A Unified Nonlinear Adaptive Approach for Detection … Unified Nonlinear Adaptive Approach for...
-
Upload
nguyenngoc -
Category
Documents
-
view
221 -
download
1
Transcript of A Unified Nonlinear Adaptive Approach for Detection … Unified Nonlinear Adaptive Approach for...
A Unified Nonlinear Adaptive Approach for Detection and Isolation of Engine FaultsDetection and Isolation of Engine Faults
Liang Tang Xiaodong (Frank) ZhangJonathan DeCastroImpact Technologies, LLC
Wright State University
www.impact-tek.com © 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data 1
2009 NASA GRC Propulsion Controls and Diagnostics (PCD) WorkshopDecember 8 – 10, Cleveland, Ohio
OutlineOutline
• Background/Motivationsg• Our approach• Simulation results• Future work
www.impact-tek.com 2© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Motivations of This Programot at o s o s og a
Many existing EHM methods have the following disadvantages:
• deal with component faults and sensor faults separatelyhigh false alarm rate or unnecessary maintenance.
• involve linearizing the engine dynamics and blending the linear health monitors through gain scheduling
overly time-consuming and complicatedoverly time consuming and complicated
• designed at a nominal health, or non-degraded, condition lose their effectiveness as the real engine degrades over timelose their effectiveness as the real engine degrades over time
• use fixed thresholds for residual evaluation false alarm especially in transient scenarios
www.impact-tek.com 3© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
p y
Our ApproachOu pp oac
faults degradation
Control inputsEngine
a systematic and unified framework for sensor faults and component faults
Key Innovations
A bank of nonlinear adaptivef lt di ti ti t
for sensor faults and component faults
a direct nonlinear approach to dealing with nonlinear engine models and faults
fault diagnostic estimators
Transferable Belief Model (TBM)
Diagnostic residuals
nominal nonlinear onboard engine model enhanced by an adaptive neural network
Fault sensitivity and robustness further enhanced by adaptive thresholdsTransferable Belief Model (TBM)
based residual evaluation
Diagnostic decisionNonlinear adaptive
y p
An architecture that naturally supports distributed parallel processing
www.impact-tek.com 4© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
pfault diagnostic method
The Developed Diagnosis/Isolation Architecturep g
NeuralNetwork
www.impact-tek.com 5© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Assumptionsssu pt o s
• Only one single subsystem (sensor/actuator/component) may be faulty at a time;faulty at a time;• Consistent with NASA IVHM goal• The approach can potentially be extended to cover multiple
simultaneous faults
• Engine health degradation is not considered as a fault, whereas a fault is an abnormal and abrupt event;
N ti t d t t ti• No assumptions on steady state operation
www.impact-tek.com 6© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Fault Isolation Logicau t so at o og c
Fault detection logic: The FDI logic indicates that “a subsystem fault is detected” if the following condition is met:
• At least one component of the diagnostic residuals generated by the FDE exceeds its corresponding adaptive threshold.
Fault Isolation Logic: The FDI logic indicates that “a fault in subsystem s is isolated” if the following two conditions are both met:
All th di ti id l t d b FIE i b l th i• All the diagnostic residuals generated by FIE s, remains below their corresponding adaptive thresholds.• At least one component of the diagnostic residuals generated by each remaining FIE exceed its corresponding threshold.
* A subsystem can be either a sensor, an actuator or a component.
g p g
www.impact-tek.com 7© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Adaptive ThresholdAdaptive Threshold
• The residual may change with the ti i t t l i t dtime-variant control inputs and dynamic operating conditions of the engine.
• Adaptive thresholds are derived based on the nonlinear adaptive estimation/observer algorithms and the bound on modeling uncertainty.the bound on modeling uncertainty.
• The thresholds automatically adapt to the changing engine operating
diti t h th b tconditions to enhance the robustness and fault sensitivity of the FDI scheme.
www.impact-tek.com 8© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Adaptive Real-time Nonlinear Engine Modelp g
• Objective: to compensate for modeling uncertainty such as normal aging and engine-to-engine variation
control inputs sensor outputscontrol inputs sensor outputs aging and engine-to-engine variation
• A neural network is trained during flight to capture changes in engine dynamics as a result of normal aging Nominal engine model
+ +-Nominal engine model+ +-FDIE
dynamics as a result of normal aging.
• The training is done separately in parallel to the fault diagnostic estimators
Nominal engine model
Neural Network
+
t k i ht
Nominal engine model
Neural Network
+
t k i htestimators.
