A New Robust Prediction Approach for Dual Response Surface Models

Project no. 32502 · Status: Kapanmış (Yayın bekleniyor) · 2024

Summary

<p class="ql-align-justify"><span style="color: rgb(0, 0, 0);">Robust&nbsp;parameter&nbsp;design&nbsp;is&nbsp;an&nbsp;important&nbsp;approach&nbsp;used&nbsp;to&nbsp;determine&nbsp;operating&nbsp;conditions&nbsp;that&nbsp;ensure&nbsp;system&nbsp;performance&nbsp;remains&nbsp;stable&nbsp;against&nbsp;variability.&nbsp;In&nbsp;response&nbsp;surface&nbsp;methodology&nbsp;(RSM),&nbsp;models&nbsp;that&nbsp;assume&nbsp;normality&nbsp;are&nbsp;commonly&nbsp;used&nbsp;to&nbsp;estimate&nbsp;the&nbsp;mean&nbsp;and&nbsp;variability&nbsp;of&nbsp;the&nbsp;response.&nbsp;However,&nbsp;when&nbsp;the&nbsp;underlying&nbsp;data&nbsp;deviate&nbsp;from&nbsp;normal&nbsp;distribution,&nbsp;these&nbsp;classical&nbsp;estimators&nbsp;may&nbsp;lead&nbsp;to&nbsp;unreliable&nbsp;conclusions.</span></p><p class="ql-align-justify"><span style="color: rgb(0, 0, 0);">In&nbsp;practical&nbsp;manufacturing&nbsp;environments,&nbsp;skewed&nbsp;or&nbsp;contaminated&nbsp;data&nbsp;structures&nbsp;are&nbsp;frequently&nbsp;encountered.&nbsp;Such&nbsp;situations&nbsp;may&nbsp;cause&nbsp;inaccurate&nbsp;estimation&nbsp;of&nbsp;both&nbsp;process&nbsp;mean&nbsp;and&nbsp;variability,&nbsp;resulting&nbsp;in&nbsp;suboptimal&nbsp;operating&nbsp;conditions.&nbsp;Therefore,&nbsp;developing&nbsp;estimation&nbsp;approaches&nbsp;that&nbsp;remain&nbsp;reliable&nbsp;under&nbsp;non-normal&nbsp;data&nbsp;structures&nbsp;has&nbsp;become&nbsp;an&nbsp;important&nbsp;research&nbsp;topic.</span></p><p class="ql-align-justify"><span style="color: rgb(0, 0, 0);">In&nbsp;this&nbsp;study,&nbsp;a&nbsp;robust&nbsp;modeling&nbsp;framework&nbsp;is&nbsp;proposed&nbsp;for&nbsp;dual&nbsp;response&nbsp;surface&nbsp;methodology&nbsp;in&nbsp;order&nbsp;to&nbsp;improve&nbsp;parameter&nbsp;estimation&nbsp;when&nbsp;the&nbsp;normality&nbsp;assumption&nbsp;is&nbsp;violated.&nbsp;The&nbsp;proposed&nbsp;approach&nbsp;incorporates&nbsp;the&nbsp;BSB2&nbsp;robust&nbsp;M-estimator,&nbsp;which&nbsp;reduces&nbsp;the&nbsp;influence&nbsp;of&nbsp;extreme&nbsp;observations&nbsp;and&nbsp;improves&nbsp;the&nbsp;stability&nbsp;of&nbsp;parameter&nbsp;estimates.&nbsp;By&nbsp;integrating&nbsp;this&nbsp;estimator&nbsp;into&nbsp;dual&nbsp;response&nbsp;surface&nbsp;modeling,&nbsp;more&nbsp;reliable&nbsp;predictions&nbsp;of&nbsp;process&nbsp;mean&nbsp;and&nbsp;variability&nbsp;can&nbsp;be&nbsp;obtained.</span></p><p class="ql-align-justify"><span style="color: rgb(0, 0, 0);">The&nbsp;performance&nbsp;of&nbsp;the&nbsp;proposed&nbsp;methodology&nbsp;is&nbsp;evaluated&nbsp;through&nbsp;a&nbsp;printing&nbsp;experiment&nbsp;data&nbsp;that&nbsp;is&nbsp;frequently&nbsp;used&nbsp;as&nbsp;a&nbsp;benchmark&nbsp;example&nbsp;in&nbsp;the&nbsp;literature.&nbsp;The&nbsp;results&nbsp;demonstrate&nbsp;that&nbsp;the&nbsp;proposed&nbsp;approach&nbsp;provides&nbsp;more&nbsp;stable&nbsp;and&nbsp;reliable&nbsp;estimates&nbsp;compared&nbsp;with&nbsp;traditional&nbsp;methods,&nbsp;especially&nbsp;under&nbsp;non-normal&nbsp;data&nbsp;conditions.</span></p><p><strong style="color: rgb(0, 0, 0);">Keywords:</strong><span style="color: rgb(0, 0, 0);">&nbsp;Dual&nbsp;response&nbsp;surface&nbsp;methodology,&nbsp;Robust&nbsp;estimators,&nbsp;Non-normal&nbsp;data,&nbsp;BS82&nbsp;M-estimator,&nbsp;Confidence&nbsp;interval-based&nbsp;modeling</span></p>

View details