Thermal Lifetime Evaluation of Electrical Machines Using Neural Network

In this study, three artificial neural network (ANN) models are developed with back propagation (BP) optimization algorithms to predict various lightning damage modes in carbon/epoxy laminates. The proposed ANN models use three input variables associated with lightning waveform parameters (i.e., the peak current amplitude, rising time, and decaying time) to predict fiber damage, matrix damage, and through-thickness damage in the composites. The data used for training and testing the networks was actual lightning damage data collected from peer-reviewed published literature. Various BP training algorithms and network architecture configurations (i.e., data splitting, the number of neurons in a hidden layer, and the number of hidden layers) have been tested to improve the performance of the neural networks. Among the various BP algorithms considered, the Bayesian regularization back propagation (BRBP) showed the overall best performance in lightning damage prediction. When using the BRBP algorithm, as expected, the greater the fraction of the collected data that is allocated to the training dataset, the better the network is trained. In addition, the optimal ANN architecture was found to have a single hidden layer with 20 neurons. The ANN models proposed in this work may prove useful in preliminary assessments of lightning damage and reduce the number of expensive experimental lightning tests.

Under the strategic framework of global carbon emission reduction, it is urgent to reduce the energy consumption of centrifugal pumps. Monitoring pump’s operational states can effectively help people to understand system’s behavior, and achieve the optimal energy management on pump equipment. In view of installation space, cost, reliability and other factors, it is necessary to find an alternative to the traditional sensor monitoring scheme. Therefore, it is significant to develop a prediction method for the operational state of centrifugal pump using sensorless technology. In this paper, a neural network estimation model based on Bayesian regularization back propagation (BRBP) algorithm is proposed. Compared with the traditional QP estimation model, the new model has better nonlinearity and higher precision. Using power and rotational frequency as input neurons, BRBP model realizes the output of flow rate and head through neuronal calculations. By comparing the predicted flow rate and head values of a multistage centrifugal pump with the measured values at different rotational frequencies, it is found that the mean relative error of flow rate of BRBP model is in the range of 1% to 4%, and the mean relative error of head is in the range of 1% to 2%. The prediction accuracy of BRBP neural network model is obviously higher than that of QP model, which meets the requirements of industrial application.

It is difficult to accurately reckon vehicle position for vehicle navigation system (VNS) during GPS outages, a novel prediction algorithm of dead reckon (DR) position error is put forward, which based on Bayesian regularization back-propagation (BRBP) neural network. DR, GPS position data are first de-noised and compared at different stationary wavelet transformation (SWT) decomposition level, and DR position error data are acquired after the SWT coefficients differences are reconstructed. A neural network to mimic position error property is trained with back-propagation algorithm, and the algorithm is improved for improving its generalization by Bayesian regularization theory. During GPS outages, the established prediction algorithm predictes DR position errors, and provides precise position for VNS through DR position error data updating DR position data. The simulation results show the positioning precision of the BRBP algorithm is best among the presented prediction algorithms such as simple DR and adaptive linear network, and a precise mathematical model of navigation sensors isn’t established. DOI: http://dx.doi.org/10.11591/telkomnika.v11i7.285 2 Full Text: PDF

For a continuous mode of operation, insulating material in an electrical machine is subject to constant thermal, electrical, mechanical and environmental stresses where thermal stress is a major cause of gradual insulation deterioration, which leads to ultimate winding failure. To guarantee a satisfactory lifetime, electrical machines are designed to operate winding temperatures well below their thermal class, which results in an oversized design. Standard methods for thermal lifetime evaluation of electrical machines are based on accelerated aging tests that require several months of testing. This paper proposes an alternative approach relying on a supervised neural network that significantly shortens the time demanded by accelerated aging tests for thermal lifetime evaluation of electrical machines. The supervised neural network is based on a feedforward neural network trained with Bayesian Regularisation Backpropagation (BRP) algorithm. The network predicts the wire insulation resistance with respect to its aging time at aging temperatures of 250°C, 270°C and 290°C, which reveals a good match of prediction outcomes against the experimental findings. The mean time-to-failure at each aging temperature is extracted using the Weibull probability plot in order to compare the Arrhenius curves for both conventional and proposed method and a relative error of 0.125% is achieved in terms of their temperature indexes. In addition, the analysis shows a time saving of 1680 hours (57% time saved of experimental test procedure) when the thermal life of the insulating material is predicted using BRP neural network.

