Study of satellite autonomous fault diagnosis is the great challenge and interest
in the field of satellite reliability research. Based on lots of related references,
combined with National Key 9th Five-Year pre-research project Fault Mode and
Demonstration System of Satellite (49.1.2) and 10th Five-Year pre-research project
Network and Component Technology of Satellite Fault Diagnosis (413200204-11),
adopting Multi-Agent technology, taking a satellite as object, the dissertation
expanded systematic and in-depth research with theory and practice on satellite fault
diagnostic technology based on Multi-Agent system and gave detail description.
The realizing of Agent autonomy and improvement of MAS structure were
in-depth studied. The BDIK(Belief, Desire, Intention and Knowledge) mental mode
and influence factor utility method, which can solve the problem of Agent
autonomous decide-making, were proposed. In order to solve the problem of task
conflict existing in Agent, task process relationship was stud ied. The series-wound 数据挖掘研究院
and shunt-wound structures of Agents, which can improve the reliability of MAS,
were presented. The diagnostic need was satisfied better through the study of half
static state coalition structure and supplement of KQLM performatives. The
theoretical foundations of satellite fault diagnosis based on Multi-Agent system were
set up.
Agent collaboration and founding of Multi-Agent diagnostic system model were
studied. SD (Static-Dynamic) task distributing method and object-oriented
negotia tion method were brought forward. These solved the problems of Agent tasks
distribution and conflict resolution. Considered the features of satellite, the model of
satellite autonomous fault diagnostic system based on MAS was founded. This
resolved the diagnostic difficult of satellite being a complicated system. The working
principle of MAS was studied. Constitution and function of manager and monitoring
Agent were researched. Study of signal denoising was carried out also. The 数据挖掘研究院
feasibility was proved by examples.
Model-based diagnostic Agent of thermal control system was studied. Main
fault modes and working modes of satellite thermal control system were analyzed.
These supplied the evidence for Agent perception. Diagnostic knowledge was
acquired through thermal system modeling and solution. Shallowing of deep
knowledge was realized. It supplied the diagnostic knowledge for Agent. The structure of autonomous diagnostic Agent was founded. The specific process of
Agent autonomous diagnosis and inference was given. Study indicated that
diagnostic Agent diagnoses exactly and strong autonomously.
Information fusion and its application in Multi-Agent diagnosis were studied.
D-S evidence theory was improved. It solved the problem of evidence’s deficiency.
Fuzzy membership grade, weights and D-S evidence theory were combined together
to fulfill Agent diagnosis. Subjective, objective knowledge of diagnosis and uncertain
information were unified successfully. Diagnostic autonomy and diagnostic ability of
uncertain information were increased. Diagnostic results’ conflicts were resolved
through applying the improved D-S evidence theory to fusion of Multi-Agent
diagnostic results. This improved diagnostic accuracy. The method’s feasibility was
proved by examples.
Development study of Multi- Agent satellite thermal control fault diagnostic
system was carried out. Combined with network technology, Multi-Agent satellite
thermal control fault diagnostic prototype system was developed. The process of
development was explained in detail. It has been indicated by diagnostic experiment
results of simulation and testing data from ground that Multi-Agent based diagnostic
method presented in this paper is feasible in increasing satellite diagnostic autonomy
and accuracy. The method has wide application prospect.
Keywords Fault diagnosis, Satellite, Multi-Agent System, Model, Fusion
