**Towards Fuzzy Logic Standardization **

*Work Group "Fuzzy Logic and Fuzzy Control" of the VDI/VDE-GMA (German Association of Mechanical and Electrical Engineers) UA451, Chairperson: Constantin von Altrock *

Citation Reference: This paper was published at the Fifth IEEE International Conference on Fuzzy Systems, held in New Orleans September 1996. The Fuzzy Logic Application Note series is published by Inform Software Corporation on its Internet server to promote the use of fuzzy logic technologies in applications.

*In 1991, the German Association of Mechanical and Electrical Engineers (VDI/VDE) established the work group UA451, focusing on "Fuzzy Logic and Fuzzy Control", in its Measurement and Automation Chapter (GMA). The group is made up of a number of European pioneers in fuzzy logic applications, as well as leading fuzzy logic software and hardware suppliers. Its company members include: Allen-Bradley, Foxboro/Eckardt, Inform Software, Intel, Moeller, Microchip Technologies, National Semiconductor, SGS-Thomson, Siemens Semiconductor, and Texas Instruments. To promote fuzzy logic in industrial applications, the group focuses on two activities: providing information platforms for designers to exchange project experiences and fuzzy logic standardization. This paper presents some of the intermediate results of the standardization activities. *

**1. Why Fuzzy Logic Standardization? **

In 1990, a broad level of interest in fuzzy technologies was triggered in Europe by the advent of numerous Japanese products that successfully exploit fuzzy logic. One group of interested parties within the control and engineering community promoted a "belief" type discussion about the pros and cons of the technology [2]. A second group, a number of "pioneers", started to use fuzzy logic in their applications. The success story of these applications is responsible for the current situation in Europe, where fuzzy logic is now a well-accepted technology in engineering. For detailed descriptions of the history of the European applications see [4] and for technical details on these applications see [5].

A detailed market research study counted 684 successful applications of fuzzy logic in Europe by May 1994. Most demonstrate that a number of substantial benefits were achieved through the use of fuzzy logic. Having identified the technical potential of fuzzy logic in these applications, the technology is ready for introduction to a broader variety of developers of control solutions. The pioneering designers of fuzzy logic guarded against the distraction of a large number of differing glossaries, methods, and algorithms utilized and proposed in scientific literature. To promote a broad spectrum of successful fuzzy logic applications, however, standardization becomes a prerequisite. The work group UA451 has identified five areas where an immediate push for standardization has emerged:

- Description formats for fuzzy logic systems
- Performance measures for fuzzy logic systems
- Development methodology for fuzzy logic systems
- Fuzzy logic solutions for standard applications
- Adaptation of fuzzy logic systems

The remainder of this paper presents the individual areas of work, details the goals for standardization in the respective areas, and discloses the current level of work.

**2. Description Formats for Fuzzy Logic Systems **

Currently, more than 30,000 publications on fuzzy set theory exist. Many propose methods, algorithms, and terminology on how to utilize fuzzy technologies in solutions. The lack of standardization has resulted in many inconsistencies among these publications. The same terminology frequently refers to different methods, and identical algorithms are often referred to by different names [6].

To promote fuzzy logic for mainstream use by practitioners, fuzzy logic terminology and methodology demand standardization. In cooperation with other groups, the work group UA451 shall propose a standard terminology. Along with this definition, a "basic fuzzy logic functionality" common to the majority of applications shall be defined. If software tools and hardware platforms support this "basic fuzzy logic functionality", porting of systems becomes easier and practitioners can use different products for their application with the same training.

The proposed standard, however, will restrict itself to terminology and methods. It will not cover any implementation details of the methods or algorithms. This shall remain a degree of freedom for hardware and software vendors. Also, the proposed standard only covers a "basic fuzzy logic functionality". This keeps the proposed standard open for specific extensions by the suppliers and at the same time maintains portability for practitioners that wish to restrict their application to the "basic fuzzy logic functionality".

