A Neuro-inspired Approach for a Generic Knowledge Management

System of the Intelligent Cyber-Enterprise

Caramihai Simona, Dumitrache Ioan, Moisescu Mihnea, Saru Daniela and Sacala Ioan

Automation and Computer Science Faculty, University “Politehnica”of Bucharest, 313 Splaiul Independentei,

Bucharest, Romania

Keywords: Cyber-Physical Systems, Intelligent Cyber-Enterprise, Perception.

Abstract: The paradigm of Cyber-Physical Systems may be successfully applied to a large number of case-studies, but

the most challenging of them are focusing on large scale systems whose dynamics is adapting to various

functional scenarios and environmental conditions as energy networks, traffic systems and especially

different kinds of cooperative networks of enterprises. However, the large spectrum of possible

configurations of such processes raises many issues with respect to the identification of problems to be

solved, and furthermore, of the solving method itself, making the efficient use of available knowledge a real

challenge. This paper is presenting a neuro-inspired approach for the design of a knowledge management

system dedicated to complex networking enterprises organized as Cyber-Physical Systems, and whose

functioning is implying dynamical reconfiguration in response to environmental changes, based on the

gathering and use of large flows of information, which are referred to as Intelligent Cyber-Enterprises. The

proposed ideas are based on a human brain model of reasoning and learning.

1 INTRODUCTION

The tremendous development of the technology in the

last 50 years has allowed the solving of many of the

socio-economical problems of the world. We may

produce many goods faster and transport them at

larger distances than whenever in our history, find out

rapidly large amounts of information about almost

every topic imaginable and we are interconnected by

different kinds of networks of information and socio-

economical relations.

Manufacturing is one of the fields that have been

faced both with an exceptional technological

evolution and with a set of extremely complex

challenges to face.

Customization of products, scarcity of resources,

environment awareness, competitivity, globalization

of markets are only some of the factors that are

making the management of a manufacturing

enterprise an increasingly difficult problem to solve.

Among the tools that are used in this respect,

information technologies, implied in control

engineering and in knowledge management are the

most important. On the other hand, the experience of

the last 50 years is showing that there are levels of

complexity in certain types of enterprises that require

new models and paradigms to be successfully

addressed. (Dumitrache, 2013).

Globalization and sustainable development

impose a new vision on the economy, considering the

social and environmental impact by creation of new

sustainable business and production systems. The

industry of the future encompasses the concept of an

enterprise built on the basis of connectivity – between

services, organizational structures, machines, between

machines and their human operators, into a network

of suppliers, transporters and customers – by means

of information as well as material flows.

The high-performance wireless sensors and

actuators networks, the Internet of Things and

Services, the Cyber-Physical Systems paradigm are

only a few of the conceptual and technological drivers

for next industrial revolution.

New materials, new technologies, new approaches

in control engineering of complex process as

networks of embedded systems integrating advanced

soft computing technologies and distributed

intelligence are a real support for the next generation

of enterprises. Advances in the communication

technology - M2M and H2M - allow the creation of a

real collaborative environment whose components,

Simona, C., Ioan, D., Mihnea, M., Daniela, S. and Ioan, S.

A Neuro-inspired Approach for a Generic Knowledge Management System of the Intelligent Cyber-Enterprise.

DOI: 10.5220/0008348603670374

In Proceedings of the 11th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2019), pages 367-374

ISBN: 978-989-758-382-7

367

machines and humans, communicate in a smart

society of intelligent agents.

The real link between the components of such a

heterogeneous system, encompassing processes and

products that have to be continuously adapted to the

dynamics of the environment, is through information

and knowledge.

Not only knowledge was already recognized at the

end of the previous century as one of the most

important assets of an enterprise (Davenport 1998),

but the capability of rapidly and reliably identifying

problems and then the adequate knowledge necessary

for their solving is crucial for its surviving.

As such, the development of an effective learning

knowledge management system is crucial.

