Artificial Intelligence for the Diagnostics of Gas Turbines—Part I: Neural Network Approach

2006 ◽  
Vol 129 (3) ◽  
pp. 711-719 ◽  
Author(s):  
R. Bettocchi ◽  
M. Pinelli ◽  
P. R. Spina ◽  
M. Venturini
Keyword(s):  
Neural Network ◽  
Measurement Uncertainty ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Measurement Errors ◽  
Learning Algorithm ◽  
Computational Time ◽  
Training Phase ◽  
Neural Network Approach ◽  
Single Output

In the paper, neural network (NN) models for gas turbine diagnostics are studied and developed. The analyses carried out are aimed at the selection of the most appropriate NN structure for gas turbine diagnostics, in terms of computational time of the NN training phase, accuracy, and robustness with respect to measurement uncertainty. In particular, feed-forward NNs with a single hidden layer trained by using a back-propagation learning algorithm are considered and tested. Moreover, multi-input/multioutput NN architectures (i.e., NNs calculating all the system outputs) are compared to multi-input/single-output NNs, each of them calculating a single output of the system. The results obtained show that NNs are sufficiently robust with respect to measurement uncertainty, if a sufficient number of training patterns are used. Moreover, multi-input/multioutput NNs trained with data corrupted with measurement errors seem to be the best compromise between the computational time required for NN training phase and the NN accuracy in performing gas turbine diagnostics.

Artificial Intelligence for the Diagnostics of Gas Turbines: Part I — Neural Network Approach

2005 ◽  
Author(s):  
R. Bettocchi ◽  
M. Pinelli ◽  
P. R. Spina ◽  
M. Venturini
Keyword(s):  
Neural Network ◽  
Measurement Uncertainty ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Measurement Errors ◽  
Learning Algorithm ◽  
Computational Time ◽  
Training Phase ◽  
Neural Network Approach ◽  
Single Output

In the paper, Neural Network (NN) models for gas turbine diagnostics are studied and developed. The analyses carried out are aimed at the selection of the most appropriate NN structure for gas turbine diagnostics, in terms of computational time of the NN training phase, accuracy and robustness with respect to measurement uncertainty. In particular, feed-forward NNs with a single hidden layer trained by using a back-propagation learning algorithm are considered and tested. Moreover, Multi-Input/Multi-Output NN architectures (i.e. NNs calculating all the system outputs) are compared to Multi-Input/Single-Output NNs, each of them calculating a single output of the system. The results obtained show that NNs are robust with respect to measurement uncertainty, if a sufficient number of training patterns are used. Moreover, Multi-Input/Multi-Output NNs trained with data corrupted with measurement errors seem to be the best compromise between the computational time required for NN training phase and the NN accuracy in performing gas turbine diagnostics.

Artificial Intelligence for the Diagnostics of Gas Turbines—Part II: Neuro-Fuzzy Approach

2006 ◽  
Vol 129 (3) ◽  
pp. 720-729 ◽  
Author(s):  
R. Bettocchi ◽  
M. Pinelli ◽  
P. R. Spina ◽  
M. Venturini
Keyword(s):  
Measurement Uncertainty ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Fuzzy Inference ◽  
Combined Cycle ◽  
Computational Time ◽  
Health State ◽  
Training Phase ◽  
Inference System ◽  
Neuro Fuzzy

In the paper, neuro-fuzzy systems (NFSs) for gas turbine diagnostics are studied and developed. The same procedure used previously for the setup of neural network (NN) models (Bettocchi, R., Pinelli, M., Spina, P. R., and Venturini, M., 2007, ASME J. Eng. Gas Turbines Power, 129(3), pp. 711–719) was used. In particular, the same database of patterns was used for both training and testing the NFSs. This database was obtained by running a cycle program, calibrated on a 255MW single-shaft gas turbine working in the ENEL combined cycle power plant of La Spezia (Italy). The database contains the variations of the Health Indices (which are the characteristic parameters that are indices of gas turbine health state, such as efficiencies and characteristic flow passage areas of compressor and turbine) and the corresponding variations of the measured quantities with respect to the values in new and clean conditions. The analyses carried out are aimed at the selection of the most appropriate NFS structure for gas turbine diagnostics, in terms of computational time of the NFS training phase, accuracy, and robustness towards measurement uncertainty during simulations. In particular, adaptive neuro-fuzzy inference system (ANFIS) architectures were considered and tested, and their performance was compared to that obtainable by using the NN models. An analysis was also performed in order to identify the most significant ANFIS inputs. The results obtained show that ANFISs are robust with respect to measurement uncertainty, and, in all the cases analyzed, the performance (in terms of accuracy during simulations and time spent for the training phase) proved to be better than that obtainable by multi-input/multioutput (MIMO) and multi-input/single-output (MISO) neural networks trained and tested on the same data.

