You are on page 1of 9

Intelligent System for Knowledge Management in Petroleum Industries

CENPES, PETROBRAS Research and Development Centre, Cidade Universitária, Quadra 7, Ilha do Fundão, CEP: 21949-900, Rio de Janeiro, RJ E-mails:{Kaku}{mcampos}

KREBS Cidade Universitária, Quadra 7, Ilha do Fundão, CEP: 21949-900, Rio de Janeiro, RJ E-mails:

Applied Artificial Intelligence Lab. PUC-RJ, Rua Marquês de São Vicente, 225, CEP: 22453-900, Rio de Janeiro, RJ E-mails:,

Artificial intelligence, knowledge management, expert system, intranet

Consolidation and maintenance of knowledge basis are considered important, but are not treated adequately as a tool to enable changing in common practices inside a company. Trying to keep the experience of a company in a “former document” is not a trivial practice. Reports, calculation sheets and manuals are often considered reference documents for future consulting in projects and technical support services. Trying to extract the main points, and make them available in a fast and easy manner to the technician, is still a challenge. The loss of specialists interrupts the improvements originated from experience feedback and inhibits innovation. This prevents maturation of the technical staff and thus the growth of an institution. The strength of an institution is based on a solid external recognition of the potentialities of its staff and the capacity to improve solutions based on its past experiences. This paper presents the development of a methodology and a standard environment to consolidate the knowledge of specialists related to instrumentation, control and automation activities, focusing on the design and technical consulting demands inside a petroleum refining company. The environment for knowledge management takes advantage of the facilities of the Corporate Intranet to extend this environment to all technicians of the company. This knowledge management system has been operating since February 2002 in Petrobras, the Brazilian Oil Company, Brazil.

Business operates in a fiercely competitive environment where the key to survival is strategic management of knowledge [Desouza, 2001]. This knowledge comes in many forms. Database, trade secrets, the know-how of employees, patents, copyrights and process are just a few examples of the intellectual assets that enterprises hold. As Davenport and Pruzac (1998) define: “Knowledge is a fluid mix of framed experiences, values, contextual information, and expert insight that provides a

framework for evaluating new experiences and information. It originates and is applied in the minds of workers. In organization, it is often embedded not only in documents or repositories but also in organizational routines, processes, practices and norms.” Nevertheless, company’s knowledge is not concentrated. Knowledge workers spend much of their time searching for data and analyzing the discovered relationships, before doing something with the analysis (i.e. making decisions, implementing strategies, designing a system, etc.) [Marin-Llanes et al., 2001]. With widespread and easy access to internet and intranet, the current challenge for users is not the lack of documents and information, but the lack of ways to order and select the most important information. So, knowledge workers suffer of a flood of information, that takes too much time to select and analyze, for decision-making. To help to deal with the information overload various techniques have emerged for knowledge gathering and its management, for example [Desouza, 2001]: • Use of statistics (regression and cluster analysis on data) • Use of Artificial-intelligence (AI) based techniques (neural networks, genetic algorithms, intelligent agents). A prime example of Intelligent Agents is in search engines of various Internet Websites [Medina et al., 1997]. This Agents help in retrieval of documents based on user-defined criterion. The field of knowledge management is still in its evolutionary stage [Desouza, 2001]. This paper describes the development of a system for gathering, organizing and disseminating a specific company’s knowledge related to instrumentation, control and automation activities. The objective is to preserve and formalize the knowledge of experts, so this information can be shared with others and reused in solving new and recurring problems. It’s a challenge because there are much tacit or implicit knowledge of experts, which are difficult to obtain and organize. The paper is structured as follow. The instrumentation, control and automation activities are described in Section 2. The project and the first approach to develop the knowledge management system using an expert system is discussed in Section 3. The final knowledge management system is described in Section 4.

