Docstoc

research and application of distributed fusion system based on

Document Sample
research and application of distributed fusion system based on Powered By Docstoc
					Research and Application of Distributed Fusion System Based on Grid Computing
Yu SU, Hai ZHAO, Wei-ji SU, Gang WANG, Xiao-dan ZHANG
School of Information Science & Engineering Northeastern University 110004 Shenyang, P.R. China webants@163.com

Abstract. Grid computing is used to realize the connection of distributed, heterogeneous computing resources through grid, assist in accomplishing computing task. Focusing on the weakness of centralized information fusion, put forward the disturbed fusion model utilizing grid computing node to realize local decision fusion on the basis of distributing information fusion architecture, and transform the decision to global fusion center to make global decision, and achieve more exact situation assessment and trend estimate, and utilize this model to equipment fault diagnosis system as an attempt of grid computing.

1 Introduction
Fusion is the integrating and processing of multi-sensor information gathered which adopt the technology of the computer to obtain the integrating and processing of the observe technology information [1]. Adopt distributed multi-sensor fusion system can obtain more concrete and more accurate estimation and determination. Grid computing accomplish share by connecting many disturbed computing resource with high-speed network in order to accomplish compute task together [2]. For this reason, it is an effective attempt to utilize the sharing resource ability of grid computing to accomplish disturbed fusion technology. In this paper using grid computing to accomplish disturbed fusion structure was discussed, and grid structure and disturbed fusion technology were researched.

2 Distributed Information Fusion Technology
Information fusion can integrate observed outside information and prophetic knowledge. Multi-sensor information fusion system fully utilized data gathered by sensors, through assembling mutually supplementary and redundant information on space and time according to the rule, and produce the intact coherence explanation and description to the observed target. Multi-sensor information fusion can improve

the capability of the multi-source heterogenic sensor system, and can eliminate information losing caused by the single sensor’s imperfect detection. According to different structure, disturbed information system can be classified three sorts: centralized fusion system, layered fusion system and horizontal fusion system. In them, centralized fusion system could flexibly select fusion algorithm, but transmission spending is very large; Local node of layered fusion system has fusion ability, it can alleviate the computing intensity of global fusion center, but local decision cannot reflect global status. In horizontal fusion system each local fusion node is parallel. The condition of other nodes could be considered while fusing, and so the local decision could reflect global status.

3 Grid System and its Architecture
Grid system is a synthetic architecture of hardware and software, and is composed of disturbed, heterogenic, different hardware and software system which can make resources share and utilize effectively, so that user can use this resource solve all kinds of complex computational problems pellucidly, effectively and unrestricted. Grid architecture can be looked on as a method that can decentralize essential module of system, appoint the purpose and function of system module, illuminate the mutual action among modules [3]. At present, there are two architectures: Forster’s five layers sandglass architecture and Open Grid Services Architecture (OGSA). Five layers sandglass architecture would disperse the operation and management of share resources according to the distance from each component to share resources. Different from five layers sandglass architecture that center on protocol, OGSA centers on service. In OGSA, service conception includes all kinds of compute resource, storage resource, network, program, database and so on. In short, everything is service. In order to make service thinking more tangible and more concrete, OGSA defined the conception-“Grid Service”. Grid Service provides a set of interface to follow specific tradition to settle service detection, dynamic service establishment, lifecycle management and information etc. The model of OGSA which centers on grid service, can achieve grid service through providing uniform kernel interface.

