Methods Robustification

1 methods

1.1 experimental
1.2 analytical
1.3 numerical

methods

there 3 distinct methods of robustification, practitioner might use mix provides best in results, resources , time.

experimental

the experimental approach known. involves identification of variables can adjusted , variables treated noises. experiment designed investigate how changes nominal value of adjustable variables can limit transfer of noise noise variables output. approach attributed taguchi , associated taguchi methods. while many have found approach provide impressive results, techniques have been criticised being statistically erroneous , inefficient. also, time , effort required can significant.

another experimental method used robustification operating window. developed in united states before wave of quality methods japan came west, still remains unknown many. in approach, noise of inputs continually increased system modified reduce sensitivity noise. increases robustness, provides clearer measure of variability flowing through system. after optimisation, random variability of inputs controlled , reduced, , system exhibits improved quality.

analytical

the analytical approach relies on development of analytical model of system of interest. expected variability of output found using method propagation of error or functions of random variables. these typically produce algebraic expression can analysed optimisation , robustification. approach accurate model developed , can difficult if not impossible complex systems.

the analytical approach might used in conjunction kind of surrogate model based on results of experiments or numerical simulations of system.

numerical

in numerical approach model run number of times part of monte carlo simulation or numerical propagation of errors predict variability of outputs. numerical optimisation methods such hill climbing or evolutionary algorithms used find optimum nominal values inputs. approach typically requires less human time , effort other two, can demanding on computational resources during simulation , optimization.