Dynamic Gas Turbine Condition Monitoring Scheme With Multi-Part Neural Network

2020 ◽  
Author(s):  
Zhouzheng Li ◽  
Kun Feng ◽  
Binbin Yan
Keyword(s):  
Neural Network ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Health Management ◽  
Fault Analysis ◽  
Performance Data ◽  
Sensor Data ◽  
Integrated Network ◽  
The Neural Network

Abstract Gas turbines are high value industrial assets with significant roles in various kinds of industrial processes, health management systems are therefore important for helping maintaining gas turbines’ stability in long-term operations. With more and more performance data able to be collected by sensors and the new machine learning methods developed, researchers are able to build more powerful digital models to monitor the gas turbine. This paper introduces a performance parameters alarm scheme for gas turbine using an adaptive state following model. The proposed scheme consist of 3 parts: Part 1, a dynamically adaptive multi-part neural network trained using performance data that can simulate different parts of gas turbine and output “normal” sensor data to make comparison with the actual data collected; Part 2, a group of thresholds set according to system noise that flags sudden failures by sensing performance parameter outliers, this also decides which data should be used to update the neural network; Part 3, a recorder for “reference point” outputs that can reflect change of the gas turbine’s status and detect long-term degradation. Unlike traditional approaches, the proposed adaptive states following model separates long term degradation and short term sudden failure, therefore both faults can be detected more accurately. The core of the proposed method is that physical properties are embedded into the neural network as constraints to regulate training and make the model more interpretable. In our scheme, a gas turbine is divided into 4 parts referencing the equipment’s physical mechanism, they are simulated digitally by 4 sub corresponding networks, which are then combined into the proposed integrated network. The proposed scheme achieves an overall pleasing result and shows potential in gas turbine fault analysis.

Real Time Prognostic Strategies: Application to Gas Turbines

2011 ◽  
Vol 312-315 ◽  
pp. 601-606
Author(s):  
M. Yadegari
Keyword(s):  
Real Time ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Health Management ◽  
System Behavior ◽  
Industrial Sectors ◽  
Healthy Working ◽  
The World ◽  
Short And Long Term

Gas turbines are increasingly deployed throughout the world to provide electrical and mechanical power in consumer and industrial sectors. A health management system can incorporate prognostic algorithms to effectively interpret and determine the healthy working span of a gas turbine. The research project’s objective is to develop real-time monitoring and prediction algorithms for simple cycle natural gas turbines to forecast short and long term system behavior.

Gas Turbine Machinery Diagnostics: A Brief Review and a Sample Application

2017 ◽  
Author(s):  
Cody W. Allen ◽  
Chad M. Holcomb ◽  
Mauricio de Oliveira
Keyword(s):  
Machine Learning ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Health Management ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Sensor Data ◽  
Industry Standard ◽  
Industrial Gas Turbines ◽  
Sample Application

The intersection of machine learning methods and gas turbine sensor data has expanded rapidly in the last decade to include numerous applications of regression, clustering, and even neural network algorithms. Learning algorithms have pushed traditional engine health management into the realm of prognostic health management. This paper starts with a review of several common computational methods used to monitor the condition of gas turbines currently employed by both industry and academia. Sources of application of machine learning algorithms from outside the gas turbine industry are also brought in. Focus is generally placed on industrial gas turbines with an industry standard monitoring system. The authors explore beyond gas path analysis with a novel use of machine learning algorithms to engine component classification. The paper concludes with a case study of applying learning algorithms to machine data to identify different fuel valves.

Design of a Multivariable Neural Controller and Its Application to Gas Turbines

1997 ◽  
Vol 119 (3) ◽  
pp. 565-567
Author(s):  
Q. Song ◽  
M. J. Grimble
Keyword(s):  
Neural Network ◽  
Gas Turbine ◽  
Discrete Time ◽  
Gas Turbines ◽  
Neural Controller ◽  
Turbine Plant ◽  
The Neural Network ◽  
Gas Turbine Plant ◽  
Multivariable Controller ◽  
Compensation Signal

The algorithm for a multivariable controller using neural network is based on a discrete-time fixed controller and the neural network provides a compensation signal to suppress the nonlinearity. The multivariable neural controller is easy to train and applied to an aircraft gas turbine plant.

scholarly journals Forecast of Waterway Cargo Turnover Volume Based on Genetic Algorithm to Optimize Neural Network Parameters

2021 ◽  
Vol 2083 (3) ◽  
pp. 032010
Author(s):  
Rong Ma
Keyword(s):  
Neural Network ◽  
Genetic Algorithm ◽  
Bp Neural Network ◽  
Forecast Model ◽  
Network Parameters ◽  
The Neural Network ◽  
Target Effect ◽  
The Empirical Analysis ◽  
Better Than

Abstract The traditional BP neural network is difficult to achieve the target effect in the prediction of waterway cargo turnover. In order to improve the accuracy of waterway cargo turnover forecast, a waterway cargo turnover forecast model was created based on genetic algorithm to optimize neural network parameters. The genetic algorithm overcomes the trap that the general iterative method easily falls into, that is, the “endless loop” phenomenon that occurs when the local minimum is small, and the calculation time is small, and the robustness is high. Using genetic algorithm optimized BP neural network to predict waterway cargo turnover, and the empirical analysis of the waterway cargo turnover forecast is carried out. The results obtained show that the neural network waterway optimized by genetic algorithm has a higher accuracy than the traditional BP neural network for predicting waterway cargo turnover, and the optimization model can long-term analysis of the characteristics of waterway cargo turnover changes shows that the prediction effect is far better than traditional neural networks.

