類神經網路在Gompertz修正模式上之應用與分析

Application and Analysis of Neural Network in A Modify Gompertz Model

高竟茵(Ching-Yin Kao)

指導教授 ： 陳雲岫

元智大學/工程學院/工業工程與管理學系/碩士(2002年)

摘要

自1962年Duane提出Duane可靠度成長模式後，學者一直偏重於探討可靠度成長模式的參數估計與提出新模式。傳統上，在參數估計方面常用統計學的方法分析可靠度成長模式，但因對非線性系統以及變數間有交互作用的系統統計方法之評估效果較差。Gompertz 模式為眾多可靠度成長模式中應用性較好，在1996年Keceecioglu針對Gompertz 的初期預測不精準提出一個修正模式，謂之Gompertz修正模式。本研究的具體目標為發展一個以類神經網路為基之預測系統，以克服Gompertz修正模式的在建構時之困難與耗時。我們採用平均絕對誤差(MAPE)比較預測效果。本研究中採用ReliaSoft Corporation網頁下的Gompertz範例進行模式驗證。本論文亦根據可靠度成長模式模擬數據，進行類神經網路建構可靠度成長模式之驗證，並與修正模式做比較。本研究在模擬模式假設可靠度的趨勢會依照Gompertz 模式，加入一隨機變數，以及採用不同線性遞減誤差比例尋找Gompertz 修正模式之最佳參數解。最後，由實例與模擬結果證明BPNN的建構系統比Modify Gompertz與Gompertz技術所建構的模式均有較好的預測能力。

關鍵字

可靠度成長模式 ； 倒傳遞類神經網路 ； Gompertz修正模式 ； Gompertz模式

並列摘要

The reliability growth model has been proposed by Duane since 1962. Many different approaches to estimate parameters or new models are proposed since them, Duane model is the most popular one to be applied to describe reliability. However, poor accuracy in the early stage is a drawback of the Gompertz model, thus Keceecioglu proposed a modified version of the Gompertz model in 1996. The trade-off of increasing precision in early stage for modified Gompertz model is complex solution and long computing time. In this thesis, we propose a neural network-based approach BPNN mosel to overcome the weakness of modify Gompertz model. The mean absolute error percent (MAPE) is used as a performance index. A Gompertz Data from web page of Reliabsoft Corporation is quoted to verify model our proposed model. Also the comparing among Gompertz, modified Gompertz and BPNN growth model is studied by simulated data. Simulation results show that BPNN growth model performs better than Modify Gompertz and Gompertz models.

並列關鍵字

Reliability Growth ； Back Propagation Neural Network ； Modify Gompertz Model ； Gompertz Model

參考文獻

3. 尹華誠，「類神經網路在產品研發階段可靠度管理模式之建構與分析」，碩士論文，元智大學工業工程與管理研究所， 2000。

11. Barlow R. E.and E. M. Scheuer, “Reliability Growth During a Development Test Program,” Technometrics, Vol. 8, 53-60, 1966.

12. Bhattacharyya G.K., A. Fries and R.A. Johnson, “Properties of Continuous Analog Estimators for a Discrete Reliability-Growth Model,” IEEE Transactions on Reliability, Vol. 38, No. 3, 1989.

13. Corcoran W.J, H. Weingarten and P.W. Zehna, “Estimating Reliability after Corrective Action,” Management Science, Vol. 10, No. 4, 786-798,1964.

14. Dark J.E, “「Discrete Reliability Growth Models Using Failure Discounting, 」 MS Thesis (unpublished), US Naval Postgraduate School, 1988.