Many advances have been made during the last

two decades in information acquisition, storage,

sharing and retrieving. Knowledge management

systems have evolved in terms of goals and tools,

(Dalkir, 2005) but, however, there still remain some

issues:

- Knowledge codification and sharing is dependent

on semantic technologies and on field ontology,

so heterogeneous agents networking and

information interoperability are still far from

reliable

- Problem identification remains difficult as it

implies a delicate balance between “ sufficient

information” and “ too much information”

The most successful control system that exists at

this moment, able to deal with reliably retrieving

heterogeneous information and selecting the “right

amount” for identifying and solving a problem, while

running the smooth dynamic reconfiguration of its

process as required by the adaptation to the

environmental changes is the human brain. (Bannat,

2011)

Therefore, this paper proposes a behavioral model

of the perception-reasoning-learning functional loop

on which the human brain is supposed to found its

functioning and then to apply it in a knowledge

management system architecture intended to support

the enterprise of the future, as based on the Cyber-

Physical System concept.

The second section will briefly address the

concept of the Intelligent Cyber-Enterprise, as a

socio-technical complex system. The third section

will present the perception-reasoning-learning (PRL)

loop, from a functional point of view and with a

special emphasis on the awareness concept.

The last section will present some principles of

implementation and key issues for a knowledge

management system based on the PRL loop.

2 INTELLIGENT

CYBER-ENTERPRISE

The concept of Intelligent Cyber-Enterprise (ICE)

(Dumitrache, 2013); is based on the largely known

paradigm of Cyber-Physical Systems (CPS).

CPS (Baheti, 2011) are representing a research

area and a paradigm that considers complex systems

formed by the tight interconnection of physical

processes (usually of a different nature) and

information objects (usually conceptualized as

information agents) that are fulfilling a common goal

by emergent behavior, responding adaptively, by

dynamical reconfiguration, to contextual changes.

The link between the physical nature of processes

and their cybernetic counterpart is far more important

than simple modeling and reflects exactly the process

behavior as hardware-in-the loop and human in the

loop modules.

A Cyber-Enterprise is thus composed by physical

objects as machines, by knowledge objects as process

models (workflows) and control algorithms, by

humans (operators, designers, engineers a.s.o) and by

information objects (software modules,

communication processes, databases, a.s.o). All these

objects have to interact and cooperate in order to

fulfill enterprise goals, that are specified in terms of

production demands, cost, time and resource

constraints, working in a dynamic environment

represented by customers, suppliers and market

evolution.

The proper interaction between these components

is ensured by a dimension of intelligence, both in

terms of the large amounts of heterogeneous

information flows available through the enterprise and

by the emergent intelligent behavior of the enterprise

(adaptive, reactive, pro-active, optimization), through

a global knowledge management system.

(Dumitrache, 2014).

The ICE functioning is modeled by problem

solving, a problem being defined by a goal to be

fulfilled by at least one of the enterprise components,

at the lowest level of complexity. At the highest level

of complexity, problems to be solved are related to

complex goals at the operational level and strategic

decisions that imply more or less all the enterprise

resources.

From the CPS oriented modeling results that every

problem to be solved is implying at least a control

loop and at least a single enterprise resource. More

complex problems imply interconnected multiple

control loops and different resources. At the lowest

level of complexity, control loops are working in real

time, implement clear algorithms and rules using well

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

368

defined data; more complex problems use

heterogeneous information structures and heuristics.

An intelligence-based model for the ICE is

represented in figure 1.

Figure 1: CPS-based enterprise model.

Each component of the ICE architecture should

have both the capacity of independent action under

pre-defined specifications and the capacity of

communication, allowing the transfer of meta-data,

information and knowledge in order to provide

reliable context oriented behavior.

Also, there is the possibility to connect

subsystems in clusters, based on functional

requirements. For instance, machines and robots can

be grouped in manufacturing cells, by means of

communication protocols (software modules) and

synchronization rules embedded in their respective

control algorithms (process models). Specified

functionalities may be added to products by

manufacturing cells by means of control algorithms

(product models). Manufacturing cells may be

connected in a (temporary) manufacturing system for

fulfilling a given order by means of emerging

product/ process models (workflows) and their

software representations (cases and agents). Order

dynamics, as well as the current state of resources

result in dynamic reconfiguration of the ICE.

