BUSINESS PROCESS ANALYTICS

07/02/2024
1.ANALYTICS
WHAT IS A BUSINESS PROCESS:
- They are processes and
- There are several definitions:
o Davenport (1993): “a structured, measured set of activities designed to produce a
specified output for a particular customer or market; [...] a specific ordering of work
activities across time and place, with a beginning, an end, and clearly identified
inputs and outputs: a structure for action”.
o Rosemann (2001): “the self-contained, temporal and logical order of those
activities, that are executed for the transformation of a business object with the
goal of accomplishing a given task”.
o Sharp and McDermott (2001): “a collection of interrelated work tasks, initiated in
response to an event that achieves a specific result for the customer of the process”.
- How analytics can help the decision-making process
WHAT ARE ANALYTICS:
- It is the science of using data to build models that add value to decisions made by
companies, institutions and individuals.
- Data, models, science, value, decisions, individuals, ... several ingredients to be put
harmonically together to get knowledge from the (real) world
- Note that data are crucial but is only one piece of the puzzle. The challenge is to be able to
create suitable frameworks capable to fruitfully use data.

We use analytics to analyses business process.

WHY ANALYTICS?
“The world is rapidly moving into an era of unprecedented data and computational power. [...]
There are tremendous opportunities to be realized by combining the realms of data, modelling,
and decisions.”
- Technology adoption rates are dramatically huge. Industry 4.0 and IoT is based on
technologies which require the use of analytical tools.
- The availability of data (often incomplete and of bad quality) and the complexity of real
world problems require models to help the decision maker in taking good decisions.

A BROAD CATEGORIZATION OF ANALYTICS:

- DESCRIPTIVE ANALYTICS  describe, report on, and visualize the real situation by means of
data; representation of the world as it is (past and present).
- PREDICTIVE ANALYTICS anticipate trends and identify relationships in the data;
description of possible to be scenarios (future)
- PRESCRIPTIVE ANALYTICS what u can do, the best thing to do with a course of action
given the data you have and the trends you expect. Transform data analysis in
improvement actions.

1
 2.DECISION SUPPORT SYSTEM
DECISION SUPPORT SYSTEM (DSS) is a methodology based on analytical tools used to support
(semi-structured = Based on variables and dimensions which are not perfectly known or totally
deterministic. In this latter case, a totally automated system can be used) decisional processes. It is
a blend of data, information technologies and personal knowledge of the decision maker. The DSS
supports the entire decision process.

Decision related to some managerial problem. Ex: Short/long time horizon,

Financial/Resources/Production/Sales …, Project based or Routines, old/new business…
Support: a methodology designed to provide help in taking decision. Ex It is not intended to
substitute the (human) decision maker; not completely autonomous.
System: it involves many (possibly complex) processes and components. Ex Based on
different technologies (from human competences to completed automated processes).

COMPONENTS:
o DATABASE: collect all the several information and knowledge about the problem
o MODEL: the technology (black-box?) used to support the decision maker
o USE INTERFACE: the interface used to insert input and extract output

PHASES:
To properly build and make available a DSS, several phases are necessary:
1. Intelligence phase: careful analysis of the status-quo: data gathering and classification.
2. Design phase: build a model, define suitable scenarios and collect plausible alternative
solutions.
3. Choice phase: selection of the best alternative.
4. Implementation: adoption of the selected alternative and test phase on the field
Data is only one component of the picture. Choosing the correct model and implement it in a
consistent way are also of paramount importance to find the best solution to the original problem.

EXAMPLE
Examples of business processes and frameworks where the use of DSS is crucial:
- Enterprise Resource Planning (ERP)  are package software applications that connect and
manage information flows within and across complex organizations, allowing managers to
make decisions based on information that truly reflects the current state of their business.
- Inventory management Supply Chain Management (SCM)
- Customer Relationship Management (CRM)
- Business Operations Management

3. AI IN MANAGEMENT
AI: refers to the development of computer systems or software that can perform tasks that
typically require human intelligence, such as learning, reasoning, problem-solving, and
understanding natural language.

2
THREE APPLICATIONS OF AI IN MANAGEMENT
1. Decision Support Systems: AI is employed in management for decision support systems
that assist executives in making informed decisions. These systems use machine learning
algorithms to analyze vast amounts of data, identify patterns, and offer insights.
2. Predictive Analytics for Resource Optimization: AI applications in predictive analytics help
in forecasting trends and optimizing resource allocation. In management, this is particularly
valuable for inventory management, workforce planning, and supply chain optimization. AI
algorithms can analyze historical data to predict future demands, enabling managers to
optimize resources and minimize costs.
3. Chatbots for Customer Relationship Management (CRM): AI-powered chatbots are used in
management for enhancing customer interactions and support. These chatbots can handle
routine customer queries, provide information, and even assist in resolving issues.
AI is related to analytics
1. Support in daily operations: reporting, email, data visualization, translators
2. Automation in customer care: BOT and personal assistant, customization, instant
messaging, automatic response
3. Support decisions
- Analytics and forecasting / CRM
- Inventory management and SC
- Accounting figures and fraud detection
From point 1 to 3 there is an increasing of complexity of process involved, from operative to
strategic.

SUMMARIZING:

08/02/2024

4. ERP AND MRP SYSTEM

HISTORY

ERP is one the fundamental tool to organize and use

data in a proper way. It involves the entire company.
Its history started in the ’60 with a simple tool to
deal with inventory: Material requirement planning
(MRP I) what are the resources need for run the
company.
In the 70’ was introduced: MRP II
In late 80 the MRP II developed in ERP

3
MRP I:
 Proactive management methodology aimed at programming the needs of materials and to
control the inventory.
 Only two functions of the company are involved: inventory and manufacturing. Everything
is related to the operations
 It is based on the planning of supply (input) and orders (demand)
MRP II:
 Compared to MRP I, It considers also capacity and availability of human and non-human
resources (e.g., machines).
 It contributes to a more realistic production plan and makes the response to market
demand effective.
 Integrates several functions into the planning process (production, budgeting, operations,
human resources). Many departments are involved
ERP (Enterprise Resource System):
Definition: Software architecture that facilitates the flow and collection of information between all
the functions within the company such as manufacturing, logistics, finance, human resources, ...
Goals:
(i) Integrate all the information related to the different functions of the company,
providing a vehicle to share information at the right time, at the right person and in the
right way, across the company and its partners.
(ii) Allow for the creation of connections between the different phases of the value chain
at a lower cost and at a faster scale than traditional systems (which is not base on
analytics).
Model: The ERP operates through a common database that interacts with several modules to
avoid redundancy and ensure the integrity of the entire system. This system is made up of various
functional modules set up for all the "business units “of the company.

This is a visual representation of the ERP. U can

see all the stakeholders, and everything is
involved in the architecture. U can see all the
function involved.

This is a list of modules, and it is taken by

the first commercial ERP, made by the
SAP.

