Research On Faults Modeling And Reasoning Methods For Stacker-cranes Based On Fault Tree And Bayesian Networks

Stacker-cranes have complex system structures, and present high randomness, uncertainty and other features when in operation. In this paper a new faults modeling and diagnosis method based on fault tree and Bayesian Networks was proposed to address the issues that exist in traditional diagnosis techniques.With the use of materials such as faults handbook, faults modeling is to analyze the common faults and classify them into four parts: travelling mechanisms failure, task interruption faults, communication modules failure and safety module faults. Then a fault tree whose top event is that stacker-crane could not work properly was build. The diagnosis and faults reasoning included two parts: faults monitoring and diagnosing reasoning. The former was to apply techniques such as PLC and OPC to monitor the key components of stacker-cranes, locate the fault fact when system fails and gives an alarm. The latter was to diagnose the faults with the hybrid mechanism of fault tree-based rule reasoning and Bayesian Networks-based probability calculation, and finally give a diagnosis report.The failure probabilities of events are the foundation to do rule-based reasoning as well as Bayesian Networks-based probability computing. Due to the fact that stacker-cranes are in the early stage of the whole life cycle currently, sufficient failure data of the system as well as its components are unavailable. Thus with the domain experts’ subject experience, for the uncertainty of basic events’ failure probabilities in a certain fault tree, the Fuzzy set theory was used to map experts’ linguistic judgements to the corresponding fuzzy numbers, solve them to get the wanted explicit values; for the child nodes’ conditional probabilities uncertainty caused by their parents’ multi-states in a Bayesian Network, the Subjective Bayesian method was utilized to do probabilities estimation.A prototype system of faults modeling and diagnostic inference for stacker-cranes was developed by using C# and SQL SERVER 2008, in which the modules of knowledge base management, facility performance monitoring and the diagnostic reasoning were emphasized. At last the communication module faults were taken to do case analysis. The results showed that our proposed method is effective.