Nikolinka Christova

Growing RBF Network Models For Solving Nonlinear Approximation And Classification Problems.

In this paper a multi-step learning algorithm for creating a Growing Radial Basis Function Network (RBFN) Model is presented and analyzed. The main concept of the algorithm is to gradually increase by one the number of the Radial Basis Function (RBF) units at each learning step thus gradually improving the total model quality. The algorithm stops automatically, when a predetermined (desired) approximation error is achieved. An essential part in the whole algorithm plays the optimization procedure that is run at each step for increasing the number of the RBFs, but only for the parameters of the newly added RBF unit. The parameters of all previously RBFs are kept at their optimized values at the previous learning steps. Such optimization strategy, while suboptimal in nature, leads to significant reduction in the number of the parameters that have to be optimized at each learning step. A modified constraint version of the particle swarm optimization (PSO) algorithm with inertia weight is developed and used in the paper. It allows obtaining conditional optimum solutions within the range of preliminary given boundaries, which have real practical meaning. A synthetic nonlinear test example is used in the paper to estimate the performance of the proposed growing algorithm for nonlinear approximation. Another application of the Growing RBFN model is illustrated on two examples for data classification, one of them from the publicly available red wine quality data set.

A Solution for Integrated Manufacturing Operation Management

In this paper, an industrial software solution for integrated manufacturing operation management will be presented. The developed products seamlessly connect users, teams and knowledge so the people can work more efficiently by taking advantage of relevant information across business processes. The system is aimed to equip managers with the necessary tools for implementing a true just-in-time, Six Sigma and lean manufacturing methodologies. The software suite is based on standard products and provides services to external systems in order to cover the whole process of manufacturing execution. A brief description of the created system will be given as well as further implementations will be discussed.

Production Planning for Complex Plants using Fuzzy Cognitive Maps

Abstract This work emphasizes on the implementation of soft computing techniques for the production planning of complex manufacturing plants. A hierarchical control structure has been assumed to control and to optimize the chemical process of a complex plant with successful results and at the highest hierarchical level the production planning should meet the fluctuating demands in optimal way. This level integrates the operational and business management requirements where the soft computing technique of Fuzzy Cognitive Maps is proposed to model these tasks. The strategic planning for the hierarchical integrated system realizes the optimal total management at the corporate level. The obtained simulation results prove the applicability of the proposed methodology and the advantages of using soft computing modeling techniques for the sophisticated high level of hierarchical systems.

Modeling and multilevel control of complex multistage processes with recycles and intermediate storages

Abstract A multilevel hierarchical control system for a single end product multistage plant is under consideration. It is shown that a variety of models and/or hybrid modeling are reasonable for inferred key variable forming, monitoring and control. Because of the complexity, uncompleted information, large uncertainties and mixture of batch, semi-batch and continuous processes different control strategies are appropriate for each hierarchical level. Two level reactive scheduler is proposed. Integration of operational and business management is carried out at the top hierarchical level using Fuzzy Cognitive Maps (FCM) approach. Some results of practical application of the system are reported.

Partial fuzzy grid models and application to condition monitoring of photovoltaic systems

This paper presents a computational strategy for condition monitoring of multi-mode processes on the example of real data from a photovoltaic system. The concept uses a new type of fuzzy models with partially activated set of fuzzy rules on a pre-determined grid, called partial fuzzy grid models. Each such model represents one specific operating condition of the process, which is saved in the Model Base. Then the (unknown) operation data are processed by all of the partial models and produce their own estimations. The model with the best estimation shows the most similar operation to the newly presented one. The novelty in the paper is focused on the concept and the learning algorithm for the partial fuzzy grid models as well as on the calculation steps for the condition monitoring.

Incremental RBF network models for nonlinear approximation and classification

In this paper a multistep learning algorithm for creating a novel incremental Radial Basis Function Network (RBFN) Model is presented and analyzed. The proposed incremental RBFN model has a composite structure that consists of one initial linear sub-model and a number of incremental sub-models, each of them being able to gradually decrease the overall approximation error of the model, until a desired accuracy is achieved. The identification of the entire incremental RBFN model is divided into a series of identifications steps applied to smaller size sub-models. At each identification step the Particle Swarm Optimization algorithm (PSO) with constraints is used to optimize the small number of parameters of the respective sub-model. A synthetic nonlinear test example is used in the paper to analyze the performance of the proposed multistep learning algorithm for the incremental RBFN model. A real wine quality data set is also used to illustrate the usage of the proposed incremental model for solving nonlinear classification problems. A brief comparison with the classical single RBFN model with large number of parameters is conducted in the paper and shows the merits of the incremental RBFN model in terms of efficiency and accuracy.

On-line control performance monitor with robust properties

A control performance monitor (CPM) as a software agent is developed for assessment of the SISO control loops behaviour. A dedicated system for data pre-processing is applied in order to guaranty robust properties of the extracted process features. Using a fuzzy and rule based evaluation procedure an overall performance index is also calculated from the estimated indicators. The CPM provides early warning information when an inadmissible performance is detected. Some results of experimental verification of the CPM are reported.

An integrated approach to intelligent modeling of industrial plants

In this paper an integrated hierarchical soft computing methodology for modeling of industrial plants by aggregating models of different types is presented. The problem of designing adequate and reliable models for non-linear plants with large uncertainties in under consideration here. The proposed approach has the ability to model system behavior under different circumstances and it is especially efficient for complex industrial systems with immeasurable process variables and large uncertainties. A fuzzy cognitive map (FCM) is used to aggregate multiple models and to create a hybrid model, which makes a selection between the different models, according to the current operational conditions of the industrial process. The proposed methodology is considered as a promising way to cope with the modeling of a real industrial plant.

Multi-dimensional Fuzzy Modeling with Incomplete Fuzzy Rule Base and Radial Basis Activation Functions

A new type of a fuzzy model is proposed in this paper. It uses a reduced number of fuzzy rules with respective radial basis activation functions. The optimal number of the rules is defined experimentally and their locations are obtained by clustering or by PSO optimization procedure. All other parameters are also optimized in order to produce the best model. The obtained model is able to work with sparse data in the multidimensional experimental space. As a proof a synthetic example, as well as a real example of a 5-dimensional sparse data have been used. The results obtained show that the PSO optimization of the fuzzy rule locations is a better approach than the clustering algorithm, which utilizes the distribution of the available experimental data.

Optimization strategies for tuning the parameters of radial basis functions network models

In this paper the problem of tuning the parameters of the RBF networks by using optimization methods is investigated. Two modifications of the classical RBFN, called Reduced and Simplified RBFN are introduced and analysed in the paper. They have a smaller number of parameters. Three optimization strategies that perform one or two steps for tuning the parameters of the RBFN models are explained and investigated in the paper. They use the particle swarm optimization algorithm with constraints. The one-step Strategy 3 is a simultaneous optimization of all three groups of parameters, namely the Centers, Widths and the Weights of the RBFN. This strategy is used in the paper for performance evaluation of the Reduced and Simplified RBFN models. A test 2-dimensional example with high nonlinearity is used to create different RBFN models with different number of RBFs. It is shown that the Simplified RBFN models can achieve almost the same modelling accuracy as the Reduced RBFN models. This makes the Simplified RBFN models a preferable choice as a structure of the RBFN model.