Integrated Computer-Aided Engineering

Consideration about the possibility to integrate vague uncertainty notions into numerical simulation modeling tools may be a very interesting research field. In this way, it will be possible to exploit more efficient and robust modeling evaluation tools in the study of high productivity and flexibility production systems. In literature, few works investigated on the possibility to cope with the lack of numerical models able to deal with ill-defined uncertainty. In particular, if it is possible to describe uncertainty by statistical distribution, the methods of classical discrete event simulation theory are able to model the considered system thoroughly and may be regarded as an exhaustive tool. Otherwise, if uncertainty can not be described by statistical distribution, no robust methods tools are available to model and analyze discrete complex dynamic systems. In this work, the integration of Fuzzy Sets in discrete event simulation theory is analyzed. Firstly, uncertainty is considered from different point of views and all the necessary issues to introduce fuzziness in discrete event simulation models are illustrated. Then, a possibility-based approach is considered and fuzzy set theory concepts have been introduced in such a context. The soundness of simulation mechanisms has been formally established by considering the new questions arising from the description of system variables as fuzzy sets. Finally, the application of the proposed methodology to a simplified test case is showed and the obtained results are presented.

During recent years, fuzzy-neural networks have found extensive applications in numerous engineering areas. It is well known that the fusion of neural networks and fuzzy logic can overcome their individual drawbacks and benefit merits from each other. However, current fuzzy-neural networks often have complex structures and training algorithms. In addition, some of them cannot deal with fuzzy knowledge directly. Inspired by the α-level cut representation of fuzzy numbers, we propose a simple neural networks-based approach to approximating fuzzy rules in this paper. Using numerical simulations, our scheme is illustrated capable of coping with fuzzy input and output without any need for a new network topology or learning algorithm.

This paper presents a new fuzzy modeling technique via the so-called sector nonlinearity concept. To fully take advantage of the sector nonlinearity concept, we propose a new type of Takagi-Sugeno fuzzy model and develop an algorithm to identify model parameters. The algorithm consists of two steps. The purpose of the first step is to determine sector coefficients from input-output data. The second part identifies membership functions from the determined sector coefficients and the input-output data. Identification examples illustrate the utility of this approach.

This paper presents an evolutionary learning process for linguistic modeling with weighted double-consequent fuzzy rules. These kinds of fuzzy rules are used to improve the linguistic modeling, with the aim of introducing a trade-off between interpretability and precision.

The use of weighted double-consequent fuzzy rules makes more complex the modeling and learning process, increasing the solution search space. Therefore, the cooperative coevolution, an advanced evolutionary technique proposed to solve decomposable complex problems, is considered to learn these kinds of rules. The proposal has been tested with different problems achieving good results.

Neural networks with a structure adaptation capability that are equivalent to fuzzy systems are investigated with the goal of designing hardware architectures for application in time critical classification problems. The consecutive steps adopted to build a realtime prototype of a structure adapting neural network include the modification of learning algorithms making them suitable for hardware implementation, the development of a hardware architecture, and the system implementation and test. Structure adaptation must be integrated into the learning algorithm implementable in hardware. Fusion of information hidden in data with the existing knowledge expressed in rules should also be possible. Three types of structure adapting algorithms have been mapped onto the proposed hardware architecture. Performance of the system implementation is evaluated comparing the speeds using software implementations for benchmarks and real applications.

This paper presents a quantitative analysis of evolvability with evolutionary activity statistics in an evolutionary fuzzy system. In general, one can estimate the performance of an evolved fuzzy controller by its fitness. However, it is difficult to explain how its fitness or adaptability has been obtained. Evolutionary activity is used to measure the evolvability of fuzzy rules and explain why salient rules have higher evolvability. A genetic algorithm is used to construct a fuzzy logic controller for a mobile robot in simulation environments. The quantitative analysis shows that sufficient evolvability is maintained during the evolution and that it contributes to the construction of the optimal controller.

Association rules in data mining are used in order to discover latent information hidden in databases. The class of apriori algorithms provide a simple way to compute association rules. Despite the simplicity of the algorithm its computational complexity is quite high. Specifically, the computational performance degrades due to intense computational requirements during the phase of frequent itemset generation.

In this paper we have proposed a novel procedure to extract frequent itemsets. The itemset is modeled as a cell of a hypercube, and this cell is represented as a lexicographical order combination. The procedure described in this paper reduces the number of database passes to a single pass to extract frequent itemsets. Thus, the procedure is computationally attractive.

The article addresses the problem of adaptive learning in a neuro-fuzzy network based on Sugeno-type fuzzy inference. A new learning algorithm for tuning of both the antecedent and consequent parts of the fuzzy rules is proposed. The algorithm is derived from the Hartley and Marquardt methods. A characteristic feature of the proposed algorithm is that it does not include time-consuming matrix inversion operations. Simulation results prove the high performance of the algorithm and illustrate its application to time series forecasting.