• The weights of neural network in the on-board engine model are updated after a number of flights
network weightsadaptive laws
adaptive real-time reference engine model
network weightsadaptive laws
adaptive real-time reference engine model
after a number of flights.
Two NNs: one being trained, another one with fixed weights used in FDIEs.
www.impact-tek.com 9© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Nonlinear FDI: System Model
• Known nominal dynamics:
C SS ( )
• Modeling uncertainties: • State equation:
Enhanced C-MAPSS (healthy)
• State equation: • Output equation:
F lt d l• Fault models: • Component/actuator faults: • sensor faults:
www.impact-tek.com 10© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Nonlinear FDI: Fault Models• Fault time profiles :
• Sensor fault • F is the known fault distribution vector, • θ(t) is the unknown magnitude of the time--varying sensor biasθ(t) is the unknown magnitude of the time varying sensor bias.
• Process fault :For the purpose of fault isolation, we consider a fault class given byp p g y
•
•
estimated by the p-th FIEpiθ
www.impact-tek.com 11© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Adaptive thresholds are appropriately designed such that only residualsof the corresponding FIE always remain below.
Nonlinear FDI: Assumptions
• Assumption 1. The modeling uncertainties are unstructured but bounded by given known functions, i.e.,
| ( , , ) | ( , , )x u t y u tη η≤ | ( , , ) | ( , , )d x u t d y u t≤
Based on worst case scenario of normal engine degradation in a
M d li / t ( d th i b d ) d l d
Based on worst case scenario of normal engine degradation in anumber of flights
• Modeling/measurement errors (and their bounds) are modeled as a function of state and input.
• Enables the implementation of adaptive thresholds.• Improves FDI performance for transient scenarios
• Assumption 2. The rates of change of the component fault parameter vector and the sensor bias are uniformly bounded respectively
www.impact-tek.com 12© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
vector and the sensor bias are uniformly bounded, respectively.
Fault Detection Method• Fault detection estimator
• Fault detection decision schemeA fault is detected when at least one component of the modules of the output estimation error exceeds its corresponding threshold given byestimation error exceeds its corresponding threshold given by
Implemented as:
• Method based on nonlinear adaptive estimate techniques• Reduced to Kalman Filtering approach in case of a linear system• Adaptive threshold is a function of states/outputs and inputs
www.impact-tek.com 13© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Adaptive threshold is a function of states/outputs and inputs
Fault Isolation Method• Sensor fault isolation estimators
• Adaptive Thresholds for fault isolationAdaptive Thresholds for fault isolation
• Fault parameters (severity) are estimated by the FIEsAd ti th h ld i f ti f t t / t t d i t
www.impact-tek.com 14© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
• Adaptive threshold is a function of states/outputs and inputs
Engine Application Using C-MAPSSg e pp cat o Us g C SS
Fan
BypassNozzleInlet
VBV
HPC HPT
CoreNozzle
FuelVSV
VBV
Combustor
LPC LPT
AmbientConditions
Core Shaft
Fan Shaft
20 2421 4130 48 501310 70 90
(state)(state)(state)
www.impact-tek.com 15© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Simulation Results
www.impact-tek.com 16© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Structure of the Fault Detection and Isolation E ti t (I l t d i Ph I C MAPSS)Estimators (Implemented in Phase I on C-MAPSS)
Engineu y
FDEFault Detection
Decision Scheme(w/ adaptivethresholds)
FIE #1 (Nf)
thresholds)activation
