Process Safety and Control – Lecture 2

Prof. Dr.-Ing. Andreas Bück

Institute of Particle Technology, FAU Erlangen-Nürnberg

ABSTRACT

In this lecture we start with the description and analysis of complex net-
works as they appear in a multitude of applications: telecommunication, world
wide web, power grids, material and energy flow between process units, trans-
portation networks, or metabolical networks. We will introduce a mathematical
structure that allows to describe and to reason about all these different networks
in a unified framework: the graph.

Depending on the number of nodes and

Basic terminology and concepts links, highly complex networks can be
In almost every process of reasonable com- formed. Our aim is to analyse all these appli-
plexity, we can can find smaller sub-pro- cations in a unified way, e.g., to identify vital
cesses that are linked to each other. Over components of a network. We can achieve
these links, information, matter, momentum this by using a specific mathematical object,
or energy is exchanged between the process the graph.
units. The ratio of flows, number of links, the This lecture is an introduction to the
units that are connected and the way they are, basics of graph theory. Many introductions
determine the overall function of the process. to the topic are available, for example G.
Processes that can be mapped to this abstract Chartrand’s Introductory Graph Theory
description are called networks. A network (Dover Publications) or R. Diestel’s Graph
consists of node (the process units or entities) Theory (published by Springer).
which are connected by links. We start with a few definitions and
Typical examples of networks are: then continue with a first analysis of graph
transportation systems (entities: buses; links: properties. [Examples and visualisations are
roads), water and sewage systems (entities: given in the lecture.]
pumps or drains; links: pipes), communica- Graph: consists of a non-empty, finite
tion networks (entities: sender or receiver; set of vertices (nodes) V = {v 1 , v 2 , . . .} and a
links: (wireless) communication channel), set of edges e, E, denoting all links between
power grids (entities: power station, PV any two nodes. We denote an edge by
plant, wind turbine; link: power cables), reac- ei,k = (v i , v k ). The set E can be empty.
tion networks (entities: component concen-
Directed/undirected graph: In a di-
trations, links: reaction paths) or social net-
rected graph, or digraph, the direction of the
works (entities: people; links: social interac-
link is of importance. Each edge is
tion).

24 April 2023
 -2-

interpreted as an ordered pair ei,k = (v i , v k ) Cut-vertex: By G − v we mean the

with the first component denoting the start graph G from which node v with all its inci-
and the second denoting the end node. dent edges has been removed. The vertex v is
Simple graphs: do not contain loops, a cut-vertex of G, if G loses connectivity by
i.e., single edges that start and end in the removing node v. If G is connected and v a
same vertex (node). cut-vertex, then G − v contains two or more
components (subgraphs).
Random graph vs regular graph: In a
regular graph, edges between nodes are deter- Bridge: By G − e we mean the graph G
mined by a deterministic rule resulting in from which edge e has been removed. The
regular pattern. If edges are randomly placed, edge e is a bridge, if G is disconnected after
the resulting graph is a random graph. removal of e. If G is connected and e a
bridge, then G − e contains exactly two com-
Visualisation of graphs: Nodes are rep-
ponents (subgraphs).
resented by dots, edges by either (directed)
line segments or arcs. Edges may intersect Shortest path: The shortest path be-
without consequence for the overall structure tween two nodes u and w minimises the sum
of the graph. of the edge weights. (I.e., edge weights are
interpreted as costs moving along this edge.)
Order of a graph: the total number of
In an undirected, unweighted graph, all edge
vertices in the graph, |V |.
weights are equal to +1.
Size of a graph: the total number of
Adjacency matrix: The adjacency ma-
edges in the graph, |E|.
trix A of a graph is a |V | × |V |-matrix. The el-
Degree of a node: Number of edges ement aij has a non-zero value, if (source)
starting or ending in a node. In digraphs, in- node v i is directly linked to (sink) node v j .
coming and out-going are counted separately, The matrix A is symmetric for an undirected
i.e., incoming degree and outgoing degree are graph, aij = a ji . In an unweighted graph, the
counted. element value is +1, if a link exists. In a
Adjacency: Two nodes v i and v k are weighted graph, the value of element aij cor-
adjacent, if they are directly connected by responds to the edge weight. – The adjacency
one edge, i.e., ei,k exists. matrix contains all structural information
about the graph, i.e., analysis of the matrix
Incidence: A node is incident to an
properties allows reasoning about the graph.
edge, if the edge either starts or ends in this
This is of particular importance, if the graph
node.
is very complex, i.e., when a visual inspec-
Edge weight: To each edge, a weight tion is insufficent.
can be assigned, for example, to denote ca-
Incidence matrix: The incidence matrix
pacity, cost or penalty of the link. Weights do
I is an alternative complete codification of a
not need to be positive.
graph structure and equivalent to the adja-
Connectivity: Two nodes u and v of a cency matrix. The matrix I is a |V | × |E|-ma-
graph G are connected, if u = v or, if u ≠ v, a trix; the element i jk has a non-zero value, if
path between u and v exists. A path is an al- node v j is incident to the edge e k , i.e., if one
ternating sequence of nodes and edges. If no of the components of the pair e is v j . – De-
node is repeated, then it is called a trail, oth- pending on the graph structure and the type
erwise it is a walk. A graph G is connected, if of analysis, operation with matrix A or I is
any two nodes are connected, otherwise it is more convenient. The matrix A is always
disconnected. A trail in which u = v and that square, I is typically rectangular. A is often a
contains at least three nodes is called a cyle. dense matrix (high number of non-zero

24 April 2023
 -3-

elements), but symmetric. I is typically a

sparse matrix (low number of non-zero ele-
ments), very efficient to store and to manipu-
late.

Outlook
In the next lecture, we will dive into the
analysis of complex networks represented by
graphs. We will reduce the work to the analy-
sis of the adjacency matrix and incidence ma-
trix.

24 April 2023