Fault Diagnosis Of The Vibration Isolation Platform Based On Imprecise Probabilities

With the vigorous development of modern industry,people’s living and working conditions are becoming more inseparable with various types of machinery and equipment.As the mechanical equipment becomes more large-scaled and highly-automatized,once fail,they will be seriously damaged and cause grave harm.Therefore,studying on the fault diagnosis of the mechanical equipment has always been an important work.This paper investigates the existing fault diagnosis methods,in which one of the most widely-used methods is multi-sensor information fusion method because it reflects the system’s working states more comprehensively by using multi-source information.Data collected when fault occurs is not always fixed and immutable but contains substantial uncertainty.In order to obtain reliable diagnosis results from multi-source data processing analysis,DS evidence theory often used as its theoretical basis for uncertain reasoning.Through analyzing the common failure diagnosis framework based on DS evidence theory,it turns out that this diagnostic framework still has some deficiencies in evidence generation and decision making.As another uncertainty reasoning method,imprecise probability theory is a relatively more generic model,its expression of the uncertain information conforms more to our actual demands,and its data processing conforms more to our thinking habits,and its multiple decision criteria also apply to systems with different precision requirements.Based on this,this paper presents a fault diagnosis method based on imprecise probabilities,and verifies its feasibility with vibration isolation platform as object for experiment.The fault diagnosis framework based on imprecise probability theory is first constructed.In chapter 3,the method of generating the lower probabilities by modifying traditional evidence-generating method is improved.The diagnostic value function which can represent experts’ decision tendencies so as to transform failure diagnosis problems into comparison decision problems of the expected value of the cost functions under the lower probabilities is designed.The similarities and differences between the feature-level fusion and decision-level fusion are compared,and the feature-level fusion is chosen to fuse the data.Finally,the advantages and disadvantages of multiple decision criteria is analyzed by an example study.In chapter 4,the propulsion window is used to analyze the data,a great deal of analyses and comparison on different characteristic parameters by constructing multiple failures with the three-way sensor’s sampling data as the original data and the vibration isolation platform as the research object is conducted.The diagnoses of types of the fault are realized.In chapter 5,on the basis of the chapter 4,the proposed method is used to verify the diagnoses of time.When the characteristic parameters have poor quality,imprecise probability diagnosis method may cause a misjudgment,then the method of combination with DS evidence theory and the imprecise probability theory is proposed,this method can reduce misjudgment and timely find the fault.The result of the experiment shows that the proposed method is feasible and effective for fault diagnosis.