MODIFIED DESIRABILITY FUNCTIONS FOR MULTIPLE RESPONSE OPTIMIZATION

Desirability functions have been used extensively to simultaneously op timize several responses. Since the original formulation of these func tions contains non-differentiable points, only search methods can be u sed to optimize the overall desirability response. Furthermore, all re sponses are treated as equally important. We present modified desirabi lity functions that are everywhere differentiable so that more efficie nt gradient-based optimization methods can be used instead. The propos ed functions have the extra flexibility of allowing the analyst to ass ign different priorities among the responses. The methodology is appli ed to a wire bonding process that occurs in semiconductor manufacturin g, an industrial process where multiple responses are common.