4
What are the main reasons leading to the use of a dedicated decision support system in planning
material requirements?
• The reactive management method (typical of pre-MRP companies) made it difficult to
deal with uneven market demand: it does not allow the correct production planning /
scheduling.
• The coexistence of several non-integrated information systems allowed for discharge of
responsibility and did not provide the management with a clear picture of the actual/true
situation.
To properly run the company, there is the need of high-quality information, meaning exact
information: any "formal “ERP must represent reality. If the formal system is not trustworthy, the
staff will disregard it and develop its own informal informative system to perform the assigned
tasks. When a formal information system (such as ERP) and informal information systems (such as
experience-based practice) coexist, or if several information systems are present, and they are not
organized under a unique umbrella, then the company will be in serious difficulties in correctly
planning its activities and reaching the desired targets.
Input:
• Bill of materials (BoM): identifies the exact
components needed to create each item and
final deliverable.
• Master production schedule (MPS): the
number of items to be produced during a
specific time span. It is based on available
orders and forecasted future demand.
• Inventory record files: units of the goods
on hand or on order.

Output:
• Production levels for each item.
• Delivery dates and reorder points for sub-components.
• Other reports

5.DECISION SUPPORT SYSTEM: the Bill of Material and the MRP

Bom: Bill of material
It is one of the important components of ERP.
DEFINITION:
The bill of material is a tree-based structure that contains information about the components and
processing times needed to produce all the components and from them the final product.
Its "explosion" allows the scheduling backward in time of the whole production related to a final
product. It indicates the latest time at which the single components must be available in order
not to delay the production.

Definition from APISC: “A listing of all the subassemblies, intermediates, parts, and raw materials
that go into a parent assembly showing the quantity of each required to make an assembly. It is

5
used in conjunction with the master production schedule (MPS) to determine the items for which
purchase requisitions and production orders must be released.”
More engineering definition: “A list of all the materials needed to make one production run of a
product, by a contract manufacturer, of piece parts/components for its customers.”

STRUCTURE OF THE BUILD OF MATERIAL

U have to build one bike, and u have the

components to create it. Here is not reported
the TIME u need to produce it and run the
activity. For example, the time need to put
together the first three objects.
The basic components: how many
components are need and the time need to
build / buy the material.

This is a tree structure, with the number of

items needed for each category and there is the indication
of the LEAD TIME: time needed to build component number
one or have it from the suppliers. It indicates the availability
times of the subassembly once the components have
become available

EXAMPLES:

1 day= means that we need 1 day to assemble the

three components in the first level
#n= number of components needed to build the final
product and then we have the indication of the LEAD
TIME needed to build / have the components.
In order to build the component 2 we need sub-
components.
The MRP systems, relying on the information available on the BoM, allows us to compute using a
backward procedure, when the components must be necessarily available.

EXMPLE 1:
Suppose one finite item is due at
(calendar) date 10. When do we need to
start the production of all components
needed? By looking at the BOM, we must
schedule the production to satisfy this
obligation.
6
C1= component 1 it takes 5 days
C2= component 2 it takes 4 days, so we need to start producing at number 6
The days counted for the sub-components need to start to be counted from the 6 day and before.

The black number represents the final

demand to be delivered.
The blue number is the production
order for one item to be assembled at
date 10.
The red and the green numbers
represent the starting point of
production for the single items (first
and second level).
SUMMARY OF THE EXAMPLE
To have the product ready at date 10, we must:
• start working at C1 on day 5 (because to compose it, once the components are available,
it takes 5 days);
• start working at C2 and C3 on day 6.
In order to start working at C2 on day 6, we must:
• have the component C2a available at day 3,
• C2b at day 4.
• C3a and C3b on day 5.

Suppose now to have the following demand: 1 item per day from date 10 to date 13. How is the
picture changing?

And so on:

• The deadlines of the MRP must be set sufficiently far in time so that they can be met;
• the lead times of the suppliers must also be considered (raw materials may not be available);
• the process analyzed so far assumes infinite availability of production resources and does not
consider warehouses (materials already available)!

7
POSSIBILE SOLUZIONE ESERCIZIO

09/02/2023
SUMMING UP
 An MRP is more sophisticated compared to what suggested from the previous example.
 As seen, it must consider three important dimensions: demand, technical aspects of
production, availability of materials.
 In the next example, we consider a situation in which all the three inputs seen before are
considered: MPS, BoM and Inventory
For a given time horizon under analysis, ranging from 1 (initial date) to T (delivery date), we have:
 MPS: Demand (planned and forecasted orders). It is the core at the MRP system
 BoM: Technical details about the release of the final product (and its components)
 Inventory: Stock available at date 1 for all components required.

In order to produce C3 we need at least 5 day (4+1

form C3b)
FORMULA : C3=LTC3 + max {LTC3, LTC3b}
MORE GENERAL FORMULA:
C2= LTC2+max {LTC2a, LTC2b} = 4+3 (three is the
maximum between the two components
FINAL PRODUCT= LTFP + max {LT1, LTC2, LTC3}

S1= the set of successors (the set of product i)

EX: S1 = (2,3,4)

GENERAL FORMULA: β i=¿i +max { β j : j ∈ Si }

What is missing in the previous example? The INVENTORY

Ex: LD= 2, demand=10, delivery =5, on- hand =3

8
 GENERAL FORMULA: I t +1=I t – Dt +1 +Order t +1

Given available stock at date 1, we can compute availability at the end of date 2 for each
component:

Recursively, we can compute all the remaining projected availabilities for t=3, 4, …,

Given available stock (on hand) at date t and the Gross Requirements (demand) at date t+1, we can
compute the Net Requirements:
- Without any contractual obligations, the minimum possible! This is exactly the Net
Requirements. This strategy is called Lot-for-lot.
- Under contractual obligations, the minimum quantity allowed by the contract
matching the Gross Requirements. Usually, we have a minimum order quantity with
suppliers. This strategy is called Minimum order quantity

SUMMARIZING:

We do demand – inventory if the

result is positive

9
EXAMPLE 2:
A simple example of the use of the flow equation and of the planned order receipt formula. For
this simple example, we do not employ a proper BoM. We only look at availability (stock) and we
order what needed.
Inputs are as follows:
• Demand to be satisfied: 470 items for product A.
• Delivery date: 6
• Lead time for product A: 2 periods (lot for lot)
• Inventory at date 1: 70 items in house for A.
• The BoM is trivial, simply: product A, lead time 2 weeks

EXAMPLE 3:
A final product, called «B», has to be delivered at date 6
Inputs are as follows:
• Scheduled demand to be satisfied: 470 items for B
• BoM for B (below)
• Inventory at date 1:
• 70 items in house for B
• 35 items for C (component)
• 280 items for D (component)
• If we order for C, we have a minimum order: 2000
• If we order for D, we have a minimum order: 5000
B BoM: MRP:

The MRP is
based on a
simple and clear
principle: when
needed, order the minimum quantity necessary to satisfy demand! This quantity corresponds to
the Net Requirement if no contractual obligations with suppliers are in place. Note that, in
planning orders, we did not consider any optimization scheme such as the minimization of a cost
function. We only looked at the requirements, the availability and the BOM.
10
14/02/2024
U want to order the minimum item u need as late as possible. U try to fulfill demand trying to have
0 inventory, u are not worry about how it cost to order the material, but u think how much u need
to store them. U need to consider the tradeoff between storage of materials or order them.

