Simulation of Compressor Transient Behavior Through Recurrent Neural Network Models

2005 ◽  
Vol 128 (3) ◽  
pp. 444-454 ◽  
Author(s):  
M. Venturini
Keyword(s):  
Measurement Errors ◽  
Learning Algorithm ◽  
Back Propagation ◽  
Network Models ◽  
Transient Behavior ◽  
Optimal Number ◽  
Measured Data ◽  
Outlet Temperature ◽  
Neural Network Models ◽  
Hidden Layer

In the paper, self-adapting models capable of reproducing time-dependent data with high computational speed are investigated. The considered models are recurrent feed-forward neural networks (RNNs) with one feedback loop in a recursive computational structure, trained by using a back-propagation learning algorithm. The data used for both training and testing the RNNs have been generated by means of a nonlinear physics-based model for compressor dynamic simulation, which was calibrated on a multistage axial-centrifugal small size compressor. The first step of the analysis is the selection of the compressor maneuver to be used for optimizing RNN training. The subsequent step consists in evaluating the most appropriate RNN structure (optimal number of neurons in the hidden layer and number of outputs) and RNN proper delay time. Then, the robustness of the model response towards measurement uncertainty is ascertained, by comparing the performance of RNNs trained on data uncorrupted or corrupted with measurement errors with respect to the simulation of data corrupted with measurement errors. Finally, the best RNN model is tested on field data taken on the axial-centrifugal compressor on which the physics-based model was calibrated, by comparing physics-based model and RNN predictions against measured data. The comparison between RNN predictions and measured data shows that the agreement can be considered acceptable for inlet pressure, outlet pressure and outlet temperature, while errors are significant for inlet mass flow rate.

Simulation of Compressor Transient Behavior Through Recurrent Neural Network Models

2005 ◽  
Author(s):  
M. Venturini
Keyword(s):  
Measurement Errors ◽  
Learning Algorithm ◽  
Back Propagation ◽  
Network Models ◽  
Transient Behavior ◽  
Optimal Number ◽  
Measured Data ◽  
Outlet Temperature ◽  
Neural Network Models ◽  
Multi Stage

In the paper, self-adapting models capable of reproducing time-dependent data with high computational speed are investigated. The considered models are recurrent feed-forward neural networks (RNNs) with one feedback loop in a recursive computational structure, trained by using a back-propagation learning algorithm. The data used for both training and testing the RNNs have been generated by means of a non-linear physics-based model for compressor dynamic simulation, which was calibrated on a multi-stage axial-centrifugal small size compressor. The first step of the analysis is the selection of the compressor maneuver to be used for optimizing RNN training. The subsequent step consists in evaluating the most appropriate RNN structure (optimal number of neurons in the hidden layer and number of outputs) and RNN proper delay time. Then, the robustness of the model response towards measurement uncertainty is ascertained, by comparing the performance of RNNs trained on data uncorrupted or corrupted with measurement errors with respect to the simulation of data both uncorrupted and corrupted with measurement errors. Finally, the best RNN model is tested on field data taken on the axial-centrifugal compressor on which the physics-based model was calibrated, by comparing physics-based model and RNN predictions against measured data. The comparison between RNN predictions and measured data shows that the agreement can be considered acceptable for inlet pressure, outlet pressure and outlet temperature, while errors are significant for inlet mass flow rate.

scholarly journals Stock Market Prediction Using Artificial Neural Networks

2012 ◽  
Vol 6-7 ◽  
pp. 1055-1060 ◽  
Author(s):  
Yang Bing ◽  
Jian Kun Hao ◽  
Si Chang Zhang
Keyword(s):  
Neural Network ◽  
Prediction Models ◽  
Learning Algorithm ◽  
Stock Exchange ◽  
Back Propagation ◽  
Composite Index ◽  
Network Models ◽  
Back Propagation Neural Network ◽  
Neural Network Models ◽  
The Neural Network

