Design of a reference architecture
in intelligent warehouse supply logistics through the use of Industry 4.0
technologies. Case of retail Warehouses in the city of
Pilar.
ABSTRACT
The article presents a reference architecture for
intelligent warehouse supply logistics using Industry 4.0 technologies in
retail warehouses in the city of Pilar. The proposed architecture is based on
the integration of information and communication systems, such as the Internet
of Things, artificial intelligence and a DSS, to improve the efficiency and
productivity of the supply chain. In addition, some case studies are presented
that demonstrate the effectiveness of the architecture in optimizing lead time
and reducing operational costs. Overall, the proposed reference architecture
can be a valuable tool for companies seeking to improve their supply logistics
and remain competitive in the marketplace.
Keywords: Industries 4.0; Warehouses; Procurement; Architecture;
Logistics.
Diseño de una arquitectura de
referencia en la logística de abastecimiento inteligente de almacenes mediante
el uso de tecnologías de la Industria 4.0. Caso Almacenes minoristas de la
ciudad de Pilar.
RESUMEN
El
artículo presenta una arquitectura de referencia para la logística de
abastecimiento inteligente de almacenes, que utiliza tecnologías de la
Industria 4.0 en los almacenes retail de la ciudad de Pilar. La arquitectura
propuesta se basa en la integración de sistemas de información y comunicación,
como el Internet de las cosas, la inteligencia artificial y un DSS, para
mejorar la eficiencia y la productividad de la cadena de suministro. Además, se
presentan algunos casos de estudio que demuestran la efectividad de la
arquitectura en la optimización del tiempo de entrega y la reducción de costos
operativos. En general, la arquitectura de referencia propuesta puede ser una
herramienta valiosa para las empresas que buscan mejorar su logística de
abastecimiento y mantenerse competitivas en el mercado.
Palabras clave: Industrias 4.0; Almacenes;Abastecimiento;
Arquitectura; Logística.
Article received:-22-junio-2023
Accepted for publication:-22-julio-2023
INTRODUCTION
This scientific paper deals with the design of a reference
architecture in intelligent warehouse supply logistics, making use of Industry
4.0 technologies. In particular, it focuses on the case of retail warehouses in
the city of Pilar. In these warehouses it has been determined that there is a
problem of stock management and handling of goods with delivery logistics. The
problem to be solved is what model or architecture based on these technologies
should be developed to improve the stock management of these large warehouses.
This research proposes a transformation of the current stock management models
to improve the competitiveness of distribution logistics, using disciplines
such as Data Mining and Artificial Intelligence to improve the information available
for decision making. The implementation of this model could improve
distribution and logistics in the sales chain, helping to reduce costs and
distribution times of the final product in the region, which represents an
important contribution to improve competitiveness in the storage and stock
management of retail stores in the city of Pilar. Digital transformation is a
trend that has been growing in the business world and Industry 4.0 has played a
fundamental role in this process. This industry is characterized by the
incorporation of intelligent technologies in the production process, which has
allowed for greater automation and control in the management of operations. In
this sense, the logistics sector has been one of the most benefited by the implementation
of these technologies, as it has allowed a more efficient and effective
management of processes.
In the case of retail stores in the city of Pilar, the
implementation of a reference architecture in intelligent supply logistics
through the use of Industry 4.0 technologies can be a solution to the problem
of product stock management. The high level of product turnover, the large
number of references and the need for rapid replenishment are some of the
challenges faced by retail stores in this city. The implementation of an
inventory management model supported by an intelligent decision making system
can significantly improve process efficiency and market competitiveness.