6.THE EQ (Economic quantity) MODEL

PIZZA EXAMPLE
d (demand) =10
1) Q=10 (u buy 10 pizza)
2) Q=20
3) Q=5

h= unitary holding cost per period

A= reordering costs cost u pay to go to the store
Q= lot size
C= total cost per unit of time

ASSUMPTIONS OF THE MODEL:

- Demand is constant and uniform this assumption is questionable
- Reordering cost are constant over time
- Holding and storage cost are constant over time
- The whole lot is delivered at the same time

We want to minimize the average of period cost:

d/Q how many times u go to the store on

average
We need to find a trade off between:
- HOLDING COST increase
with Q
- REORDERING COST  decreases with Q
In order to find the minimum:

This is the OPTIMAL Q=

√
2 Ad
h
The largest is the demand, the more u buy.
The largest is A, the largest is Q. If H
increases, Q decreases

Now we want to define the OPTIMAL COST:

11
 OPTIMAL COST:
√ 2 hAd
This means that, under the
optimal conditions, holding
costs are exactly equal to the
reordering costs.

GRAPHICAL
REPRESENTATION OF THE TOTAL COST

EOQ, TYPICAL SITUATION

Q∗¿
OPTIMAL REORDER PONT  when to reorder to minimize the total cost T*= d
¿

EXAMPLE
Data:
d = 100 kg/day,
12
h = 0.15 Euro/Kg,
A = 30 Euro.
Compute the optimal reorder point T*.
First, we need to determine Q* (optimal lot size), so to minimize C (total cost per unit of time).
Using the previous formulae, we have:

7.HEURISTICS
It is a rule based more on intuition; u don’t want to find the optimal strategy but the best thing
that u can do with the local one.
Each heuristic tries to focus on one KPI.
The problem of lot sizing (how much should I order or how much should I produce?) requires to
find the best compromise between reordering costs (or set-up costs in case of production) and
holding costs.
Differently from EOQ, demand is not constant in time!
Heuristic methods (require exact knowledge of the short-term demand):
• Silver-meal Used to find a solution which (locally) minimizes the average cost per
period
• least unit cost Find a solution which (locally) minimizes the cost per unit of item
• part period balancing Find a solution which makes the holding costs equal to the
ordering costs
Exact methods (require exact knowledge of the long-term demand):
 Wagner-Within.
SILVER MEAL
h= holding cost (unitary)
k=fixed reordering cost
demand is not constant, demand at each period: r1, r2, r3, r4,r 5
Now the model is discrete in time. We have to choose the length of the reordering period
j= number of weeks I want to cover with my first order

k +hr 2+2 hr 3+… ..+ ( j−1 ) hrj

GENERAL FORMULA= G(j)= j

13
2hr3= the weeks needed to store the material needed at the 3th week.

N.B: The reorder policy is set up in a way to satisfy the demand for the next j0 periods, where j0 is
the first local minimum. This means that j0 is the smallest number such that:

EXAMPLE 1
K=1000 H=1 r2= 210 rj=252 for i > 2
G(1) = 1000,
G(2) = (1000 + 210)/2 = 605
G(3) = (1000 + 210 + 2 * 252)/3 = 571.3 G(4) = (1000 + 210 + 2 * 252 + 3 * 252)/4 = 617.5
The local minimum is at j0 = 3

EXAMPLE 2
Suppose that:
G(1) = 450
G(2) = 500
G(3) = 250
G(4)= 600
The local minimum is at j0 = 1 WE NEED TO CONSIDER THE FIRST LOCAL MINIMUM

LEAST UNIT COST

Consider the following assumptions:
K: fixed ordering cost;
h: unitary holding cost per period;
ri : demand to be met during the i-th period;

We can compute G(T), the unitary cost (per period) if the current order spans the next T periods

The reorder policy is set up in a way to satisfy

the demand for the next j0 periods, where j0
is the first local minimum. This means that j0 is the smallest number such that:

EXAMPLE
K = 1000, h = 1, r1 = 189, r2 = 210, ri = 252, i > 2
G(1) = 1000/189 = 5.3,
G(2) = (1000 + 210)/(189 + 210) = 3
G(3) = (1000 + 210 + 2 * 252)/(189 + 210 + 252) = 2.6

14
G(4) = (1000 + 210 + 2*252 + 3*252)/(189 + 210 + 252 + 252) = 2.7
The local minimum is at j0 = 3 we order for 3 period

15/02/2024
PART PERIOD BALANCING
Consider the following assumptions:
K: fixed ordering cost;
h: unitary holding cost per period;
ri : demand to be met during the i-th period

We can compute G(T), the holding cost if the current order spans the next T periods.

The objective is to find the j that minimize the difference

between j and k

The reorder policy is set up in a way to satisfy the demand for

the next j0 periods, where j0 is the first local minimum of the
quantity.

EXAMPLE
K = 1000, h = 1, r2 = 210, ri = 252, i > 2

G(1) = 0, |0 - 1000| = 1000

G(2) = 210, |210-1000| = 790
G(3) = 210 + 2 * 252 = 7c, |714-1000| = 286
G(4) = 210 + 2*252 + 3*252 = 1470, |1470-1000)| = 470

The local minimum is at j0 = 3.

We are not sure that this is the global minimum.

REMARKS:
• In this example, all heuristics provide the same result. Silver-Meal is generally the one that
provides the best results, but Part Period Balancing is the most used because of its simplicity.
• In the examples, only the size of the first order was calculated and the data were used for the
next few days. In reality the heuristics must be rebalanced from the first day for which there are no
stocks available.
• The order is computed to exactly meet the demand (future orders) for an exact number of
periods. It can be proved that this is always the best heuristic under these assumptions.

8. THE WAGNER WHITHIN

This is an exact method that requires the exact knowledge of the demand for the entire horizon
with respect to which we want to optimize.
15
Consider the following assumptions:
K: fixed ordering cost;
h: holding cost (per period);
ri : demand to be satisfied at period i;
cij: ordering cost and holding cost related to a lot that satisfies the demand for periods
ranging from i to j-1

Let T be the length of the time horizon we are interested in. We want to determine the sequence
of costs cij that minimizes the sum:
This problem corresponds to the identification
of the minimal path along an acyclic network. It can be solved by means of a mathematical tool
called dynamic programming.

We find the minimum compared to

BEST STRATEGY: buy from C13 to

C36. The optimal strategy is: at time
1 buy for period 1&2 and at time 3
buy for period 3,4 & 5
EX N. 2

16
 BEST STRATEGY:
buy from C14 to
C47. The optimal
strategy is: at
time 1 buy for
period 1-2-3 and
at time 4 buy for
period 4-5-6-7.

16/02/2024
9.BUSINESS PROCESS
Davenport (1993): “a structured, measured set of activities designed to produce a specified output
for a particular customer or market; […] a specific ordering of work activities across time and place,
with a beginning, an end, and clearly identified inputs and outputs: a structure for action".
Rosemann (2001): "the self-contained, temporal and logical order of those activities, that are
executed for the transformation of a business object with the goal of accomplishing a given task".
Sharp and McDermott (2001): "a collection of interrelated work tasks, initiated in response to an
event that achieves a specific result for the customer of the process".

MAIN WORDS:
"Achieves a specific result"
• the result can be physical goods or immaterial services;
• it should be possible to identify, asses and evaluate the results.
"for the customer of the process"
• any process has a customer, who can be internal or external to the organization;
17
 • the customer should be asked to provide feedbacks to the realization of the ongoing
process.
"initiated in response to a specific event"
• any business process originates from a event;
• the event can be identified with the request put in place to obtain the final result of the
process.
"work tasks"
• the Business Process collects a series of several tasks, clearly specified, for which a
responsible person (or group of persons, unit, or division, …) can be identified;
• it is NOT formed by a list of random (or unclear) tasks;
• each task can be formed by several finer steps, clearly specified, for which a responsible
person (or group of persons) can be identified.
"a collection of interrelated …"
• all these steps or tasks are not necessarily sequentially ordered; they could be possibly
carried on at overlapped times. There is the need of clearly understand the logical structure
linking the single tasks to tackle in a efficient way the business process.