The specific problem addressed in this paper is related to instrumentation, control and automation activities, focusing on the design and technical consulting demand inside a petroleum company. The engineers in these areas face many problems: • Design the requirements for instrumentation of specific Process Plants; • Create all strategies related to control, safety instrumented systems and automation for many equipment and subsystems of these Plants; • Insert into this stage, the company past design, operation and maintenance experience; • Support and solve any problem according to these specialties in any operating Plant. There are many different Plants in a petroleum industry. Production facilities in offshore platform, Fluid Catalytic Cracking, Delayed Coking, Distillation, and Hydrotreating Units, in Refineries, are some examples. Each one of these Units has specific characteristics and thus specific design, maintenance and operation requirements related to its devices. For example, a Resid FCC Unit converts gas oil and vacuum resid into a range of hydrocarbon products with high market-value, like gasoline. The Figure 1 shows a picture of a 12000 m3/day Resid FCC plant of a PETROBRAS refinery. In this Unit there is a solid fluidized catalyst flowing through the equipments. Consequently the instruments must be protected against erosion. Instrument impulse lines shall also be protected against blocking. Many details to avoid these potential problems shall be considered during the design of instrumentation concept, specification and installation. Plant operation and instrument maintenance procedures must also consider the peculiarity of this Unit.

Figure 1 – RFCC plant overview. In a different scenery, operating conditions may vary in a wide range for any Plant according to its process design basic data. For example, in an ethylene production Unit, the hydrocarbons are passed at a high speed through a high temperature (+860 0C) cracking coil. Before entering in the first distillation column the cracked gas is cooled down about –100 0C through several stages of a refrigeration system. So instruments shall be consistent with this wide range of temperatures. These characteristics associated to equipment or systems of a Plant shall be considered during a design activity. Control, safety instrumented and automation systems are closely related to process and operation knowledge of industrial Plants. Different control strategies for same process Plants shows different design considerations which sometimes implies in technological gaps that shall be clearly documented. In this field, the operation experience feedback also makes a difference in design considerations. The basic engineering associated to maintenance and field operation experience are key activities in a petroleum company, and this important feature is optimized when information sharing facilities among these sectors are available. In technical support activities, engineers face to situations where fast diagnostic and solution are required for a field problem. Execution time is reduced and uniformity in final products are searched by the development of knowledge integration facilities. Segregate particular information about Plants is not trivial. There are many conventional ways to make information available to technical staff, like manuals, reports, calculation sheets, specification documents, etc. None of these resources considers an objective easy-to-restore facility to access main information inside a wide space of search. Loss of knowledge also brings a tremendous damage to companies. Loss of specialists interrupts the implementation of improvements originated from experience feedback and inhibits innovation. This prevents maturation of the technical staff and thus the growth of an institution. Another important feature related to design activity is the need to consolidate practices and criteria not described in international or company standards. Discussions and technical meeting conclusions must be translated to internal procedures, and this knowledge, when translated to conventional information media, is difficult to update and has great potential for loss of its contents during time.

The knowledge management environment designed by Petrobras specialists tries to capture and organize relevant knowledge required for a future best design and Unit operation. This environment was not designed as a manual, technical report or other conventional information document, but it was built as a platform to retain the experience, to promote knowledge sharing, to identify opportunities to consolidate criteria, and to create a specific forum of good practices for all company. Training and expertise integration are complementary focus of this tool.

The main critical point observed during the development of a knowledge base is associated with its modeling. Translate the information that is available in a diffuse form to an objective and structured form is not trivial. Even specialists with great experience have difficult in transfer its knowledge to rules or summarized texts. The use of all available computer software resources like expert systems and multimedia tools brings a new approach for this information handling. The beginning of the project was established with a joint development between Petrobras R&D Centre (CENPES) with “Applied Artificial Intelligence Laboratory” at PUC-RJ University, to design an expert system to create this knowledge based information system. This project had a financial support from CTPETRO/FINEP/ANP, the major government petroleum regulation agency. "A Knowledge-Based-System (KBS) is a computer system which embodies knowledge about a specific problem domain and can be used to apply this knowledge to solve problems from that domain." [Smith, 1996]. A KBS may be designed to apply expertise procedures when solving real-life specific problems by the representation of its reasoning in the form of rules. Applications of some KBS functions include the activities of design, diagnosis, debugging, monitoring, interpretation, prediction, instruction, repair and control. A KBS does not copy the structure of the human mind nor the mechanism for general intelligence, but it can emulate the thought and reasoning process of human expert for decision-making. It is simply a practical computer system that utilizes strategies developed by humans to solve specific classes of problems [Hughes, 1991]. A KBS is best applied to problems with the following characteristics [Gibbsons and Scott, 1997]: • • There is a non-linear relationship between the cause and effect; Engineering knowledge is available.