In this sense, the proposal of this work is of great
relevance both academically and socially. From the academic point of view, the
research proposes an innovative solution to improve inventory management in
retail warehouses, through the use of Industry 4.0 technologies and the
implementation of an intelligent decision making system. This model is
supported by disciplines such as Data Mining and DSS, which represents an
important contribution to the development of new solutions in inventory
management. From a social point of view, the implementation of this model could
contribute significantly to improve distribution and logistics in the sales
chain, reduce costs and distribution times of the final product and improve
competitiveness in the storage and stock management of retail stores in the
city of Pilar. This research is supported by similar analyzed works such as
those of (González J. L., 2007) where logistics costs in Latin America are analyzed, with respect
to other countries, and the lack of infrastructure that leads to an increase in
operating costs and directly influences the competitiveness of products. The
ABC model is a business management tool used to analyze and control a company's
costs. It is based on the idea that not all costs are equal and that it is
important to identify and control the most important costs in order to improve
the company's efficiency and productivity. At (Arrieta Gutiérrez, 2018)
details some important elements about this methodology that will be implemented
in the research.
We find in (Arrieta Gutiérrez, 2018) a study
of the crucial processes demand/supply management, the research evidenced
inefficient practices and recognized a problem in the control of projection
against demand, generating planning errors, delivery at times out of budget,
inventories in stocks generating additional costs and damage to goods in the
Colombian textile sector. This material serves as a reference in the possible
variables to be taken into account for the development of the model to be
proposed for its application at the regional level in Paraguay.
(Bartodziej, 2017) mentions a new performance of industries in the world, Industries
4.0; this paradigm deals with the combination of different physical forces with
digital ones. Industry 4.0 technologies enable the adaptation of warehouses and
stored products to changes in their processes. Industry 4.0 technologies can
move to an inclusive robotic technology. This enables automated systems to
solve problems that occur when working with people through current
technological innovations. (Gonzalez, 2021) applies data
mining techniques to determine the behavioral patterns of retail store
customers. These patterns can later be used for intelligent replenishments and
avoid unnecessary stocks in the company. This research serves as a reference
for the analysis of purchases, so, unlike it, the research work conducted in
the stores of the city of Pilar focused on the location of products in the
warehouse, its impact on delivery logistics and the improvement of waiting
lines through the construction of a reference model for its application and the
improvement of the search and delivery processes. (Abad, 2020) we analyzed the
possible techniques and algorithms to be used for the desired purpose. A
detailed review of the algorithms used by the selected software for the
implementation of Data Mining has been carried out. The techniques used are
fully tested throughout different situations so they guarantee a reliable
result.
This research work makes a significant contribution to
retail stores in the city of Pilar, as it provides a product management model
based on AI and supported by a DSS that can improve the product storage system
and processes taking into account the sales season.
The main objective of the work is the intelligent
management of large retail stores in the city of Pilar through Decision Support
Systems and Industry 4.0 technologies, leaving as a legacy a reference model or
architecture for the precise implementation of the same.
METHODOLOGY
With regard to the research, based on the object of study,
the problem posed and the objectives set, this work was carried out
quantitatively, developing theoretical models and measurable variables for the
formulation of a hypothesis and its subsequent validation.
The universe is composed of resources and processes of the
current inventory distribution systems and the logistics chain of retail
warehouses located in the southern region of Paraguay, specifically in the city
of Pilar, department of Ñeembucú. The unit of analysis is composed of the
current distribution systems of the logistics chain of retail stores in the
city of Pilar and the management models of large warehouses. We work with the
scenarios indicated in the specific objectives.
The work carried out is of a theoretical type in the stage
of development and verification of the current models. Once the method has been
identified, a decision support scheme with computer support was sought to
validate and improve the suggestion made to achieve the objective proposed in
this research. As a consequence of the analysis, it was necessary to modify the
proposed decision models, which started a new cycle of validations.
During the research process, different data collection and
analysis techniques were taken into account, it was necessary to consider the
methods, techniques and instruments as those elements that make up the
empirical fact of the research, that is, the basic phase of the research
experience.
A systematic inquiry was made to study the significant
aspects of the facts and situations that occur in the current context of
distribution and organization of product inventories in warehouses.
The advantage of this data collection methodology was that
the data were studied as they occur at the time and without intermediaries.
This technique made it possible to obtain bibliographic
data on current models applied in other regions in similar processes. The
documentary compilation made it possible to study the inventory organization
methods and schemes used by companies in other countries or regions to improve
the cost of the product through inventory management.