2.1 Intelligent Entities

The Intelligent Entity is a concept designed to

represent the components of an ICE architecture

based mainly on information flows. (Figure 2)

Figure 2: Generic Architecture of ICE.

Intelligent Entities have the following components:

Semantic Interface – provides a unique way of

accessing both virtual and physical resources of the

system. This interface will allow both service and

event-based interactions between the system’s

components. Components are semantically enhanced

and include: services, event sinks and event sources.

Services may be considered as methods to solve

well defined problems, which usually arise in the life-

cycle of the system.

Events are defining usually perceived inputs that

may form a model for a new particular problem.

Business Adapters – Each Intelligent Entity can host

a series of specialized components that will connect to

existing enterprise components. Components will

require connection mechanisms and rules for the

semantic enrichment of data.

Business adapters are providing ways for

particularizations of generic problems represented by

services.

Physical Adapters – Intelligent Entities can also host

a set of adapters for the physical devices. Each

Physical Adapter will need to handle all aspects of

device discovery, configuration, communication and

data transformation.

Behavior Execution Engine – instances of intelligent

Entities will be able to perform behaviors expressed

in various languages or according to different

ontology. For each supported language, the Intelligent

Entity instance will have a corresponding execution

engine. A semantic interface will provide appropriate

input and output for the generic behavior models.

Lifecycle Manager – handles all aspects of the

Intelligent Entity’s lifecycle - initializing, clean-up,

monitoring the state of the components hosted on the

current instance and deploying new behaviors or

adapters. The lifecycle manager can expose a set of

operations through the semantic interface, such as a

A Neuro-inspired Approach for a Generic Knowledge Management System of the Intelligent Cyber-Enterprise

369

services to deploy / un-deploy components running on

the current instance, or an event source through which

it will generate events regarding the change in the

state of components.

2.2 The Biological Perspective

Interpreting an ICE from biological perspective has to

take into account the stepwise functional

decomposition of an enterprise from one holistic

system, with its own behavior, goal and environment

towards a set of (heterogeneous in nature and

behavior) atomic components (machines, people,

abstract objects) which are intricately networked

together by a material and by an information flow.

Those component could be either physical in

nature – or subjected to very clear and unbreakable

rules (machines, robots, tools), conceptual (control

strategies, models, guidelines), informational

(implementation of control strategies, product agents,

software modules) or biological (humans).

An ICE, by its nature, has already this modular

organization, and from the intelligence point of view,

every component may be embodied as one of the

Intelligent Entities described above, represented as

such by an agent in the cybernetic world.

Furthermore, it is inherent for enterprises that no

atomic component is working as such: they are linked

together by communication and functionalities, in

clusters: organizational structures, services,

manufacturing cells a.s.o. Every cluster has an

emergent behavior resulted by the way in which its

components are linked and may reconfigure itself in

order to respond to a specific order or environment

state. Depending on the cluster nature, they may be

similar in functioning with tissues or organs of a

body. The material flow sustains the functionality and

the information flow directs it and connects

components together, in order to obtain emergent

behavior and cooperation.

It results then that the information flow among

intelligent entities may be considered as a nervous

system into an organism.

More than that, this nervous system has different

purposes, with respect to the different levels of

control it serves: real-time, operational, and strategic.

At Real-time Level, it gathers data and gives

(simple) commands for atomic components, whose

local functioning is ensured by their own control

systems – these are the cells. Commands are usually

of the type of “start”/ “stop” events, which trigger

embedded and well known functionalities, answering

to known problems. It is the goal of operational levels

to decompose “their” problems in functionalities and

to synchronize them according to patterns. This level

may embody the cell functioning into an organism.

Data gathered are with respect to the actual state of

each component. Control is similar with reflex/

unvolitional actions. There is no real awareness in the

reasoning process.

At the Operational Level, problems are solved

with respect to external stimuli (orders, for instance),

by complex algorithms with a strong adaptive

component, reflected in the flexibility of the solution

and in the possibility to reconfigure the set of atomic

components and their respective functionalities to be

involved. Here fuzzy and rule-based reasoning

becomes involved, as usually there may exist several

solutions to the same problem or very similar

problems – and choosing either the best solution or

the appropriate algorithm are crucial. The information

to be gathered has to be sufficient, relevant, reliable.

