Neuro-fuzzy Inference System Research Articles

A Novel Neuro-Fuzzy Inference System with Multi-Level Membership Function for Classification Applications

In this paper, a novel neuro-fuzzy inference system with multi-level membership function (NFIS_MMF) for classification applications is proposed. The NFIS_MMF model is a five-layer structure, which combines the traditional Takagi-Sugeno-Kang (TSK). Layer 2 of the NFIS_MMF model contains multi-level membership functions, which are multilevel activation functions. A self-constructing learning algorithm, which consists of the self-clustering algorithm (SCA), fuzzy entropy, and the backpropagation algorithm, is also proposed to construct the NFIS_MMF model and perform parameter learning. Simulations were conducted to show the performance and applicability of the proposed model.

Prediction of sulfate expansion of PC mortar using adaptive neuro-fuzzy methodology

The prediction of sulfate attack is essential for concrete structures since it causes drastic decrements in strength and in expansion attributes of cementitious systems. In this study, the nonlinear mapping among sulfate expansion of PC mortar and some selected parameters (C 3A content, C 3S/C 2S ratio, sulfate concentration and mineral admixture substitution level) was simulated using adaptive neuro-fuzzy system. Experimental data that had been previously collected for various levels of accounted parameters were treated in the analyses. In neuro-fuzzy inference system, Sugeno-type inference technique and linear output function were used to perform approximate reasoning of fuzzy input variables. In addition, hybrid learning algorithm, combining backpropagation learning and linear least-squares estimator, were preferred for the adaptation of free parameters. Consequently, neuro-fuzzy model was compared with results obtained using linear and nonlinear multiple regression methodologies to make comparison among different techniques. Outcomes indicated that neuro-fuzzy model exhibits superior performance.

Extraction of foetal ECG by combination of singular value decomposition and neuro-fuzzy inference system

Extraction of foetal ECG that is embedded in the maternal ECG is a challenging problem. This paper presents a combined system to extract foetal ECG from maternal abdominal ECG. The system uses a combination of singular value decomposition (SVD) and a neuro-fuzzy inference system. The SVD is used to construct two reference signals, while the fuzzy system is used as an adaptive canceller. The algorithm is applied on synthetic as well as real data and the results are presented. In addition, the paper presents an example of using the same system as a noise removal tool.

A neuro-fuzzy inference system for modeling and prediction of heart rate variability in the neuro-intensive care unit

In the neurological intensive care unit (NICU), prediction of impending changes in patient condition would be highly beneficial. In this paper, we employ a neuro-fuzzy inference system (NFIS) for short-term prediction of heart rate variability in the NICU. An NFIS was selected because it allows for a “gray-box” approach through which a system identification procedure is used in conjunction with fuzzy modeling. The NFIS is described in detail and is compared to an auto-regressive moving average (ARMA) model for its ability to model both simulated and measured data from NICU patients. We found that the NFIS is capable of predicting changes in heart rate to a reasonable extent, and that the NFIS has both advantages and limitations over the ARMA model. The NFIS may therefore be a reasonable technique to consider for more extensive prediction purposes in ICU settings.