DEFINITION OF BUSINESS PROCESS A business process (BP) is a collection of activities linked

together (in time and space) and implemented by using the resources of a company (goods,
human resources, capital, …). Starting from inputs, it produces an output (goods and/or service)
for a well-specified customer. A business process is a complex management cycle based on
multiple (inter-functional) skills.

The actual configuration of a BP is determined by a series of cross-functional properties:

• flow of activity (process cycle);
• organizational structure (defines who is responsible for the BP and who is in charge to run
the activities);
• human resource skills;
• technology used (decision support systems, business intelligence)
• measurement and control system in terms of costs, time, quality and services.
BP are inter-functional activities, as they span multiple business functions

If we have a process, there is a starting

process, many functions are involved and
the process flows in time and stay for a
certain period in one function and there
group, individuals and QIT involved to see
if the process perform well or not.

An example: Business Process: Production

Planning System we have to plan production where many function are involved:
1. Sales and production forecast (Sales function) forecast demand and future order
2. Budgeting (Sales, Marketing, Accounting functions)
3. Production planning (Operations)
18
 4. Material requirement planning (MRP) (Operations & Logistic)
5. Inventory management (Logistic)
6. Resource planning (Operations, HR, Logistic)

The basics steps:

1. BEGINNING (inputs)
2. TRASFORMATION PHASE
3. OUTPUT that can be a good but also
services etc.
There is a flow that goes in the right and in the
left.

10.FROM MRP TO THE “NEW NORMAL”

PRODUCTION PLANNING
As seen before it involves many functions.

MRP (1975) is an old system, still used as a benchmark. Now the world has change.

Here we can see how the company

environment has changed.

Consider the following data and the relative

solution based on the classical MRP
methodology. Lead time is 2.
If the MRP works correctly the available
material is 0 the material arrives in time
exactly when u needed and this is the ideal situation.
Suppose that, for external reasons, the flow is interrupted for two consecutive months. This means
that two shipping, scheduled at Month 3 and Month 4, will not arrive on time.

19
We can see that the orders places at period 1 and 2 do not arrive and the order receipt is= 0. The
MRP reacts because we have to fulfill the demand, so we place a higher order to what u expected
before in order to adjust the situation. In this calculation we have not considered that sooner or
later our material will arrive because we ordered it and we paid for it. In fact we discover that in
the month 6 the materials arrive now we can se that there is an ECCESSIVE STOCK (400+630)
As a consequence, we pass from understock to overstock, compromising the pace of the inventory
and the ideal zero-stock situation.
Although very basic, this example shows us a typical effect of warehouses, the so-called bullwhip
effect = An extreme change in the supply position upstream in a supply chain generated by a small
change in demand downstream in the supply chain. Inventory can quickly move from being

backordered to being excess.

This may cause rapid oscillation between two
extreme situations: from understock to
overstock. The pace of the inventory is broken.
Moreover, the scheduler overreacted to the
lack of material causing the overstock in the
warehouse. This trait is called nervousness.
According to APICS, the nervousness can be
defined a: «the characteristic in an MRP system when minor changes in higher level records or
master production schedule cause significant timing or quantity changes in lower-level schedules
or orders»
How to limit the impact of nervousness on the pace of the inventory? BUSINESS PROCESS
REENGINEERING=> Business Process Reengineering (BPR), with the support of Information
Technology and DSS, allows companies to improve their processes; not only to automate the
existing routines but also to enable new ways of doing business.

We should “reengineer” our businesses: use the power of modern information technology to
radically redesign our business processes in order to achieve dramatic improvements in their
performance.

Innovation process: BPR is composed by a series of actions through which a company moves from
situation A (standard way) to situation B.
Organizational innovation: BPR acts on the company at the organizational level and not at a
production level. The BPR modifies "how a company operates" and not "what a company
produces".
Innovation focused on BP: the BPR acts mainly on business processes, i.e., on the flow of activities
carried out by the company.

According to Hammer:

20
 • organizational change: to improve the performance, it is necessary to integrate
technology innovation with organizational innovation;
• disruptive innovation is essential in «mature markets» in which competitors are able to
offer customers valid substitutes. Competitive advantage is related to performance
improving, cost reduction, speed in fulfilling demand, quality of the service;
• radical redesign: reengineering is not a simple incremental improvement; it is rather a
reinvention of the way of doing business; a review of the rules and hypotheses that are at
the origin of the way to manage the company

Different phases to run the BPR:

1. analysis of the present situation (AS-IS) what are the problems, the risks involved, who
are the people involved?
2. identification of the main problems; comparison with competitors; market analysis and
other external factors;
3. redesign (TO-BE);
4. implementation
11.DEMAND DRIVEN MRP:
a modern data-driven and prudential approach to inventory
BRP: from MRP to DDMRP
We present a possible way to overcome some weaknesses of the classical MRP. This method is
called Demand Driven MRP (DDMRP).
DDMRP proposes to drastically revise the Production Planning System by implementing the
following innovations:
• Displace strategically warehouses along the BoM (decoupling we try to put points in
which we cut what is before and what is after).
• Use technology and data to set optimal levels of the inventory (buffer levels) this is a
new heuristic on how to deal with inventory.
• Use semi-automated controls to dynamically adjust inventory levels and plan orders
based on real data.  this should avoid bullwhip (with decoupling) and nervousness (with
the buffer levels)
• Consider variability of demand and service levels.
DECOUPLING
Oscillations of warehouses due to supply and demand variability are detrimental for the pace of
the inventory. Variability can be managed, but it will never be totally eliminated. One possibility to
limit the nervousness and the bullwhip effect of the MRP is to decoupling supply and demand.
Decoupling means: «Creating independence between supply and use of material. Commonly
denotes providing inventory between operations so that fluctuations in the production rate of the
supplying operation do not constraint production or use rates of the next operation.»

DECOUPLING POINTS AND BUFFERS

Decoupling points displaced along the BoM allow to create zones of influence and avoid
propagation of waves. Determining their optimal location is a crucial and strategic decision and
they represent a point along the BoM where to put a warehouse. Once decoupling points are set,
each of them is endowed with a buffer (the modelling of a warehouse). A buffer is defined as the
quantity of stock needed to optimally decouple supply and demand. Buffers are quantities of
21
inventory that provide customers with reliable availability and, at the same time, enable
aggregation of demand orders, creating a more stable and efficient pace of orders.
In this image we can see that in the first part,
we have a long BoM and at some point we
have problems related to the bullwhip effects.
In order to avoid it, we can put some
decupling points which interrupt the flow. So
what is before and what comes after that
point is independent and the fluctuation are
much more smaller.
The buffer is represented by the green, yellow
and red image.