The documentary collection helped us to detect and consult
other materials based on other knowledge and/or information of any reality, so
that they can be useful for the purpose of the study.
Due to its technical characteristics, this type of
technique does not take into account the possible reactions of the subjects
under investigation.
Simulations were carried out with the proposed decision
models and the computer decision models to analyze the result of the
suggestions and whether they met the necessary conditions for their subsequent
application. The simulations showed that the proposed model can be applicable
and would help to improve the entire logistics chain management process, thus
reducing the current distribution costs to the points of sale.
Data Mining was used as a tool for the analysis of the
information. The proposed technique is intended to be applied to any item of
retail warehouses in Pilar or in any city and organization with the same
characteristics.
Data mining helps to deal with a large amount of
information that is produced in different areas, in this case, we deal with the
information produced from the sales of retail stores and the outputs that they
cause in their warehouses. This information is important because it generates
patterns of customer behavior and specific products (sales, seasonal changes,
sales decline, sales peaks, etc.).
We call this type of analysis descriptive mining, since it
analyzes and describes a situation found in a group of data.
On the other hand, also thanks to Predictive Data Mining
and Artificial Intelligence we have determined the different situations that
can be foreseen for the location of products in a given warehouse. Through
descriptive mining we obtain behavioral information and through predictive
mining we propose results to improve the current situation of retail stores in
the city of Pilar.
RESULTS AND DISCUSSION
In order to carry out the proposals and analysis of the
methodology to be implemented, the current situation of the retail stores in
the city of Pilar was verified and the following results were obtained:
·
Retail logistics warehouses in the city of Pilar
currently lack an intelligent merchandise location system.
·
Nowadays, the location is favored by the moment
and the free spaces.
This information is verified through observation and data
collection by means of an interview and a simple questionnaire to the owners of
retail stores in the city of Pilar.
This questionnaire is a set of questions designed to
collect information from stores in the city of Pilar. Questionnaires can be
used in a wide variety of contexts, from market research and customer satisfaction
evaluation to worker performance evaluation and business decision making, in
this case it was structured to know exactly what is the real situation and how
warehouse managers organize their warehouses.
The results of this interview indicate that 53.3% of
retail store managers classify according to the order of arrival of items,
33.3% do so according to available space and 13.3% do so by determining
priority products.
None of the warehouse managers use the ABC method for
product classification and organization. This information was collected through
visits to retail stores in the city, identifying those that actually have
product warehouses with stock and their location in an area larger than 100 m2.
During the visit to the warehouses, the products and their classification in
the inventory were observed. In addition, the following survey was made to the
warehouse managers.
The brief survey was conducted in the city of Pilar,
targeting retail warehouses located in the area. This survey was applied in
order to obtain data on the current situation of product storage in the
warehouses located in the city of Pilar.
Taking into account the data collected, the information
was analyzed and a reference model for intelligent stock management was
developed using Industry 4.0 tools.
It is then determined that the products with the greatest
movement should have a privileged location. This will allow a quick and smooth
reaction from the stocker.
Response time in the distribution chain is paramount,
every second counts and adds up in the final calculation of time spent on
delivery to the customer or carrier. The less downtime in the search and
replenishment of the product, the lower the cost charged to each sale.
The ABC inventory management model is a method of
categorizing inventories, used as a rudimentary prioritization mechanism to
concentrate efforts and resources on items that are most important to the
business (Vermorel, 2020) mentions that this method is based on the empirical observation of
a small fraction of items or SKUs that generally represent an important part of
the business. It is also mentioned that since 2000, this method is mainly used
as a data visualization method and as a way of prioritization for supply chain
managers who must regularly review their replenishment configurations.
If we add the technological features of Industry 4.0, such
as AI and DSS, to this proven method, which is already widespread worldwide, we
have a great opportunity to improve on current inventory management standards.
Artificial Intelligence can predict the items with the
best movement in certain months of the year, thus leaving aside the fact that
items are static for a long period of time, as currently analyzed.
