model that imitates the behaviour of a real and
complex system (or process) over time and from a given point of view.
Simulation models are used to: Investigate, analyse, predict, and optimise the performance of the
system under various conditions (decision variables).
Why simulation? Simulation is useful since construction processes are:
There are the following reasons to simulate: - Complex
- Duration and scheduling of activities - Of a time-varying nature
- Material inflow and outflow - In largely uncontrolled (random) environment
- Safety-related issues (evacuation) - Interdependent
- Influenced by many external factors
- Unique but repetitive
What is optimisation?
Optimisation is the practice of: Using simulation models to find the set of controllable variables that
yield the best result/performance from a certain perspective under certain constraints.
A model is a representation of reality based on a set of available information It is used to capture,
analyse, and predict the behaviour and characteristics of a real system.
The basic definition of a real system:
- A process that changes a set of inputs to a set of outputs
- An integrated set of components working towards a goal
- A set of interacting objects working towards a goal
The basic definition of an environment:
- Context (World) that hosts the model
- External factors that affect the system
- Boundaries between the system and the environment
In a model, there are different components with a different definition. An activity is a continuous
process over time, while an event is a point in time when something is happening (for example that
start of emptying the bucket the excavator when the truck arrived). An event triggers a change in the
system, while the system state is a point in time when all activities are frozen (a snapshot of the
system). In this example, the amount of soil can also be classified under objects.
Model Components:
(1) Entities/objects
(2) Attributes of entities (e.g., bucket size)
(3) Events (e.g., truck arrival)
(4) Activities (e.g., excavator digging)
(5) System state (snapshot of the system)
(6) Initial state (e.g., number of excavators)
Page 1 of 20
,Simulation and Optimization of Construction Processes (2020 – 2021)
Summary of the lectures
Simulation models can be classified into static and dynamic models. A static model is for example
tossing a coin, time is not influencing the process/outcome, it is about winning the game. In a dynamic
model, the system changes when the time moves ahead.
Static Model: Dynamic Model:
- Time-invariant system - System behaviour is a function of time
- E.g., Monte-Carlo simulation - E.g., Model of water retaining behind a
- E.g., Estimating the probability of dam
winning a game
The simulation models can also be classified into deterministic and stochastic models. In a
deterministic model, the if-then rule is applied. For example, if I throw a dice and six is the output then
I will do this (the same thing is always done for a given outcome). A deterministic model is often used
in construction simulation. In a stochastic model, you might get a different output with the same input.
The stochastic model is more accurate than a deterministic model.
Deterministic Model: Stochastic Model:
- Output is determined with certainty - Model generates different outputs from
- Input 𝑥 leads always to output 𝑦 the same input in different runs/iterations
- Output is determined by a probability
function
In simulation models, there can be made a difference between a continuous state and discrete state,
and between continuous-time and discrete-time.
Continuous state is for example filling a glass with water and taking snapshots, there are infinite
possibilities (no predefined or limited number of states) you can stop at any point in time and take a
snapshot.
Discrete state is for example the game of hopscotch (hinkelen), you jump from square to square and
there is a predefined number of squares (1 to 10). If the process is frozen at any point in time, you can
only be in one of the predefined squares (squares 1 to 10). There are no infinite possibilities.
Continuous time is used for System Dynamics (SD). There are no intervals, but time is running
continuously.
Discrete time is used for Agent-based Simulation (ABS) and Discrete Event Simulation (DES). You only
look at the system at certain time intervals (fixed intervals or intervals determined by events) For
example check the state of the system every two minutes, between the two minutes is not simulated.
A simulation built in the following seven steps:
(1) Observation of real system
(2) Determine the purpose of modelling
(3) Develop a conceptual model
(4) Data collection
(5) Build a simulation model
(6) Model verification
(7) Model validation
Verification building the model correctly (correctly implemented with good input and structure).
Purpose: ensure the conceptual model is reflected accurately in the computerised representation.
Page 2 of 20
, Simulation and Optimization of Construction Processes (2020 – 2021)
Summary of the lectures
Validation building the correct model (an accurate representation of the real system). Purpose: ensure
that the simulation model is a “fair” representation of the real system (considering the modelling
purpose)
Two approaches to validate simulation models: subjective and objective methods.
Observable System Non-observable System
Subjective - Comparison using graphical displays - Explore model behaviour
Approach - Explore model behaviour - Comparing to other models
Objective - Comparison using statistical tests - Comparison to other models using
Approach statistical tests
Subjective methods:
- Face validation: the user/expert analyses the results of the model. Sensitivity analysis can be
used.
- Graphical/Animations: Graphically, the results of the simulation are compared to common
sense/historic data.
- Turing Test: Testing whether or not a human agent can determine the difference between the
results from the actual system and the model.
Objective methods:
- Predictive testing: The strongest method of validation. The predicted results from the
simulation are tested in the real system. Is this always feasible?
- Statistical Analysis: Using historical data to compare the performance of the mode.
o Confidence interval
o T-test
Operation and process simulation:
- A construction operation results in the placement of a specific segment of construction that
has some technologic processes and work tasks (for example earthwork operation).
- A construction process is the unique collection of work tasks related to each other through a
technology structure and sequence (for example digging process).
Depending on the nature of the processes and the purpose, three types of simulation are common in
the construction industry: Discrete Event Simulation (DES), Agent-based Simulation (ABS), and System
Dynamics (SD)
Page 3 of 20