EEG Data Mining Using PCA

2011 ◽  
pp. 161-180 ◽  
Author(s):  
Lenka Lhotská ◽  
Vladimír Krajca ◽  
Jitka Mohylová ◽  
Svojmil Petránek ◽  
Václav Gerla
Keyword(s):  
Neural Network ◽  
Data Mining ◽  
Principal Components ◽  
The Neural Network ◽  
Eeg Data ◽  
Eigen Decomposition ◽  
Electroencephalogram Eeg ◽  
Epileptic Spike ◽  
Eeg Signal Processing

This chapter deals with the application of principal components analysis (PCA) to the field of data mining in electroencephalogram (EEG) processing. The principal components are estimated from the signal by eigen decomposition of the covariance estimate of the input. Alternatively, they can be estimated by a neural network (NN) configured for extracting the first principal components. Instead of performing computationally complex operations for eigenvector estimation, the neural network can be trained to produce ordered first principal components. Possible applications include separation of different signal components for feature extraction in the field of EEG signal processing, adaptive segmentation, epileptic spike detection, and long-term EEG monitoring evaluation of patients in a coma.

scholarly journals Real-Time Maritime Traffic Anomaly Detection Based on Sensors and History Data Embedding

Sensors ◽  
2019 ◽  
Vol 19 (17) ◽  
pp. 3782 ◽  
Author(s):  
Julius Venskus ◽  
Povilas Treigys ◽  
Jolita Bernatavičienė ◽  
Gintautas Tamulevičius ◽  
Viktor Medvedev
Keyword(s):  
Neural Network ◽  
Data Processing ◽  
Processing Time ◽  
Heavy Traffic ◽  
Sensor Data ◽  
Identification System ◽  
Automated Identification ◽  
Maritime Traffic ◽  
Abnormal Movements ◽  
The Neural Network

The automated identification system of vessel movements receives a huge amount of multivariate, heterogeneous sensor data, which should be analyzed to make a proper and timely decision on vessel movements. The large number of vessels makes it difficult and time-consuming to detect abnormalities, thus rapid response algorithms should be developed for a decision support system to identify abnormal movements of vessels in areas of heavy traffic. This paper extends the previous study on a self-organizing map application for processing of sensor stream data received by the maritime automated identification system. The more data about the vessel’s movement is registered and submitted to the algorithm, the higher the accuracy of the algorithm should be. However, the task cannot be guaranteed without using an effective retraining strategy with respect to precision and data processing time. In addition, retraining ensures the integration of the latest vessel movement data, which reflects the actual conditions and context. With a view to maintaining the quality of the results of the algorithm, data batching strategies for the neural network retraining to detect anomalies in streaming maritime traffic data were investigated. The effectiveness of strategies in terms of modeling precision and the data processing time were estimated on real sensor data. The obtained results show that the neural network retraining time can be shortened by half while the sensitivity and precision only change slightly.

Research and Development of Practical Industrial Cogeneration Technology in Japan

2000 ◽  
Author(s):  
Toshiaki Abe ◽  
Takashi Sugiura ◽  
Shuji Okunaga ◽  
Katsuhiro Nojima ◽  
Yasukata Tsutsui ◽  
...  
Keyword(s):  
Research And Development ◽  
Gas Turbine ◽  
High Temperatures ◽  
Gas Turbines ◽  
High Efficiency ◽  
Development Project ◽  
Term Operation ◽  
Resistance To Heat

This paper presents an overview of a development project involving industrial cogeneration technology using 8,000-kW class hybrid gas turbines in which both metal and ceramics are used in parts subject to high temperatures in order to achieve high efficiency and low pollution. The development of hybrid gas turbines focuses mainly on the earlier commercialization of the turbine system. Stationary parts such as combustor liners, transition ducts, and first-stage turbine nozzles (stationary blades) are expected to be fabricated from ceramics. The project aims at developing material for these ceramic parts that will have a superior resistance to heat and oxidation. The project also aims at designing and prototyping a hybrid gas turbine system to analyze the operation in order to improve the performance. Furthermore, the prototyped hybrid gas turbine system will be tested for long-term operation (4,000 hours) to verify that the system can withstand commercialization. Studies will be conducted to ensure that the system’s soundness and reliability are sufficient for industrial cogeneration applications.

