|
|
|
|
|
|
|
Intelligent Optimal Design Of Complex Systems
TO GIVE AN OPTIMAL SOLUTION TO YOUR PROBLEMS EVEN WHEN YOU KNOW THAT YOU DO NOT KNOW THE WHOLE SOLUTION. |
||
|
Objectives The designers of products are often compelled to make decisions under uncertain conditions, seeking solutions to problems which do not have known solutions. These include constitutive modelling, fatigue life prediction, wear, friction, or prediction of errors of inelastic analysis of structures. With rapidly intensifying competition, it has become absolutely crucial for biomedical sciences companies to achieve shorter time-to-market, while insuring their optimal production. In the development of a new viable car with high safety and reliability, how to design or select the componants at the lowest cost ? We need to treat problems using the available expert knowledge, experimental data and simulation/computational tools. The most important specificity is to make the FUSION of the data coming from various sources but also the FUSION of the parameters to be able to extract knowledges even from a very small data base. The people at CADLM are experts in optimization of processes and formulation of material input, direct coupling with CAD Systems, optimal design of structures for random or cyclic loading, control of processes, non-destructive tests and fluid/structure interactions, civil engineering problems. They are transferring the special approach that was built at Ecole Polytechnique by J. Zarka’s team, to industries. ADVANCED INTELLIGENT DESIGN OF COMPLEX SYSTEMS This approach was proposed in 1986 and, since 1991, it has been involved in many applications where we want to know how to use any model and how to use limited data … In engineering practice, we are faced with problems in the design, testing, inspection, and maintenance of the products. Usually these problems are very complex and the knowledge of issues involved (inputs) is not complete. Thus, the problems are not fully understood. Moreover, the available data may be not be statistically representative (i.e. be in limited number), fuzzy, qualitative and missing in part. The Intelligent Optimal Design of Complex Systems takes the actual best knowledges of the researchers/experts and mixes them intelligently with the results of experiments or real returns. It was successfully applied to many real industrial systems showing explicitly the multi-level modelling and multi-disciplinary optimal design Automatic Learning One AUTOMATIC LEARNING EXPERT SYSTEMS GENERATOR is able to automatically extract the rules from the raw examples base given by the experts and thus generate an expert system. The experts know they do not know the full solution but they are able to build an examples base, for which the solution is known experimentally or numerically with sometimes some fuzzy or missing information. The main problem is to provide a good description of such an examples base. (By analogy, we can say that the data base is the program, the learning tool is the compiler and the execution gives the knowledge). Basically such a learning system includes five main functions : PREPARE : to transform the example files from user format (ASCII, dBase, Excel, ...) into the own format of the system and to handle the discretization of the descriptors and the splitting of the initial data base into a training set and a test set. LEARN : to automatically extract a rules base from the training set according to the quality of available information (noise, sparseness of the training set,..). TEST : to experimentally evaluate the quality of the extracted rule on the test set. INCLEAR : to allow the expert to visualize IN CLEAR the extracted rules with the initial user format and to say what descriptors are kept. CONCLUDE : from the description of a new case, to deliver a conclusion based on the extracted rules. In all problems, it is necessary to consider one conclusion which may be a class or any continuous real number. Moreover, often, several conclusions may be considered together. The rules have to be automatically generated for each one of them. Then an optional but fundamental sixth function, OPTIMISE, based on genetic algorithms and other special optimization techniques, may be used to solve the inverse problem, i.e. when some conclusions and some descriptors have to belong to some given sets (or constraints), what are the possible solutions and, in some particular cases, what is the best solution if an objective function is given (cost, weight, efficiency ..) PRINCIPLES OF THE ADVANCED INTELLIGENT DESIGN OF SYSTEMS In this new framework, it is needed for each particular problem: i) to build a DATABASE of examples, i.e. to obtain some experimental, real or simulated results where the EXPERTS indicate all variables or descriptors that may take a part. This is, at first, done with some PRIMITIVE descriptors x, which are usually in a different number and with a different meaning for each example. Then, the data are transformed with the introduction of some INTELLIGENT descriptors XX, with the actual whole knowledge thanks to beautiful (but often insufficient) theories and models. These descriptors may be numbers, Boolean, strings, names of files which give access to data bases, or treatments of curves, signals and images. But for all examples, their number and their type are now always the same. This is the proposed one way to allow the fusion of data coming from various sources and to reduce the number of the parameters and thus to use the limited data. The results or conclusions may be classes (good, not good...) or numbers. ii) to generate the RULES with any Automatic Learning Tool. Each conclusion is explained as function or set of rules for some among the input intelligent descriptors with a known reliability or accuracy. iii) this allows the virtual design, i.e for any new set of the input parameters, without making any new test or any new simulation, the properties or conclusion are known. iv) to optimize at two levels (Inverse Problems or Optimal Design). • Considering the intelligent descriptors as independent; it is possible to get the OPTIMAL SOLUTION satisfying the special required properties and allowing the DISCOVERY OF NEW MECHANISMS, • Considering the intelligent descriptors linked to primitive descriptors for a special family; it is possible to obtain the optimal solution that is technologically possible. So, not only a Practical Optimal Solution is obtained but also the Experts may learn the missing parts, may build models or theories based only on the retained intelligent descriptors and guided by the shapes of the rules or relationships. |
||
Copyright © 2003-2008 CADLM. Tous droits réservés.
437197 - 2