DDMRP: history
A brief history of Demand Driven MRP.
• 2011: a DDMRP pilot is proposed by Carol Ptak and Chad Smith as an alternative formal
logic of planning and control. The Demand Driven Institute
(http://www.demanddriveninstitute.com) is founded.
• 2018: The Demand Driven Institute launched its expertise to support the development of
the methodology.
• 2022: Many companies use this methodology to run inventory management (e.g.,
Michelin, CCBA – Coca-Cola bottling partner, NLMK Russia, Shell)

DDMRPR: 5 consecutive phases

DDMRP rests on 5 consecutive phases, based on the idea of Position, Protect and Pull.
1. Strategic Inventory Positioning: where to position decoupling points (Position).
2. Buffer Profiles and Levels: how to
optimally dimension buffers (Protect).
3. Dynamic Adjustments: how to modify
buffer levels dynamically (Protect).
4. Demand Driven Planning, how to
generate optimally the scheduling of
orders (Pull).
5. Visible and Collaborative Execution, how to manage the flow of materials and information
(Pull).
12.DEMAND DRIVEN MRP: dynamic adjustment of buffers
BUFFER SETTING
Once the decoupling points have been placed within the supply chain, how do we set the buffer
levels? An interesting question is: is on-hand stock an asset or a cost?
A cushion level of stock may serve in case of extra (unexpected) demand or in case of (unexpected)
delays in the supply chain. On the other hand, inventory is costly.
Summarizing, an optimal level of available resource is vital. DDMRP helps the decision maker to
set optimally the buffer levels. To this aim, different thresholds have to be determined.

22
To avoid understock and overstock, the methodology proposes a strategic planning based on three
zones:
This image represent the quantity of materials and it is
increasing from the red to the green. The line between
green and yellow, will give us an idea of when it is time
to reorder. The line between yellow and red is the
cushion level, if u go below that level, it is time to think
that there are problems.

THE PACE OF THE FLOW

The buffer is the heart of the planning system in DDMRP. All relevant information relating to
demand, supply and stock availability are combined in the buffer to produce a net flow position
(NFP) that determines the generation of the supply order. The net flow position considers both on-
hand and on-order materials. In classical systems, on-order quantities are neglected by the
decision maker, and this may cause nervousness.
Net flow position is determined as:
NFP(t) = on-hand(t) + on-order(t) – demand(t)

The buffer setting depends on few and clear dimensions:

• DLT: Decoupled lead time (the expected lead time). DLT is a constant integer number. It is
the time between two buffers, it is one portion of the lead time.

• ADU: Adjusted daily usage (real short-term demand). (similar of the heuristic). It is an
average. It is not constant in time.

ADU(t) = (d(t) + d(t+1) + … + d(t+τ))/ τ

where τ is the number of periods considered.
U need to consider more or less the double number of the lead time.

• RF: Risk factor (depends on variability of demand and lead times and on the risk aversion
of the decision maker). RF is a constant, usually a number 0 < RF < 1.

Based on the previous quantities, we can determine the buffer thresholds.

We can determine 3 thresholds: the green, yellow and red. They

will adjust in time depending on the data.

In all the three formulas there is always a common part:

ADU(t)*DLT which is what we want to keep in our inventory.
RED CUSHION LEVEL it is the safety zone

23
YELLOW when we want to reorder and u want to reach the green level.
GREEN It is called “top of green” and it
is the highest possible level we want to
reach for the NFP in our inventory. When
it reaches the yellow part, we want to
reorder in order to reach the G.
U need to consider the inventory and the
shipping (what it is arriving).
The reordering rule is as follows:
it means that I put an order only
when, the NFP goes below the
yellow. And I reorder exactly what I
need to reach the green level. The NFP moves between the yellow and the green line in the
following graph, and automatically, this will make the available material (black), ranging between
the red and the yellow. We want to stay above
the red line, if we go below, it means that there
are problems.
Suppose that the situation changes, how can I
change the dimension of the buffer? The
threshold are not fixed, they adjust to the
environment.

21/02/2024
QUESTIONS
What is the BP under analysis? PRODUCTION PLANNING
What is the AS IS that we consider? THE MRP
What are the main divers of change? Lead time risks and unexpected demand

22/02/2024
13.PLANNING LARGE PROJECTS
What is the link between Project management and business process management?
Project and processes have different peculiarities and different use in the context of management.
They also share some common traits, especially from the point of view of their representation and
their measurements.
PROJECT= is a well-defined series of tasks commissioned by a customer with pre determined
contractual specifications and standards.

EX: development of a new software for an improved business process or a construction of

an bridge, build or a ship, or the organization after a natural disaster or a setting of the new
marketing campaign.
Definition form the Project management institute: Temporary endeavor undertaken to create a
unique product, service, or result”.
24
TEMPORARY NATURE the project has longevity (short or large) it has a beginning and an end
OUTCOME OF THE PROJECT unique product, service or result that cannot be met with an
ongoing working effort
The project is NOT a routine operation, but it is a SPECIFIC SET OF OPERATIONS designed to
accomplish a singular goal.

THE THREE MAIN DIMENSION IN PM

The project management is a blend of methodologies:
application of knowledge, skills, tools and techniques,
whose goal is to improve the ability to plan, organize,
implement and control different cross functional activities
balancing the project requirements of time, costs and
quality in order to realize the scope of the project

TECHNIQUES FOR PM
1. Critical Path Analysis
a) Identification of activities and their attributes
b) Relationships, links… diagram representation (on nodes and on arches)
c) Setting durations: critical path analysis (on nodes and on arrows)
2. Basics of Project Risk Management (PERT analysis)
a) How to assess risk measurements
b) Estimate the probability of adverse scenarios

All the main ingredients to consider when u have a complex project: such as a construction/
marketing campaign:
• Construct the scheduling of the involved activities, identifying ties among them.
• Estimate duration and cost of the single activities in order to forecast the total time taken
to complete the whole project and the total cost.
• Identify the best scheduling in terms of times and costs.
• Perform risk analysis related to the duration of the project.
• Find out if you can reduce time by adding resources optimally, i.e., limiting the impact on
costs. We will explore some of the techniques used to deal with all these goals.

13.1 CPA ANALYSIS- activities, durations, and ties

CRITICAL PATH ANALYSIS (CPA)
1. Break down project into simple activities.
2. Estimate duration of each activity.
3. Decide on the logical sequence of activities and the ties among different activities.
4. Draw an informative diagram.

A simple example:
Company ABC is launching a new product on the market.

25
The marketing department has agreed on the 9 activities that need to be done before the national
campaign is launched.
The marketing manager has now to propose to the CEO the best scheduling for the activities.
Now we need to schedule these activities: ACTIVITY DESCRIPTION

If u sum al the duration u have 39 which is NOT the

duration of the project, because some activities can be
run at the same time. For this we need to add a new
information: the IMMEDIATE PREDECESSOR
activities which we know must be run before starting a
new activity.

A has to be run before B.

E needs B and C to be completed

GRAPH REPRESENTATION
We will represent the flow of the project through an oriented graph. There are two approaches:
1) activities on nodes (AON);
2) activities on arrows (AOA).
In this class we will rely only on AON method: Activities are represented as nodes on the graph.
Arches represent relations among them: given two nodes x and y, we represent the arch (x, y) if
activity x is an immediate predecessor of activity y. This representation is also called Critical Path
Analysis.

NUMBERS: duration
LETTERS: activities

Exercise:

26
Are we sure the project can be completed in a finite time?
A little bit of Maths…
Theorem: A finite oriented graph, without cycles, has (at least one) initial node (a node without
entrant arches) and (at least one) final node (a node without exit arches).
Proof: We now show that the graph has final nodes. Take any node xQ1. Two possibilities apply:
either it is a final node (and we can stop) or there exists at least one arch departing from xQ1. In
this latter case, call xQ2 one terminal node for the arch departing from xQ1. If xQ2 is a terminal
node, we stop. Otherwise, call xQ3 the arrival node of the arch departing from xQ2. This procedure
ends in a finite number of passages since, by assumption, the graph is finite. The proof of the fact
that the graph has a initial node is equivalent.

