First of all EVOP is much easier to apply than traditional experimental techniques such as response surfaces, factorial designs and even Taguchi methods. It is also more accurate. A second order polynomial is the best approximation to the response surface other methods can make. This assumption limits the accuracy of the final answer. Another problem with conventional methods is the complexity with the number of factors. The table below shows the minimum number of trials required for a response surface design as a function of the number of factors.
In addition, it usually takes several response surface designs (each modeling a smaller portion of the region) to obtain a satisfactory solution.
With sequential simplex optimization the number of trials in the initial simplex is k+1 where k is the number of experimental factors. Optimization is achieved by evaluating the slope of the immediate area and moving in the best direction.
Another advantage of EVOP is the effect on current production. The factor settings for the initial simplex can be set very close to current production settings. This allows the initial experimental trials to be run with either no impact or a limited impact on either scrap or downtime. Additional experimental trials improve existing production output over time. EVOP is a great way to pursue continuous improvement!
Traditional experimental techniques are difficult to use when yield is the response variable. How can conduct an experiment to reduce scrap for 1000 parts per million? At least 1000 trials are required to expect to find one defect. EVOP is perfect in this situation because the initial trials are conducted with factors setting very close to current production levels. This allows an experimental setting to be replicated for weeks if necessary because there is either an improvement in scrap or a slight degradation.