FIE #2 (Nc)
FIE #3 (T24)Fault Isolation
Decision Scheme
FIE #4 (Wf)
FIE #5 (VSV)
(w/ adaptivethresholds)
www.impact-tek.com 17© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
FIE #6 (HPC)
Fault Isolation Logic (Nf Sensor Bias Example)
00FIE1 (Nf) OR
100
FDE OR1
trigger00
FIE1 (Nf)
100
FIE2 (Nc) OR0 Nf f lt
Isolation indicator
0
110
FIE3 (T24) OR
0
1
1
Nf sensor fault
Isolation indicator:
Only one “0”: fault isolated100
FIE4 (Wf) OR1
1
Only one “0”: fault isolatedNo “0”: unknown faultMultiple “0”s: fault not isolable
100
FIE5 (VSV) OR
1
1
www.impact-tek.com 18© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
100
FIE6 (HPC) OR
Fault Scenarios & Implementationau t Sce a os & p e e tat o
• 6 faults (bias)Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) (ver. 2.0) Plot
resluts
• Sensors• Nf• Nc T 2_sens
Nc_sens
Nf_sens
imevec .DVSV' datavec .DVSV0:123
[timevec .Wf' datavec .Wf']0:326
[timevec .VBV' datavec .VBV']0:342
[timevec .TRA ' datavec .TRA ']0:543
0:56
mevec .DTamb ' datavec .DTamb0:55
[timevec .Alt ' datavec .Alt ']0:54 altitude
dTamb
Nf _sens
Nc_sens
T2_sens
TRA
DVSV
VBV
Wf cmd
• T24• Actuators
• WfVSV
OL / CL switch
OL Wf cmd
CL Wf cmdWf cmd
Ps30_sens
P2_sens
T 48 _sens
T 24 _sens
T 48 sens
T 24 _sens
T 2_sens
Nf_sens
Nc_sens
[timevec .MN' datavec .MN']0:56 Mach
Wf cmd
VSV cmd
VBV cmd
T24_sens
T48_sens
P2_sens
Ps30_sens
Nf
Nc
T2
T24
T48
VSV cmd
• VSV• Component
• HPCHealth Parameters
P50 _sensT 48 _sens
Ps30_sens
P2_sens
P50 _sens
Nf _sens
Nc_sens
T2_sensaltitude
T24 sens
Engine
VBV cmd P50_sens
Controller
T48
Ps30
P2
P50
VBV cmd
Simulation Outputs Output Scopes Initializer
T24_sensT48_sens
dTamb
P2_sensPs30_sens
P50_sens
MachWf cmd
VSV cmd
www.impact-tek.com 19© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
FDI
VBV cmd
Sensor Noise and Dynamics ModuleSe so o se a d y a cs odu e
Sensor Variance SLS Full value
Gaussian Noise Variance:
Nf 0.0 RPM2 2388
Nc 0.0 RPM2 9048
T2 1.0 degree R2 519 • Variance values provided by GRCT2 1.0 degree R 519
T24 5.0 degree R2 642
T48 5.0 degree R2 2070
P2 0.001 PSI2 14.6
• Small variance due to raw data processing/filtering
• Algorithm works with larger noise
Ps30 8.0 PSI2 523
P50 0.08 PSI2 19
Time constant = 1; time step = 0 015Parameters for sensor dynamics: Time constant = 1; time step = 0.015Parameters for sensor dynamics:
www.impact-tek.com 20© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Transient and Degradation Scenarioa s e t a d eg adat o Sce a o
• PLA is ramped from 60o to 69o in 1 second. After steady-state operation, PLA is ramped back to 60o in 1 second. The altitude is 25000 ft, and the Mach number is 0.62.
• Simulated engine degradation: 0.2% degradations in LPC efficiency, LPC flow capacity, HPC efficiency, and HPC flow capacity
www.impact-tek.com 21© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Fault Scenario 1: Nf sensor fault
20
40 residual-Nfthreshold-Nf
20
25F IE 1 (N f) b ias es t im at ion
• Bias = 20 (~1%); fault time = 35;Nf sensor bias estimation (from FIE1)Fault detected (Td=35.1)
0 10 20 30 40 50 60 70 800
0
10
20 residual-Ncthreshold-Nc 10
15
0 10 20 30 40 50 60 70 800
0 10 20 30 40 50 60 700
5
10
15FDE residuals
residual T24threshold
35 40 45 50 55 60 65 70 75 800
5
Thresholds change as operating condition changes0 10 20 30 40 50 60 70
0
10
20
30FIE 2 (NC)
residual-Nfthreshold-Nf
0
10
20
FIE 1 (Nf)
residual-Nfthreshold-Nf
X
35 40 45 50 55 60 65 70 75 800
35 40 45 50 55 60 65 70 75 800
50
100
150
residual-Ncthreshold-Nc
35 40 45 50 55 60 65 70 75 800
35 40 45 50 55 60 65 70 75 800
20
40
60
residual-Ncthreshold-Nc
www.impact-tek.com 22© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
35 40 45 50 55 60 65 70 75 80
35 40 45 50 55 60 65 70 75 800
5
10
