A direct method for the determination of learning curve parameters from historical data

SUMMARY A direct method for estimating the parameters of the learning curve time constant model from historical data is presented. Compared with present methods, an assumption on asymptotic performance is unnecessary, and it is much simpler to implement than least error squares curve fitting. An example shows that these advantages are obtained without serious penalties in accuracy The method uses the experimental data to estimate the impulse response of the model. By expanding the model transfer function as an infinite series, the model parameters are obtained by summing the zero and first time moments of the impulse response, thus requiring simple multiplication and no iteration.