This is the biological level of volitional actions, as the

result of a perceived situation which calls for a

(previously learned) solution. Reasoning has a degree

of awareness (problem identification; solution

search), but stimuli perception does not imply

awareness, as the once identified solution is applied as

a pattern. Again, the commands are in terms of

“start”/ “stop” events, triggering algorithms, but also

triggering procedures that check the consistency

between the estimated results of an action

This is also the level where, in the framework of

the ICE, human operators enter the information flow

as agents (human in the loop). As (Davis, 2012)

presents, here are the kind of problems where the

human capacity of perceiving situations is far

exceeding the capability of software control systems

and the most recent developments in enterprise

control are focusing on H2M interactions in order to

obtain the best efficiency.

Finally, The Strategic Level is that where the

existent knowledge is used in new ways, to solve new

problems that present similitude with old ones, or

where new knowledge is gathered in order to

substantiate completely new problems solving. Here

is the completely self-aware reasoning as well as self-

aware stimuli search and perception; and usually here

are mostly people that decide, only assisted by

knowledge management and business intelligence

systems. Here, the main problems are, firstly to be

certain that the available knowledge is appropriately

used and secondly, to decide what kind of new

knowledge is necessary.

A way of ensuring the smooth interaction of

human in the loop systems is to design the

information and knowledge flows (and as it is, a

knowledge management system) inspired by the

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

370

human brain functioning. It should be noted that the

conceptual knowledge management structure

proposed by this paper is dedicated for such systems,

as their structure and complexity requires new

approaches for their overall management.

As a control system of an ICE, the human brain

has a limited capacity (even if extremely large),

reflected mostly in the number of volitional tasks it

can deal with efficiently. It uses also the capability to

work with simple algorithms for managing basic

functions of the organism, without implying the

operational and strategic levels, and may adapt to a

number of (learned) scenarios, reflected by the state

of the environment and of the organism.

It may change the resource allocation (focus)

either deliberately (when encountering a new

situation) or unconsciously (when confronted with an

unexpected result from an usual action).

This is the reason that the awareness/ volitional

nature of reasoning/ perception and acting/ control is

a key concept of the approach proposed in this paper.

3 PERCEPTION, REASONING

AND LEARNING

3.1 Some Considerations on Brain

Structure and Functioning

The relevant aspects which are associated to the idea

of the human brain as a cognitive system may be

conceptualized through the paradigm of information

processing into a system of maximal complexity from

the point of view of the actual knowledge.

At microscopic level, the brain is constituted from

more than 10

neural processing units, each one

having around 10

-10

input-output axonal

connections and practically the same number of

dendritic and somatic connections. At the brain level,

the connectivity includes hundreds of trillions of

connections. Structurally, the brain is hierarchically

organized, being formed from recurrent

interconnected networks. These networks are

fulfilling the different specialized functions of the

brain (Baldassano, 2015).

Organizational principles for cognitive functions

are based on connectivity patterns distributed among

functionally specialized brain regions. In (Friston,

2003) is suggested that every area has a unique input-

output connectivity pattern and a pattern

corresponding to a task dependent on functional

connectivity. The essential characteristics of brain

information and knowledge processing are

determined by the interconnectivity of cortical

networks and by the distributed and parallel

functional integration. In order to realize a large area

of dynamic behaviors, every network has to have a

functional reaction sustained by recurrent anatomical

connectivity.

The input-output connectivity specifications and

the local architecture of a given brain region are

representing the determining factors for the region’s

behavior and for its cognitive significance. They are

indicating its functional specialization and its field of

options with respect to the functional integration with

other brain regions. The dynamics of interfaces

between functionally specialized brain regions is

characterizing, at least partially, the specificity of the

functional integration in a given processing context.

Additional determining factors of the functional

architecture are the mechanisms allowing processing

modules to incorporate adaptive changes, to provide

the system with the capability to learn, as a functional

consequence of processing. It may thus be considered

that the overall processing system is parameterized

with respect to the adaptability factors of the network

which is characterizing the learning process.