Remark
If correctly analyzed, a project should be represented by a directed, finite graph without cycles. In
this case, there must exist initial and terminal nodes and the project can be realized in a finite
time. By convention, we add to the graph two auxiliary activities:
• activity START preceding all the (initial) nodes;
• activity END following all terminal nodes.

CPA ANALYSIS- Total duration of the project

27
GRAPHICAL METHOD
STEPS:
1. Make a forward pass through the network, calculating the earliest start time (EST) and the
earliest finish time (EFT) for each activity.
2. Make a backward pass through the network, calculating the latest start time (LST) and the latest
finish time (LFT) for each activity.
3. Determine the minimum time that the project can be completed in, the critical paths and the
floats of all the non-critical activities.

ACTIVITY- START AND FINISH TIMES

1. FORWARD PASS

2.

BACKWARD PASS

The activities that can not be delated are the ones where the earlies start date and the latest start
date have the same number. For this activity C is critical, D is NOT critical we can take some delay
equal to 1 week (8-7). This period is called SLACK of the activity D.

What are the critical activities?

28
Critical activities are those activities that cannot be delayed: a delay would increase the total time.
The float (or slack) of an activity is the difference between the latest start time and earliest start
time i.e. LST – EST.
A: 0; B: 1; C: 0; D: 1; E: 0; F: 1 G: 0; H: 0; I: 0
Critical activities are those activities whose float is zero: A, C, E, G, H, I

The critical activities form a critical path. The critical path is the longest (in terms of duration!)
path through the network. (A-C-E-G-H-I). There can be more than one critical path BUT THEY ALL
HAVE THE SAME DURATION. Any delay to a critical activity will automatically delay the project. If
you delay a non-critical activity by more than its float the project will again be delayed.

COMPUTING THE DURATION OF THE PROJECT

Call X= {0, 1, …, n} the set of nodes, where 0 denotes START and n denotes END.
For all k = 1, 2, …, n, call Pk the set of all the immediate predecessors for activity k. For sure, Pk is
not empty; moreover, h∈k if and only if the arch (h, k) is part of the graph (i.e., activity h is an
immediate predecessor of k).
For all k=0, 1, 2, …, n-1, call Sk the set of all the immediate successors of activity k. For sure, Sk is
not empty; moreover, l∈Sk if and only if the arch (k, l) belongs to the graph (i.e., activity l is an
immediate successor of h).

The aim is to find a way to execute all the activities as soon as possible, considering all the ties
among them.
Call [ak, βk] the time window in which activity k is executed. We need to
determine Dk e Ek for all k. Obviously, if we denote by dk the duration of
the k-th, activity, we have:

βk is called earliest finish time (EFT) for k, meaning that activity k cannot be completed before this
date.
ak is called earliest start time (EST) for k, meaning that activity k cannot start being executed
before this date.

EFT (βk) is computed as follows:

29
EXAMPLE: suppose we have 2 activities:
βk is the max between β1 and β2, but we have to add the duration of
the project

Activity k can start as soon as ALL the predecessors are completed; put differently, when the latest
among all the immediate predecessors are completed. We also have to add dk.
EST is obtained as:

Note that, βn is the total duration of the project.

30
EXAMPLE

EXMPLE: graph
ASSUMPTIONS that we need to be right:
- The graph is ORIENTED there are arrows pointing from one activity to another
one
- WITHOUT CYCLES we don’t have this:

- There is a STARTING and TEMINNALE node

31
Algebraic solution:

32
13.2 PERT ANALYSIS
PERT= Program Evaluation and Review Technique
Given a large project, how to control for the risks of exceeding a certain duration?
• What’s next: perform some risk analysis related to the duration of the project.
• Again, it is convenient to use an “activity-based perspective”: split aggregate risk into single
activity's risks
HANDLING UNCERTAINTY I N TIME ESTIMATES
• Can we trust time estimate?
• We perform a Schedule Risk Analysis
• What is the simplest distribution (to be estimated) which is not uniform?

3-POINT ESTIMATES
• The simplest one is the triangular distribution. It is based only based in three estimates that
have very clear interpretations.
• Choose a left threshold (best case) and a right threshold (worst case) + a «mode» (most likely
event)
• Eg. Activity B estimated time was 5 weeks.
We now better specify it with a 3-point time estimate:
– Optimistic duration: to = 24 days
– Pessimistic duration: tp = 36 days
– Most likely duration: tm = 30 days (5 weeks)
VISUAL REPRESENTATION OF THE density function

x = duration
y= likelihood of the durations
Durations smaller than 24 days and bigger than
36 have a likelihood equal to 0

Average

33
The problem is that, now, we are putting a lot of weight on extreme events! (to and tp contribute
to the mean exactly as the most likely time).
A better guess for a simple and more accurate distribution has been made: PERT distribution

According to PERT theory, we than compute the

average time for the
activity:
In this case we have a 4 that multiply the most likely, in
order to put more mass on that event.
We can even estimate the risk involved in this
calculation. An approximation of the standard deviation (error) is in fact:

It gives a simple measure of variability. The standard deviation is a measure of variability but also
of risk
How to use these new data?

TOTAL DURATION REVISITED

 Create a new table collecting the data (tp, to, tm).
• Compute the average time for each activity and its SD.
How to compute the aggregate average duration?
• Suppose that the critical path is formed by activities 1, 2, …, n, whose durations are X1,
X2, …, Xn (these are random variables)
• We recall from statistics that, if T=X1+X2+ … +Xn, then E[T]=E[X1]+E[X2]+ … +E[Xn].
• Therefore, the total duration T of the project is a random variable, whose average is given
by the sum of the average times of the activities on the critical path.

EXAMPLE: Given the data about 3-point estimates for durations of single activities, compute the
total duration of the project. In solving this problem, consider predecessors as in Example 1.

After doing the mean of

each activity, we need to
use mean in order to find
the new duration of the
project. It will have the
same structure but a
different duration:

34
What about the aggregate risk?
 Consider the previous situation where a critical path is formed by activities 1, …, n, with
durations X1, …, Xn.
 If T=X1+X2+ … +Xn, and X1, X2, …, Xn are statistically uncorrelated, then
Var[T]=Var[X1]+Var[X2]+ … +Var[Xn].
 Therefore, we have an estimate of the total variance of the duration of the project. Recall
also that
Var(X1)=[SD(X1)]^2
 The total standard deviation of the project:

SUMMARIZING
Find one of the critical paths and compute the total (average) duration on it:

where 𝑡𝑖 is the average time for activity 𝑖 on the critical path.

• Compute the total SD:

where 𝑆𝐷𝑖 is the SD for activity 𝑖 on the critical path!

• In case of multiple critical paths, the variances can differ. Choose the one with the highest
variance!

ESTIMATION OF THE LIKELYHOOD OF EVENTS

• Back to Example 1, suppose you want to compute the probability that the total time exceeds x =
28 weeks:
35
Prob(𝑻𝒄𝒓𝒊𝒕 > 28)

• Strong assumption: the number of activities is large enough so that a proper «limit theorem» is
in place. We approximate the duration using a Gaussian estimate.
• More precisely, we state that:

where Z is Gaussian with mean 0 and variance 1.