터보제트엔진의 퍼지제어기 설계 및 다목적함수 만족기법을 통한 제어성능 향상에 관한 연구

터보제트엔진 모델에 대한 제어에 있어서, 비교적 잘 설계된 PI 제어기 성능결과를 바탕으로 Takagi-Sugeno형 뉴로-퍼지 추론계를 통한 플랜트 모델의 제어 시스템을 규명함으로서, PI형 T-S 퍼지규칙들을 퍼지제어기를 설계하였다. 이렇게 설계된 제어기의 성능을 향상시키기 위하여, 각 퍼지규칙들을 퍼지 C-Means Algorithm으로부터 각각의 목적 함수군으로 분류한 후, 각 분류군에 대해 규칙간의 가중치가 각 목적함수의 만족도에 부합되도록 하는 기법을 제시하였고, 이를 잘 설계된 T-S형 퍼지제어기에 적용하여 성능을 향상시킴으로써 그 유용성을 보였다. In the study of control technique for a turbojet engine model, the Takagi-Sugeno fuzzy logic controller has been designed based on the model identification by the well designed PI controlled system through T-S neuro-fuzzy inference system. To enhance this designed controller, those procedures are proposed that certainty factors are adopted to each rule of objective groups which are classified by the fuzzy C-Means algorithm and the satisfaction degrees are matched to meet the objectives. This proposed technique shows its feasibility by upgrading performances of the previously well-designed T-S fuzzy controller.

The local paradigm for modeling and control: from neuro-fuzzy to lazy learning

The composition of simple local models for approximating complex nonlinear mappings is a common practice in recent modeling and control literature. This paper presents a comparative analysis of two different local approaches: the neuro-fuzzy inference system and the lazy learning approach. Neuro-fuzzy is a hybrid representation which combines the linguistic description typical of fuzzy inference systems, with learning procedures inspired by neural networks. Lazy learning is a memory-based technique that uses a query-based approach to select the best local model configuration by assessing and comparing different alternatives in cross-validation. In this paper, the two approaches are compared both as learning algorithms, and as identification modules of an adaptive control system. We show that lazy learning is able to provide better modeling accuracy and higher control performance at the cost of a reduced readability of the resulting approximator. Illustrative examples of identification and control of a nonlinear system starting from simulated data are given.

Pin power reconstruction for CANDU reactors using a neuro-fuzzy inference system

A neuro-fuzzy inference system has been developed for reconstructing fuel pin powers from Canada deuterium uranium (CANDU) core calculations performed with a coarse-mesh finite difference diffusion approximation and single-assembly lattice calculations. The neuro-fuzzy inference system is trained by a genetic algorithm and a least-squares method using the partial core calculation results of two 6/spl times/6 fuel bundle models. Verification tests have been performed for two partial core benchmark problems composed of other 6/spl times/6 and 3/spl times/3 fuel bundles. The reconstructed pin powers are compared with the reference solutions obtained with the detailed collision probability calculations using the HELIOS lattice analysis code. The results indicate that the proposed reconstruction algorithm is accurate, yielding the error due to the reconstruction scheme of less than 0.5%.

Restoration scheme of mobility databases by mobility learning and prediction in PCS networks

This paper proposes a restoration scheme based on mobility learning and prediction in the presence of the failure of mobility databases in personal communication systems (PCSs). In PCSs, mobility databases must maintain the current location information of users to provide a fast connection for them. However, the malfunction of mobility databases may cause some location information to be lost. As a result, without an explicit restoration procedure, incoming calls to users may be rejected. Therefore, an explicit restoration scheme against the malfunction of mobility databases is needed to guarantee continuous service availability to users. Introducing mobility learning and prediction into the restoration process allows systems to locate users after a failure of mobility databases. In failure-free operations, the movement patterns of users are learned by a neuro-fuzzy inference system (NFIS). After a failure, an inference process of the NFIS is initiated and the users' future location is predicted. This is used to locate lost users after a failure. This proposal differs from previous approaches using a checkpoint because it does not need a backup process nor additional storage space to store checkpoint information. In addition, simulations show that our proposal can reduce the cost needed to restore the location records of lost users after a failure when compared to the checkpointing scheme.