In this study we apply back propagation Neural Network models to predict the daily Shanghai Stock Exchange Composite Index. The learning algorithm and gradient search technique are constructed in the models. We evaluate the prediction models and conclude that the Shanghai Stock Exchange Composite Index is predictable in the short term. Empirical study shows that the Neural Network models is successfully applied to predict the daily highest, lowest, and closing value of the Shanghai Stock Exchange Composite Index, but it can not predict the return rate of the Shanghai Stock Exchange Composite Index in short terms.

scholarly journals Using artificial neural network models for the prediction of thrust force and torque in drilling operation of Al7075

2018 ◽  
Vol 178 ◽  
pp. 01008
Author(s):  
Panagiotis Kyratsis ◽  
Nikolaos Efkolidis ◽  
Daniel Ghiculescu ◽  
Konstantinos Kakoulis
Keyword(s):  
Thrust Force ◽  
Network Models ◽  
Cutting Parameters ◽  
Drilling Process ◽  
Neural Network Models ◽  
Cutting Velocity ◽  
Artificial Neural ◽  
Hidden Layer ◽  
The One ◽  
Tool Diameter

This study investigates the thrust force (Fz) and torque (Mz) in a drilling process of an Al7075 workpiece using solid carbide tools (Kennametal KC7325), depending on the effects of crucial cutting parameters such as cutting velocity, feed rate and tool diameter of 10mm, 12mm and 14mm. Artificial neural networks (ANN) methodology is used in order to acquire mathematical models for both the thrust force (Fz) and torque (Mz) related to the drilling process. The ANN results showed that the best prediction topology of the network for the thrust force was the one with five neurons in the hidden layer, while for the case of Mz the best network topology for the prediction of the experimental values was the one with six neurons in the hidden layer. Based on the results acquired, the ANN models achieved accuracy of 1,96% and 1,95% for both the thrust force and torque measured, while the R coefficient for the prediction model of the thrust force is 0.99976 and 00.99981 for the torque. As a result they can be considered as very accurate and appropriate for their prediction.

Pruning of rat cortical taste neurons by an artificial neural network model

1995 ◽  
Vol 74 (3) ◽  
pp. 1010-1019 ◽  
Author(s):  
T. Nagai ◽  
H. Katayama ◽  
K. Aihara ◽  
T. Yamamoto
Keyword(s):  
Learning Algorithm ◽  
Back Propagation ◽  
Quantitative Measure ◽  
Response Patterns ◽  
Input Unit ◽  
Basic Taste ◽  
Relative Contribution ◽  
Quinine Hydrochloride ◽  
Taste Discrimination ◽  
Hidden Layer

1. Taste qualities are believed to be coded in the activity of ensembles of taste neurons. However, it is not clear whether all neurons are equally responsible for coding. To clarify the point, the relative contribution of each taste neuron to coding needs to be assessed. 2. We constructed simple three-layer neural networks with input units representing cortical taste neurons of the rat. The networks were trained by the back-propagation learning algorithm to classify the neural response patterns to the basic taste stimuli (sucrose, HCl, quinine hydrochloride, and NaCl). The networks had four output units representing the basic taste qualities, the values of which provide a measure for similarity of test stimuli (salts, tartaric acid, and umami substances) to the basic taste stimuli. 3. Trained networks discriminated the response patterns to the test stimuli in a plausible manner in light of previous physiological and psychological experiments. Profiles of output values of the networks paralleled those of across-neuron correlations with respect to the highest or second-highest values in the profiles. 4. We evaluated relative contributions of input units to the taste discrimination of the network by examining their significance Sj, which is defined as the sum of the absolute values of the connection weights from the jth input unit to the hidden layer. When the input units with weaker connection weights (e.g., 15 of 39 input units) were "pruned" from the trained network, the ability of the network to discriminate the basic taste qualities as well as other test stimuli was not greatly affected. On the other hand, the taste discrimination of the network progressively deteriorated much more rapidly with pruning of input units with stronger connection weights. 5. These results suggest that cortical taste neurons differentially contribute to the coding of taste qualities. The pruning technique may enable the evaluation of a given taste neuron in terms of its relative contribution to the coding, with Sj providing a quantitative measure for such evaluation.