As a subsequent step, the work group shall propose a fuzzy systems description language. This description language facilitates the direct transfer of a fuzzy system from one hardware platform or software tool to another. To allow inter-platform portability, the description language shall be an ASCII type format that can be manually read and written. Definitions of this language shall only cover the basic functionality, yet allow free expansions. Vendors of fuzzy logic software and hardware can either base their products directly on this fuzzy systems description language, or just this representation for export and import while employing a proprietary format internally.

The work group shall deliver the fuzzy systems description language definition in a form that can be used directly by parser/scanner-building software tools and make it freely available.

**3. Performance Measures for Fuzzy Logic Systems **

To compare the computing performance of different hardware platforms, various performance measures (benchmarks) are used. For fuzzy logic systems, measures such as MFLIPS (million fuzzy logic inferences per second) and others have been proposed in the past. However, it has become apparent that these performance measures are of limited practical value. An MFLIPS number for example does not satisfactorily answer the question of how much computing time a specific fuzzy logic system requires on a certain hardware platform. Since many different definitions exist for what "a fuzzy logic inference" is and what computational steps it comprises, some vendors go as far as only counting a single multiplication as "a fuzzy logic inference". Furthermore, in some fuzzy logic systems, fuzzification and defuzzification can require more computation time than the rule inference [1].

To create a more practical performance measure, a set of "typical fuzzy logic systems" that spans the entire range from simple to complex and is representative of various application areas shall be proposed by the work group UA451 as a benchmark suite. Practitioners seeking a ballpark estimation of the computing resources required to process their fuzzy logic solution can identify the fuzzy logic system closest to their application from the benchmark suite. Total computation time and required code size of "typical fuzzy logic systems" then becomes a meaningful measure to compare the performance of different hardware platforms. To ensure that benchmark results lead to fair comparisons, definition of the test procedure is also necessary (resolutions, input patterns, etc..).

The work group shall create a set of such "typical fuzzy logic systems", and make the definitions freely available using the fuzzy systems description language.

**4. Development Methodology for Fuzzy Logic Systems **

In fuzzy logic design, no formally defined development procedure already exists. The work group UA451 shall define such a development methodology based on the ISO9000 general system development guidelines. It will define and describe necessary steps for completing a fuzzy logic design. These steps will be both procedural and technological in nature. Procedural steps define auditions and reviews; technological steps deal with the design decisions relevant to each stage of the design. The guidelines will also classify the various fuzzy logic methods and algorithms to help with these design decisions.

Such a formalized development methodology should guide novice practitioners along the most efficient track to a fuzzy logic solution. This lowers the risk of failure simply because of problems navigating through the abundance of existing fuzzy set theory. For experienced practitioners, the proposed development methodology shall provide a formal frame to follow when creating and documenting real-world fuzzy logic solutions.

For simple fuzzy logic applications, the proposed development methodology shall even include "cookbook-like" recipes. This includes, for example, the choice of membership functions, inference algorithms, and defuzzification routines. In addition, guidelines for stability analysis and verification of fuzzy logic systems shall be proposed [3].

The work group shall create this development methodology consistent with the general ISO9000 development guidelines.

**5. Fuzzy Logic for Standard Applications **

While some applications require a custom structured fuzzy logic system, many applications fall into standard categories such as temperature control, positioning, or speed control. For such "standard applications", frames of fuzzy logic systems shall be proposed which have generally proven to be successful.

These fuzzy logic solutions for "standard applications" shall be by the work group UA451 so that practitioners can use them as an initial prototype for their fuzzy logic solution. In an optimization phase, the prototypes will require adaptation and refinement for the specific applications. To facilitate optimization, the work group will detail rules in a "cookbook-like" fashion, which explain how to tune the fuzzy logic solutions for the "standard application".

**6. Adaptation of Fuzzy Logic Systems **

In general, the design of adaptive control systems is more difficult when compared to the design of the non-adaptive variety. On the other hand, adaptive system design can expedite an acceptable solution.