Artificial Neural Network (ANN) based tap changer control of closed primary bus and cross network connected parallel transformers has demonstrated potential use in power distribution system. In those research works the proposed ANN for application in this control were developed using various algorithms and concluded that a network trained by Bayesian Regularization (BR) backpropagation algorithm produced the best performance measured in terms of correct tap changing decisions. However, further improvement of ANN based transformer tap changer operation is always desirable. A general rule for obtaining good generalization is to use the smallest network that solves the problem. In this paper, we show that a small sized ANN is obtainable for further improvement of transformer tap changer operation by modifying the standard Cascade-Correlation algorithm. The modification incorporates weight smoothing of output layer weights in Cascade-Correlation learning using Bayesian frame work. Experimental results demonstrate that significant improvement in performance is achieved when an ANN is trained by modified Cascade-Correlation algorithm instead of standard Cascade-Correlation or Bayesian Regularization backpropagation algorithm. A comparison of performances of different algorithms in application to transformer tap changer operation is analyzed and the results are presented.

This paper investigates the effect of sizing agent molecular weight on carbon fiber (CF) surface properties and the effect of thermal aging time on mechanical properties of CF/epoxy composites. The surface properties of three CCF800 CF samples with varying sizing agent molecular weight were characterized by surface morphology, surface roughness, chemical functional groups, and element composition. The results showed that the sample with low molecular weight exhibited low roughness and high proportion of activated carbon atoms. The flexural strength, flexural modulus, and interlaminar shear strength of CCF800/5228 composites were measured at 25°C and 150°C by thermal ageing time 0, 100 h, 250 h, 500 h, and 1000 h. The results showed that the thermal aging time up to 1000 h had little effect on the flexural modulus, and the interlaminar shear performance at 150°C showed a trend of increasing at the beginning and then decreasing.

In applying neural network to identification, back propagation (BP) algorithm is usually trapped to a local optimum and has a low speed of convergence, whereas particle swarm optimization (PSO) is advantageous in terms of global optimal searching. In this paper, a new algorithms which combines guaranteed convergence particle swarm optimizer (GCPSO) algorithm with BP (Back Propagation) algorithm is introduced in this paper and applied to BP neural networks to optimize the parameters of BP neural networks so as to improve the convergence speed and precision of BP neural networks. Compared with the BP algorithm and PSO-BP algorithm, the results of the simulation show that the algorithm of BP Neural Network Model Based on Guaranteed Convergence Particle Swarm Optimization speeds up the convergence process and enhances the accurate rate in pattern recognition.

A novel method for determining the stress-strain constitutive relationship of high-grade pipeline weld zone material

Drying paddy with low-pressure superheated steam (LPSS) can effectively increase the γ-aminobutyric acid content in paddy. This study aimed to investigate the characteristics and mathematical models (MMs) of thin-layer drying of paddy with LPSS. The experimentally obtained data were fitted by nonlinear regression with 5 MMs commonly used for thin-layer drying to calculate the goodness of fit of the MMs. Then, the thin-layer drying of paddy with LPSS was modeled with two machine learning methods as a Bayesian regularization back propagation (BRBP) neural network and a support vector machine (SVM). The results showed that paddy drying with LPSS is a reduced-rate drying process. The drying temperature and operating pressure have a significant impact on the drying process. Under the same pressure, increasing the drying temperature can accelerate the drying rate. Under the same temperature, increasing the operating pressure can accelerate the drying rate. The comparison of the model evaluation indexes showed that 5 common empirical MMs (Hederson and Pabis, Page, Midilli, Logarithmic, and Lewis) for thin-layer drying can achieve excellent fitting effects for a single experimental condition. However, the regression fitting of the indexes by calculating the coefficient(s) of each model showed that the empirical MMs produce poor fitting effects. The BRBP neural network-based model was slightly better than the SVM-based model, and both were significantly better than the empirical MM (the Henderson and Pabis model), as evidenced by a comparison of the training root mean square error (RMSE), testing RMSE, training mean absolute error (MAE), testing MAE, training R2, and testing R2 of the Henderson and Pabis model, the BRBP neural network model, and the SVM-based model. This result indicates that the MMs established by the two machine learning methods can better predict the moisture content changes in the paddy samples dried by LPSS. Keywords: paddy, low-pressure superheated steam, drying, mathematical model, characteristic DOI: 10.25165/j.ijabe.20231601.7810 Citation: Li Y, Che G, Wan L, Zhang Q L, Qu T Q, Zhao F Z. Characteristics and mathematical models of the thin-layer drying of paddy rice with low-pressure superheated steam. Int J Agric & Biol Eng, 2023; 16(1): 273–282.