Fatigue Strength and Metal Condition in the Context of Gas Turbine Buckets Life Extension

2012 ◽  
Author(s):  
Sergey A. Ivanov ◽  
Alexander I. Rybnikov
Keyword(s):  
Fatigue Strength ◽  
Service Life ◽  
Gas Turbine ◽  
Fatigue Resistance ◽  
Gas Turbines ◽  
Life Extension ◽  
Term Operation ◽  
Life Estimation ◽  
Metal Properties

Criteria for remaining life estimation and methods for enhancing fatigue resistance of heavy-duty gas turbine bucket metal are based on the analysis of changes in the structure and properties of metal after long-term operation. High-cycle fatigue (HCF) resistance is shown to be a decisive characteristic in the residual life estimation of turbine buckets after operation over 100,000 hours. The tests of the buckets from cast and wrought nickel-based alloys after long-term operation demonstrated decreasing of fatigue strength by up to 25%. The metal structure in operation undergoes notable deterioration mainly in phase redistribution. The size and configuration of metal phases are changing also. It caused the changes in metal properties. The decrease of the bucket fatigue strength correlates with the decrease of metal ductility. The reconditioning heat treatment resulted in restoring mechanical properties of metal. The fatigue resistance also increased nearly to the initial level. The influence of operational factors on bucket fatigue strength deterioration has been established. The mechanical damages on bucket airfoil may decrease the fatigue resistance. We found the correlation of endurance limit and damages depth. The procedures for metal properties recovering and buckets service life substantial extension have been developed. It has resulted in the extension of the buckets service life by up to 50% over the assigned life in gas turbines operated by Gazprom.

Long-Term Degradation of Ceramics for Gas Turbine Applications

2007 ◽  
Author(s):  
Mark van Roode ◽  
Mattison K. Ferber
Keyword(s):  
Water Vapor ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Protective Coatings ◽  
Ceramic Matrix Composites ◽  
Oxide Ceramic ◽  
Component Design ◽  
Temperature Capability ◽  
Silicon Based

A study has been conducted to establish the effect of long-term (30,000+ hours) properties of monolithic ceramics (Si3N49 SiC), SiC/SiC and oxide/oxide ceramic matrix composites (CMCs), and protective coatings on component life in gas turbine engines with pressure ratios (PRs) ranging from 5:1 to 30:1. A model has been presented that shows the interaction between two major long-term degradation modes of ceramics, creep and degradation from water vapor attack in the ceramic hot section. Water vapor attack is the most severe mode overshadowing creep for long-term (∼30,000 hours) gas turbine operation, and its impact on component durability becomes more severe as PR increases. Components in the turbine hot section, downstream from the combustor (blades, integral turbine rotors, nozzles), fabricated from Si3N4 without protective coatings, have a temperature limitation of ∼800°C for gas turbines with PR ranging from 5:1 to 30:1. These ceramic components afford little, if any, advantage over metallic components for improving gas turbine performance. The application of a BSAS-type Environmental Barrier Coating (EBC) would improve temperature capability of turbine nozzles and rotating parts to ∼1100–1200°C. For small low-PR (5:1) engines, thick (∼10 mm) uncoated monolithic silicon-based combustor liners can be used up to ∼1360°C and thinner (∼3 mm) SiC/SiC CMCs up to ∼1100°C. These temperatures are reduced for higher-PR engines. The incorporation of a BSAS-type EBC improves temperature capability of silicon-based ceramic combustor liners. Oxide/oxide CMCs with protective coatings have a predicted temperature capability of ∼1220-∼1380°C over the range of PR range studied. They can be used as structural materials for combustor liners and other stationary turbine hot section components. As PR increases the durability of coated oxide/oxide CMCs improves relative to that of silicon-based monolithics and CMCs. As expected, ceramic component durability increases for shorter component design lives, making these materials more acceptable for shorter-term applications, such as automotive transportation (∼3,000 hours/150,000 km).

scholarly journals Using neural networks to describe tracer correlations

2004 ◽  
Vol 4 (1) ◽  
pp. 143-146 ◽  
Author(s):  
D. J. Lary ◽  
M. D. Müller ◽  
H. Y. Mussa
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Chemical Models ◽  
The Neural Network ◽  
Pressure Time ◽  
Fortran Code ◽  
Time Of Year ◽  
Hidden Layer ◽  
And Training

Abstract. Neural networks are ideally suited to describe the spatial and temporal dependence of tracer-tracer correlations. The neural network performs well even in regions where the correlations are less compact and normally a family of correlation curves would be required. For example, the CH4-N2O correlation can be well described using a neural network trained with the latitude, pressure, time of year, and CH4 volume mixing ratio (v.m.r.). In this study a neural network using Quickprop learning and one hidden layer with eight nodes was able to reproduce the CH4-N2O correlation with a correlation coefficient between simulated and training values of 0.9995. Such an accurate representation of tracer-tracer correlations allows more use to be made of long-term datasets to constrain chemical models. Such as the dataset from the Halogen Occultation Experiment (HALOE) which has continuously observed CH4  (but not N2O) from 1991 till the present. The neural network Fortran code used is available for download.

Sign in / Sign up

Export Citation Format

Share Document