An expert system, which is a knowledge-based-system, was used to implement the knowledge management system. This system started focusing on the diagnosis problem. Many (if-then) rules were established relating problems in instrumentation area. A generic application tool to configure these rules and construct user interfaces was developed. Figure 2 shows an interface associated to this application, which displays rules and returns engineering and maintenance recommendations. This system was conceived as a stand alone expert system software and was developed in “Delphi” language. This knowledge representation model seemed to be adequate to treat problems concerning less number of parameters, since the number of rules grows exponentially with it. Consequently, the use of expert systems is limited to a restricted domain of problems. Some problems, which can be solved using expert systems, are: • • • Diagnostic on measurement problems in Industrial Plants Evaluation for applicability of measurement technologies Former construction of design criteria based on if-then rules

Figure 2: Rules Based Diagnosis Expert System Environment Interface

In order to make a more general tool to handle all knowledge representation forms, Petrobras staff designed a new environment to treat every problem related to plant design and technical support services. The proposal for this new environment is to replace conventional documents and formal procedures during execution of technical activities. Some of these procedures can be listed as follows: • • • • • Look and compare specific documents (calculation sheets, manuals, etc.) according to past designs; Reasoning about old experiences; Site visit for installation details; Technical consulting with experienced staff; Technical discussions related to design criteria.

It was evident that, having a large amount of information (books, reports, calculation sheets, manuals, etc.) without good organization to help users to find what they really need, could make information completely useless. So, the new system should be designed to be both easy to use and accessible to all employees on the company intranet. The second goal: to foster the development of a system capable of bringing people together to share information and delivering precise information with speed and efficiency. The ultimate purpose of this knowledge management environment is to support human activity, i.e., engineers during basic design, detailed design, plant construction and maintenance.

A multidisciplinary team was formed to develop this project, and many instrumentation and control engineers of the company were contacted and involved. The environment development was organized in three main stages: • • • Knowledge extraction Environment design: Knowledge modeling and representation Environment implementation and tests

Knowledge extraction In this stage, technical requirements were identified and information was extracted from many different ways: technical discussion forums, technical meetings, site visit, specific studies, technician interviews, etc. Operators and engineers of many refineries were contacted in order to extract their field experience in these areas. Knowledge modeling and representation The great challenge in knowledge base construction is associated to its modeling and representation. Synthesis and objective information substituted detailed explanation and basic concepts, since the main objective was to concentrate the company “know-how”, and not general information. The focus on the problem and how to characterize the solution become key factors for the success of the knowledge modeling and representation. The search for a functional information structure was done by making an index according to the “natural way” as a technician searches for the information in his mind. So, close integration between environment developers and final user was done and became important for the practicality of the knowledge base. Web site search structures are a world wide easy-to-use information sharing tool. In this environment, there is not a requirement for a rigid pre-defined information structure. The process for information search is made according to each users point of view. Considering the diversity of human behavior, we can propose that technicians search information dynamically and close to a random procedure. The facility to structure a dynamic information handling environment using hyperlinks in web design tools provides a potential environment which is easy to configure and make constant improvements and maintenance. So, the model of the environment was structured based on actual internal practices during each stage of engineering works. This model tried to simulate each step of the work flow of a technician during the design of an industrial plant to make clear the association between technicians work and the knowledge environment. Basically, the knowledge environment was classified in two scenarios: • • knowledge about all relevant aspects associated to equipment and system related to each process technology inside a refinery; knowledge about specific elements which compose instrumentation, control and automation systems, constructing a guideline for internal engineering practices which are not covered by standards;

Examples of the environment are showed on figures 3 to 8. User selects the process technology and the specific system or equipment which he is interested inside this plant. All relevant information related to instrumentation, control and automation of the selected system is presented by the use of all resources available (free text, pictures, P&I diagrams, installation details, equipment drawings, etc) for a consistent information handling.