In (Starzyk, 2004) is suggested the possibility of

developing dynamically reconfigurable models which

are incorporating functional neural clusters. One may

consider the existence of adaptive agents which

connect depending on context and on the complexity

of the cognitive goal; the dynamic reconfiguration of

such groups of agents being automatically realized.

The cognitive functions are emergent due to the

global dynamics of agents (as sub-networks in terms

of neural Grouping operator) and their interactions.

Interactive processing of information and knowledge,

as well as fault-tolerance and high robustness are

ensuring the emergence at the overall brain level, if

considered as a dynamical auto-organizing (large-

scale) system.

Each cognitive function implies cooperation from

several neural structures and the contribution of

different interconnected functional modules. It may be

considered that, at brain level, there is an

interconnection process of several structural and

functional modules in parallel-distributed

configurations that are generating emergent behavior.

3.2 The Cognitive Perspective

Perception (https://www.cognifit.com) is the ability

of the brain to capture, process, and actively make

sense of the environmental information.

From a cognitive point of view, it is the active

process that makes possible to interpret environment

A Neuro-inspired Approach for a Generic Knowledge Management System of the Intelligent Cyber-Enterprise

371

with the stimuli received throughout sensory organs.

It requires both of the two processing approaches:

 "bottom-up" or non-aware (passively receiving

stimuli from all sensorial channels)

 "top-down" or aware (anticipating and

eventually selecting certain stimuli – perception

control)

Reasoning is the (brain) capability of solving

problems. A reasoning process is based on problem

identification, followed by its categorization and,

eventually, by the identification of the appropriate

solution.

The Problem Solving process implies reasoning

and acting. Feedback of the acting is triggering a new

perception/ reasoning stage.

The categorization phase is the one that

determines how the following phases are performed:

 Selection (of the solving patterns – if any)

 Focusing (on relevant input information)

 Planning, estimation, validation: an internal loop

whose execution depends on the category of the

identified problem, but whose goal is to advance

towards the solution by a stepwise decomposition

of the problem

 Evaluation of success (achieving the desired goal

or estimation of the distance towards it)

Learning is the process by which a piece of

knowledge modeling a solved problem is stored in a

way that makes its retrieval, (eventually as a

volitional action), possible.

There may be also defined a deep learning, by

which a piece of knowledge, used and validated for a

given number of times may be retrieved “reflexively”

i.e. without a volitional act.

It is easy to observe that perception is the key

element of reasoning, as it offers the necessary

information both for problem identification and for

the internal loop (by validation).Perception is subject

to training ability – which targets the optimization of

information capture/ selection, ensuring that the “right

amount of data” is received.

Experiments have proved that the brain has the

capability to estimate the kind of information is

expected in a certain situation and is either focusing

on it (neglecting other information) or interpolating

data with respect with an existing pattern, when real

information is presenting gaps (the phenomenon of

“we see what we expect to see”). It is a capability that

allows the brain to take decisions with a greater

speed, based on previous experience, but has the

drawback that sometimes does not allow to correctly

identifying a new problem when it presents a high

degree of similitude with a known one.

On the contrary, when faced with a recognized

new problem, large amounts of (sometimes possible

irrelevant) information are gathered, in order to select,

during the process of reasoning, the relevant one. This

is hindering the capacity of taking informed decisions

with speed, but allows for the optimization of the

solution.

With respect to these aspects, the cognitive

sciences have identified three phases of perception,

according to the following gestalt principle: every

person has a role in one’s perception process,

designating a three stage sequence:

Step 1: First hypothesis about what one is about to

perceive. This will guide the selection, organization

and interpretation of the stimuli.

Step 2: Entrance of the sensory information.

Step 3: Contrast the first hypothesis with the sensory

information obtained

So, the efficiency of perception and, subsequently,

the reasoning and learning are related to the capability

of constructing an adequate “problem model” which

includes both a generic knowledge about the structure

of the problem and a knowledge about information to

be gathered in order to correctly define the problem.

The degree of adequacy of the model depends on

the appropriate balance between the awareness/

volitional aspects of perception and reasoning.

The design guidelines and the functionality of the

proposed Knowledge Management Architecture will

be derived from these considerations.

