Data processing: artificial intelligence – Knowledge processing system – Knowledge representation and reasoning technique
Reexamination Certificate
1999-02-26
2003-08-05
Follansbee, John A. (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
Knowledge representation and reasoning technique
C703S002000
Reexamination Certificate
active
06604092
ABSTRACT:
BACKGROUND FIELD OF THE INVENTION
This invention relates generally to expert systems and more particularly to an expert system for the design of protocols.
BACKGROUND OF THE INVENTION
The statistical methods of multivariable testing, also known as Design of Experiments (DOE), have been used in industrial process design for decades. However, it has not been embraced in the scientific community despite the significant advantages these techniques offer. One notable explanation for this is that DOE methods are perceived as formidably complex.
Scientists commonly design experiments using the traditional one-variable at a time approach. More specifically, all but one variable are held constant while the one under investigation is varied. The test variable is then fixed at some “good” value based on the results and another variable is modified.
This traditional method can be costly in both time and physical resources, particularly in cases where there is a wide variability in assay precision and linearity (a direct relationship) between variables. Additionally the traditional method does not evaluate the interactions among variables.
The statistical methods of DOE are very powerful techniques that can significantly enhance the effectiveness of an experimental design. DOE methods can simultaneously consider interactions between many variables. DOE matrices can reduce the number of test configurations, reduce defects, improve experimental times, reduce expenses, improve the quality of experimental results, and greatly increase the odds of identifying hard-to-find solutions to difficult quality problems. Thus it can be seen that DOE methods are cost effective in both time and physical resources. Further, quality experimental information translates into more reliable decisions and, ultimately, shorter times to product introductions.
Genichi Taguchi carried out significant research with DOE techniques in the late 1940's. His effort has been to make this powerful experimental technique more user friendly and apply it to improve the quality of manufactured products. Most of Taguchi's orthogonal arrays are easier-to-use rearrangements of earlier DOE designs. Interactions can be designed in and analyzed more easily, and the arrays can be modified for mixed-level designs with simple-to-follow steps.
Additionally, classical DOE does not specifically address quality. DOE using the Taguchi approach attempts to improve quality, which is defined as the consistency of performance. The prime motivation behind the Taguchi experiment design technique is to achieve reduced variation, also known as robust design. Robust designs, using ideas derived from Taguchi, allow the user to simultaneously study the controllable factors and reduce the effect of uncontrollable environmental variables. This technique, therefore, is focused on attaining the desired quality objectives in all steps.
Dr. Taguchi's standardized version of DOE, popularly known as the Taguchi method or Taguchi approach, was introduced in the USA in the early 1980's. Because of their simplicity and success in industrial process design, the Taguchi methods offer a cost-effective strategy involving interactions between wide ranging variable combinations. Today it is one of the most attractive quality building tools used by all types of engineers in the manufacturing industries.
The Taguchi philosophy of design of products and/or processes identifies three design stages: systems design, parameter design, and tolerance design. In the first stage, systems design, the designer draws upon his/her knowledge of the process in question to produce an initial design of a product or process. The use of experimentation may be irrelevant during this phase, but will become an essential element at the next stage, parameter design. The objective of parameter design is to choose suitable values for the parameters of the product or process. In the third stage, tolerance design, inexpensive components are replaced by better ones to achieve quality within the desired tolerance.
Current DOE tools assume that the user will have sufficient information to effectively define the first two stages; systems design and parameter design. That is, that the user is familiar with the nuances of the particular plan of the scientific experiment or treatment (protocol method) being employed. They also assume that the user will be able to select appropriate constants, variables, and variable value ranges.
In addition, the user is expected to have a sophisticated knowledge of statistical design and analysis; many programs provide no guidance in the analysis of the results. Often, the user is presented with a lot of statistical output that requires substantial effort to translate results data into relevant answers.
In sum, the use of DOE tools can be complex, daunting, and can require a significant amount of time and effort to master.
These limitations make the use of current DOE design applications unattractive to those who could otherwise make productive use of these powerful statistical design a protocol.
Other inventors have created several types of expert systems for protocol design, employing DOE methods. However, none integrate a combination of a simple user interface, measurement unit conversion tools, specialized learning knowledge bases, a data structure for storing user tested protocol methods, a rule set which is used to process saved data and incorporate it into the knowledge bases, a hierarchy of parameter selection rules, robust experimental design and analysis tools, display the experiment design analysis in a way which is easily understood, and a feedback method for the refinement of the protocol method.
In addition many of these inventions are of such a sophisticated nature that their implementation is limited to hardware and software systems with specialized tools and are thus limited to a small group of users who have access to such facilities. U.S. Pat. No. 4,472,770 (Li, Sep. 18, 1984), U.S. Pat. No. 4,710,864 (Li Dec. 1, 1987) and U.S. Pat. No. 4,910,600 (Li Mar. 20, 1990), “Self-optimizing method and machine”, make use of statistical design matrix for automated experiment design and testing an object but assumes that this object is well defined, determines the test designs without human control or interaction, and does not integrate a knowledge base, nor does it have the ability to save results for future reference by others.
Patent JP7200662 (Hiroko, Aug. 04, 1995), “Experiment Plan Production Support Systems Having Design Knowledge Base”, requires that the product, the results of a completed process, has already been generated and that the relevant parameters of the initial process that produced the resulting product are known. These results are required before the experiment plan can be generated. In addition, it does not provide for an feedback loop.
U.S. Pat. No. 5,107,499 (Lirov, et al. Apr. 21, 1992), “Arrangement for automated troubleshooting using selective advice and a learning knowledge base”, interactively communicates between a user and utilizes a learning knowledge base but it does so in a complex fashion and does not incorporate DOE design methods.
U.S. Pat. No. 5,253,331 (Lorenzen, et al. Oct. 12, 1993), “Expert system for statistical design of experiments”, defines a method for interacting with a user to specify an experimental design. However, it does not utilize a knowledge base, nor provide a feedback method after the experiment has been completed, and involves complex interactions between multiple layers of programming language tools and is thus is restrictive in the type of computer hardware and software platforms on which it can be developed.
REFERENCES
“Statistics in Research and Development”, Second Edition, Ronald Caulcutt, Chapman & Hall, 1991
Cobb, B. D. and J. M. Clarkson (1994), “A Simple Procedure for Optimizing the Polymerase Chain Reaction (PCR) Using Modified Taguchi Methods,” Nucleic Acids Research, Vol. 22, No. 18, pp. 3801-3805.
Briones, P., “Experimental Design: A useful Tool for PCR Optimization”, Bi
Follansbee John A.
Hirl Joseph P.
Stewart Lisa E.
LandOfFree
Expert system utilizing a knowledge base and design of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Expert system utilizing a knowledge base and design of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Expert system utilizing a knowledge base and design of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3117285