This methodology aims to dynamize and provide intelligent
management of Pilar's retail stores.
The ABC model is normally prepared on the basis of 1 year
of consumption or movement of items in the warehouse.
The premise that we put forward is differentiated by means
of an aggregate plus with the implementation of Artificial Intelligence as
sales data analysis, thus forming the following model:
Essential
tables and basic data structure.
The following list of fields is an example of the data
that the database must necessarily have for standard operation. To the list
below, the particular fields of each warehouse must be added.
Article
master
|
|
SKU
|
Numeric or
Alphanumeric
|
DESCRIPTION
SKU
|
Alphanumeric
|
ITEM
|
Alphanumeric
|
Table 1.
Proposed table for article master. Own source
Stock of
items
|
|
SKU
|
Numeric or
Alphanumeric
|
Current
Stock
|
Numeric
|
Average
daily sales
|
Numeric
|
Average
seasonal sales
|
Numeric
|
Fully variable
cost price
|
Numeric
|
Sales price
|
Numeric
|
Table 2.
Proposed table for the stock of articles. Own source.
Process
Description:
-
Preparation
of the stock list of retail items
-
Preparation
of the daily sales list for at least 1 year
o
This
time period will be able to guarantee the AI's path through the entire sales
process and relate them to the seasons.
o
The
longer the period of time, the better the result of the analysis.
-
Categorize
the company's most profitable items. This will allow identifying them as a
select group in the analysis process.
-
Categorize
the list of items by items (if they are of different colors, different sizes
or different styles).
o
Ex.1.:
Category Polo shirts: Red polo shirt - Black polo shirt - White polo shirt.
o
Ex.2.:
Categroy Shirts: Green Shirt - White Shirt - Pink Shirt
The determination of the
seasons will be established according to the existing relationship with the
products sold in the retails, in this case, the seasons are: spring, summer,
autumn and winter.
We also present some
tests carried out to validate the proposed model.
Test
1 - Sales scenario by line of business - 1 year period.
a)
Breakdown:
·
List
of stock items. Demo-simulations
·
Consumption
of items by season (spring, summer, fall, winter). Demo-simulations
b)
Proposal of product
classification zones in the warehouses.
The GREEN ZONE is the zone of filler products, which have
no priority, the YELLOW ZONE is the middle zone and corresponds to products
with medium demand. However, the RED ZONE is determined by the season and
corresponds to the high priority products, which normally comprises 20% of
them.
Test results:
The tree resulting from
this test with real data is a tree with too many nodes, due to the amount of
information we loaded into the analysis tool, which is why we will proceed to
analyze the most relevant ones.
In the case of a
decision tree with many leaves, we are talking about a tree with a large number
of branches representing different possible outcomes or classifications. This
type of tree can be useful in situations where detailed and precise
classifications are needed, although they can also be more difficult to
interpret and analyze than simpler, more representative trees.
The resulting tree of
the fitted sheets gives us a classification where it is interpreted that sales
in the spring season are less than 88 units, however, in summer are the highest
sales, therefore, we have an idea of the season of highest sales in units of
fitted sheets, and the season of highest movement of this same product.
This tree indicates that
56% of spring sales were less than 88 units, however, 46% of summer sales were
greater than 88 units. It is important to clarify that this information indicates
that there were more sales tickets in the spring season, thus defining the
adjustable as a product with a lot of movement in this season.
The analysis of Basic
sheets shows that during the spring, 59% of the sheets are sold in units less
than 31, however, the summer season stands out with 57% of sales between 31 and
85 units and in winter with 26% of sales greater than 85 units.
Although in the spring
season they are sold in smaller quantities, there are more movement tickets,
therefore, in this season the Basic sheets should be in a high demand or red
zone. During the summer season in the medium demand or yellow zone and in the winter
and fall, in the green zone.
The analysis of the
classic products node show us that in the spring season there are the highest
ticket movements for these products, with 57% of the tickets reaching almost 95
units; however, summer and winter sales indicate higher quantities, but less
movement.