Figure 3: Information search structures

Figure 4: Identification of all system inside a plant

Figure 5: Instrumentation installation details

Figure 6: Control diagrams for specific systems

Figure 7: Construction specifications

Figure 8: Special equipment overview

The knowledge environment also guides technicians with all documentation, which shall be generated for the specific process plant design, including the scope of each document and the procedure for its execution. Figures 9 to 12 show how the environment can potentially concentrate design criteria and store statistical data for future engineering evaluation. Information considered useful for decision making during design stage are available for on line consulting during a technical support service.

Links to Intranet and Internet sites were implemented to enable access to technical standards and manufacturers web site. Search inside these sites was made to direct access on specific informations.

Figure 9: Engineering practical equations

Figure 10: Statistical data for safety systems

Figure 11: Links to technical standards Implementation and Tests

Figure 12: Links to manufacturer sites

The implementation of the knowledge environment was done by instrumentation, control and automation specialists using a web design tool. Interface construction was continuously discussed and improved since new requests formated new functionalities. Therefore, new solutions were found to present a user-friendly interface that enabled technicians to obtain the required information without the need to be faced with complicated mechanisms. Web consulting services were contacted to improve multimedia functionalities and implementation of information security services and staffs access control. Tests were conducted with the environment installed at Petrobras R&D (CENPES) Intranet server enabling remote access from refineries staff.

This paper presented a system to knowledge management related to instrumentation, control and automation activities. This system have been operating since February/2002 on the company Intranet.

The community of specialists inside the company has considered this project the most relevant internal development during workshops and technical meetings. The benefits of the environment are enormous. Some of these are listed below: • Consolidation of company experience; • Establishment of tool for better and faster decision making; • Integration of technical communities creating synergies and patterns of communication; • Training; • Consolidation of Intranet and Internet investments. This knowledge management system also helps to create collaboration among people from different departments in the company. The system provides a basis to create a virtual organization, and helps the employees to respond quickly and effectively to their tasks. The possibility to share in a simple manner restricted data, information, and so on, with other users over a company network shows the full power of web and multimedia applications.

Desouza, H., “Intelligent Agents for Competitive Intelligence: Survey of Applications”, Competitive Intelligence Review, Vol. 12 (4), 57-63, 2001. Davenport, T. and Pruzac, L., “Working Knowledge”. Boston: Harvard Business School Press., 1998. Gibbons, W. and Scott, T., “Using a Knowledge-Based-Systems tool to enhance process control, quality and productivity”, EUFIT’97, September 8-11, 2278-2283, 1997. Hughes, J., “Object Oriented databases”, Prentice Hall, 1991. Marin-Llanes, et al., “Information Analysis Techniques for Competitive Intelligence Process”, Competitive Intelligence Review, Vol. 12 (1), 32-40, 2001. Medina, J.M., Pons, O., Cubero, J.C., and Vila, M.A., “On the Use of Fuzzy Techniques to Improve Internet Search”, EUFIT’97, September 8-11, 1186-1190, 1997. Passos, Claudio Azevedo, Ambiente para Desenvolvimento de Sistemas Especialistas: Edição e Prototipação. – Tese de Mestrado – IME, 1997. Smith, P., “An Introduction to knowledge engineering”, International Thomson Computer Press., 1996. Teixeira, Cezar Bezerra - GSEOO, Gerador de Sistemas Especialistas com Bases de Conhecimento que suportam Herança. – Tese de Mestrado – IME, 1995.