4 Realizing Distributed Fusion Arithmetic and Application
In distributed information fusion system, each local fusion node made decision on the local sensor data, and transmit the decision to the global center to make decision combination and global reasoning. In the grid computing based fusion system, adopt grid node instead of local fusion node to provide function of local fusion decision, and the key point is how to combine the decision criterion from local nodes. Considering the duality presumable H0 normal, H1 abnormal, the priori probabilities are P0 and P1. Each local fusion node make local decision, the decision criterion is, H={H0=true, f(xi)<β; H1=true, f(xi)≥β}.According to the criterion, the

local decision ui is, ui ={1,H1=true; 0,H0=true}. Global fusion center make decision on the local decision, δ=δ({ui}Ni=1). In them, δ is the probability of the decision result equaling to H1, and the global fusion decision result is, u ={1, p=δ; 0,p =1-δ}. The local decision is association with false alarm ratio and detection probability, detection probability Pd is the probability value of making H1 decision when the real result is H1, and false alarm ratio is the probability value of making H1 decision when the real result is H0. In the process of fusion, the usual method is to obtain the higher detection probability and lower false alarm ratio Pd, so Neyman-Pearson method is adopted. Stability criterion is,
⎧ ⎪1 ⎪ ⎪ ) = ⎨r ⎪ ⎪ ⎪0 ⎩ if if if

∑ ∑

M

u u u

δ = δ ({ u i } iN= 1

j =1 M j =1 M j=1

j

> K = K < K

(1)

j

∑

j

Therefore, global criterion is,
⎧ ⎪1 ⎪ ⎪ u = ⎨1 ⎪ ⎪ 0 ⎪ ⎩ p = 1 , if p = r , if

∑

M

u u

∑

j =1 M j =1

j

> K = K

(2)

j

other

Applying the distributed fusion structure to Jilin Fengman hydro-electric fault diagnosis system, setting multi-sensors to gather the gas parameters dissolved in oil of the transformer equipment, and the device situation is assessed according to the range of many parameters. In order to realize the function of grid computing, Globus software package and Web service interface is adopted to form a simple grid system. Factory Registry The grid system includes interfaces (Factory, Registry, GridService, Mapper Factory Mapper), which can offer and create temporary services, search services and Grid Grid obtain services, the service structure of Service Service the grid system shows in figure 1. Factory is the unit of providing grid computing, and Registry is the instance of registering Factory service, Mapper is used to map the grid computing to Physical Resource Environment the physical resource. Multi-sensors gather the gas parameter dissolved in Fig. 1. Service structure of grid system the oil, and by analyzing the component and the content of the gas dissolved in the oil, the serious degree

of the fault can be discovered, and also can bring forward the corresponding solution. Here utilizing the grid system to realize the computing node to make local decision.

5 Conclusion
Via the research of disturbed information fusion, put forward utilizing grid computing to realize the architecture of disturbed information fusion system. And after the comparison of the architecture and performance of five layers sandglass architecture and OSGA, adopting OSGA architecture and utilizing Globus software package and WebService to construct grid system realizing providing grid computing function. The distributed fusion model utilizing grid computing node to realize local decision fusion function, and transform the decision to global fusion center to make global decision, and achieve more exact situation assessment and trend estimate.

Acknowledgements
This research was sponsored by the Natural Science Fund of P.R.China under Grant No. 69873007.

References
1. He Y,Wang G H: Multisensor information fusion with applications [M].Beijing: Publishing House of Electronics Industry.2000 156-160. 2. Xiao N, Ren H, Gong S L: Resource directory technology based grid system software design and application. Computer Research and Development, 2002,92(8) 3. ZHAO H: Studying of FCS Architecture and Designing/Implementing of one Model System Based on Data Fusion Theory [D]. Shenyang: Northeastern University.1994.36-45 4. Hall D: Mathematical Techniques in multisensor data fusion [M]. Artech House Inc, Boston, London, 1992. 235–238 5. Foster I, Kesselman C: The Grid: Blueprint for a New Computing Infrastructure[D] . Morgan Kauf2man , 1999. 6. Xu L Y,Du X D: Application of neural fusion to accident forecast in hydropower station [A]. Proceedings of the Second International Conference on nformation Fusion[C]. 1999, 12:1166-1171


				
DOCUMENT INFO
Shared By:
Stats:
views:8
posted:12/18/2009
language:English
pages:4
Description: research and application of distributed fusion system based on