IV. Polimerización: reacción debida al entrecruzamiento entre dos átomos de carbono de grasas insaturadas o a los enlaces de oxígeno
2.2. Aditivos de mejora de las cualidades sensoriales
This section analyzes the application of Neuro-fuzzy technique in the domain of fault diagnosis.
Salahshoor et al. [144] have devised an innovative data-driven fault detection and diagnosis methodology on the basis of a distributed configuration of three adaptive neuro-fuzzy inference system for an industrial power plant steam turbine. Each neuro-fuzzy classifier has been developed for a dedicated category of four steam turbine faults. A proper selection of four measured variables has been configured to feed each classifier with the most influential diagnostic information. A diverse set of test scenarios has been carried out to illustrate the successful diagnostic performances of the proposed fault detection system. Sadeghian et al. [145] have used nonlinear system identification method to predict and detect process fault of a cement rotary kiln. To identify the various operation points in the kiln, locally linear neuro- fuzzy model trained by LOLIMOT algorithm has been adapted by the authors. Then, using this method, they have obtained three distinct models for the normal and faulty situations. At the end, they have checked the proposed technique with the validation data. Eslamloueyan et al. [146] have proposed a hierarchical artificial neural network (HANN) for isolating the faults of the Tennessee–Eastman process which is the simulation of a chemical plant created by the Eastman Chemical Company to provide a realistic industrial process for evaluating process control and monitoring methods. Fuzzy clustering algorithm has been used by them to divide the fault patterns space into a few sub-spaces. They have developed supervisor network along with the special neural networks to diagnose the fault present in the system. Simon et al. [147] have describes the pattern recognition based data analysis of an existing industrial batch dryer, and the comparison of three artificial intelligence techniques suited to perform classification tasks: neural networks, neuro-fuzzy and Takagi–Sugeno fuzzy models. They have found that the neural networks trained with the Bayesian regularization have shown the most robust classification performance with respect to other two methods. They have claimed that since the proposed method for pattern recognition is not case specific it can be directly used for the monitoring of a large variety of drying processes. Quteishat et al. [148] have proposed a modified fuzzy min-max network for improved performance when large hyper boxes are formed in the network. This methodology is used to facilitate the extraction of rule set from FMM to justify the predictions. The results from the developed
FMM have been authenticated with the sensor measurements collected from a power generation plant for fault diagnosis. Topcu et al. [149] have studied the optimum uses of pozzolans as supplementary cementing material for blended cement production. They have developed a system based on artificial neural network and fuzzy logic for predicting the strength parameters for different types of cement motars. Tran et al. [150] presented a fault diagnosis technique based on adaptive neuro-fuzzy inference system in combination with classification and regration tree. The ANFIS model has been trained with the results obtained from the least square algorithm. They have observed that the developed ANFIS model has the potential for fault diagnosis of induction motors. Fang et al. [151] have explored performance of a structural damage defection technique based on frequency response and neural network. In this paper they have investigated a tunable steepest discount algorithm using heuristics approach for improving the converging speed. From the analysis of the result of the proposed method for a cantilever beam they have concluded that the neural network technique can estimate the damage condition with high accuracy. Beena et al. [152] have proposed a new approach for fault detection in structural system using fuzzy logic technique and neural network based on hebbin-learning. They have used the continuum mechanics and finite element method to measure the vibration parameters because of structure damage. The developed technique works quite well for structural damage even in the presence of noise. Kuo et al. [153] have presented a symbiotic evolution based fuzzy neural diagnostic system for fault detection of a propeller shaft used in the marine propulsion system. The system auto-generates its own optimal fuzzy neural architecture for fault diagnosis. They have stated that the results from the hybrid fuzzy neural system have been found to be more closure with the real conditions than the other traditional methods. Ye et al. [154] have developed a new online diagnostic algorithm to find out the mechanical fault of electrical machine using wave let packet decomposition method and adaptive neuro fuzzy inference system. According to them the new integrated fault diagnostic system significantly reduces the seal complexity, and computational time of the system. They have validated results from the diagnostic technique for a 3-phase induction motor drive system. Kuo [155] has proposed a fault detection system using data acquisition, feature extraction and pattern recognition for detecting faults of blades by applying multiple vibration sensors. The feature extraction algorithm has been developed based on back propagation artificial neural network. The fuzzy
logic technique has been employed to speed up the training speed. According to him the results from the system are very close to the results obtains from the experimental analysis. Zio et al. [156] have presented a fault diagnostic problem using neuro fuzzy approach. They have used this approach for the purpose of high rate of correct classification and to obtain an easily interpretable classification model. The efficiency of the approach has been verified by applying to a motor bearing system and the results obtained are quite encouraging. Wang et al. [157] have presented the comparison of the performance for two fault diagnosis system that is recurrent neural networks and neuro fuzzy systems using two benchmark data sets. As described by them, it is found that the neuro fuzzy prognostic system is more reliable for machine health condition monitoring than the neural network fault diagnostic system. Zhang et al. [158] have proposed a bearing fault detection technique based on multi scale entropy and adaptive neuro fuzzy inference system (ANFIS) to measure the nonlinearity existing in a bearing system. They have conducted experiments on electrical motor bearing with three different fault categories and the results obtained from the experimentation have been used to design and train the ANFIS system for fault diagnosis.
2.3.3.5.2 Genetic-fuzzy technique
The research papers reviewed from the domain of application of Genetic-fuzzy technique for crack and fault detection in structural and mechanical systems are presented in this section. Wu et al. [159] have presented a new version of fuzzy support vector machine to diagnose faults in automatic car assembly. The input and output variables have been described by them as fuzzy numbers in the fuzzy based system. They have shown that the modified GA helps the fuzzy support vector classifier machine to seek optimized parameters. The results from their methodology in car assembly for fault diagnosis confirm the feasibility and the validity of the diagnosis method. Pan et al. [160] have analyzed the effect of random delays in network controlled system by using fuzzy PID models. They have tuned the models by minimizing the time multiplied absolute error and squared model output with stochastic algorithms viz. the GA and particle swarm optimization. After analyzing the performance of the algorithm they have shown that random variation in network delay can be handled efficiently with fuzzy logic based PID models over other techniques as mentioned in the
paper. Pawar et al. [161] have devised a structural health monitoring methodology using genetic fuzzy system for online damage detection. They have used displacement and force based measurement deviations between damage and undamaged condition for building the rules and data pool for the fuzzy and genetic system respectively. The developed methodology has been applied for composite rotor blades and the results are found to be encouraging. Yuan et al. [162] have proposed an artificial immunization algorithm (AIA) to optimize the parameters obtained from support vector machines (SVM) generally used as machine learning tool for fault-diagnosis. They have used the proposed fault diagnosis model for a turbo pump rotor and found that the SVM optimized by AIA gives higher accuracy than the normal SVM.