Simulation of a Mobility Prediction Scheme Based on Neuro-Fuzzy Theory in Mobile Computing

Mobility prediction is an important issue for efficient management in mobile computing. Whenever users move to a new location, mobile systems should know about their location change so that the connection is not in terrupted. In this paper, we focus on the simulation of a user mobility prediction scheme based on neuro-fuzzy theory. Our scheme predicts future locations from the movement patterns of users accumulated over several days or months, i.e., the user's location history. In or der to model users' movements, we use a neuro-fuzzy inference system. Whenever a user moves to a new lo cation, the system autonomously learns various move ment patterns. The prediction of the future location to which users will move is obtained by the approximate reasoning of the neuro-fuzzy inference system. This ap proach can derive an appropriate future location from learned location information, even if the mobile systems do not know the actual location history. Simulation re sults show that our prediction scheme has high accuracy for various users' patterns. In addition, our scheme is applicable, with considerable prediction accuracy, to users whose history has not been learned by the neuro- fuzzy inference system.

The Use of Routine Weather Observations to Calculate Liquid Water Content in Summertime High-Elevation Fog

This paper represents a stage within a larger project to estimate acid ion deposition from cloud impacting on high-elevation forests. Acid ion deposition depends principally on three factors: the liquid water content (LWC), the ion concentration(s) in fog or cloud water, and the efficiency of the deposition process. In the present paper, the objective is to estimate LWC on Roundtop Mountain in southern Quebec from routine meteorological measurements at the Sherbrooke weather station. After describing preliminary efforts, the methodology that was found to work best is presented. This scheme was a hybrid of applications of two statistical nonlinear regression schemes. First, the classification and regression trees (CART) algorithm was applied to predict the occurrence or nonoccurrence of fog at Roundtop. The algorithm produced by this application permitted the elimination of a large proportion of the data records for which fog was very unlikely to occur at Roundtop. The remaining data were then processed by a second application of CART to determine the predictors that are important for estimating LWC at Roundtop. Finally, these same remaining data were processed by the neuro-fuzzy inference systems (NFIS) algorithm to derive the final prediction algorithm. This hybrid method (CART‐CART‐NFIS) achieved a correlation coefficient of 0.810, with accuracies of 0.962 and 0.664 for the no-fog and fog events, respectively. (Corresponding threat scores were 0.916 and 0.530, respectively.) These measures of skill were significantly better than those obtained from initial estimates or from schemes that used CART alone. Although optical cloud detector and LWC data are necessary for derivation of the fog-occurrence and LWC prediction algorithms, in the end those algorithms are applied to only the predictor data. Fog-occurrence and LWC data are not required, except for verification purposes. The algorithms and list of predictors still need to be tested to determine how widely applicable they are.

Empirical-statistical reconstruction of surface marine winds along the western coast of Canada

CANFIS, an empirical-statistical technique, is used to reconstruct continuous daily surface marine winds at 6-hourly intervals at 13 Canadian buoy sites along the western coast of Canada for the 40 yr period 1958-1997. CANFIS combines Classification and Regression Trees (CART) and the Neuro- Fuzzy Inference System (NFIS) in a 2-step procedure. CART is a tree-based algorithm used to optimize the process of selecting relevant predictors from a large pool of potential predictors. Using the selected predictors, NFIS builds a model for continuous output of the predictand. In this project we used CAN- FIS to link large-scale atmospheric predictors with regional wind observations during a learning phase from 1990 to 1995 in order to generate empirical-statistical relationships between the predictors and buoy winds. The large-scale predictors are derived from the NCAR/NCEP 40 yr reanalysis project while the buoy winds come from the Canadian Atmospheric Environment Service buoy network. Vali- dation results with independent buoy wind data show a good performance of CANFIS. The CANFIS winds reproduce the independent buoy winds with greater accuracy than winds reconstructed with a stepwise multivariate linear regression technique. In addition, they are better than the NCEP reana- lyzed winds interpolated to the buoy locations. The reconstructed statistical winds recover more than 60% of the observed wind variance during an independent verification period. In particular, correlation coefficients between independent buoy wind time series and CANFIS wind time series vary between 0.61 and 0.98. Our results suggest that CANFIS is a successful downscaling method. It is able to recover a substantial fraction of the variation of surface marine winds, especially along coastal regions where ageostrophic effects are relatively important.