A large number of applications have shown that fuzzy logic technology accommodates the efficient design of adaptive systems. This is primarily because fuzzy logic allows the adaptation strategies to be defined linguistically rather than by a mathematical model.

In designing adaptive fuzzy logic systems, it is important to understand which part of the system shall be adapted for what result. Non-experienced practitioners find themselves overwhelmed with the many degrees of freedom of how adaptation can be applied with fuzzy logic systems. The work group UA451 shall identify and list different approaches to adaptive fuzzy logic system design. Typical application areas, as well as pros and cons of the different approaches, shall be discussed. While generic recipes cannot be generated for the use of adaptive techniques in fuzzy logic system design, the consequences of making the individual parts of a fuzzy logic system adaptive shall be discussed in detail.

**7. Current Work and Cooperatives **

To create the standardization proposal, the work group UA451 has divided itself into five teams that each cover one of the five areas of work. A first draft of a standardization proposal shall be generated within a year. This proposal shall be further published to open the discussion to interested groups.

While the scope of this standardization proposal is rather broad in nature, and thus will require some time to complete, the work group UA451 also supports other standardization activities. In cooperation with the work group DKE-AK-962.2.5, the UA451 has generated a working draft of an extension to the International Standard IEC 1131 , responding to a new work item proposal of the Industrial-Process Measurement and Control Sub-Committee No. 65B from July 1993.

The scope of this effort is restricted to the integration of fuzzy logic with the existing IEC1131 standard for PLC programming. The proposed extension will standardize the functionality and a description language for fuzzy logic systems in the application area of industrial automation. The proposed definitions will form a subset of the general standardization effort described in this paper.

**8. What Does Standardization Contribute to the Practical Use of Fuzzy Logic? **

The overall goal of the work of UA451 is to make fuzzy logic a worldwide accepted technology, available and usable by every practitioner seeking a solution for his/her problem. In detail, the work group targets the following goals:

- Resolve the confusion over terminology
- Resolve the confusion over methodology
- Inform practitioners about what they REALLY need to know from the abundance of fuzzy set theory
- Inform practitioners on how to design fuzzy logic systems - cookbook-like, if necessary
- Provide control system hardware vendors with a thorough definition of fuzzy logic functionality so they can implement fuzzy logic functionality on EVERY target system
- Provide design software vendors with a thorough definition of fuzzy logic functionality so they can implement portable fuzzy logic functionality
- Free software tools from target hardware systems by making fuzzy system designs portable
- Establish transparent performance measures to stimulate further optimization of the fuzzy logic implementations

**Literature **

[ 1 ] | Adcock, T., "DSP + Fuzzy Logic: A whole new level of performance", Computer Design Conference on Fuzzy Logic in San Diego (1994). |

[ 2 ] | Mamdani, E. H., "Twenty years of fuzzy control: experiences gained and lessons learnt", Second IEEE International Conference on Fuzzy Systems, ISBN 0-7803-0615-5, p. 339 - 344. |

[ 3 ] | Tanaka, K. and Sano, M., "Concept of stability margin for fuzzy systems and design of robust controllers", Second IEEE International Conference on Fuzzy Systems, ISBN 0-7803-0615-5, p. 29 - 34. |

[ 4 ] | von Altrock, C., "Industrial Applications of Fuzzy Logic in Europe", in Yen and Langari, "Industrial Applications of Fuzzy Control and Intelligent Systems", IEEE Press (1994). |

[ 5 ] | von Altrock, C., "Fuzzy Logic and NeuroFuzzy Applications Explained", Prentice Hall, ISBN 0-13-368456-2 (1995). |

[ 6 ] | Zimmermann, H.-J., "Fuzzy Set Theory -- and its applications", Second Revised Edition (1991), Boston, Dordrecht, London, ISBN 0-7923-9075-X. |