This result indicates that clasic products should have a
privileged location in the spring season.
EM clustering with the
same data. 8,583 actual sales records.
Green Zone - Non-priority
|
Red Zone - Priority Zone
|
-
Bags
-
Curtains
-
Covers
-
Necessaires
-
Shopping Bags
-
Quilts
|
-
Adjustable
-
Pillows
-
Quilts
-
Blankets
-
Sheets
-
Towels
|
Table 3. Product
classification table according to EM Clustering results. Own source.
According to
the resulting grouping, it is easy to differentiate the items that will NOT be
a priority in each season, marked in "green" color, and those that,
due to sales excellence, should be significantly closer to the Red Zone of
product location, marked in red color in the graph.
Through data mining and
artificial intelligence applied with Weka, we obtain important information for
the identification of products and their peak seasons.
The J48 and Clustering
EM algorithms are very useful artificial intelligence tools for classifying
products and sales in retail warehouses. These algorithms are particularly
effective in the context of the city of Pilar, as they allow classifying and
analyzing large amounts of data quickly and efficiently.
These two tools are very
effective in classifying products and sales in retail stores in the city of
Pilar. These algorithms allow analyzing large amounts of data quickly and
efficiently, which can help stores improve their inventory management, optimize
their marketing strategies and improve the customer experience.
The file structure
needed for the analysis has been established and we have determined that the
relationship of the season and the sale of the star products of the retails in
the city of Pilar is demonstrated through the analysis carried out with the
information added to the system.
This analysis is
extremely important to determine the actions to be taken when determining the
location of each product and its priority on the warehouse shelves.
The same data analysis
has been performed using logs generated by IoT tools and the results have been
similar, thus proving the performance and feasibility of application of the
proposed model for AI-assisted seasonal product storage improvement.
The following Use Case
and DFD proposal is also presented for DSS guidance that will support decision
making, subsequent to the analysis of the Data Mining tool.
The base DFD should
contain the following processes and movements to fit the proposed definitions
and architectures. In the following, we describe a series of processes that
should be contained in the DSS systems that help decision making for the
reference architecture in intelligent warehouse supply logistics through the
use of Industry 4.0 technologies.
CONCLUSIONS
Industry 4.0 emerges as a revolutionary method of
implementing technological tools for process improvement, production and sales
know-how. The retail stores in the city of Pilar do not escape this innovative
regime of new technologies that involve the use of knowledge, science and data.
It is precisely the data that provide the trend towards the use of modern
industry ingenuity and that is combined with the use of new software along with
traditional models
For the implementation of this architecture or model we
suggest the use of emerging technologies in Industry 4.0 such as: Weka, for Big
Data analysis and Python for DSS.
The implementation of tools such as Weka in this study
makes it possible to classify product items by their average sales. The average
sales and seasons will determine their location in the storage areas.
This relationship is the fundamental premise of the
architecture presented and its existence confirms that the use of Industry 4.0
tools such as Data Mining, Artificial Intelligence and Decision Support Systems
provide a competitive advantage to those who take advantage of it and apply it
in their storage and product replenishment operations.
This competitive advantage provided by these tools and the
proposed reference architecture model positions the organization at a higher
level than those who do not use it and ensures customer loyalty.
LIST OF REFERENCES
Abad,
F. M. (2020). Application of data mining techniques with Weka software.
Salamanca, Spain.
Arrieta
Gutiérrez, C. A. (2018). Software as a supply and demand optimizer in the
textile sector.
Available
at: http://revistas.unisimon.edu.co/index.php/innovacioning.
Bartodziej,
C. J. (2017). The concept Industry 4.0.
González,
J. L. (2007). "Latin America: Addressing High Logistics Costs".
González,
R. (2021). Data Mining Techniques Applied to Item Rotation Analysis.
Asunción.
Vermorel,
J. (March 2020). LOKAD. Retrieved from
https://www.lokad.com/es/definicion-analisis-abc-(inventory)