residual T24threshold
35 40 45 50 55 60 65 70 75 80
35 40 45 50 55 60 65 70 75 800
5
10
residual T24threshold
FIE 3 (T24) FIE 4 (Wf)
Fault Scenario 1: Nf sensor fault (cont’d)X
35 40 45 50 55 60 65 70 75 800
20
40FIE 3 (T24)
residual-Nfthreshold-Nf
20
35 40 45 50 55 60 65 70 75 800
20
40FIE 4 (Wf)
Nf: residualNc: threshold
40
XX
35 40 45 50 55 60 65 70 75 800
10
20
residual-Ncthreshold-Nc
20 residual-T24
35 40 45 50 55 60 65 70 75 800
20
40
Nc: residualNc: threshold
10
id l T24
35 40 45 50 55 60 65 70 75 800
10
residual-T24threshold-T24
FIE 5 (VSV)
35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
XX
35 40 45 50 55 60 65 70 75 800
20
40FIE 6 (HPC)
Nf: residualNf: threshold
FIE 635 40 45 50 55 60 65 70 75 80
0
20
40FIE 5 (VSV)
Nf: residualNf: threshold
40
XX
35 40 45 50 55 60 65 70 75 800
20
40FIE 6
Nc: residualNc: threshold
35 40 45 50 55 60 65 70 75 800
20
40
Nc:residualNc: threshold
www.impact-tek.com 23© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
Fault Scenario 2: Wf actuator faultBi 200 ( 3%) f lt ti 35
5
10 residual-Nfthreshold-Nf
• Bias = -200 (~3%); fault time = 35 Wf actuator bias estimation (from FIE4)Fault detected (Td=36.1)
5 0F IE 4 (W f) b ia s e s t im a t io n
0 10 20 30 40 50 60 70 800
0 10 20 30 40 50 60 70 800
20
40
residual-Ncthreshold-Nc
-1 0 0
-5 0
0
0 10 20 30 40 50 60 70 80
0 10 20 30 40 50 60 700
5
10
15FDE residuals
residual T24threshold 3 5 4 0 4 5 5 0 5 5 6 0 6 5 7 0 7 5 8 0
-2 0 0
-1 5 0
35 40 45 50 55 60 65 70 75 800
10
20
FIE 1 (Nf)
residual-Nfthreshold-Nf
35 40 45 50 55 60 65 70 75 800
100
200FIE 2 (NC)
residual-Nfthreshold-NfX
35 0 5 50 55 60 65 0 5 80
35 40 45 50 55 60 65 70 75 800
20
40
60
residual-Ncthreshold-Nc
35 0 5 50 55 60 65 0 5 80
35 40 45 50 55 60 65 70 75 800
5
10x 10
9
residual-Ncthreshold-Nc
X
www.impact-tek.com 24© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data35 40 45 50 55 60 65 70 75 80
0
5
10
residual T24threshold
35 40 45 50 55 60 65 70 75 800
5
10
residual T24threshold
Fault Scenario 2: Wf actuator fault (cont’d)
35 40 45 50 55 60 65 70 75 800
5
10
15FIE 4 (Wf)
Nf: residualNc: threshold
35 40 45 50 55 60 65 70 75 800
5
10FIE 3 (T24)
residual-Nfthreshold-Nf
X
35 40 45 50 55 60 65 70 75 800
20
40
Nc: residualNc: threshold
35 40 45 50 55 60 65 70 75 800
10
20
30
residual-Ncthreshold-Nc
35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
10
20
residual-T24threshold-T24
0
10
20FIE 5 (VSV)
Nf: residualNf: threshold
35 40 45 50 55 60 65 70 75 800
5
10FIE 6 (HPC)
Nf: residualNf: threshold
XX
35 40 45 50 55 60 65 70 75 800
35 40 45 50 55 60 65 70 75 800
20
40
Nc:residualNc: threshold
35 0 5 50 55 60 65 0 5 80
35 40 45 50 55 60 65 70 75 800
20
40FIE 6
Nc: residualNc: threshold
www.impact-tek.com 25© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
5
10
residual-T24threshold-T24
Fault Scenario 3: HPC component fault• Efficiency loss = 0 02; Flow capacity loss = 0 02; fault time = 35
5
10 residual-Nfthreshold-Nf
Efficiency loss 0.02; Flow capacity loss 0.02; fault time 35Fault severity estimation (from FIE6)Fault detected (Td=35.7)
0F IE 6 (H P C ) b ias es t im at ion
0 10 20 30 40 50 60 70 800
0
20
40 residual-Ncthreshold-Nc -0 .015
-0 .01
-0 .005
0 10 20 30 40 50 60 70 800
0 10 20 30 40 50 60 700
5
10
15FDE residuals
residual T24threshold
35 40 45 50 55 60 65 70 75 80-0 .03
-0 .025
-0 .02
35 40 45 50 55 60 65 70 75 800
10
20FIE 2 (NC)
residual-Nfthreshold-Nf
35 40 45 50 55 60 65 70 75 800
10
20
FIE 1 (Nf)
residual-Nfthreshold-Nf
X35 0 5 50 55 60 65 0 5 80
35 40 45 50 55 60 65 70 75 800
50
100
150
residual-Ncthreshold-Nc
35 40 45 50 55 60 65 70 75 80
35 40 45 50 55 60 65 70 75 800
20
40
60
residual-Ncthreshold-Nc
X
www.impact-tek.com 26© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data35 40 45 50 55 60 65 70 75 80
