1 ufrgs design of an embedded system for the proactive maintenance of electrical valves luiz f....
TRANSCRIPT
![Page 1: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/1.jpg)
1
UFRGS
Design of an Embedded System for the
Proactive Maintenance of
Electrical Valves
Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo S. Lubaszewski
Natal, August 31st to September 3rd, 2009
![Page 2: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/2.jpg)
2
OUTLINE
Introduction Proactive maintenance Fault detection and classification Embedded system design Conclusions
![Page 3: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/3.jpg)
3
INTRODUCTION
![Page 4: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/4.jpg)
4
INTRODUCTION
Recent advances in: Electronics Computing
Proactive ≠ corrective, preventive or predictive
To automate and integrate proactive(also know as intelligent) maintenance
tasks into embedded system
That are based either on post-failure correction or on off-line periodic
system checking
Focuses on fault prediction and diagnosis based on
component lifetimes and on system on-line monitoring
![Page 5: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/5.jpg)
5
MAINTENANCE
Systems
or equipments
On-line monitoring
Signals of vibration, temperature, torque, and others
Quantify the performance degradation and determine the
remaining system lifetime
Proactive maintenance
scheme
Signal processing
Statistical analysis
Artificial intelligent
![Page 6: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/6.jpg)
6
PROACTIVE MAINTENANCE
Embedded system for the proactive maintenance
Electrical valves
Model
Signals of torque and position
Quantify the degradation and diagnose the fault location
Oil distribution network
Proactive maintenance
scheme
Wavelet packet transform
Self-organizing maps
(prototyped using FPGAs)
![Page 7: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/7.jpg)
7
PROACTIVE MAINTENANCE
![Page 8: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/8.jpg)
8
MAINTENANCE SCHEME
Proposed system for electrical actuators
Mathematical model
Wavelet packet transformSelf-organizing maps
![Page 9: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/9.jpg)
9
MATHEMATICAL MODEL
Electrical actuator
Model
Main components Forces
Differential and algebraic equations
Fault injection
![Page 10: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/10.jpg)
10
SIGNAL PROCESSING
Wavelet packet transform Preserves timing and spectral information Suitable for the analysis of non-stationary signals Capability of decomposing the signal in frequency bands
Energy (spectral density) Torque Position
The energy is used by the self-organizing maps
Divided into N frequency bands
The WPT runs in a PC station during the
training plhase
During on-line testing, the WPT shall be part of the embedded system
![Page 11: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/11.jpg)
11
SELF-ORGANIZING MAPS (SOM)
SOM or Kohonen maps (class of neural networks) Unsupervised learning paradigm based on:
Competition (searcher the winner neuron) Cooperation (identifying the direct neighbors) Adaptation (updating the synaptic weights)
The goal of a SOM is after trained, mapping any input data from a Rn space representation into R2 lattice-like matrix
Energy vector
Synaptic weight vector
![Page 12: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/12.jpg)
12
SELF-ORGANIZING MAPS
Competition step (Euclidean distance):
Using these equation, searches in the map the neuron with the weights (WBMU) that best matches the energy vector (E)
Thus, two maps are trained: Fault detection: considering only typical situations of normal operation Fault classification: considering normal, degraded and faulty situations
The misbehaviors are simulated by injecting parameter deviations into the valve mathematical model
• For detection the min(Dkj), or the Quantization Error, is computed• For classification one map is colored based in the distance betwen W and E
![Page 13: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/13.jpg)
13
FAULT DETECTION AND CLASSIFICATION
![Page 14: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/14.jpg)
14
EXPERIMENTAL RESULTS
To train the fault detection map, a lot of simulations are performed to obtain typical values of torque and opening position under normal valve operation
To train the fault classification map, fault simulation is needed, some parameters are gradually incremented
KR simulate the degradation of the internal valve worm
gear, till it breaks
KM deviations simulate the elasticity loss of the valve
spring along time
100 operation cycles
![Page 15: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/15.jpg)
15
MODEL RESULTS
Fault simulation
0 10 20 30 40 50 60 70 80 90 1000
50
100
150
200
250
300
Time [s]
Tor
que
[Nm
]
Normal
Faulty
a)
• • •
•
•
•
•
•
• • • ••
•
•
•
•
• • • • •• • •
•
•
0 10 20 30 40 50 60 70 80 90 1000
10
20
30
40
50
60
70
80
90
100
Time [s]Po
sitio
n [%
]
Faulty
Normal
b)
• • • ••••
••
•• • •
••
•••• • • •
• • •
•
Torque Position
![Page 16: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/16.jpg)
16
DETECTION RESULTS
Fault detection
0 50 100 150 200 250 3000
0.01
0.02
0.03
0.04
0.05
Operation cycle
Qua
ntiz
atio
n er
ror
a)
0 50 100 150 200 250 3000
0.01
0.02
0.03
0.04
0.05
Operation cycle
Qua
ntiz
atio
n er
ror
b)
Faulty KR Faulty KM
Fault free
Degradation
Fault
![Page 17: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/17.jpg)
17
CLASSIFICATION RESULTS
Fault classification map of faults in KR and KM
Normal (N)
Degraded KR (DR)
Degraded Km (DM)
Faulty Km (KM)
Faulty KR (KR)
SOM Visualization Neurons Labels
KM
KM
N
N
N
N
KR
KR
KR
KR
DM
DM
DM
DM
DM
DR
DR
DM
DM
DM
DM
DM
DM
DR
DR
DR
DR
DM
DM
DM
DM
DM
DM
DM
DM
DM
DR
DR
DR
M
M
Each cluster is assigned a
different color
During the on-line testing phase, a winner neuron computed for a measured input vector can be
easily located in this map
![Page 18: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/18.jpg)
18
EMBEDDED SYSTEM DESIGN
![Page 19: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/19.jpg)
19
EMBEDDED SYSTEM DESIGN
For the on-line monitoring, the computation of the input vector WPT, winner neuron and quantization error, shall be embedded into the valve
These functions was implement using an FPGA platform (XUP Virtex2 PRO Xilinx)
The WPT computation runs in a Microblaze processor synthesized for the FPGA device (runs at a 100MHz)
Virtex2 PRO
![Page 20: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/20.jpg)
20
ARCHITECTURES
Parallel higher consumption of area
Series greater consumption of time
Time is not so important
The area is limited
![Page 21: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/21.jpg)
21
OTHER SOLUTIONS
1. Takeshi Yamakawa, Keiichi Horio and Tomokazy Hiratsuka. Advanced Self-Organizing Maps Using Binary Weight Vector and Its Digital Hardware Design. 9th International Conference on Neural Information Processing, v.3, 2002.