Artificial Intelligence for the Diagnostics of Gas Turbines: Part II — Neuro-Fuzzy Approach

2005 ◽  
Author(s):  
R. Bettocchi ◽  
M. Pinelli ◽  
P. R. Spina ◽  
M. Venturini
Keyword(s):  
Measurement Uncertainty ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Fuzzy Inference ◽  
Combined Cycle ◽  
Computational Time ◽  
Health State ◽  
Training Phase ◽  
Inference System ◽  
Neuro Fuzzy

In the paper, Neuro-Fuzzy Systems (NFSs) for gas turbine diagnostics are studied and developed. The same procedure used previously for the set up of Neural Network (NN) models was used. In particular, the same database of patterns was used for both training and testing the NFSs. This database was obtained by running a Cycle Program, calibrated on a 255 MW single shaft gas turbine working in the ENEL combined cycle power plant of La Spezia (Italy). The database contains the variations of the Health Indices (which are the characteristic parameters that are indices of gas turbine health state, such as efficiencies and characteristic flow passage areas of compressor and turbine) and the corresponding variations of the measured quantities with respect to the values in new and clean conditions. The analyses carried out are aimed at the selection of the most appropriate NFS structure for gas turbine diagnostics, in terms of computational time of the NFS training phase, accuracy and robustness towards measurement uncertainty during simulations. In particular, Adaptive Neuro-Fuzzy Inference System (ANFIS) architectures were considered and tested, and their performance was compared to that obtainable by using the NN models. An analysis was also performed in order to identify the most significant ANFIS inputs. The results obtained show that ANFISs are robust with respect to measurement uncertainty, and, in all the cases analyzed, the performance (in terms of accuracy during simulations and time spent for the training phase) proved to be better than that obtainable by MIMO and MISO Neural Networks trained and tested on the same data.

Set Up of a Robust Neural Network for Gas Turbine Simulation

2004 ◽  
Author(s):  
R. Bettocchi ◽  
M. Pinelli ◽  
P. R. Spina ◽  
M. Venturini ◽  
M. Burgio
Keyword(s):  
Neural Network ◽  
Gas Turbine ◽  
Gas Turbines ◽  
Learning Algorithm ◽  
General Procedure ◽  
Back Propagation ◽  
Computational Time ◽  
Time Simulation ◽  
Hidden Layer ◽  
Set Up

In this paper, Neural Network (NN) models for the real-time simulation of gas turbines are studied and developed. The analyses carried out are aimed at the selection of the most appropriate NN structure for gas turbine simulation, in terms of both computational time of the NN training phase and accuracy and robustness with respect to measurement uncertainty. In particular, feed-forward NNs, with a single hidden layer and different numbers of neurons, trained by using a back-propagation learning algorithm are considered and tested. Finally, a general procedure for the validation of computational codes is adapted and applied to the validation of the developed NN models.

A Modular Code for Real Time Dynamic Simulation of Gas Turbines in Simulink

2002 ◽  
Vol 128 (3) ◽  
pp. 506-517 ◽  
Author(s):  
S. M. Camporeale ◽  
B. Fortunato ◽  
M. Mastrovito
Keyword(s):  
Mathematical Model ◽  
Real Time ◽  
Gas Turbine ◽  
Dynamic Behavior ◽  
Gas Turbines ◽  
Simulation Software ◽  
Computational Time ◽  
High Fidelity ◽  
Time Step ◽  
The Mathematical Model

A high-fidelity real-time simulation code based on a lumped, nonlinear representation of gas turbine components is presented. The code is a general-purpose simulation software environment useful for setting up and testing control equipments. The mathematical model and the numerical procedure are specially developed in order to efficiently solve the set of algebraic and ordinary differential equations that describe the dynamic behavior of gas turbine engines. For high-fidelity purposes, the mathematical model takes into account the actual composition of the working gases and the variation of the specific heats with the temperature, including a stage-by-stage model of the air-cooled expansion. The paper presents the model and the adopted solver procedure. The code, developed in Matlab-Simulink using an object-oriented approach, is flexible and can be easily adapted to any kind of plant configuration. Simulation tests of the transients after load rejection have been carried out for a single-shaft heavy-duty gas turbine and a double-shaft aero-derivative industrial engine. Time plots of the main variables that describe the gas turbine dynamic behavior are shown and the results regarding the computational time per time step are discussed.