0
5
10
residual T24threshold
35 40 45 50 55 60 65 70 75 800
5
10
residual T24threshold
FIE 4 (Wf)FIE 3 (T24)
Fault Scenario 3: HPC faultX
35 40 45 50 55 60 65 70 75 800
5
10
15FIE 4 (Wf)
Nf: residualNc: threshold
40
35 40 45 50 55 60 65 70 75 800
5
10FIE 3 (T24)
residual-Nfthreshold-Nf
40
X X
35 40 45 50 55 60 65 70 75 800
20
Nc: residualNc: threshold
10
residual T24
35 40 45 50 55 60 65 70 75 800
20
residual-Ncthreshold-Nc
20 residual-T24
FIE 6 (HPC)15FIE 5 (VSV)
35 40 45 50 55 60 65 70 75 800
5
0
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
10
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
5
10FIE 6 (HPC)
Nf: residualNf: threshold
40FIE 6
35 40 45 50 55 60 65 70 75 800
5
10
15
Nf: residualNf: threshold
40
X
35 40 45 50 55 60 65 70 75 800
20
40
Nc: residualNc: threshold
10
residual T24
35 40 45 50 55 60 65 70 75 800
20
Nc:residualNc: threshold
10
residual-T24
www.impact-tek.com 27© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
35 40 45 50 55 60 65 70 75 800
5
residual-T24threshold-T24
35 40 45 50 55 60 65 70 75 800
5
threshold-T24
Adaptive Engine Model with Neural Networks
(1) The NN learns(1) The NN learns (2) NN becomesa part of
(3) Updated model
(4) Updated residual signals
FDE
(4) Updated residual signals
u y+-
FIEs -+
www.impact-tek.com 28© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Note that the NN being trained online is NOT used for FDIE.
Neural Networks Enhancement –Si l ti R ltSimulation Results
• Simulation Scenario:FanFlow loss: 0.3%; LPCFlow loss: 0.3%; HPCEff loss: 0.3%; HPCFlow loss: 0.3%
7comparions of fault detection residuals - Nf
7comparions of fault detection residuals - Nc
5
6
residual with NNresidual without NN
5
6
residual with NNresidual without NN
2
3
4
2
3
4
0 10 20 30 40 50 60 70 800
1
time (sec)
0 10 20 30 40 50 60 70 80
0
1
time (sec)
www.impact-tek.com 29© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Minimal detectable/isolable fault sizeMinimal detectable/isolable fault size
Fault TypeMinimum Detectable Fault Size (% of SLS)
Minimum Isolable Fault Size (% of SLS)
Estimated Fault Size
Nf sensor bias 5 (0.21%) 10 (0.42%) 11.2
Nc sensor bias 16 (0.18%) 25 (0.28%) 23.7
T24 sensor bias 5 (0.78%) 11 (1.6%) 11.1
Wf actuator bias ‐120 (0.49%) ‐125 (0.52%) ‐110
VSV actuator bias ‐0.20 (3.39%) ‐0.25 (4.24%) ‐0.21
HPC efficiency & flow capacity loss ‐0.009 (0.9%) ‐0.015 (1.5%) ‐0.016
Note:1. Tested on C-MAPSS simulator2. Tested with the selected transient scenario only
www.impact-tek.com 30© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
3. Tested with the selected noise level only4. Extensive tests will be conducted in Phase 2
Future Workutu e o
• Extend the fault modes to cover more components and sensors• Further development of on-board Neural Networks-based adaptive
engine model• Further system development to cover the full flight envelopey p g p• Comparative studies with other fault diagnostics methods
• Kalman filter• Sliding mode observerSliding mode observer
• Extensive software-in-the-loop testing and performance evaluation• Further development on OEM engine model
R l ti h d i th l t ti d f l ti• Real-time hardware-in-the-loop testing and performance evaluation with OEM support
www.impact-tek.com 31© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
Related publicationsp
• X. Zhang, M. M. Polycarpou, and T. Parisini, "Design and analysis of a fault isolation scheme for a class of uncertain nonlinear systems " IFAC Annual Reviews in Control pp 107-121 2008
Liang Tang, Xiaodong Zhang, Jonathan DeCastro and Donald L. Simon, A Unified Nonlinear Adaptive Approach for Detection and Isolation of Engine Faults, submitted to ASME TURBO EXPO 2010, June 14-18, 2010, Glasgow, UK.