STUDY ON CHEMOTAXIS BEHAVIORS OF C. ELEGANS USING DYNAMIC NEURAL NETWORK MODELS: FROM ARTIFICIAL TO BIOLOGICAL MODEL

2010 ◽  
Vol 18 (spec01) ◽  
pp. 3-33 ◽  
Author(s):  
JIAN-XIN XU ◽  
XIN DENG
Keyword(s):  
Neural Network ◽  
Sensory Neurons ◽  
Motor Neurons ◽  
Learning Algorithm ◽  
Network Models ◽  
Biological Models ◽  
Neural Network Models ◽  
Dynamic Neural Network ◽  
C Elegans ◽  
Logical Functions

With the anatomical understanding of the neural connection of the nematode Caenorhabditis elegans (C. elegans), its chemotaxis behaviors are investigated in this paper through the association with the biological nerve connections. The chemotaxis behaviors include food attraction, toxin avoidance and mixed-behaviors (finding food and avoiding toxin concurrently). Eight dynamic neural network (DNN) models, two artifical models and six biological models, are used to learn and implement the chemotaxis behaviors of C. elegans. The eight DNN models are classified into two classes with either single sensory neuron or dual sensory neurons. The DNN models are trained to learn certain switching logics according to different chemotaxis behaviors using real time recurrent learning algorithm (RTRL). First we show the good performance of the two artifical models in food attraction, toxin avoidance and the mixed-behaviors. Next, six neural wire diagrams from sensory neurons to motor neurons are extracted from the anatomical nerve connection of C. elegans. Then the extracted biological wire diagrams are trained using RTRL directly, which is the first time in this field of research by associating chemotaxis behaviors with biological neural models. An interesting discovery is the need for a memory neuron when single-sensory models are used, which is consistent with the anatomical understanding on a specific neuron that functions as a memory. In the simulations, the chemotaxis behaviors of C. elegans can be depicted by several switch logical functions which can be learned by RTRL for both artifical and biological models.

scholarly journals Radyal Tabanlı Yapay Sinir Ağlarının Kurutma Kinetiğinin Modellenmesinde Kullanımı

2020 ◽  
Vol 8 (2) ◽  
pp. 511
Author(s):  
Adil Koray Yıldız ◽  
Muhammed Taşova ◽  
Hakan Polatcı
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Network Models ◽  
Drying Kinetics ◽  
Hot Pepper ◽  
Neural Network Models ◽  
Drying Time ◽  
Drying Behavior ◽  
Artificial Neural ◽  
Hidden Layer

Drying method is preferred in agricultural products since it provides advantages in many processes such as increasing the strength of products, transporting and storing. It is necessary to estimate the drying behavior of the products in order to achieve the best drying without reducing the product quality. For this reason, many numerical drying models have been developed to estimate the drying kinetics of the products. Recently, artificial neural networks have been widely used for the development of these models. Artificial neural networks are mathematical models that work in a similar way to natural neuron cells. Radial based artificial neural networks are radial based activation functions in the transition to the hidden layer unlike other networks. In this study, modeling of drying kinetics with radial based networks was investigated. For the experiment, red hot pepper (Capsicum annuum L.) was dipped in boiled water and microwave pretreatments and, then dried in the oven at 65°C. The absorbable moisture values were calculated during the drying period. The radial based artificial neural network models were trained with the drying time values as input and the absorbable moisture values as output. The study was carried out with two data sets including all data and only the average. In trainings with all data, R value of the best model was calculated as 0.9566. R was calculated as 0.9998 with average data.

scholarly journals Learned Prediction of Compressive Strength of GGBFS Concrete Using Hybrid Artificial Neural Network Models

