Reasearchonscintillator Detectorfault Diagnosis Basedonsignalsfeature Extraction Classication

The nuclear detector is the source of acquiring the nuclear signals,its accuracy and precision are particularly important.Owing to the terrible working environment,the detector is prone to aging,failure and even break down.When they are in faults,it will cause the problems of the radiation fields safety as well as the occupational staff's safety.During the maintenance of nuclear detectors,the staff are required to have good psychological quality and professional skills.Based on the scintillator nuclear detector,this paper simulates the output signals in the normal and different faults of the scintillator detector.So that a fault diagnosis method based on feature classification is proposed.For the signal,when different parts of the scintillation detector were in malfunction,the eigenvectors of the output signals are also different.In the eigenvectors,how to classify the normal signal from the fault signals is the key point of this thesis.For a certain number of nuclear signals,the fault signal could be sort out automatically.The support vector machine algorithm is a learning algorithm based on statistical learning machine,can be regarded as a kind of learning machine,based on the training samples(including normal signals and fault signals)learning prediction and classification of unknown samples,so as to achieve the function of fault diagnosis.Finally,the scintillator detector signal,the measurement of pulse parameters,the extraction feature vector and the signal classification are simulated.The experimental results show that the method is effective for fault diagnosis.At the end of this paper,the problems and development direction of the fault diagnosis of nuclear detectors are discussed.