4 KNOWLEDGE MANAGEMENT

SYSTEM OF THE ICE

4.1 General Considerations

The knowledge management system (KMS) of an

ICE will be based on “problem models” and oriented

towards problem solving. Its functioning will

combine the principles of feed-forward and feedback

control, as presented in the following fig.3. The

feedback may address either a step in the perception/

reasoning loop or the consequences of actions taken

as results of a reasoning step.

The infrastructure of the KMS is composed by

databases, knowledge bases, communication and

interface modules, and data buses as similar with the

neural networks and clusters that compose the brain.

A problem model will include:

- a Structural part (a pattern) – which will underline

the generic model of the problem, as a workflow

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

372

composed either by functions of atomic components

or by references to other problem models;

- a Parametric part: a list of data and information

(external stimuli) that have to be gathered in order to

allow the proper perception of the problem

The perception process is defined according to the

following phases:

- Selection: The number of external stimuli usually

exceeds processing capacity. Consequently, the

perceived information is selected and eventually

filtered with respect to criteria as experience (list

of information), necessities (problem pattern) and

preferences.

- Organization: Selected stimuli are gathered in

groups in order to give them meaning. According

to Gestalt principles, stimuli organization is not

random but instead it follows specific criteria with

respect to the actual goals and functioning of the

ICE

- Interpretation: When all the selected stimuli have

been organized, they will receive meaning (will

enter the pattern as parameters), completing the

perception process.

Functions of atomic components (actions

performed by a single resources according to a fixed

ontology of the ICE) are considered as “reflex

actions” that do not necessitate any kind of reasoning

awareness, as they are solved according to well

defined, non-adaptive algorithms whose applicability

as a whole (structure and parameters) follows a binary

decision (they may or may not be applied). The

perception necessary for this kind of problems is tacit

(non-aware). Their solving implies the lower level of

the knowledge management architecture. They are not

learned following a problem solving successful

process, but are pre-defined. However, there may be

conceived a reinforcement mechanism reflecting their

adequacy.

They are stored in a manner that allows their

retrieval with precision.

Knowledge management systems supporting this

king of enterprise behavior may be used for any kind

of organization; they are not specific for ICE, but

represents a necessary condition for building ICE

structures.

The architecture of the knowledge management

system of ICE may develop in time with Problem

Solving (PS) higher levels dealing with:

- New problems to be solved, based on reflex

actions

- New problems to be solved based on known

(similar) problems – that have to be adapted,

either as parameters (stimuli) or even as

structural sub-modules

- New problems to be solved, with new contexts

A problem solved (a problem model) may become

a reflex action if successfully applied for a number of

times (becomes a repetitive occurrence).

Each kind of problem (level of PS) will

necessitate a different kind of capturing, storing,

sharing and retrieving information procedures

(different kind of memory).

Problem models may thus migrate from higher to

lower levels of PS, as the perception and reasoning

implied in their solving necessitates less awareness

and supervision.

The feedback loop in figure 3 is modeling the

volitional part of the reasoning mechanism; when a

problem is dealt with at a higher PS level, the

feedback loop is used at every step of problem

solving.

Figure 3: A control model of reasoning.

When it migrates towards the level of reflex

actions, the feedback loop becomes less important,

and the problem solving process is said to be un-

volitional. As it skips some of the phases in its

development, the process becomes faster.

4.2 The Design Approach

The design approach of the Knowledge Management

System will combine the top-down (behavioral) and

bottom-up (formal) modeling.

The design process will have the following

phases:

1. Issues identification

2. Framework design

3. Identifying correlations between PS processes at

different levels

4. Consistency check of functioning specifications

5. Building algorithms for every phase of

perception

6. Find appropriate technology for implementation

7. Validation

Among the issues to be addressed are:

A Neuro-inspired Approach for a Generic Knowledge Management System of the Intelligent Cyber-Enterprise

373

- Conscious (self aware) vs. Unconscious (non-

aware or tacit) perception/ reasoning - and the

appropriate switching mechanisms

- Retrieving information in reasoning and learning

(networks of networks of knowledge)