Numerical Assessment of SCAP: A Passive System for Preventing Thermoacoustic Oscillations in Gas Turbine Annular Combustors

2007 ◽  
Author(s):  
Stefano Tiribuzi
Keyword(s):  
Gas Turbine ◽  
Gas Turbines ◽  
Research Unit ◽  
Combined Cycle ◽  
Computational Time ◽  
Computational Domain ◽  
Ambient Conditions ◽  
Pressure Oscillations ◽  
Operation Conditions ◽  
Thermoacoustic Oscillations

ENEL operates a dozen combined cycle units whose V94.3A gas turbines are equipped with annular combustors. In such lean premixed gas turbines, particular operation conditions could trigger large pressure oscillations due to thermoacoustic instabilities. The ENEL Research unit is studying this phenomenon in order to find out methods which could avoid or mitigate such events. The use of effective numerical analysis techniques allowed us to investigate the realistic time evolution and behaviour of the acoustic fields associated with this phenomenon. KIEN, an in-house low diffusive URANS code capable of simulating 3D reactive flows, has been used in the Very Rough Grid approach. This approach permits the simulation, with a reasonable computational time, of quite long real transients with a computational domain extended over all the resonant volumes involved in the acoustic phenomenon. The V94.3A gas turbine model was set up with a full combustor 3D grid, going from the compressor outlet up to the turbine inlet, including both the annular plenum and the annular combustion chamber. The grid extends over the entire circular angle, including all the 24 premixed burners. Numerical runs were performed with the normal V94.3A combustor configuration, with input parameters set so as no oscillations develop in the standard ambient conditions. Wide pressure oscillations on the contrary are associated with the circumferential acoustic modes of the combustor, which have their onset and grow when winter ambient conditions are assumed. These results also confirmed that the sustaining mechanism is based on the equivalence ratio fluctuation of premix mixture and that plenum plays an important role in such mechanism. Based on these findings, a system for controlling the thermoacoustic oscillation has been conceived (Patent Pending), which acts on the plenum side of the combustor. This system, called SCAP (Segmentation of Combustor Annular Plenum), is based on the subdivision of the plenum annular volume by means of a few meridionally oriented walls. Repetition of KIEN runs with a SCAP configuration, in which a suitable number of segmentation walls were properly arranged in the annular plenum, demonstrated the effectiveness of this solution in preventing the development of wide thermoacoustic oscillations in the combustor.

Early Warning of Gas Turbine Failure by Nonlinear Feature Extraction Using an Auto-Associative Neural Network Approach

2011 ◽  
Author(s):  
Thomas Palme´ ◽  
Peter Breuhaus ◽  
Mohsen Assadi ◽  
Albert Klein ◽  
Minkyo Kim
Keyword(s):  
Neural Network ◽  
Gas Turbine ◽  
Early Warning ◽  
Linear Method ◽  
Data Driven ◽  
Operational Parameters ◽  
Neural Network Approach ◽  
Data Set ◽  
Nonlinear Pca ◽  
Auto Associative Neural Network

This study investigates the application of nonlinear Principal Component Analysis (PCA), implemented through the use of Auto-Associative Neural Network (AANN), for early warning of impending gas turbine failure. The study is based on a real operational data set that includes a compressor failure. The analyzed data set consists of measured operational parameters whose identity are unknown, hence this study presents a purely data driven approach to the problem of early warning. In this case study, the use of AANNs for early detection of abnormal engine behavior could have provided the operator with a warning a few days prior to the fully developed failure, which resulted in a forced shut-down and extensive maintenance. Furthermore, a comparison is made between the nonlinear PCA by AANNs and the standard PCA model, which is an inherently linear method. The result shows that the AANN provides a more reliable detection of the failure by a higher residual generation during failure mode as well as fewer false indications prior to the failure. Consequently, this study shows that nonlinear PCA as performed with AANNs can be a valuable data driven tool for early warning of gas turbine failure.