uncertain nonlinear systems, IFAC Annual Reviews in Control, pp. 107 121, 2008.• X. Zhang, T. Parisini, and M. M. Polycarpou, "A sensor bias fault isolation scheme for a class of nonlinear
systems", IEEE Transactions on Automatic Control, vol. 50, no. 3, pp. 370- 376, March 2005.• X. Zhang, T. Parisini, and M. M. Polycarpou, "Adaptive fault-tolerant control of nonlinear systems: a diagnostic
information-based approach ", IEEE Transactions on Automatic Control, vol. 49, no.8, pp. 1259-1274, August 2004, (Nominated for the Axelby Best Paper Award).
• X Zhang M M Polycarpou and T Parisini "A robust detection and isolation scheme for abrupt and incipientX. Zhang, M. M. Polycarpou, and T. Parisini, A robust detection and isolation scheme for abrupt and incipient faults in nonlinear systems", IEEE Transactions on Automatic Control, vol. 47, no. 4, pp. 576-593, April 2002.
• X. Zhang, M. M. Polycarpou, and T. Parisini, "Robust fault isolation of a class of nonlinear input-output systems", International Journal of Control, vol. 74, no.13, pp. 1295-1310, September 2001.
• DeCastro, J.A., Litt, J.S., and Frederick, D.K., “A Modular Aero-Propulsion System Simulation of a Large Commercial Aircraft Engine”, Proceedings of the 44th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Hartford, CT, July 21–23, 2008Exhibit, Hartford, CT, July 21 23, 2008
• DeCastro, J.A., “Rate-Based Model Predictive Control of Turbofan Engine Clearance,” J. Propulsion and Power, Vol. 23, No. 4, July–August 2007.
• L. Tang, M. Roemer, S. Bharadwaj, and C. Belcastro, “An Integrated Aircraft Health Assessment and Fault Contingency Management System, AIAA Guidance,” Navigation and Control Conference and Exhibit, Honolulu, Hawaii, 18 - 21 Aug 2008.
• L. Tang, M. J. Roemer, and G. J. Kacprzynski, “Real-Time Health Estimation And Automated FaultL. Tang, M. J. Roemer, and G. J. Kacprzynski, Real Time Health Estimation And Automated Fault Accommodation For Propulsion Systems,” GT2007-27862, Proceedings of ASME Turbo Expo 2007 Power for Land, Sea and Air, May 14-17, 2007, Montreal, Canada.
• L. Tang, M. J. Roemer, G. J. Kacprzynski, and J. Ge, “Dynamic Decision Support and Automated Fault Accommodation for Jet Engines,” IEEE Aerospace Conference, Big Sky, Montana. Mar. 3-10, 2007.
• L. Tang, A. Saxena, M. E. Orchard, G. J. Kacprzynski, G. Vachtsevanos, and A. Patterson-Hine, “Simulation-based Design and Validation of Automated Contingency Management for Propulsion Systems,” IEEE
www.impact-tek.com 32© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data
g g y g p y ,Aerospace Conference, Big Sky, Montana. Mar. 3-10, 2007.
• M. J. Roemer, L. Tang, G. J. Kacprzynski, J. Ge, and G. Vachtsevanos, “Simulation-Based Health and Contingency Management,” IEEE Aerospace Conference, Big Sky, Montana. Mar. 4-11, 2006.
Thank you!Thank you!
Acknowledgement:
This work was supported by NASA under SBIR Phase I Contract NNX09CC71P The authors acknowledge theContract NNX09CC71P. The authors acknowledge the contributions of Don Simon and Jeffrey Armstrong from NASA Glenn Research Center and Dr. Al Volponi from Pratt & Whitney for his support on this program.
www.impact-tek.com 33© 2009, Impact Technologies, LLC – FAR 52.227-14(h) Rights in Data