Investigators: Charles P. Sturrock, Carole D. Flanigan, and Bijan Mashayekhi (NACE Research Associate)

Technical Description:

In 1978, NIST and Battelle Columbus Laboratories estimated that corrosion costs the U.S. economy 4.2% of the gross domestic product. Today, this would be over $200 billion per year or more than $1000 per person each year. Also, it was estimated that a significant portion of this cost could be prevented by using existing knowledge. To make this knowledge more readily available to industry, NIST joined with NACE International (formerly the National Association of Corrosion Engineers) in 1982 to form the NACE-NIST Corrosion Data Program (CDP). The over-riding goal of the CDP is to enhance the cost effective practice of corrosion prevention by collecting, organizing, evaluating and disseminating in computerized form, corrosion and materials performance information.

Technical Objectives:

• Develop materials performance informatics that are:
  - accurate and practical in technical content,
  - relevant to industry's engineering needs,
  - convenient and easy to use,
  - representative of the latest advances in computing technology, and
  - available at reasonable cost to the user.

• Organize workshops and symposia to bring the materials performance information community towards consensus on materials performance informatics and related standards.

• Collect experimental data (from laboratory and field exposure tests) and examine for trends via automated induction, using such tools as connectionist models, machine learning, pattern recognition, etc.

Anticipated Outcome:

• Reliable materials performance informatics that are used by industry to improve the design of their products and production processes strengthening their competitive position while improving customer and employee safety.

• Standards and recommended practices for the development of materials performance informatics.

• Identification of previously unknown patterns and/or correlations implicit within materials performance data.

Accomplishments for FY 1995:

• Five materials advisory expert systems from the CHEMCOR series addressing the handling and storage of chemical process industry corrosives were released for public distribution in FY 1995. These expert systems were developed with funding and corrosion expertise provided by the Materials Technology Institute, a consortium of some 40 companies that sponsors nonproprietary research in materials performance of interest to the chemical process industries.

  CHEMCOR 8: Materials selection and performance for the handling and storage of once-through waters
  CHEMCOR 9: Materials selection and performance for the handling and storage of hydrogen fluoride
  CHEMCOR 10: Materials selection and performance for the handling and storage of dilute sulfuric acid
  CHEMCOR 11: Materials selection and performance for the handling and storage of nitric acid
  CHEMCOR 12: Materials selection and performance for the handling and storage of ammonia

• The first prototype of an advanced materials performance expert system available on microcomputers, known as the AUSCOR program, was completed and released for review and evaluation by the project sponsor, the Materials Technology Institute. The AUSCOR program can predict the performance of some 28 stainless steel and nickel- based chromium-bearing alloys in actual service across a wide range of aqueous environments. AUSCOR is considered an advanced system because part of its reasoning is based on mixed potential theory, which provides for variations in oxidizing capacity, temperature and velocity of the electrolyte, galvanic coupling of the alloy in question with other metals, changes in the active-passive behavior of the alloy, and numerous other factors generally considered only by the most informed and experienced corrosion experts in the industry.

• Two software modules from the POWERCOR series addressing corrosion issues in the electric power industry were released for public distribution in FY 1995. These modules were developed with funding and corrosion expertise provided by the Electric Power Research Institute, a consortium of some 600 utility companies.

  POWERCOR 5: Corrosion in wet flue gas desulfurization systems
  POWERCOR 6: Corrosion in service water systems

• A variety of inductive learning methods were evaluated using three typical corrosion databases. Of the methods studied, which included classical and nonparametric statistical methods, connectionist models, and methods originating in machine learning, the decision tree approach was found to provide the best combination of accuracy and scrutability in predicting an output from a set of known values of inputs.

• Two papers were prepared on this work and submitted for publication.

• One invited presentation was made on this work.

Impact and Technical Highlights:

• One of the member companies of the Materials Technology Institute reported that its occasional use of one of the 12 CHEMCOR expert systems saved two to three times its investment in the entire series.

Back to Table of Contents

Last modified: Mon Jan 06 09:46:15 1997 Metallurgy Webmeister