Formal derivation:

Then we do an
approximation where we say that the previous probability, is approximal equal to the probability
that a normalized random variable Z, is bigger than the same quantity that we have on the
previous equation. And then , form statistics we can remember that when we have standardized a
random variable, the probability that Z in bigger than a certain value, is equal to the probability of
Z being smaller than the opposite value:

Then we need to remember:

 Φ is the Cumulative Distribution Function of a Standard Normal Random Variable. It cannot
be computed in closed form, but we have numerical methods.
• Using excel, Φ
is computed
with the
command NORMSDIST
Suppose you want to compute the probability that the total time exceeds 28 weeks. In days, we
have x=168 (28 weeks = 28*6 (working days) = 168 days). What is the probability that the duration
of the project exceed 168 day?
Consider the quantity

t= average time of the project

x= 168 the figure we want to check
SD= standard deviation

Using excel, compute NORMSDIST(𝑧) .

36
 we sum al the
variances= 16, we
compute the
standard deviation
of it (4).
Tcrit are and the
critical path we
founded before.
Noe we have to
estimate the
probability of
exceeding 27/28
weeks :

EXERCISE
Given the table below of activities, durations and predecessors, identify the critical path and its
expected total duration. Compute also the total standard deviation of the duration of the project.
STEPS:
1. Compute the average, the SD and rhe variance for each activity
2. Run the CPA analysis for the total expected duration of the project and the standard
deviation of it

37
29/02/2024
PART FOR THESIS
MACHINE LEARNIG APPROACH u have a problem and u try to use the experience accumulated in
the past, an u train the algorithm to suggest u the best way to do things. And this is a cycle.

Depending on the data that u have u need to find the correct β.

In questa lezione abbiamo fatto esercizi

01/03/2024
MODELLING DISTRIBUTED SYSTEMS: PETRI NETS
The traffic light can be seen as an operating machine with three states and three transitions.

38
 The states are circles, the rectangles
are actions, activities, movements. The
two of them are related.
Now we recognize 3 states/places
(circles), 3 events/transitions
(rectangles), and also the connections
(arrows) among them. We will call this
kind of system a distributed system
represents a (finite) collection of
business activities (tasks, operations, services, …) orchestrated into a precise logical sequence.

Our goal:
• Introduce models to visually represent a DS.
• Provide support to understand the fundamental characteristics/properties of the system.
• Study the dynamics of the system.

Petri net is a tool to visualize, validate and simulate the dynamics of a DS and any complex discrete
system (supply chains, production routines, …).

EXAMPLE: u receive claims from customers and u process them

The process can take 2

different paths.
There are 2 different types
of objects, and there is
something that stay in time,
the token.

EXAMPLE 2: it is a bicycle line production.

The entry arrows

into the first
rectangle are the
input, what u need
for the
specific
activity such as
worker or a spot
available in the
machines.

39
If we trigger the activity in the rectangle we use the worker to do something.

The basic components are states (CIRCLES) and change of states or transitions. The general
principles of Petri nets theory are the following:
• states and transitions are distributed entities of the system;
• each transition is related to a specific (and well determined) area of influence in the
system;
• a transition can take place only if the current situation allows for it

EXAMPLE
For an information system, possible states are: the state of accounting documents, the situation of
the current production activity, the situation of warehouses ...
On the other hand, possible transitions are: the issuance of an invoice, the starting of a production
line, the withdrawal of a quantity of raw material from the warehouse ...
A transition changes the state of the system. However, the issuing of an invoice relates to some
accounting documents without directly influencing the start of a machine tool. As said, each
transition has its own area of influence, although the universe of the company is still a unitary
entity to be studied as a unique dynamic system.

PRODUCER- CONSUMER SYSTEM

We give labels to each circles (states) and rectangles (transaction):

This is the STATIC

REPRESENTATIONS

Now we add the situation given by the token:

Conditions active at a certain stage (marking/tokens)
At this stage, the producer
is ready to produce and the
consumer is ready to
receive.
This situation is called
marking it is the picture
with a situation
After c1 is done we cancel the token there, we move to e1 and then to c2. According the
information, at e3 we can give the pizza to the customer in c5 and also produce another pizza in c1

40
We want to understand possible evolutions of marking, at
some point there will be an equation describing the
evolution of the marking:

Note that, e2 causes the activation of two tokens. They

signal that the producer is ready to produce and the deposit
of completed items is full.
Here we can find the all possible solutions

CONDITIONS-EVENT NET
A system of condition-events (C/E) is defined by a directed and bipartite graph G = (N, A), in which
the set of nodes N is the union of two finite and disjoint sets named, respectively, C (conditions)
and E (events). To each event we associate two groups of arches: input arches (I) and output arches
(O):

(INPUT: from a condition to and action OUTPUT: from an action to a condition)

with the following rule: every condition is connected to some events and every event is connected
to some conditions (there are no isolated conditions or events).
A marking of the graph is a function

where 0 and 1 denote the absence or the presence of

the token within the condition, respectively

A marked net is defined by R = (C, E, I, O, M0), where

M0: C  {0, 1}
is the initial marking (the starting point of the system).
Given the event e  E, the set of its entry-conditions
(preconditions), •e, and the set of its exit-conditions
(postconditions), e• , are given by:
•e = { c  C | (c, e)  I },
e• = { c  C | (e, c)  O }.
41
Any condition-event net must satisfy the following constraint:
for any e  E : •e  e• = .
A condition cannot be both a entry and exit condition for an event e. In this case, the net is said
pure.

An event e  E is called enabled if all c 

•e contain one token and all c  e• are
empty; put differently, all preconditions
are TRUE, whereas all postconditions are
FALSE.
For example, in Fig. 2 (quella in
alto”producer customer system), marking
M2 enables both event e1 and e3;
marking M3 does not enable e2 because
of the token active in c3 (the producer
cannot deliver since the deposit is busy).
In this case, we say that e2 is blocked.
An enabled event can fire, i.e. modify the
marking. The firing of an event e  E
moves all the tokens in all precondition to
all postcondition. In other words, all
preconditions move from TRUE to FALSE
and all postconditions from FALSE to
TRUE.
A condition-event system is defined by a
pure marking net and by a firing rule. The
marking also specifies the state of the system.
Let M be a marking and e  E an enabled event for M. If M' is the marking produced by the firing
of the event e, it is common to denote the transition between the two markings as:

Now we will see how to represent algebraically the arches I (arches pointing to events) and O
(arches outgoing from events) by means of a matrix W. We call it the incidence matrix:

42
A marking M is represented by a column vector, whose dimension is the number of conditions. Its
entry number i, called M(ci ), is 1 or 0 (1: token active; 0: token absent). Consider the producer-
consumer system as in Fig. 1 b). Let C = (c1, c2, c3, c4, c5) be the possible conditions and E = (e1,
e2, e3, e4) the events of the system. The
incidence matrix W and the initial marking
are, respectively,

+1 POST (rispetto a e)
-1 PRE

It is important to have at least one output and one input in each column. We need to use the
matrix to represent the evolution of the system.
Now we need to use a vector representing e1. Remember that the number of columns need to be
equal of the number of raw, so we need to build the characteristic vector with 4 dimensions.