They proposed a new learning algorithm of the SOM (using Hamming distance and parallel processing) is proposed (processing time = 0.02s)
![Page 22: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/22.jpg)
22
OTHER SOLUTIONS
2. Paolo Ienne, Patrick Thiran and Nikolaos Vassilas. Modified Self-Organizing Feature Map Algorithms for Efficient Digital Hardware Implementation. IEEE Transactions on Neural Networks, v.8, n.8, 1997.
They describes three variants (parallel and series) of the SOM algorithm (update the weights after presentation of a group, or all, of input vectors)
![Page 23: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/23.jpg)
23
OTHER SOLUTIONS
3. Pino, B., Pelayo, F. J., Ortega, J. and Prieto, A. Design and Evaluation of a Reconfigurable Digital Architecture for Self-Organizing Maps. 7th International Conference on Microelectronics for Neural, Fuzzy and Bio-Inspired Systems, Granada, p.395–402, Apr.,1999.
They presents a modular and reconfigurable digital architecture for SOM algorithm (using Manhattan distance)
![Page 24: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/24.jpg)
24
ALGORITHM
for k=1 to K / (input vector numbers)for j=1 to J / (neuron numbers)
C1=C2=C3=0for n=1 to N / (element numbers by vector)
C1 = Enk - Mn
j
C2 = C1 * C1
C3 = C3 + C2
end
Dkj = √ C3 / (vector of J lines)end
[value , position] = min(Dkj)end
Min (Dkj) MBMU
![Page 25: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/25.jpg)
25
ARCHITECTURE
IP-core architecture proposed for the computation of WBMU e min Dkj and the hardware used to obtain Dkj :
Most important and most area consuming block in the FPGA
is the memory (BRAM)
The operator of the Euclidean distance circuity shall be based on a floating point representation (IEEE
754)
![Page 26: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/26.jpg)
26
AREA EVALUATION
Experiments performed with this system have show that map sizes of 90 neurons, with synaptic weight vectors of 20 elements of 32 bits each, are quite to reach the accuracy required
FPGA area evaluation
![Page 27: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/27.jpg)
27
PERFORMANCE EVALUATION
The gate valve takes 100s to travel the full range, from the closed to the 100% open position
The whole travel is performed in 20 incremental steps, lasting 5s each
WBMU and min Dkj computation time for a single (Th, a) measurement:
In this application, t=5s is the time
limit for all computations
![Page 28: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/28.jpg)
28
CONCLUSIONS
![Page 29: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/29.jpg)
29
CONCLUSIONS
Proactive maintenance scheme is proposed for the detection and diagnosis of faults in electrical valves, used for flow control in an oil distribution network
This is the first attempt to apply a proactive maintenance methodology to this sort of actuators
A hardware implementation of self-organizing maps is proposed to solve the valve maintenance problem
An embedded system implements these maps for the detection and classification of faults
The maps are trained using a mathematical model (fault injection)
The embedded system computes the best matching
between an acquired measure and the neurons of the map
![Page 30: 1 UFRGS Design of an Embedded System for the Proactive Maintenance of Electrical Valves Luiz F. Gonçalves, Jefferson L. Bosa, Renato V. B. Henriques, Marcelo](https://reader036.vdocuments.us/reader036/viewer/2022062408/56649e965503460f94b9a016/html5/thumbnails/30.jpg)
30
CONCLUSIONS
The embedded system was prototyped using an XUP Virtex2 PRO Xilinx FPGA Development Board
The results obtained for memory, requirements, area and performance point out to a low cost, promising solution for embedding maintenance in valves
Acknowledgements CNPq CAPES Petrobrás