MPID Control Tuning for a Flexible Manipulator Using a Neural Network

2010 ◽  
Vol 22 (1) ◽  
pp. 82-90 ◽  
Author(s):  
Tamer Mansour ◽  
◽  
Atsushi Konno ◽  
Masaru Uchiyama
Keyword(s):  
Neural Network ◽  
Vibration Control ◽  
Learning Algorithm ◽  
The Other ◽  
Gradient Algorithm ◽  
Flexible Manipulator ◽  
Computational Time ◽  
Speed Of Response ◽  
Control Gain ◽  
The Neural Network

This paper studies the use of neural networks as a tuning tool for the gain in Modified Proportional-Integral-Derivative (MPID) control used to control a flexible manipulator. The vibration control gain in the MPID controller has been determined in an empirical way so far. It is a considerable time consuming process because the vibration control performance depends not only on the vibration control gain but also on the other parameters such as the payload, references and PD joint servo gains. Hence, the vibration control gain must be tuned considering the other parameters. In order to find optimal vibration control gain for the MPID controller, a neural network based approach is proposed in this paper. The proposed neural network finds an optimum vibration control gain that minimizes a criteria function. The criteria function is selected to represent the effect of the vibration of the end effector in addition to the speed of response. The scaled conjugate gradient algorithm is used as a learning algorithm for the neural network. Tuned gain response results are compared to results for other types of gains. The effectiveness of using the neural network appears in the reduction of the computational time and the ability to tune the gain with different loading condition.

On Test Uncertainties in Field Performance Tests

2015 ◽  
Author(s):  
Rainer Kurz ◽  
Klaus Brun
Keyword(s):  
Measurement Uncertainty ◽  
Gas Turbine ◽  
Field Test ◽  
Measurement Errors ◽  
Equations Of State ◽  
Accurate Determination ◽  
Field Performance ◽  
Significant Implication ◽  
Test Results

Field testing of gas turbine or electric motor driven compressor packages requires the accurate determination of efficiency, capacity, head, or power consumption in sometimes less than ideal working environments. Nonetheless, field test results have significant implication for the compressor and gas turbine manufacturers and their customers. Economic considerations demand that the performance and efficiency of an installation are verified to assure the return on investment for the project. Thus, for the compressor and gas turbine manufacturers, as well as for the end-user, an accurate determination of the field performance is of vital interest. This paper discusses a method to determine the measurement uncertainty and, thus, the accuracy, of test results under the typical constraints of a site performance test, for compressors capable of variable speed operation. Namely, a method is presented which can be employed to verify the validity of field test performance results. Results are compared with actual field test results, using redundant methods. Typical field test measurement uncertainties are presented for different sets of instrumentation. The effect of different equations of state on the calculated performance is also discussed. Test parameters that correlate to the most significant influence on the performance uncertainties are identified and suggestions are provided on how to minimize their measurement errors. Results show that compressor efficiency uncertainties can be unacceptably high when some basic rules for accurate testing are violated. However, by following some simple measurement rules and maintaining commonality of the gas equations of state, the overall compressor package performance measurement uncertainty can be limited and meaningful results can be achieved.

scholarly journals Neural network approach to reduce dynamic measurement errors

ACTA IMEKO ◽  
2016 ◽  
Vol 5 (3) ◽  
pp. 24 ◽  
Author(s):  
Andrei Sergeevich Volosnikov ◽  
Aleksandr L. Shestakov
Keyword(s):  
Neural Network ◽  
Measurement Errors ◽  
Measurement Data ◽  
Dynamic Measurement ◽  
Network Approach ◽  
Neural Network Approach ◽  
Experimental Data Processing ◽  
The Neural Network ◽  
Mathematical Processing ◽  
Processing Of Measurement Data

<p>The neural network inverse model of a sensor with filtration of the sequentially recovered signal is considered. This model effectively reduces the dynamic measurement errors due to deep mathematical processing of measurement data. The result of the experimental data processing of a dynamic temperature measurement validates the efficiency of the proposed neural network approach to reduce dynamic measurement errors.</p>

Sign in / Sign up

Export Citation Format

Share Document