43
We call i the characteristic vector associated to event i.
i is a row vector, whose dimension is the number of events. Its entries are 0 except for position i,
which is 1. For example, 1 = (1, 0, 0, 0) represents event e1.
Consider a transition M1 [e M2, moving the system from M1 to M2. Assume that  represents
exactly the event e enabled in M1. The transition can be algebraically expressed in terms of W and
:
M2 = M1 + W ^T,

where ^T denotes the transpose of .

If M1 [e1 M2 e M2 [e2 M3 are valid transitions, the transition from M1 to M3 is denoted by M1
[e1 e2 M3.
l
Generally speaking, a
sequence e1, e2, ...,
en of events moves
the system from a
marking M to a new
M':
M [e1 e2 ... en M'.
If (1, 2, ..., n ) is
the sequence of
vectors associated to
e1, e2, ..., en, M' is
represented by

Define s = 1+2+...
+n. It is called the
characteristic vector
associated to the
sequence of events
e1, e2, ..., en.
It can be shown that
This is called
fundamental
equation of the
system.

The fundamental equation allows us to analyze nets by using tools from linear algebra.
For example, the evolution from M0 to M3 as in Fig. 2 can be written as M0 [e1 e2 e1 M3. (This
tell us that we are moving from M0 to M3 because we are following the sequence e1, e2 and e1.
Remember that e1 is the event, it is not the sequence of events)
It can be computed thanks to the fundamental equation using s = (2, 1, 0, 0). (By doing this
aggregation we are losing the information about the order of the operation.)
44
Vice-versa, given any two markings M e M', we say that M' is reachable (attainable) by starting
from M if there exists (at least) one valid vector s such that

Two remarks: s must have non-negative entries. Even if non-negative, it could be related to a
solution of the equation which is non-attainable.

By solving the fund. equation, we find that s =

(2, 2, 0, 0) is a possible vector of firings leading
to M. This non-negative solution does not
represent a valid sequence of events since the
marking M requires two tokens at the same
position!

NEW EXAMPLE

SUMMARIZING: in this example,

depending on the value chosen for a,
we find different solutions (as for as a>=
2). Precisely there are INFINITE
solutions. The minimum one, the one
involving less operations is S= (2,1,0,0),

45
EXERCISE FOR HOME (LECTURE 13)

We conclude that the fundamental equation is used to check for the existence of a valid sequence
of firings leading M to the new marking M'.
Summarizing, we say that a marking M' is attainable starting from M if there exists a valid
sequence of firings such that M' = M + W sT.
A sequence s of firings is valid if, by starting from M, M' = M + W sT where M' is a marking. Note
that, s does not give any information about the exact sequence of firings leading to M'; rather, only
information about the number of firings.

07/03/2024
CONFLICT EVENTS
There are situations in which a certain marking enables two or more events, but only one of them
can fire. In this case, we say that they are in conflict: firing one of them inhibits the second.
An example of conflict is related to a system of traffic lights. Suppose to have a crossroad. If two
traffic lights are controlling two conflicting directions, at least one of them must be Red.
Go back to the traffic light…

How to represent a system of two traffic lights? As said, at least one of the two
must be Red…

46
 In RG1 if the pre conditions happens, the
traffic light turns green. Thus RG2 can not turn
green because X is already occupy. For this
reason there is a CONFLICT the input is used
for more than one working stations

The net represents a system of two traffic

lights. They interact in such a way that one of them is always Red. For each of them, we have three
conditions (Green, Yellow, and Red) and three events: (G->Y, Y->R & R->G). Condition x ensures that
at least one traffic light is Red. If x is TRUE, both traffic lights are Red. RV1 (1 turns to Green) is
enabled if R1 is marked and x is marked. RV2 (2 turns to Green) is enabled if R1 is marked and x is
marked. Therefore, RV1 e RV2 are in conflict: only one can fire!

PRODUCTION SYSTEM WITH 2 CELLS

We want to describe a production system divided into two cells: manufacturing and measurement.
Semi-finished items are transferred from one cell to the other by a dedicated robot

EXAMPLE 1

We have a conflict because c11 needs to be

used for both parts.
The enabled events are: e1 e5.( non so se sia
giusto)
The initial marking shown in the figure is to be
read as follows: item ready to be processed,
processing machine empty, item ready for
measurement, measuring machine empty, robot
empty. It can be represented by the following
vector:

One possible (valid) evolution of the system procuces the following marking:

47
 a= number of components u move in the system
a=1 seems to be a good solutions
S= (1,1,0,0,1,1,0,0) lo ottieni sostituendo a=1
ad ogni lettera e lo metti in ordine alfabetico. Se
a=1 facesse risultare numeri negativi, NON
sarebbe andato bene, quindi dobbiamo trovare
numeri POSITIVI O UGUALI A 0.

We now know that we need to fire at least one once these events: e1, e2, e5, e6

At this stage, two items have finished the processing phase and other items the measurement
phase. The net does not impose a precise order for operations: events e3 and e7 are both enabled.
Nevertheless, only one of them can fire: c11 is a precondition for both of them. They are in
conflict: firing one of them inhibits the second.
EXAMPLE 2

Let M0 T = (1, 0, 0, 1, 0, 1, 0) be the initial state

of the system, as reported in Fig. 4. From this
initial state, we can reach the marking MT = (0,
1, 0, 0, 1, 0, 1), as reported in the following fig.

48
RISOLVILO DAL SOLA

PETRI NETS
When representing real (and complex) situations, the condition-event (C/E) approach analyzed so
far may generate huge and not very clear diagrams.
Think about the 1producer-1consumer and the 2producers1consumer cases (Fig.1 and Fig.4). To
obtain the latter, we duplicated the subnet relative to the producing unit.
Suppose to allow for many producers and many consumers: it becomes untraceable.
Condition c3 in Fig.1 captures the presence of produced items in the warehouse. However, at the
moment, it does not account for their numerosity.
We can overcome this issue, by allowing for multiple tokens to be present at the same time at a
given place.

We relax the requirement which says that the

transition cannot fire if some postconditions
are occupied (since now we allow for multiple
tokens). A plausible marking for the system as
in Fig.6 is, for example, MT = (1, 1, 6, 3, 0). It
presumes that one producer is ready to
produce, one producer is ready to deliver, the
warehouse contains 6 items and 3 consumers
are ready to receive.
Under this new representation, the node is
called place. Each place may contain at most
k tokens, where k > 0 is the capacity of the place.

We obtain a net of places and transitions (P/T) defined by a sestuple (S, T, I, O, K, M0), where S and
T are the finite set of places and transitions; I and O are arches pointing to and pointing out from
transitions. K e M0 are the capacity function of places and the initial marking:

For example, by looking at Fig. 6, we set K(p3) = n; in this case, we limit the warehouse to have at
most n items. If the capacity is not explicitly given, we assume it is infinite.

49
 CHIDI IRE

Let •t and t• , respectively the set of places preceding

and following transition t  T. We say that t is enabled
under marking M if
- for all p  •t , M(p)  1 and
- for all p  t• , M(p) + 1  K(p).
A transition t can fire if in all places •t there is at least
one token and if the tokens generated in t• make the
numerosity not exceed the capacity of the places. The
firing of one transition t, enabled under M, determines
a new marking M’. We write:
SUMMRIZING:
 Condition / event nets admit only ONE TOKES
 Place / transition nets admit MULTIPLE TOKENS
 All the rest remains unchanged!
 Note that the formal definition of marking and ENABLED transition are different. usually in
our exercise we consider infinite capacity.

EXCERCISE

50
 08/03/2024

51
52