Accelerating neural network training using weight extrapolations

Instantaneous learning is a desirable feature in neural networks. This type of learning enables the network to be trained very quickly, typically in just one or two passes of the training set as opposed to hundreds or even thousands of passes for networks trained by an iterative process, such as the error backpropagation (BP) algorithm. This dissertation first reviews several existing types of neural networks with instantaneous learning capability. It then proposes a new network called the Fuzzy Classification (FC) Neural Network that can be trained with two passes of the training samples. The first pass assigns the synaptic weights for the input and output layers. The second pass determines the radius of generalization r for each training sample. The network exhibits fuzziness in two regards: (1) by fuzzification of the location of each training vector in the input space; and (2) by assigning fuzzy memberships of output classes to new input vectors. The operation of the FC network is analyzed from different perspectives, namely, separability of patterns, curve fitting and kernel regression. It is shown that one mode of operation of the FC network can be made to behave like a Radial Basis Function (RBF) network by proper choice of membership function and network parameter. The performance of the FC network as a pattern classifier is also discussed in a statistical framework. The generalization performance of the FC network is compared against that of the CC4 Comer Classification Neural Network, the RBF network, and the Multilayer Perceptron network trained by the BP algorithm. Experiments involving a Henon map time series, a Mackey-Glass time series, and a spiral pattern are used in the comparison. The performance of the FC network is found to be comparable to that of RBF and BP networks, and better than that of the CC4 network. Its performance is also more scalable than those of CC4 and RBF networks. Further, the time taken to design an FC network for any given problem is much shorter compared to the other three networks.

an Artificial Neural Network classifier is a nonparametric classifier. It does not need any priori knowledge regarding the statistical distribution of the class in a giver selected data Source. While, neural network can be trained to distinguish the criteria used to classify easily in a generalized manner that allows successful classification the newly arrived inputs not used during training. Through this paper it is eastliblished that back propagation neural network works successfully for the purpose of classification. Back propagation suffers from getting stuck into Local Minima. Weight optimization in Back propagation can be optimized using the Genetic Algorithm (GA). The back propagation algorithm is improved by invoking Genetic algorithm, to improve the overall performance of the classifier. The performance of a fitness algorithm using the approach suggested by us is a Hybrid System that is being analyzed in this paper. In this paper the issue of improving the fitness (weight adjustment) of Back propagation algorithm is addressed. Some of the Advantages of Hybrid algorithms are: convergence speed will be increased and the local minima problem can be overcome. The proposed Hybrid Algorithm is to perform learning as a back propagation and optimize weights using GA for classification.

Recently, as a kind of transformer oil with lots of outstanding characteristics such as high ignition point, reproducible, and natural degradation, vegetable insulating oil has becomes more and more popular in the oil-filled transformers. Thermal ageing as well as thermal and electrical faults, which can lead to the decomposition of insulating oil and insulating paper, will occur in operating transformers, then dissolved gases are formed and reach equilibrium in oil through diffusion process. In this paper, camellia oil as sample oil, thermal ageing as well as thermal and electrical faults simulation experiments of transformers were conducted, main characteristic gases decomposed by camellia insulating oil were obtained, and the changing rules of gas content with the changing of thermal ageing time, thermal fault temperature and discharge energy were also analyzed. Which indicated that if we increase the thermal ageing time, thermal faults temperature and discharge energy, the decomposition processes of camellia insulating oil could also be accelerated. And also the decomposition processes of camellia insulating oil and insulation paper under thermal and electrical stress were analyzed. Simulation results explained the gas production rules of thermal ageing and thermal faults as well as electrical faults of camellia insulating oil at microcosmic level. The result of above study provides theoretic support for condition assessment and operating maintenance of vegetable insulating oil-filled transformers, and also has great theoretical reference value and practical guiding significance for ensuring safe operation of vegetable insulating oil-filled transformers.

The Sixth International Symposium on Neural Networks (ISNN 2009)

Effect of thermal aging on the microstructure and mechanical property of 410S ferritic stainless steel