運用混合演算法於田口動態特性之參數設計最佳化

Abstract

田口玄一所提倡的參數設計，主要目的是要決定產品或製程的因子水準等設定值，使其對雜音變數的敏感性最小。但多數侷限於單一品質特性之靜態特性問題，對於連續型參數無法求取真正的最佳參數解。 本研究應用混合演算法的求解程序，來協助解決田口動態特性的多品質特性問題。首先，利用類神經網路模擬出控制因子與反應值的關係，再利用啟發式演算法求取最佳參數設定值。對於多重品質特性最佳化問題，利用望想函數或指數望想函數綜合衡量多個品質特性，以克服田口方法無法考量多重品質特性同時最佳化的缺點。室內無線區域網路傳輸品質評估系統的實際案例，說明方法之可行性。本研究之結果，將可協助區域網路工程師減少網路規劃所耗費的時間與判斷上的難度。

Parameter design is critical to enhancing a system’s robustness by identifying specific control factor and their levels that make the system less sensitive to noise. Most engineers applied Taguchi methods to optimize parameter design. However, Taguchi methods can only obtain the discrete optimal solution. They cannot identify the real optimum when the parameter values are continuous. The multi-response problem is too difficult to be considered by engineers with limited statistical knowledge. This research proposes a hybrid approach for combining neural networks and meta-heuristic algorithm to optimize the continuous parameter design problem. First, neural networks are used to simulate the relationship between the control factor values and corresponding responses. Second, meta-heuristic algorithm is employed to obtain the optimal parameter settings. The desirability function (or exponential desirability function) is utilized to transform the multiple responses into a single response. The real case of the propagation system evaluation of indoor wireless LAN is presented to demonstrate the practicability of the proposed procedure.