Materials ◽  
2019 ◽  
Vol 12 (22) ◽  
pp. 3708 ◽  
Author(s):  
In-Ji Han ◽  
Tian-Feng Yuan ◽  
Jin-Young Lee ◽  
Young-Soo Yoon ◽  
Joong-Hoon Kim
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Compressive Strength ◽  
Experimental Studies ◽  
Back Propagation ◽  
Network Models ◽  
Neural Network Models ◽  
Artificial Neural ◽  
Artificial Neural Network Ann ◽  
Artificial Neural Network Models

A new hybrid intelligent model was developed for estimating the compressive strength (CS) of ground granulated blast furnace slag (GGBFS) concrete, and the synergistic benefits of the hybrid algorithm as compared with a single algorithm were verified. While using the collected 269 data from previous experimental studies, artificial neural network (ANN) models with three different learning algorithms namely back-propagation (BP), particle swarm optimization (PSO), and new hybrid PSO-BP algorithms, were constructed and the performance of the models was evaluated with regard to the prediction accuracy, efficiency, and stability through a threefold procedure. It was found that the PSO-BP neural network model was superior to the simple ANNs that were trained by a single algorithm and it is suitable for predicting the CS of GGBFS concrete.

Image Classification of Crop Diseases and Pests Based on Deep Learning and Fuzzy System

2020 ◽  
Vol 16 (2) ◽  
pp. 34-47 ◽  
Author(s):  
Tongke Fan ◽  
Jing Xu
Keyword(s):  
Fuzzy System ◽  
Insect Pests ◽  
Learning Algorithm ◽  
Network Models ◽  
Spectral Graph Theory ◽  
Neural Network Models ◽  
Pests And Diseases ◽  
Optimal Network ◽  
Blurred Image

The automatic classification of crop disease images has important value. The classification algorithm based on manual feature extraction has some problems, such as the need for professional knowledge, is time-consuming and laborious, and has difficulty extracting high-quality features. In this article, the theory of the fuzzy system is discussed. The theory of the fuzzy system is applied to the pretreatment of blurred images. A local blurred image deblurring method based on depth learning is proposed. By training convolutional neural network models with different structures, the image of diseases and insect pests is segmented using normalized segmentation algorithms based on spectral graph theory, and the segmentation knot of leaf diseases is obtained. Finally, the optimal network structure is obtained by comparing the segmentation results with the traditional machine learning algorithm. Experiments show that the segmentation results of pests and diseases obtained by this algorithm have better robustness, generalization, and higher accuracy.

System Identification for Nonlinear Maneuvering of Ships Using Neural Network

2010 ◽  
Vol 54 (01) ◽  
pp. 1-14
Author(s):  
G. Rajesh ◽  
G. Giri Rajasekhar ◽  
S. K. Bhattacharyya
Keyword(s):  
Neural Network ◽  
System Identification ◽  
Simulated Data ◽  
Network Models ◽  
Training Data ◽  
Neural Network Models ◽  
The Neural Network ◽  
Marquardt Algorithm ◽  
Hydrodynamic Derivatives ◽  
Hidden Layer

This paper deals with the application of nonparametric system identification to the nonlinear maneuvering of ships using neural network method. The maneuvering equations contain linear as well as nonlinear terms, and one does not attempt to determine the parameters (or hydrodynamic derivatives) associated with nonlinear terms, rather all nonlinear terms are clubbed together to form one unknown time function per equation, which are sought to be represented by neural network coefficients. The time series used in training the network are obtained from simulated data of zigzag and spiral maneuvers. The neural network has one middle or hidden layer of neurons and the Levenberg-Marquardt algorithm is used to obtain the network coefficients. Using the best choices for number of hidden layer neurons, length of training data, convergence tolerance, and so forth, the performances of the proposed neural network models have been investigated and conclusions drawn.

Sign in / Sign up

Export Citation Format

Share Document