| Summary: | 博士 === 國立交通大學 === 工業工程與管理系 === 91 === A precise product reliability prediction model can provide useful information enhance product quality and reduce product cost for manufacturers. Current analyzing and forecasting methods have following three drawbacks: first, complete failure data is required. Otherwise, statistical methods must be employed for analyzing the incomplete field data. Second, a large amount of historical failure data is required. Third, the seasonal effect of failure data was not considered by almost all existing methods. This study proposes three methods for analyzing and predicting reliability for repairable systems. The proposed methods require only number of units and number of repairs in unit time to predict the failure data. The first method utilizes grey system theory to construct a new predictive model for failure data. This model can forecast system’s reliability with just few historical data. The second method utilizes the seasonal autoregressive integrated-moving average（SARIMA）model to build the predictive model for failure data. The second model considers the seasonal effect of the field failure data. The third method constructs the predictive model by combining SARIMA model and neural network model. The third model can not only analyze the trends and seasonal vibration of the data, but also forecast the short and long term reliability of the system using only a small amount of historical data. Finally, a real case of the repairable system is presented to illustrate the feasibility and effectiveness of the proposed methods.
|
|---|
