The Optimal Choice of Threshold Value for Taguchi Continuous-Digital Dynamic System
The continuous-digital dynamic system defined by Taguchi are problems in which the input signals are
continuous and the output responses are discrete (digital) with two misclassification probabilities. For example, a
temperature control circuit provides a way of setting the target temperature. It compares the sensed temperature with the
target temperature and makes a decision about turning the heater ON or OFF. For this purpose, a threshold value is required
to discriminate or classify output responses into two categories: “ON” and “OFF”, or “accuracy” and “inaccuracy”. Taguchi
recommended calculating the equalized error rates based on the two errors with the same loss coefficient to achieve robust
design when the probability distributions of output responses are unknown. The paper proposes a general approach to
inferential estimation of the optimal threshold value by using the interval exponential regression to deal with unequal
variances in the Normal distributions model for the continuous-digital dynamic system based on the decision costs and
accuracy. The paper also presents a simulation study performed in order to assess the performance of the proposed
estimators. The maximum error of optimal threshold value is 0.005 compared with actual threshold value.
Keywords- Continuous-Digital Dynamic System, Threshold Value, Robust Design, Interval Exponential Regression.