- Optimization in reasoning: time vs. precision;

experience vs. innovation – the use of the

Grouping/ Focusing/ Selection operators in

information gathering

- How are new problems addressed (building

reasoning techniques)

The design framework is consisting in:

• Using two distinct perception mechanisms/

knowledge levels:

• Self-aware/ explicit: taking into account all

available external stimuli and checking their

relevance vs. Problem models – needs feedback

• Tacit: selecting only the stimuli from the list of

information attached to the problem model –

does not need feedback

• Reasoning objectives – different for every PS

level:

• Reflex Solving: known pattern, tacit perception,

non-aware reasoning (problem model retrieving

and application), un-volitional action (if

necessary), eventually reinforcement mechanism

• Solving a Known Problem: self-aware

reasoning (retrieving a problem model based on a

search criterion, comparing with the problem to

be solved, selecting the appropriate pattern), tacit

perception, un-volitional action, reinforcement

mechanism

• Solving a new problem by associative reasoning

(Building new problem models): self-aware

reasoning (decomposition, retrieving, evaluation,

estimation – feed-forward), self aware perception

(searching for stimuli, eventual volitional actions

for gathering information), evaluation of

perception results (information feedback),

volitional actions, evaluation of results

(functional feedback), validation of solution,

building problem model (volitional learning)

• Solving a completely new problem (Capturing/

Acquiring new knowledge and solving a new

problem by trial and error)

• Control approach: feed-forward/ feed-back –

with different weights

As mentioned, different types of processes

(capturing, comparing, retrieving information,

validation, comparing structures and parameters) have

to be developed for each PS level, according to

Gestalt principles and the ICE functioning.

5 CONCLUSIONS

The paper presents an approach for the design of a

generic Knowledge Management System conceived

to assist the control of an Intelligent Cyber-Enterprise.

This approach is inspired from the functioning of

the human brain, based on a loop of perception-

reasoning-learning as a Problem Solving procedure

and thus the knowledge management architecture is

organized on PS levels.

The proposed architecture is described by some

basic concepts and functionalities, issues to be solved

an correlations to be made in order to establish a

proper information and knowledge flow.

ACKNOWLEDGEMENTS

The paper was supported by the project “Engineer in

Europe” no. 114/GP/ 10.04.2019.

REFERENCES

Baheti, R. and Gill, H., 2011. Cyber-physical systems. The

impact of control technology, 12(1), pp.161-166.

Baldassano, C., 2015. Visual Scene Perception in the

Human Brain: Connections to Memory, Categorization,

and Social Cognition (Doctoral dissertation, Stanford

University).

Bannat A., Bautze T., et all, 2011, Artificial Cognition in

Production Systems, IEEE Transactions on Automation

Science and Engineering, vol 8 no 1, pages 148-173

Dalkir, K., 2005, Knowledge management in theory and

practice, Elsevier

Davenport, T.H. and Prusak, L., 1998. Working Knowledge:

How Organizations Manage What They Know. Harvard

Business School Press, Boston

Davis, J., Edgar, T., Porter, J., Bernaden, J. and Sarli, M.,

2012. Smart manufacturing, manufacturing intelligence

and demand-dynamic performance, Computers &

Chemical Engineering, 47, pp.145-156.

Dumitrache, I. and Caramihai, S.I., 2014. Intelligent cyber-

enterprise in the production context. IFAC Proceedings

Volumes, 47(3), pp.821-826.

Dumitrache, I., Caramihai, S., Stanescu, A.M., 2013. From

Mass Production to Intelligent Cyber-Enterprise. 2013

19th International Conference on Control Systems and

Computer Science, 399-404. 10.1109/CSCS.2013.100.

Friston, K.J. and Buchel, C., 2003. Functional connectivity.

Human Brain Function, 2, pp.999-1018.

Starzyk, J.A., Guo, Y. Zhu, Z., 2004,. Dynamically

reconfigurable neuron architecture for the

implementation of self-organizing learning array. In

18th International Parallel and Distributed Processing

Symposium,. Proceedings. (p. 143). IEEE.

KEOD 2019 - 11th International Conference on Knowledge Engineering and Ontology Development

374