應用類神經網路預測安全閥彈簧之可靠度

Abstract

本論文以失效分析(FMEA)推斷安全閥容易發生失效的零件，並應用類神經網路(Neural Network)的實驗數據模型的模擬預測能力，建立彈簧安全閥的壽命預測，以期能用低於傳統可靠度實驗數據的資料量亦能求得準確的結果，即為在不失去準確性的情況下，縮短可靠度實驗數據資料建立時間的方法，以期能使可靠度分析更容易普及於工程的各個環節中，以提升工程的品質，良率，安全性。

This paper is use the FMEA to analyze the Safety Valves' failure effects to determine which components are easier to fail. And use Neural Network's data model prediction ability to establish Safety Valves' MTTF. Except the method to earn the accurate result without so much testing like traditional Reliability prediction. That is the way to shorten the time of establish of Reliability Testing database without lose precision, then we can make Reliability Analysis more universal and widely use in every part of engineering and upgrade the quality, yield rate and safety.